1 2018-07-31 Martin Liska <mliska@suse.cz>
6 * profile.c (struct location_triplet): New.
7 (struct location_triplet_hash): Likewise.
8 (output_location): Do not output a BB that
9 is already recorded for a line.
10 (branch_prob): Use streamed_locations.
12 2018-07-31 Martin Liska <mliska@suse.cz>
15 * coverage.c (coverage_begin_function): Do not mark target
16 clones as artificial functions.
18 2018-07-31 Martin Liska <mliska@suse.cz>
24 * profile.c (branch_prob): Do not record GOTO expressions
25 for GIMPLE statements which locations are already streamed.
27 2018-07-31 Olivier Hainque <hainque@adacore.com>
29 * gcc.c (handle_spec_function): Accept a soft_matched_part
30 argument, as do_spec_1. Pass it down to ...
31 (eval_spec_function): Accept a soft_matched_part argument,
32 and pass it down to ...
33 (do_spec_2): Accept a soft_matched_part argument, and pass
35 (do_spec_1): Pass soft_matched_part to handle_spec_function.
36 (handle_braces): Update call to handle_spec_function.
37 (driver::set_up_specs): Update calls to do_spec_2.
38 (compare_debug_dump_opt_spec_function): Likewise.
39 (compare_debug_self_opt_spec_function): Likewise.
41 2018-07-31 Olivier Hainque <hainque@adacore.com>
43 * common.opt (nolibc): New option.
44 * doc/invoke.texi (Link Options): Document it.
45 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
46 * config/alpha/linux.h: Likewise.
47 * config/arc/elf.h: Likewise.
48 * config/arm/uclinux-elf.h: Likewise.
49 * config/arm/unknown-elf.h: Likewise.
50 * config/avr/avrlibc.h: Likewise.
51 * config/bfin/bfin.h: Likewise.
52 * config/bfin/linux.h: Likewise.
53 * config/bfin/uclinux.h: Likewise.
54 * config/darwin.h: Likewise.
55 * config/darwin10.h: Likewise.
56 * config/darwin12.h: Likewise.
57 * config/gnu-user.h: Likewise.
58 * config/lm32/uclinux-elf.h: Likewise.
59 * config/pa/pa-hpux11.h: Likewise.
60 * config/pa/pa64-hpux.h: Likewise.
61 * config/sparc/sparc.h: Likewise.
63 2018-07-31 Olivier Hainque <hainque@adacore.com>
65 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
68 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
71 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
74 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
76 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
77 terminated string literal.
79 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
81 PR rtl-optimization/85160
82 * combine.c (is_just_move): New function.
83 (try_combine): Allow combining two instructions into two if neither of
84 the original instructions was a move.
86 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
89 * doc/extend.texi (Global Register Variables): Discourage use of type
91 (Local Register Variables): Likewise.
93 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
95 PR tree-optimization/86506
96 * hwint.h (ceil_log2): Resync with hwint.c implementation.
98 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
101 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
102 hard_regno, make sure no insn between `from` and `to` clobbers it.
104 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
105 Tom de Vries <tdevries@suse.de>
107 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
108 (PTX_DEFAULT_RUNTIME_DIM): ... this.
109 (nvptx_goacc_validate_dims): Set default worker and gang dims to
110 PTX_DEFAULT_RUNTIME_DIM.
111 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
113 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
115 * config/pa/pa.c (pa_output_addr_vec): Align address table.
116 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
117 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
119 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
121 * config/rs6000/constraints.md (wG constraint): Delete, no longer
123 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
124 predicate to reflect toc fusion has been deleted.
125 (toc_fusion_mem_raw): Delete, no longer used.
126 (toc_fusion_mem_wrapped): Likewise.
127 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
129 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
130 Delete, no longer used.
131 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
132 meant to be used for toc fusion.
133 (rs6000_debug_print_mode): Delete toc fusion debugging.
134 (rs6000_debug_reg_global): Likewise.
135 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
136 fusion and secondary reload support that were never used.
137 (rs6000_option_override_internal): Delete TOC fusion, that was only
138 partially defined, and it did not work unless you also used the
140 (rs6000_legitimate_address_p): Delete TOC fusion support.
141 (rs6000_opt_masks): Likewise.
142 (fusion_wrap_memory_address): Delete function, no longer used.
143 (fusion_split_address); Delete TOC fusion support.
144 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
145 longer used with toc fusion being deleted.
146 (TARGET_TOC_FUSION_FP): Likewise.
147 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
149 (toc fusion spliter): Delete TOC fusion support.
150 (toc_fusionload_<mode>): Likewise.
151 (toc_fusionload_di): Likewise.
152 (fusion_gpr_load_<mode>): Delete generator function, this insn no
153 longer needs to be named. Rename predicate to delete TOC fusion.
154 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
155 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
156 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
157 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
158 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
160 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
162 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
163 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
164 __int128 in built-in function prototypes.
165 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
166 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
168 2018-07-27 Martin Sebor <msebor@redhat.com>
170 PR tree-optimization/86696
171 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
172 types, including enums.
173 (handle_char_store): Be prepared for the above function to fail.
175 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
177 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
178 when optimization level is lower than 2 or optimize for size.
180 2018-07-26 Martin Sebor <msebor@redhat.com>
182 PR tree-optimization/86043
183 PR tree-optimization/86042
184 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
185 (get_string_cst_length): Rename...
186 (get_min_string_length): ...to this. Add argument.
187 (handle_char_store): Extend to handle multi-character stores by
189 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
190 * tree.h (initializer_zerop): Add argument.
192 2018-07-26 Jakub Jelinek <jakub@redhat.com>
195 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
196 declare target to variables if they have always,{to,from,tofrom} map
199 2018-07-26 Martin Liska <mliska@suse.cz>
202 * lto-wrapper.c: Add linker_output as prefix
203 for ltrans_output_file.
205 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
207 PR rtl-optimization/85805
208 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
209 value for hard registers if that was written in the same mode.
211 2018-07-26 Martin Liska <mliska@suse.cz>
213 PR gcov-profile/86536
214 * gcov.c (format_gcov): Use printf format %.*f directly
215 and do not handle special values.
217 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
219 * common/config/arc/arc-common.c (arc_option_optimization_table):
220 Update default optimizations for size.
222 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
224 * config/arc/arc.md (movsf_insn): Add short instruction selection.
225 * config/arc/constraints.md (CfZ): New constraint.
226 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
227 (subsf3_fpu): Likewise.
228 (cmpsf_fpu): Likewise.
229 (cmpsf_fpu_uneq): Likewise.
231 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
233 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
235 (arc_is_uncached_mem_p): uncached applies to both the variable and
238 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
240 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
243 2018-07-25 David Malcolm <dmalcolm@redhat.com>
245 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
246 field "m_scopes" from vec to auto_vec.
248 2018-07-25 Martin Liska <mliska@suse.cz>
250 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
253 2018-07-25 Richard Biener <rguenther@suse.de>
256 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
257 special wrt context_die late.
258 (gen_subprogram_die): Re-use DIEs in local scope.
260 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
262 PR tree-optimization/86644
263 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
265 2018-07-25 Martin Liska <mliska@suse.cz>
268 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
269 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
271 2018-07-25 Martin Liska <mliska@suse.cz>
274 * params.def: Explain ASan abbreviation and provide
275 a documentation link.
277 2018-07-24 Martin Sebor <msebor@redhat.com>
279 PR tree-optimization/86622
280 PR tree-optimization/86532
281 * builtins.h (string_length): Declare.
282 * builtins.c (c_strlen): Correct handling of non-constant offsets.
283 (check_access): Be prepared for non-constant length ranges.
284 (string_length): Make extern.
285 * expr.c (string_constant): Only handle the minor non-constant
286 array index. Use string_constant to compute the length of
287 a generic string constant.
289 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
291 PR tree-optimization/86618
292 * tree-vect-stmts.c (vectorizable_call): Don't take the address
293 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
295 2018-07-24 David Malcolm <dmalcolm@redhat.com>
297 PR tree-optimization/86636
298 * json.cc (json::object::set): Fix comment. Add assertions.
299 (json::array::append): Move here from json.h. Add comment and an
301 (json::string::string): Likewise.
302 * json.h (json::array::append): Move to json.cc.
303 (json::string::string): Likewise.
304 * optinfo-emit-json.cc
305 (optrecord_json_writer::impl_location_to_json): Assert that we
306 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
307 wrapper around it. Expand the location once, rather than three
309 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
310 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
312 (optrecord_json_writer::optinfo_to_json): Likewise, in four
313 places. Fix some overlong lines.
315 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
317 * config/aarch64/aarch64-simd.md
318 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
319 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
320 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
321 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
322 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
323 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
324 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
325 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
326 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
328 2018-07-24 Jakub Jelinek <jakub@redhat.com>
331 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
332 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
333 and abs_d == d, do the power of two handling if profitable.
335 2018-07-24 Richard Biener <rguenther@suse.de>
337 * match.pd: Add BIT_FIELD_REF canonicalizations.
339 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
342 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
344 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
346 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
347 terminated STRING_CST object.
349 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
351 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
353 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
355 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
356 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
357 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
358 the elements into a register.
359 (rs6000_split_v4si_init_di_reg): Delete.
360 (rs6000_split_v4si_init): Delete.
361 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
362 (vsx_init_v4si): Rewrite as a define_expand.
364 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
366 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
367 zero_extend argument from memory): New.
369 2018-07-22 Martin Sebor <msebor@redhat.com>
372 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
373 diagnosing calls with unknown arguments unless -Walloca-larger-than
374 is restricted to less than PTRDIFF_MAX bytes.
376 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
378 * doc/gcov.texi (Invoking Gcov): Editorial changes.
380 2018-07-20 David Malcolm <dmalcolm@redhat.com>
382 * pretty-print.c (text_info::set_location): Remove redundant
383 "line_table" parameter from call to rich_location::set_range.
385 2018-07-20 Martin Sebor <msebor@redhat.com>
388 * builtins.c (expand_builtin_alloca): Adjust.
389 * calls.c (alloc_max_size): Simplify.
390 * cgraphunit.c (cgraph_node::expand): Adjust.
391 * common.opt (larger_than_size, warn_frame_larger_than): Remove
393 (frame_larger_than_size): Same.
394 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
395 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
397 * doc/invoke.texi (GCC Command Options): Document option arguments.
398 Explain byte-size arguments and suffixes.
399 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
400 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
401 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
402 * doc/options.texi (UInteger): Expand.
403 (Host_Wide_Int, ByteSize): Document new properties.
404 * final.c (final_start_function_1): Include sizes in an error message.
405 * function.c (frame_offset_overflow): Same.
406 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
407 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
408 Diagnose unbounded alloca calls only for limits of less than
410 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
411 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
413 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
414 only for limits of less than PTRDIFF_MAX.
415 * langhooks-def.h (lhd_handle_option): Change function argument
417 * langhooks.c (lhd_handle_option): Same.
418 * langhooks.h (handle_option): Same.
419 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
421 (var_type, var_type_struct): Same.
422 (var_set): Handle ByteSize flag.
423 * optc-gen.awk: Add comments to output to ease debugging. Make
424 use of HOST_WIDE_INT where appropriate.
425 * opts-gen-save.awk: Use %lx to format unsigned long.
426 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
427 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
428 arguments. Parse bytes-size suffixes.
429 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
430 (enum_value_to_arg): Same.
431 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
432 (handle_option): Adjust.
433 (generate_option): Change function argument to HOST_WIDE_INT.
434 (cmdline_handle_error): Adjust.
435 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
436 (set_option): Change function argument to HOST_WIDE_INT.
437 (option_enabled): Handle cl_host_wide_int.
438 (get_option_state): Handle CLVC_SIZE.
439 (control_warning_option): Same.
440 * opts.c (common_handle_option): Change function argument to
441 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
442 OPT_Wvla_larger_than_.
443 * opts.h (enum cl_var_type): Add an enumerator.
444 * stor-layout.c (layout_decl): Print a more meaningful warning.
445 * toplev.c (output_stack_usage): Adjust.
447 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
449 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
450 call to inline_expand_builtin_string_cmp.
451 (expand_builtin_strcmp): Likewise.
452 (expand_builtin_strncmp): Likewise.
453 (inline_string_cmp): Delete the last parameter, change char_type_node
454 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
456 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
457 the inlining expansion on target where the type of the call has same or
458 narrower precision than unsigned char.
460 2018-07-20 David Malcolm <dmalcolm@redhat.com>
462 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
463 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
464 * common.opt (fsave-optimization-record): New option.
465 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
466 * doc/invoke.texi (-fsave-optimization-record): New option.
467 * dumpfile.c: Include "optinfo-emit-json.h".
468 (struct kv_pair): Move to coretypes.h.
469 (optgroup_options): Make non-static.
470 (dump_context::end_scope): Call
471 optimization_records_maybe_pop_dump_scope.
472 * dumpfile.h (optgroup_options): New decl.
475 * optinfo-emit-json.cc: New file.
476 * optinfo-emit-json.h: New file.
477 * optinfo.cc: Include "optinfo-emit-json.h".
478 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
479 (optinfo_enabled_p): Check optimization_records_enabled_p.
480 (optinfo_wants_inlining_info_p): Likewise.
481 * optinfo.h: Update comment.
482 * profile-count.c (profile_quality_as_string): New function.
483 * profile-count.h (profile_quality_as_string): New decl.
484 (profile_count::quality): New accessor.
485 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
486 and optinfo_emit_json_cc_tests.
487 * selftest.h (selftest::json_cc_tests): New decl.
488 (selftest::optinfo_emit_json_cc_tests): New decl.
489 * toplev.c: Include "optinfo-emit-json.h".
490 (compile_file): Call optimization_records_finish.
491 (do_compile): Call optimization_records_start.
492 * tree-ssa-live.c: Include optinfo.h.
493 (remove_unused_scope_block_p): Retain inlining information if
494 optinfo_wants_inlining_info_p returns true.
496 2018-07-20 Richard Biener <rguenther@suse.de>
499 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
500 to cover -flto-partition=none.
502 2018-07-20 Martin Liska <mliska@suse.cz>
504 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
505 (get_decl_source_range): Remove unused function.
507 2018-07-20 Richard Biener <rguenther@suse.de>
509 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
510 (struct vn_phi_s): Likewise.
511 (struct vn_reference_s): Likewise.
512 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
513 for searching the slot of an entry known to be in the hash itself.
514 (vn_phi_hasher::equal): Likewise.
515 (vn_reference_hasher::equal): Likewise.
516 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
518 (optimistic_info, current_info): Remove, keeping only valid_info.
519 (vn_reference_lookup_1): Remove fallback lookup.
520 (vn_reference_lookup_2): Likewise.
521 (vn_nary_op_lookup_1): Likewise.
522 (vn_phi_lookup): Likewise.
523 (vn_nary_build_or_lookup_1): Make sure to not chain the built
525 (vn_reference_insert): Adjust, chain the inserted hash element
526 at last_inserted_ref.
527 (vn_reference_insert_pieces): Likewise.
528 (visit_reference_op_call): Likewise.
529 (vn_nary_op_insert_into): Chain the inserted hash element at
531 (vn_nary_op_insert_pieces): Adjust.
532 (vn_nary_op_insert): Likewise.
533 (vn_nary_op_insert_stmt): Likewise.
534 (vn_phi_insert): Adjust, chain the inserted hash element at
536 (process_scc): Remove clearing and copying the optimistic
537 table. Instead remove elements inserted during an optimistic
538 iteration from the single table we maintain.
539 (init_scc_vn): Adjust.
540 (free_scc_vn): Likewise.
541 (sccvn_dom_walker::record_cond): Likewise.
542 (sccvn_dom_walker::after_dom_children): Likewise.
544 2018-07-19 Martin Sebor <msebor@redhat.com>
546 PR tree-optimization/84047
547 PR tree-optimization/83776
548 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
549 (check_array_bounds): Call it.
551 2018-07-19 Martin Sebor <msebor@redhat.com>
553 * align.h (align_flags): Use member initialization.
555 2018-07-19 David Malcolm <dmalcolm@redhat.com>
557 * Makefile.in (OBJS): Add optinfo.o.
558 * coretypes.h (class symtab_node): New forward decl.
559 (struct cgraph_node): New forward decl.
560 (class varpool_node): New forward decl.
561 * dump-context.h: New file.
562 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
564 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
565 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
566 (set_alt_dump_file): Likewise.
567 (dump_context::~dump_context): New dtor.
568 (dump_gimple_stmt): Move implementation to...
569 (dump_context::dump_gimple_stmt): ...this new member function.
570 Add the stmt to any pending optinfo, creating one if need be.
571 (dump_gimple_stmt_loc): Move implementation to...
572 (dump_context::dump_gimple_stmt_loc): ...this new member function.
573 Start a new optinfo and add the stmt to it.
574 (dump_gimple_expr): Move implementation to...
575 (dump_context::dump_gimple_expr): ...this new member function.
576 Add the stmt to any pending optinfo, creating one if need be.
577 (dump_gimple_expr_loc): Move implementation to...
578 (dump_context::dump_gimple_expr_loc): ...this new member function.
579 Start a new optinfo and add the stmt to it.
580 (dump_generic_expr): Move implementation to...
581 (dump_context::dump_generic_expr): ...this new member function.
582 Add the tree to any pending optinfo, creating one if need be.
583 (dump_generic_expr_loc): Move implementation to...
584 (dump_context::dump_generic_expr_loc): ...this new member
585 function. Add the tree to any pending optinfo, creating one if
587 (dump_printf): Move implementation to...
588 (dump_context::dump_printf_va): ...this new member function. Add
589 the text to any pending optinfo, creating one if need be.
590 (dump_printf_loc): Move implementation to...
591 (dump_context::dump_printf_loc_va): ...this new member function.
592 Start a new optinfo and add the stmt to it.
593 (dump_dec): Move implementation to...
594 (dump_context::dump_dec): ...this new member function. Add the
595 value to any pending optinfo, creating one if need be.
596 (dump_context::dump_symtab_node): New member function.
597 (dump_context::get_scope_depth): New member function.
598 (dump_context::begin_scope): New member function.
599 (dump_context::end_scope): New member function.
600 (dump_context::ensure_pending_optinfo): New member function.
601 (dump_context::begin_next_optinfo): New member function.
602 (dump_context::end_any_optinfo): New member function.
603 (dump_context::s_current): New global.
604 (dump_context::s_default): New global.
605 (dump_scope_depth): Delete global.
606 (dumpfile_ensure_any_optinfo_are_flushed): New function.
607 (dump_symtab_node): New function.
608 (get_dump_scope_depth): Reimplement in terms of dump_context.
609 (dump_begin_scope): Likewise.
610 (dump_end_scope): Likewise.
611 (selftest::temp_dump_context::temp_dump_context): New ctor.
612 (selftest::temp_dump_context::~temp_dump_context): New dtor.
613 (selftest::verify_item): New function.
614 (ASSERT_IS_TEXT): New macro.
615 (ASSERT_IS_TREE): New macro.
616 (ASSERT_IS_GIMPLE): New macro.
617 (selftest::test_capture_of_dump_calls): New test.
618 (selftest::dumpfile_c_tests): Call it.
619 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
620 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
621 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
623 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
624 (dump_node, dump_bb): Move these unrelated decls.
625 (class dump_manager): Add leading comment.
626 * optinfo.cc: New file.
627 * optinfo.h: New file.
629 2018-07-19 Michael Collison <michael.collison@arm.com>
630 Richard Henderson <rth@redhat.com>
632 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
633 (subti): Handle op1 zero.
634 (subvti4, usub4ti4): New.
635 (*sub<GPI>3_compare1_imm): New.
636 (sub<GPI>3_carryinCV): New.
637 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
638 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
640 2018-07-19 Michael Collison <michael.collison@arm.com>
641 Richard Henderson <rth@redhat.com>
643 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
644 (addti3): Create simpler code if low part is already known to be 0.
645 (addvti4, uaddvti4): New.
646 (*add<GPI>3_compareC_cconly_imm): New.
647 (*add<GPI>3_compareC_cconly): New.
648 (*add<GPI>3_compareC_imm): New.
649 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
650 handle constants within this pattern..
651 (*add<GPI>3_compareV_cconly_imm): New.
652 (*add<GPI>3_compareV_cconly): New.
653 (*add<GPI>3_compareV_imm): New.
654 (add<GPI>3_compareV): New.
655 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
656 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
657 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
658 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
659 with 'comparison' operator.
660 (*add<GPI>3_compareV_cconly_imm): Ditto.
661 (*add<GPI>3_compareV_cconly): Ditto.
662 (*add<GPI>3_compareV_imm): Ditto.
663 (add<GPI>3_compareV): Ditto.
664 (add<mode>3_carryinC): Ditto.
665 (*add<mode>3_carryinC_zero): Ditto.
666 (*add<mode>3_carryinC): Ditto.
667 (add<mode>3_carryinV): Ditto.
668 (*add<mode>3_carryinV_zero): Ditto.
669 (*add<mode>3_carryinV): Ditto.
671 2018-07-19 Michael Collison <michael.collison@arm.com>
672 Richard Henderson <rth@redhat.com>
674 * config/aarch64/aarch64-modes.def (CC_V): New.
675 * config/aarch64/aarch64-protos.h
676 (aarch64_addti_scratch_regs): Declare
677 (aarch64_subvti_scratch_regs): Declare.
678 (aarch64_expand_subvti): Declare.
679 (aarch64_gen_unlikely_cbranch): Declare
680 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
681 for signed overflow using CC_Vmode.
682 (aarch64_get_condition_code_1): Handle CC_Vmode.
683 (aarch64_gen_unlikely_cbranch): New function.
684 (aarch64_addti_scratch_regs): New function.
685 (aarch64_subvti_scratch_regs): New function.
686 (aarch64_expand_subvti): New function.
688 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
690 * config/aarch64/aarch64-option-extensions.def: New entry for profile
692 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
693 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
696 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
699 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
700 address check not strict.
702 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
704 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
706 (store_pair_lanes<mode>): Likewise.
707 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
708 enum value 'ADDR_QUERY_LDP_STP_N'.
709 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
710 (aarch64_print_address_internal): Add declaration.
711 (aarch64_print_ldpstp_address): Remove.
712 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
713 (aarch64_print_operand): Change printing of 'y'.
714 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
715 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
716 'true' rather than '1'.
717 * config/aarch64/constraints.md (Uml): Likewise.
718 (Uml): Rename to Umn.
721 2018-07-19 Richard Biener <rguenther@suse.de>
723 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
725 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
726 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
727 (vn_reference_hasher): Likewise.
728 (struct vn_tables_s): Remove obstack and alloc-pool members.
729 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
730 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
731 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
732 (vn_reference_insert_pieces): Likewise.
733 (alloc_vn_nary_op_noinit): Adjust.
734 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
736 (shared_lookup_phiargs): Remove.
737 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
738 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
739 (visit_reference_op_call): Likewise.
740 (copy_nary, copy_phi, copy_reference): Remove.
741 (process_scc): Rewind the obstack when iterating. Do not
742 copy the elements to valid_info but just move them from one
743 hashtable to the other.
744 (allocate_vn_table): Adjust.
745 (free_vn_table): Likewise.
746 (init_scc_vn): Likewise.
747 (free_scc_vn): Likewise.
749 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
752 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
753 indirect_return as function type attribute.
754 (ix86_attribute_table): Change indirect_return to function
756 * doc/extend.texi: Update indirect_return attribute.
758 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
760 * wide-int.h (widest2_int): New.
761 * gimple-fold.c (arith_overflowed_p): Use it.
762 * tree.h (widest2_int_cst): New.
763 * tree-vrp.c (wide_int_binop_overflow): Rename from
765 Rewrite to work on trees.
766 (extract_range_from_multiplicative_op_1): Abstract code to...
767 (wide_int_range_min_max): ...here.
768 (wide_int_range_cross_product): ...and here.
769 (extract_range_from_binary_expr_1): Abstract overflow code to...
770 (wide_int_range_mult_wrapping): ...here.
771 * tree-vrp.h (wide_int_range_cross_product): New.
772 (wide_int_range_mult_wrapping): New.
774 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
775 Julia Koval <julia.koval@intel.com>
777 * config/i386/x86-tune-costs.h (skylake_memcpy,
778 skylake_memset): Replace rep_prefix with unrolling for size 512.
780 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
783 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle comparision with EQ_EXPR
786 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
788 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
789 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
790 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
791 previously known as "PowerPC AltiVec Built-in Functions". Move
792 some material to new subsubsections "PowerPC AltiVec Built-in
793 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
795 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
796 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
797 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
798 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
800 2018-07-18 Richard Biener <rguenther@suse.de>
802 PR tree-optimization/86557
803 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
806 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
808 * config/s390/s390.c (s390_function_profiler): Generate CFI.
810 2018-07-17 Jeff Law <law@redhat.com>
812 * config/arm/arm.c (get_label_padding): Update for recent
813 changes to label_to_alignment.
815 PR tree-optimization/86010
816 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
818 * config/mips/mips.c (vr4130_align_insns): Update for recent
819 changes to label_to_alignment.
821 * config/frv/frv.c (frv_label_align): Update for recent changes
822 to label_to_alignment.
824 * config/nios2/nios2.c (nios2_label_align): Update for recent
825 changes which dropped ALIGN_LABELS_LOG.
827 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
829 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
830 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
832 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
834 * config/arc/arc.c (arc_label_align): Use align_labels instead of
835 deprecated align_labels_log.
837 2018-07-17 Richard Biener <rguenther@suse.de>
840 * dwarf2out.c (init_sections_and_labels): Always generate
841 a debug_line_str_section for early LTO debug.
842 (dwarf2out_finish): Reset debug_line_str_hash output early.
843 Bump counter for extra dwarf5 .debug_loc labels to not conflict
845 (dwarf2out_early_finish): Output debug_line_str.
847 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
849 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
850 index register on z196 or later.
852 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
854 * config/s390/s390.c (s390_default_align): Set default function
856 (s390_override_options_after_change): Call s390_default align.
857 (s390_option_override_internal): Call s390_default_align.
858 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
860 2018-07-17 Jakub Jelinek <jakub@redhat.com>
863 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
864 to _looptemp_ clauses, other than the first two.
866 2018-07-17 Martin Liska <mliska@suse.cz>
868 * opts.c: Do not enable OPT_falign_* for -Os.
870 2018-07-17 Martin Liska <mliska@suse.cz>
872 * align.h (MAX_CODE_ALIGN): New.
873 (MAX_CODE_ALIGN_VALUE): New.
874 * common/config/i386/i386-common.c (ix86_handle_option):
875 (MAX_CODE_ALIGN): Moved to align.h.
876 * final.c (MAX_CODE_ALIGN): Likewise.
877 * opts.c (parse_and_check_align_values):
878 (MAX_CODE_ALIGN): Likewise.
879 (MAX_CODE_ALIGN_VALUE): Likewise.
881 2018-07-17 Martin Liska <mliska@suse.cz>
883 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
884 in order to fulfil coding style.
885 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
886 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
887 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
888 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
889 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
890 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
891 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
892 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
893 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
895 2018-07-17 Martin Liska <mliska@suse.cz>
898 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
899 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
900 align_flags of label_to_alignment.
901 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
903 * config/m68k/m68k.c: Do not use removed align_labels_value and
905 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
906 (LOOP_ALIGN): Likewise.
907 (LABEL_ALIGN): Likewise.
908 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
909 Remove not used macro.
910 (rs6000_loop_align): Change return type to align_flags.
911 (rs6000_loop_align_max_skip): Remove.
912 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
913 Change return type to align_flags.
914 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
915 Remove not used macro.
916 (rs6000_loop_align): Change return type to align_flags.
917 (rs6000_loop_align_max_skip): Remove.
918 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
919 * config/rx/rx-protos.h (rx_align_for_label): Make it
921 * config/rx/rx.c (rx_align_for_label): Change return type
923 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
925 into align_flags class.
926 (LABEL_ALIGN): Likewise.
927 (LOOP_ALIGN): Likewise.
928 * config/s390/s390.c (s390_label_align): Use align_flags
930 (s390_asm_output_function_label): Likewise.
931 * config/sh/sh.c (sh_override_options_after_change):
932 Use align_flags class directly without macros.
933 (find_barrier): Likewise.
934 (barrier_align): Likewise.
935 (sh_loop_align): Likewise.
936 * config/spu/spu.c (spu_option_override):
937 Use align_flags_tuple::get_value instead of removed macros.
938 (spu_sched_init): Likewise.
939 * config/spu/spu.h (GTY): Likewise.
940 * config/visium/visium.c (visium_option_override):
941 Set "8" as default secondary alignment.
942 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
943 in order to guarantee secondary alignment of 8.
944 * coretypes.h: Include align.h header file.
945 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
946 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
947 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
948 * doc/tm.texi.in: Likewise.
949 * final.c (struct label_alignment): Remove not used structure.
950 (LABEL_ALIGN): Change type to align_flags.
951 (LOOP_ALIGN): Likewise.
952 (JUMP_ALIGN): Likewise.
953 (default_loop_align_max_skip): Remove.
954 (default_label_align_max_skip): Likewise.
955 (default_jump_align_max_skip): Likewise.
956 (default_label_align_after_barrier_max_skip):
957 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
958 (LABEL_TO_MAX_SKIP): Remove.
959 (label_to_alignment): Return align_flags type instead of integer.
960 (label_to_max_skip): Remove.
961 (align_fuzz): Use align_flags type.
962 (compute_alignments): Use align_flags type and use align_flags::max
963 to combine multiple alignments.
964 (grow_label_align): Grow vec instead of C array.
965 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
966 (shorten_branches): Use align_flags type and use align_flags::max
967 to combine multiple alignments.
968 (final_scan_insn_1): Remove usage of secondary alignment that comes
969 from label alignment, but instead use proper secondary alignment
970 which is computed in grow_label_align.
971 * flags.h (struct align_flags_tuple): Move to align.h.
972 (struct align_flags): Likewise.
973 (state_align_loops): Rename to align_loops.
974 (state_align_jumps): Rename to align_jumps.
975 (state_align_labels): Rename to align_labels.
976 (state_align_functions): Rename to align_functions.
977 (align_loops_log): Remove.
978 (align_jumps_log): Remove.
979 (align_labels_log): Remove.
980 (align_functions_log): Remove.
981 (align_loops_max_skip): Remove.
982 (align_jumps_max_skip): Remove.
983 (align_labels_max_skip): Remove.
984 (align_functions_max_skip): Remove.
985 (align_loops_value): Remove.
986 (align_jumps_value): Remove.
987 (align_labels_value): Remove.
988 (align_functions_value): Remove.
989 * output.h (label_to_alignment): Change return type to align_flags.
990 (label_to_max_skip): Remove.
991 * target.def: Remove loop_align_max_skip, label_align_max_skip,
992 jump_align_max_skip macros.
993 * targhooks.h (default_loop_align_max_skip): Remove.
994 (default_label_align_max_skip): Likewise.
995 (default_jump_align_max_skip): Likewise.
996 (default_label_align_after_barrier_max_skip): Remove.
997 * toplev.c (read_log_maxskip): Use ::normalize function.
998 (parse_N_M): Remove not used argument and also call ::normalize.
999 (parse_alignment_opts): Do not pass unused arguments.
1000 * varasm.c (assemble_start_function): Use directly align_functions
1001 instead of removed macros.
1002 * system.h: Do not poison removed macros.
1004 2018-07-17 Jakub Jelinek <jakub@redhat.com>
1007 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
1008 and cond temporaries don't have reference type if iterator has
1009 pointer type. For init use &for_pre_body instead of pre_p if
1010 for_pre_body is non-empty.
1012 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1014 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
1015 double-double modes to SFmode directly directly.
1016 (trunc<mode>sf2_fprs): Delete.
1018 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1020 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
1021 for conversions between IFmode and the decimal floating point modes.
1022 (init_float128_ieee): Use the correct names for conversions between
1023 KFmode and the decimal floating point modes.
1025 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1027 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
1028 for the conversions between TDmode and IFmode.
1029 (init_float128_ieee): Use more correct names for the conversions
1030 between TDmode and KFmode.
1032 2018-07-16 Jakub Jelinek <jakub@redhat.com>
1034 PR tree-optimization/86526
1035 * builtins.c (expand_builtin_memcmp): Formatting fixes.
1036 (inline_expand_builtin_string_cmp): Likewise.
1037 (inline_string_cmp): Likewise. Use c_readstr instead of
1038 builtin_memcpy_read_str. Add unit_mode temporary.
1040 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
1043 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
1044 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
1046 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
1048 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
1049 Alphabetize prototypes of built-in functions, separating out
1050 built-in functions that are listed in this section but should be
1051 described elsewhere.
1053 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
1056 * expmed.c (emit_store_flag): Do not emit setcc followed by a
1057 conditional move when trapping comparison was split to a
1058 non-trapping one (and vice versa).
1060 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1062 * config/s390/s390.c (s390_function_profiler): Generate nops
1063 instead of profiler call sequences.
1064 * config/s390/s390.opt: Add the new option.
1066 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1068 * config/s390/s390.c (s390_function_profiler): Generate
1069 __mcount_loc section.
1070 * config/s390/s390.opt: Add the new option.
1072 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1074 * common.opt: Add the new warning.
1075 * config/s390/s390.c (s390_function_profiler): Emit "brasl
1076 %r0,__fentry__" when -mfentry is specified.
1077 (s390_option_override_internal): Disallow -mfentry for 31-bit
1079 * config/s390/s390.opt: Add the new option.
1081 2018-07-16 Richard Biener <rguenther@suse.de>
1084 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
1085 for function-local FUNCTION_DECL and RESULT_DECL immediately.
1087 2018-07-16 Martin Liska <mliska@suse.cz>
1090 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
1093 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
1095 * config/arc/arcHS.md: Update ARCHS scheduling rules.
1097 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
1099 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
1101 * config/arc/arc-cpus.def (hs4x): New cpu.
1103 * config/arc/arc-tables.opt: Regenerate.
1104 * config/arc/arc.c (arc_sched_issue_rate): New function.
1105 (TARGET_SCHED_ISSUE_RATE): Define.
1106 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
1107 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
1109 (attr tune): Add ARCHS4x tune values.
1110 (attr tune_dspmpy): Define.
1111 (*tst): Correct instruction type.
1112 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
1113 * config/arc/arcHS4x.md: New file.
1114 * config/arc/fpu.md: Update instruction type attributes.
1115 * config/arc/t-multilib: Regenerate.
1117 2018-07-16 Tom de Vries <tdevries@suse.de>
1120 * var-tracking.c (vt_initialize): Fix pre_dec handling.
1122 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1124 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
1127 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
1129 PR tree-optimization/86514
1130 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
1131 conversion to a boolean type from a type with greater precision.
1133 2018-07-16 Tom de Vries <tdevries@suse.de>
1135 * var-tracking.c (vt_initialize): Print adjusted insn slim if
1136 dump_flags request TDF_SLIM.
1138 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
1140 * fold-const.c (int_const_binop_1): Abstract...
1141 (wide_int_binop): ...wide int code here.
1142 (poly_int_binop): ...poly int code here.
1143 Abstract the rest of int_const_binop_1 into int_const_binop.
1144 * fold-const.h (wide_int_binop): New.
1145 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
1146 Remove useless PLUS/MINUS_EXPR case.
1147 (zero_nonzero_bits_from_vr): Move wide int code...
1148 (zero_nonzero_bits_from_bounds): ...here.
1149 (extract_range_from_binary_expr_1): Move mask optimization code...
1150 (range_easy_mask_min_max): ...here.
1151 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
1152 (range_easy_mask_min_max): New.
1154 2018-07-15 Jeff Law <law@redhat.com>
1157 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
1160 2018-07-14 Jim Wilson <jimw@sifive.com>
1162 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
1164 2018-07-14 Paul Koning <ni1d@arrl.net>
1166 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
1168 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
1170 * lto-streamer-out.c (copy_function_or_variable): Dump info about
1173 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
1174 Steve Munroe <munroesj52@gmail.com>
1176 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
1177 (_mm_andnot_si128): Likewise.
1178 (_mm_or_si128): Likewise.
1179 (_mm_xor_si128): Likewise.
1181 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
1184 * builtins.c (expand_builtin_memcmp): Inline the calls first
1185 when result_eq is false.
1186 (expand_builtin_strcmp): Inline the calls first.
1187 (expand_builtin_strncmp): Likewise.
1188 (inline_string_cmp): New routine. Expand a string compare
1189 call by using a sequence of char comparison.
1190 (inline_expand_builtin_string_cmp): New routine. Inline expansion
1191 a call to str(n)cmp/memcmp.
1192 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
1194 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
1196 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
1198 * config/arm/driver-arm.c: Include arm-native.h.
1199 (host_detect_local_cpu): Use auto-generated data tables.
1200 (vendors, arm_cpu_table): Delete. Move part information to ...
1201 * config/arm/arm-cpus.in: ... here.
1202 * config/arm/parsecpu.awk (gen_native): New function.
1203 (vendor, part): New CPU fields.
1204 (END): Add support for building the native CPU detection tables.
1205 * config/arm/t-arm (arm-native.h): Add build rule.
1206 (driver-arm.o): Add dependency on arm-native.h.
1208 2018-07-13 Richard Biener <rguenther@suse.de>
1211 * match.pd (addr1 - addr2): Allow either of the operand to
1214 2018-07-13 Tom de Vries <tdevries@suse.de>
1216 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
1219 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
1221 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
1222 arrays instead of numbered variables.
1224 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
1226 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
1227 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
1229 2018-07-13 Richard Biener <rguenther@suse.de>
1232 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
1233 instead of get_context_die.
1235 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
1236 Richard Biener <rguenther@suse.de>
1239 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
1240 that the loop latch destination where phi is defined.
1242 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
1244 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
1245 (riscv_expand_epilogue): Add assertion to check interrupt mode.
1246 (riscv_set_current_function): Extract getting interrupt type to new
1248 (riscv_get_interrupt_type): New function.
1249 (riscv_merge_decl_attributes): New function, checking interrupt type is
1251 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
1253 2018-07-12 Paul Koning <ni1d@arrl.net>
1255 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
1258 2018-07-12 Paul Koning <ni1d@arrl.net>
1260 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
1261 zero reference, add doloop_end instead.
1262 * doc/md.texi (decrement_and_branch_until_zero): Remove.
1263 (Looping patterns): Remove decrement_and_branch_until_zero. Add
1264 detail for doloop_end.
1266 2018-07-12 Martin Sebor <msebor@redhat.com>
1269 * attribs.c (decl_attributes): Reject conflicting attributes before
1270 calling attribute handlers.
1272 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
1274 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
1276 (gcc::dump_manager::get_dump_file_name): likewise.
1277 (dump_begin): Likewise.
1278 * dumpfile.h (dump_begin): Update prototype.
1279 (gcc::dump_manager::get_dump_file_name,
1280 gcc::dump_manager::get_dump_file_name): Update prototype.
1282 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1284 * internal-fn.h (vectorizable_internal_fn_p): New function.
1285 * tree-vect-slp.c (compatible_calls_p): Likewise.
1286 (vect_build_slp_tree_1): Remove nops argument. Handle calls
1287 to internal functions.
1288 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
1290 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1292 * fold-const.h (inverse_conditions_p): Declare.
1293 * fold-const.c (inverse_conditions_p): New function.
1294 * match.pd: Use inverse_conditions_p. Add folds of view_converts
1295 that test the inverse condition of a conditional internal function.
1296 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
1297 * internal-fn.c (internal_fn_mask_index): Handle conditional
1299 (vectorized_internal_fn_supported_p): New function.
1300 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
1301 (any_pred_load_store): Replace with...
1302 (need_to_predicate): ...this new variable.
1303 (redundant_ssa_names): New variable.
1304 (ifcvt_can_use_mask_load_store): Move initial checks to...
1305 (ifcvt_can_predicate): ...this new function. Handle tree codes
1306 for which a conditional internal function exists.
1307 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
1308 instead of ifcvt_can_use_mask_load_store. Update after variable
1310 (predicate_load_or_store): New function, split out from
1311 predicate_mem_writes.
1312 (check_redundant_cond_expr): New function.
1313 (value_available_p): Likewise.
1314 (predicate_rhs_code): Likewise.
1315 (predicate_mem_writes): Rename to...
1316 (predicate_statements): ...this. Use predicate_load_or_store
1317 and predicate_rhs_code.
1318 (combine_blocks, tree_if_conversion): Update after above name changes.
1319 (ifcvt_local_dce): Handle redundant_ssa_names.
1320 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
1321 general conditional functions.
1322 * tree-vect-stmts.c (vectorizable_call): Likewise.
1324 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1325 Alan Hayward <alan.hayward@arm.com>
1326 David Sherwood <david.sherwood@arm.com>
1328 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
1329 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
1330 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
1331 plus and minus and convert them into IFN_COND_FMA-based sequences.
1332 (convert_mult_to_fma): Handle conditional plus and minus.
1334 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1336 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
1337 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
1338 (cond_fnms_optab): New optabs.
1339 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
1341 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
1342 * internal-fn.h (get_conditional_internal_fn): Declare.
1343 (get_unconditional_internal_fn): Likewise.
1344 * internal-fn.c (cond_ternary_direct): New macro.
1345 (expand_cond_ternary_optab_fn): Likewise.
1346 (direct_cond_ternary_optab_supported_p): Likewise.
1347 (FOR_EACH_COND_FN_PAIR): Likewise.
1348 (get_conditional_internal_fn): New function.
1349 (get_unconditional_internal_fn): Likewise.
1350 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
1351 (gimple_match_op::gimple_match_op): Add a new overload for 5
1353 (gimple_match_op::set_op): Likewise.
1354 (gimple_resimplify5): Declare.
1355 * genmatch.c (decision_tree::gen): Generate simplifications for
1357 * gimple-match-head.c (gimple_simplify): Define an overload for
1358 5 operands. Handle calls with 5 arguments in the top-level overload.
1359 (convert_conditional_op): Handle conversions from unconditional
1360 internal functions to conditional ones.
1361 (gimple_resimplify5): New function.
1362 (build_call_internal): Pass a fifth operand.
1363 (maybe_push_res_to_seq): Likewise.
1364 (try_conditional_simplification): Try converting conditional
1365 internal functions to unconditional internal functions.
1366 Handle 3-operand unconditional forms.
1367 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
1368 Define ternary equivalents of the current rules for binary conditional
1370 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
1372 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
1373 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
1374 (optab): Handle them.
1375 (SVE_COND_FP_TERNARY): New int iterator.
1376 (sve_fmla_op, sve_fmad_op): New int attributes.
1377 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
1378 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
1379 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
1381 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1383 * target.def (preferred_else_value): New target hook.
1384 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
1385 * doc/tm.texi: Regenerate.
1386 * targhooks.h (default_preferred_else_value): Declare.
1387 * targhooks.c (default_preferred_else_value): New function.
1388 * internal-fn.h (conditional_internal_fn_code): Declare.
1389 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
1390 (get_conditional_internal_fn): Use it.
1391 (conditional_internal_fn_code): New function.
1392 * gimple-match.h (gimple_match_cond): New struct.
1393 (gimple_match_op): Add a cond member function.
1394 (gimple_match_op::gimple_match_op): Update all forms to take a
1396 * genmatch.c (expr::gen_transform): Use the same condition as res_op
1397 for the suboperation, but don't specify a particular else_value.
1398 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
1399 (visit_nary_op, visit_reference_op_load): Pass
1400 gimple_match_cond::UNCOND to the gimple_match_op constructor.
1401 * gimple-match-head.c: Include tree-eh.h
1402 (convert_conditional_op): New function.
1403 (maybe_resimplify_conditional_op): Likewise.
1404 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
1405 (gimple_resimplify2): Likewise.
1406 (gimple_resimplify3): Likewise.
1407 (gimple_resimplify4): Likewise.
1408 (maybe_push_res_to_seq): Return null for conditional operations.
1409 (try_conditional_simplification): New function.
1410 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
1412 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
1414 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
1416 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
1418 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
1420 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1422 (hash_tree): Do not hash DECL_FCONTEXT
1423 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
1424 Do not stream DECL_FCONTEXT.
1425 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
1426 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
1428 2018-07-12 Richard Biener <rguenther@suse.de>
1431 * dwarf2out.c (gen_block_die): Only output blocks when they have
1432 at least one !DECL_IGNORED_P variable.
1434 2018-07-12 Richard Biener <rguenther@suse.de>
1437 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
1438 Remove -mieee-fp handling.
1440 2018-07-12 Richard Biener <rguenther@suse.de>
1442 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
1443 left-over from last patch.
1445 2018-07-12 Jakub Jelinek <jakub@redhat.com>
1447 PR tree-optimization/86492
1448 * gimple-ssa-store-merging.c
1449 (imm_store_chain_info::coalesce_immediate_stores): Call
1450 check_no_overlap even for the merge_overlapping case. Formatting fix.
1452 2018-07-12 Richard Biener <rguenther@suse.de>
1455 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
1456 move possibly trapping operations into the conditional.
1458 2018-07-12 Richard Biener <rguenther@suse.de>
1460 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
1461 (vn_lookup_simplify_result): Remove recursion limit applied
1463 (vn_nary_build_or_lookup_1): Adjust.
1464 (try_to_simplify): Likewise.
1465 * gimple-match-head.c (gimple_resimplify1): Instead apply one
1467 (gimple_resimplify2): Likewise.
1468 (gimple_resimplify3): Likewise.
1469 (gimple_resimplify4): Likewise.
1471 2018-07-11 Jakub Jelinek <jakub@redhat.com>
1473 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
1474 Use __mmask64 type instead of __mmask8 for __M argument.
1475 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
1476 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
1478 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
1479 __mmask16 for __M argument.
1480 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
1481 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
1482 to __mmask16 instead of __mmask8.
1483 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
1484 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
1485 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
1486 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
1487 instead of __mmask16 for __U argument.
1488 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
1489 __mmask16 instead of __mmask8 for __U argument.
1490 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
1492 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
1494 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
1496 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
1498 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
1500 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
1502 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
1504 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
1506 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
1508 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
1510 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
1511 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
1512 return type as well as __M argument type and all casts from __mmask8
1514 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
1515 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
1516 return type as well as __M argument type and all casts from __mmask8
1518 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
1519 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
1520 return type as well as __M argument type and all casts from __mmask8
1522 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
1523 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
1524 return type as well as __M argument type and all casts from __mmask8
1526 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
1527 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
1530 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
1532 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
1533 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
1536 2018-07-11 Paul Koning <ni1d@arrl.net>
1538 * doc/md.texi (define_subst): Document how multiple occurrences of
1539 the same argument in the replacement pattern are handled.
1541 2018-07-11 Paul Koning <ni1d@arrl.net>
1543 * doc/extend.texi (Common Variable Attributes): Move "mode" into
1545 (Common Type Attributes): Add "mode" attribute.
1547 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
1549 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
1550 stream DECL_ORIGINAL_TYPE.
1551 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
1552 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
1553 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
1554 Do not walk original type.
1555 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
1557 (write_ts_decl_non_common_tree_pointers): Do not stream
1559 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
1560 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
1562 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
1564 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
1565 thread twice from the same starting edge.
1567 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
1569 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
1570 * gimple.c (gimple_call_nonnull_result_p): ...here...
1571 (gimple_call_nonnull_arg): ...and here.
1572 * gimple.h (gimple_call_nonnull_result_p): New.
1573 (gimple_call_nonnull_arg): New.
1575 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
1577 * config/arm/arm-cpus.in: Move information from fpu field of each
1578 cpu definition to the isa field.
1579 * config/arm/parsecpu.awk (fpu): Delete match rule.
1580 (gen_comm_data): Don't add bits from the CPU's FPU entry.
1582 2018-07-11 Richard Biener <rguenther@suse.de>
1585 * dwarf2out.c (init_sections_and_labels): Use
1586 output_asm_line_debug_info consistently.
1587 (dwarf2out_early_finish): Likewise.
1588 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
1591 2018-07-11 Richard Biener <rguenther@suse.de>
1593 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
1594 Rework father_bb setting in a way to avoid propagating constants
1595 multiple times on a loop body.
1597 2018-07-10 Mark Wielaard <mark@klomp.org>
1600 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
1602 2018-07-10 Richard Biener <rguenther@suse.de>
1604 * hash-map.h (hash_map::iterator::operator*): Return
1605 references to key and value.
1607 2018-07-10 Jakub Jelinek <jakub@redhat.com>
1610 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
1611 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
1612 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
1613 (gimplify_omp_for): For composite loops, move outer
1614 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
1615 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
1616 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
1617 TREE_LIST for both the original class iterator and the "last" helper
1618 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
1619 loop, remember has_decl_expr from outer composite loops for the
1620 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
1622 2018-07-09 Martin Sebor <msebor@redhat.com>
1626 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
1627 accessing implicitly initialized array elements.
1628 * expr.c (string_constant): Handle string initializers of
1629 character arrays within aggregates.
1630 * gimple-fold.c (fold_array_ctor_reference): Add argument.
1631 Store element offset. As a special case, handle zero size.
1632 (fold_nonarray_ctor_reference): Same.
1633 (fold_ctor_reference): Add argument. Store subobject offset.
1634 * gimple-fold.h (fold_ctor_reference): Add argument.
1636 2018-07-09 Paul Koning <ni1d@arrl.net>
1638 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
1639 (pdp11_insn_cost): New function.
1640 (pdp11_md_asm_adjust): New function.
1641 (TARGET_INVALID_WITHIN_DOLOOP): Define.
1642 (pdp11_rtx_costs): Update to match machine better.
1643 (output_addr_const_pdp11): Correct format mismatch warnings.
1644 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
1645 * config/pdp11/pdp11.md: General change to add base_cost and/or
1646 length attributes for use by new pdp11_insn_cost function.
1647 (MIN_BRANCH): Correct definition.
1649 (doloop_end): Use standard pattern name for looping pattern.
1650 (doloop_end_nocc): New.
1651 (movsf): Add another constraint alternative.
1652 (zero_extendqihi2): Add constraint alternatives for not in place
1654 (zero_extendhisi2): Remove.
1655 (shift patterns): Add CC handling variants.
1659 (define_peephole2): New peephole to recognize mov that sets CC for
1662 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1664 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
1665 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
1668 2018-07-09 Jakub Jelinek <jakub@redhat.com>
1671 * real.c (real_nextafter): Return true if result is denormal.
1673 2018-07-09 Martin Liska <mliska@suse.cz>
1675 * common.opt: Add back wrongly removed attribute.
1677 2018-07-09 Richard Biener <rguenther@suse.de>
1680 * dwarf2out.c (gen_block_die): For an early generated DIE
1681 always output high/low PC attributes.
1683 2018-07-09 Tom de Vries <tdevries@suse.de>
1685 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
1686 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
1687 onto VAR_DECL with abstract origin.
1689 2018-07-07 Jim Wilson <jimw@sifive.com>
1691 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
1693 2018-07-07 Tom de Vries <tdevries@suse.de>
1695 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
1697 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
1699 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
1701 (combine_bound): Use wide-int overflow calculation instead of
1703 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
1705 * fold-const.c (int_const_binop_2): Same.
1706 (extract_muldiv_1): Same.
1707 (fold_div_compare): Same.
1708 (fold_abs_const): Same.
1710 * poly-int.h (add): Same.
1714 * predict.c (predict_iv_comparison): Same.
1715 * profile-count.c (slow_safe_scale_64bit): Same.
1716 * simplify-rtx.c (simplify_const_binary_operation): Same.
1717 * tree-chrec.c (tree_fold_binomial): Same.
1718 * tree-data-ref.c (split_constant_offset_1): Same.
1719 * tree-if-conv.c (idx_within_array_bound): Same.
1720 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
1721 * tree-ssa-phiopt.c (minmax_replacement): Same.
1722 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
1723 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
1724 * vr-values.c (vr_values::adjust_range_with_scev): Same.
1725 * wide-int.cc (wi::add_large): Same.
1726 (wi::mul_internal): Same.
1727 (wi::sub_large): Same.
1728 (wi::divmod_internal): Same.
1729 * wide-int.h: Change overflow type to overflow_type for neg, add,
1730 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
1731 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
1732 mul_internal, divmod_internal.
1733 (overflow_type): New enum.
1734 (accumulate_overflow): New.
1736 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
1738 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
1739 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
1741 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
1743 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
1744 argument is checked for zero before entering loop, avoid checking again.
1746 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
1748 * gimplify.h (generic_expr_could_trap_p): Set as global function.
1749 * gimplify.h (generic_expr_could_trap_p): Likwise.
1750 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
1752 2018-07-06 Jakub Jelinek <jakub@redhat.com>
1754 PR tree-optimization/86401
1755 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
1756 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
1757 (fold_bit_and_mask): ... here. New helper function for match.pd.
1758 * fold-const.h (fold_bit_and_mask): Declare.
1759 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
1761 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
1764 * target.def (translate_mode_attribute): New hook.
1765 * targhooks.h (default_translate_mode_attribute): Declare.
1766 * targhooks.c (default_translate_mode_attribute): New function.
1767 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
1768 * doc/tm.texi: Regenerate.
1769 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
1770 (rs6000_translate_mode_attribute): New function.
1772 2018-07-06 Paul Koning <ni1d@arrl.net>
1774 * doc/md.texi (define_split): Document DONE and FAIL.
1775 (define_peephole2): Ditto.
1777 2018-07-05 Jeff Law <law@redhat.com>
1779 PR tree-optimization/86010
1780 * tree-ssa-dse.c (compute_trims): More aggressively trim at
1781 both the head and tail of mem* and str* calls.
1783 2018-07-05 Jim Wilson <jimw@sifive.com>
1785 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
1787 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
1789 * config/aarch64/aarch64-simd.md: correct flags text for
1790 MIN_EXPR replacement.
1792 2018-07-05 James Clarke <jrtc27@jrtc27.com>
1794 * configure: Regenerated.
1796 2018-07-05 Carl Love <cel@us.ibm.com>
1798 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
1799 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
1800 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
1801 VSX_BUILTIN_DOUBLEL_V4SF.
1803 2018-07-05 Martin Sebor <msebor@redhat.com>
1806 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
1807 than its domain to compute its the upper bound of a char array.
1809 2018-07-05 Nathan Sidwell <nathan@acm.org>
1811 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
1812 * doc/cpp.texi: Update comment.
1813 * doc/tm.texi: Rebuilt.
1814 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
1815 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
1816 * doc/extend.texi (Backwards Compatibility): Clarify it is system
1817 headers affected by extern "C".
1818 * system.h: Poison NO_IMPLICIT_EXTERN_C.
1819 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
1820 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
1821 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
1822 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
1823 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
1824 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
1825 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
1826 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
1827 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
1828 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
1829 config/sparc/sp64-elf.h, config/spu/spu.h,
1830 config/stormy16/stormy16.h, config/v850/v850.h,
1831 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
1832 define NO_IMPLICIT_EXTERN_C.
1833 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
1835 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1838 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1839 instead of GET_MODE_SIZE when comparing Units.
1841 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1844 * rtlanal.c (set_noop_p): Constrain on mode change,
1845 include hard-reg-set.h
1847 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1849 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
1851 2018-07-05 Jakub Jelinek <jakub@redhat.com>
1854 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
1857 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
1859 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
1861 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
1864 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
1866 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
1868 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
1871 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
1873 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
1875 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
1876 neg_*_op* variables.
1878 2018-07-04 Martin Liska <mliska@suse.cz>
1880 * tree-switch-conversion.c: Define
1881 max_ratio_for_speed and max_ratio_for_size constants.
1883 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
1884 Martin Liska <mliska@suse.cz>
1889 * common.opt: Rename align options with 'str_' prefix.
1890 * common/config/i386/i386-common.c (set_malign_value): New
1892 (ix86_handle_option): Use it to set -falign-* options/
1893 * config/aarch64/aarch64-protos.h (struct tune_params): Change
1894 type from int to string.
1895 * config/aarch64/aarch64.c: Update default values from int
1897 * config/alpha/alpha.c (alpha_override_options_after_change):
1899 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
1900 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1901 max skip conditionally.
1902 * config/i386/freebsd.h (SUBALIGN_LOG): New.
1903 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1904 max skip conditionally.
1905 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1906 max skip conditionally.
1907 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
1908 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1909 max skip conditionally.
1910 * config/i386/i386.c (struct ptt): Change type from int to
1912 (ix86_default_align): Set default values.
1913 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
1914 max skip conditionally.
1915 * config/i386/iamcu.h (SUBALIGN_LOG): New.
1916 (ASM_OUTPUT_MAX_SKIP_ALIGN):
1917 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
1918 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1919 max skip conditionally.
1920 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
1921 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
1922 * config/i386/x86-64.h (SUBALIGN_LOG): New.
1923 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1924 max skip conditionally.
1925 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1926 * config/ia64/ia64.c (ia64_option_override): Set default values
1927 for alignment options.
1928 * config/m68k/m68k.c: Handle new str_align_* options.
1929 * config/mips/mips.c (mips_set_compression_mode): Change
1931 (mips_option_override): Set default values for options.
1932 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
1934 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1936 * config/rx/rx.c (rx_option_override): Likewise.
1937 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
1938 (LABEL_ALIGN): Use align_labels_log.
1939 (LOOP_ALIGN): Use align_loops_align.
1940 * config/s390/s390.c (s390_asm_output_function_label): Use new
1942 * config/sh/sh.c (sh_override_options_after_change):
1943 Change type of constants.
1944 * config/spu/spu.c (spu_sched_init): Likewise.
1945 * config/sparc/sparc.c (sparc_option_override): Set default
1947 * config/visium/visium.c (visium_option_override): Likewise.
1948 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
1949 emit p2align format with last argument if it's not needed.
1950 * doc/invoke.texi: Document extended format of -falign-*.
1951 * final.c: Use align_labels alignment.
1952 * flags.h (struct target_flag_state): Change type to use
1954 (struct align_flags_tuple): New.
1955 (struct align_flags): Likewise.
1956 (align_loops_log): Redefine macro to use new types.
1957 (align_loops_max_skip): Redefine macro to use new types.
1958 (align_jumps_log): Redefine macro to use new types.
1959 (align_jumps_max_skip): Redefine macro to use new types.
1960 (align_labels_log): Redefine macro to use new types.
1961 (align_labels_max_skip): Redefine macro to use new types.
1962 (align_functions_log): Redefine macro to use new types.
1963 (align_loops): Redefine macro to use new types.
1964 (align_jumps): Redefine macro to use new types.
1965 (align_labels): Redefine macro to use new types.
1966 (align_functions): Redefine macro to use new types.
1967 (align_functions_max_skip): Redefine macro to use new types.
1968 (align_loops_value): New macro.
1969 (align_jumps_value): New macro.
1970 (align_labels_value): New macro.
1971 (align_functions_value): New macro.
1972 * function.c (invoke_set_current_function_hook): Propagate
1973 alignment values from flags to global variables default in
1975 * ipa-icf.c (sem_function::equals_wpa): Use
1976 cl_optimization_option_eq instead of memcmp.
1977 * lto-streamer.h (cl_optimization_stream_out): Support streaming
1979 (cl_optimization_stream_in): Likewise.
1980 * optc-save-gen.awk: Support strings in cl_optimization.
1981 * opth-gen.awk: Likewise.
1982 * opts.c (finish_options): Remove error checking of invalid
1984 (MAX_CODE_ALIGN): Remove.
1985 (MAX_CODE_ALIGN_VALUE): Likewise.
1986 (parse_and_check_align_values): New function.
1987 (check_alignment_argument): Likewise.
1988 (common_handle_option): Use check_alignment_argument.
1989 * opts.h (parse_and_check_align_values): Declare.
1990 * toplev.c (init_alignments): Remove.
1991 (read_log_maxskip): New.
1992 (parse_N_M): Likewise.
1993 (parse_alignment_opts): Likewise.
1994 (backend_init_target): Remove usage of init_alignments.
1995 * toplev.h (parse_alignment_opts): Declare.
1996 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
1998 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
1999 * tree.c (cl_option_hasher::equal): New.
2000 * varasm.c: Use new global macros.
2002 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
2004 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
2005 Use a simpler align directive also if MAXSKIP = ALIGN-1.
2006 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2007 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2008 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2009 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
2010 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
2011 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
2012 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
2013 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2014 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2015 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2016 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2018 2018-07-04 Martin Liska <mliska@suse.cz>
2019 Jonathan Wakely <jwakely@redhat.com>
2021 * coverage.c: Use correct type.
2022 * doc/invoke.texi: Language correction.
2024 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
2027 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
2028 ENDBRANCH for non-tail call which may return via indirect branch.
2029 * doc/extend.texi: Document indirect_return attribute.
2031 2018-07-03 Martin Sebor <msebor@redhat.com>
2033 PR tree-optimization/86274
2034 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
2036 (format_floating): Correct handling of infinities and NaNs.
2038 2018-07-03 Martin Sebor <msebor@redhat.com>
2040 * print-tree.c (print_real_cst): New function.
2041 (print_node_brief): Call it.
2042 (print_node): Ditto.
2044 2018-07-03 Jeff Law <law@redhat.com>
2046 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
2047 into a single pattern.
2049 * config/h8300/h8300.md (ors code_iterator): New.
2050 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
2051 a single pattern and single splitter.
2052 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
2053 (iorqi3_1, xorqi3_1): Likewise.
2054 (iorqi3, xorqi3 expanders): Similarly.
2056 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
2057 (movmd_internal) into a single pattern using the P mode iterator.
2058 (movmd splitters): Similarly.
2059 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
2060 (movsd splitters): Similarly.
2062 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
2063 ADDB, ADDW and ADDL into a single ADD attribute which selects the
2064 right table based on the size of the operand.
2065 * config/h8300/h8300.md (length_table): Corresponding changes. All
2066 references to "addb", "addw" and "addl" changed to "add".
2067 (btst patterns): Merge two variants into a single pattern.
2068 (tstqi, tsthi): Likewise.
2069 (addhi3_incdec, addsi3_incdec): Likewise.
2070 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
2071 (mulhi3, mulsi3): Likewise.
2072 (udivhi3, udivsi3): Likewise.
2073 (divhi3, divsi3): Likewise.
2074 (andorqi3, andorhi3, andorsi3): Likewise.
2076 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
2079 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
2080 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
2082 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2084 PR tree-optimization/85694
2085 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
2086 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
2088 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
2089 (<u>avg<mode>3_ceil): New patterns.
2091 2018-07-03 David Malcolm <dmalcolm@redhat.com>
2093 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
2094 scan-tree-dump directive.
2095 * gcc.dg/vect/slp-perm-2.c: Likewise.
2096 * gcc.dg/vect/slp-perm-3.c: Likewise.
2097 * gcc.dg/vect/slp-perm-5.c: Likewise.
2098 * gcc.dg/vect/slp-perm-6.c: Likewise.
2099 * gcc.dg/vect/slp-perm-7.c: Likewise.
2100 * gcc.dg/vect/slp-perm-8.c: Likewise.
2102 2018-07-03 Marek Polacek <polacek@redhat.com>
2105 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
2107 2018-07-03 Richard Biener <rguenther@suse.de>
2110 * tree-ssa-structalias.c (find_func_clobbers): Properly
2111 handle indirect calls.
2113 2018-07-03 Jeff Law <law@redhat.com>
2115 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
2116 (shifts): New code iterator.
2117 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
2118 expander. Fix HImode handling on H8/SX.
2119 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
2120 (subqi3, subhi3, subsi3 expanders): Likewise.
2121 (andqi3, andhi3, andsi3 expanders): Likewise.
2122 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
2123 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
2124 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
2125 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
2126 (zero_extendqihi2, zero_extendqisi2): Likewise.
2127 (extendqihi2, extendqisi2): Likewise.
2128 (rotlqi3, rotlhi3, rotlsi3): Likewise.
2129 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
2130 (rotlqi3_1, rotlhi3_1): Likewise.
2131 (logicalhi3_sn, logicalsi3_sn): Likewise.
2132 (logicalhi3, logicalsi3): Likewise.
2134 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2136 * tree-vect-patterns.c (vect_recog_rotate_pattern)
2137 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
2138 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
2139 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
2140 type to append_pattern_def_seq instead of creating a stmt_vec_info
2142 (build_mask_conversion): Likewise. Remove vinfo argument.
2143 (vect_add_conversion_to_patterm): Likewise, renaming to...
2144 (vect_add_conversion_to_pattern): ...this.
2145 (vect_recog_mask_conversion_pattern): Update call to
2146 build_mask_conversion. Pass the vector type to
2147 append_pattern_def_seq here too.
2148 (vect_recog_gather_scatter_pattern): Update call to
2149 vect_add_conversion_to_pattern.
2151 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2153 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
2154 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2155 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
2156 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
2157 STMT_VINFO_PATTERN_DEF_SEQ to null here.
2158 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
2159 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
2160 append_pattern_def_seq instead of new_pattern_def_seq.
2161 (vect_recog_divmod_pattern): Do both of the above.
2162 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
2165 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2167 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
2168 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
2169 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
2170 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
2171 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
2172 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
2173 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
2174 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2175 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2176 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
2177 parameter with a single stmt_vec_info.
2178 (vect_recog_func_ptr): Likewise.
2179 (vect_recog_gather_scatter_pattern): Likewise, folding in...
2180 (vect_try_gather_scatter_pattern): ...this.
2181 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
2182 the stmt_vec_info of the statement to be matched. Don't clear
2183 STMT_VINFO_RELATED_STMT.
2184 (vect_pattern_recog): Update call accordingly.
2186 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2188 PR tree-optimization/85694
2189 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
2190 (uavgM3_ceil): Document new optabs.
2191 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
2192 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
2194 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
2195 (savg_ceil_optab): New optabs.
2196 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
2197 (vect_vect_recog_func_ptrs): Add it.
2198 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
2199 constant directly from the associated lhs.
2201 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2203 * tree-vect-patterns.c (vect_split_statement): New function.
2204 (vect_convert_input): Use it to try to split an existing cast.
2206 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2208 * poly-int.h (print_hex): New function.
2209 * dumpfile.h (dump_dec, dump_hex): Declare.
2210 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
2211 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
2212 min_input_precision, operation_precision and operation_sign.
2213 * tree-vect-patterns.c (vect_get_range_info): New function.
2214 (vect_same_loop_or_bb_p, vect_single_imm_use)
2215 (vect_operation_fits_smaller_type): Delete.
2216 (vect_look_through_possible_promotion): Add an optional
2217 single_use_p parameter.
2218 (vect_recog_over_widening_pattern): Rewrite to use new
2219 stmt_vec_info infomration. Handle one operation at a time.
2220 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
2221 (vect_truncatable_operation_p, vect_set_operation_type)
2222 (vect_set_min_input_precision): New functions.
2223 (vect_determine_min_output_precision_1): Likewise.
2224 (vect_determine_min_output_precision): Likewise.
2225 (vect_determine_precisions_from_range): Likewise.
2226 (vect_determine_precisions_from_users): Likewise.
2227 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
2228 (vect_vect_recog_func_ptrs): Put over_widening first.
2230 (vect_pattern_recog): Call vect_determine_precisions.
2232 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2234 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
2235 statements that have been replaced by further pattern statements.
2236 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
2238 2018-07-03 Richard Biener <rguenther@suse.de>
2240 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
2241 always set *dt. Dump vectype in vectype overload.
2242 * dumpfile.h (dump_gimple_expr): New function.
2243 (dump_gimple_expr_loc): Likewise.
2244 * dumpfile.c (dump_gimple_expr): New function.
2245 (dump_gimple_expr_loc): Likewise.
2247 2018-07-02 Jeff Law <law@redhat.com>
2249 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
2250 the H8/300, H8/300H and H8/S variants into a single pattern.
2251 (movhi_h8300, movqi_h8300hs): Similarly.
2252 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
2253 (QHI mode iterator): New.
2255 * config/h8300/h8300.md: Remove trailing whitespace.
2257 2018-07-02 Jim Wilson <jimw@sifive.com>
2259 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
2260 instead of emit_insn for interrupt returns.
2261 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
2262 (riscv_sret, riscv_uret): Likewise.
2264 2018-07-02 David Malcolm <dmalcolm@redhat.com>
2266 * pretty-print.c (selftest::test_pp_format): Move save and restore
2267 of quotes to class auto_fix_quotes, and add an instance.
2268 * selftest.c: Include "intl.h".
2269 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
2270 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
2271 * selftest.h (selftest::auto_fix_quotes): New class.
2273 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
2275 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
2276 (aarch64_sve_prepare_conditional_op): Remove.
2277 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
2278 Allow aarch64_simd_reg_or_zero as select operand; remove
2279 the aarch64_sve_prepare_conditional_op call.
2280 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
2281 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
2282 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
2283 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
2284 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
2285 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
2286 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
2287 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
2288 and a splitters to match all of the *_any patterns.
2289 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
2291 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
2292 (SVE_COND_FP_BINARY_REV): Remove.
2293 (sve_int_op_rev, sve_fp_op_rev): New.
2294 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
2295 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
2296 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
2297 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
2298 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
2299 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
2300 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
2301 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
2302 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
2304 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
2305 Remove match_dup 1 from the inner unspec.
2306 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
2308 * config/aarch64/aarch64.md (movprfx): New attr.
2309 (length): Default movprfx to 8.
2310 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
2311 (*madd<SVE_I>, *msub<SVE_I): Likewise.
2312 (*<su>mul<SVE_I>3_highpart): Likewise.
2313 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
2314 (*v<ASHIFT><SVE_I>3): Likewise.
2315 (*<su><MAXMIN><SVE_I>3): Likewise.
2316 (*<su><MAXMIN><SVE_F>3): Likewise.
2317 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
2318 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
2319 (*div<SVE_F>4): Likewise.
2321 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
2323 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
2326 2018-07-02 Richard Biener <rguenther@suse.de>
2328 PR tree-optimization/86363
2329 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
2330 memset argument refers to a non-variable address.
2332 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
2334 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
2335 {PLUS,MINUS}_EXPR code to...
2336 (adjust_symbolic_bound): ...here,
2337 (combine_bound): ...here,
2338 (set_value_range_with_overflow): ...and here.
2340 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
2342 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
2344 (extract_range_from_abs_expr): ...here.
2346 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
2348 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
2349 -fno-omit-frame-pointer when not optimizing.
2351 2018-07-02 Martin Liska <mliska@suse.cz>
2354 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
2355 (propagate_nothrow): Likewise.
2357 2018-07-02 Martin Liska <mliska@suse.cz>
2360 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
2362 2018-07-02 David Malcolm <dmalcolm@redhat.com>
2364 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
2365 function in r262149, changing "loc" param from source_location to
2366 const dump_location_t &.
2367 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
2368 declaration, as above.
2370 2018-07-01 Paul Koning <ni1d@arrl.net>
2372 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
2373 -munit-asm, -mgnu-asm, -mdec-asm.
2374 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
2375 (pdp11_output_labelref): New.
2376 (pdp11_output_def): New.
2377 (pdp11_output_addr_vec_elt): New.
2378 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
2379 %# and %@ format codes.
2380 (pdp11_option_override): New.
2381 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
2382 (pdp11_output_ident): New.
2383 (pdp11_asm_named_section): New.
2384 (pdp11_asm_init_sections): New.
2385 (pdp11_file_start): New.
2386 (pdp11_file_end): New.
2387 (output_ascii): Use .ascii/.asciz for -mdec-asm.
2388 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
2389 %o, like %c but octal.
2390 (pdp11_option_override): New.
2391 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
2393 (DATA_SECTION_ASM_OP): Ditto.
2394 (READONLY_DATA_SECTION_ASM_OP): New.
2395 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
2396 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
2397 (ASM_OUTPUT_LABELREF): Ditto.
2398 (ASM_OUTPUT_DEF): Ditto.
2399 (ASM_OUTPUT_EXTERNAL): New.
2400 (ASM_OUTPUT_SOURCE_FILENAME): New.
2401 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
2402 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
2403 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
2404 %# and %@ format codes.
2405 * config/pdp11/pdp11.opt (mgnu-asm): New.
2406 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
2407 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
2408 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
2410 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
2412 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
2413 dereferencing path[] beyond its length.
2415 (debug_all_paths): New.
2416 (rewire_first_differing_edge): New.
2417 (adjust_paths_after_duplication): New.
2418 (duplicate_thread_path): Call adjust_paths_after_duplication.
2420 (thread_through_all_blocks): Add new argument to
2421 duplicate_thread_path.
2423 2018-06-30 Jim Wilson <jimw@sifive.com>
2425 * config/riscv/predicates.md (p2m1_shift_operand): New.
2426 (high_mask_shift_operand): New.
2427 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
2428 pattern using p2m1_shift_operand.
2429 (lshsi3_zero_extend_3+2): New combiner pattern using
2430 high_mask_shift_operand.
2432 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2434 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
2436 (vect_recog_rotate_pattern): ...here.
2437 (vect_convert_input): Try to insert casts of invariants in the
2439 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
2440 preheader to be empty.
2442 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2444 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
2445 vector type. If given, install it in the new statement's
2447 (vect_element_precision): New function.
2448 (vect_unpromoted_value): New struct.
2449 (vect_unpromoted_value::vect_unpromoted_value): New function.
2450 (vect_unpromoted_value::set_op): Likewise.
2451 (vect_look_through_possible_promotion): Likewise.
2452 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
2453 (vect_widened_op_tree, vect_convert_input): Likewise.
2454 (vect_convert_inputs, vect_convert_output): Likewise.
2455 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
2456 to handle the optional cast of the multiplication result and
2457 vect_widened_op_tree to detect the widened multiplication itself.
2458 Do not require the input and output of promotion casts to have
2459 the same sign, but base the signedness of the operation on the
2460 input rather than the result. If the pattern includes two
2461 promotions, check that those promotions have the same sign.
2462 Do not restrict the MULT_EXPR handling to a double-width result;
2463 handle quadruple-width results and wider. Use vect_convert_inputs
2464 to convert the inputs to the common type.
2465 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
2466 to handle the optional cast of the ABS result. Also allow a sign
2467 change or a sign extension between the ABS and MINUS.
2468 Use vect_widened_op_tree to detect the widened subtraction and use
2469 vect_convert_inputs to convert the inputs to the common type.
2470 (vect_handle_widen_op_by_const): Delete.
2471 (vect_recog_widen_op_pattern): New function.
2472 (vect_recog_widen_mult_pattern): Use it.
2473 (vect_recog_widen_shift_pattern): Likewise.
2474 (vect_recog_widen_sum_pattern): Use
2475 vect_look_through_possible_promotion to handle the promoted
2478 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2480 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
2481 the containing gimple_seq *.
2482 * gimple-iterator.h (gsi_for_stmt): Declare it.
2483 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
2484 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
2485 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
2486 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
2487 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
2489 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
2491 (vect_mark_pattern_stmts): ...here. Handle cases in which the
2492 statement being replaced is part of an existing pattern
2493 definition sequence, inserting the new pattern statements before
2495 (vect_pattern_recog_1): Don't return a bool. If the statement
2496 is already part of a pattern, instead apply pattern matching
2497 to the pattern definition statements. Don't clear the
2498 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
2499 (vect_pattern_recog): Don't break after the first match;
2500 continue processing the pattern definition statements instead.
2501 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
2503 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2505 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2506 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2507 (vect_recog_widen_sum_pattern): Use it.
2509 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2511 * tree-vect-loop.c (vectorizable_reduction): Assert that the
2512 phi is not a pattern statement and has not been replaced by
2513 a pattern statement.
2514 * tree-vect-patterns.c (type_conversion_p): Don't check
2515 STMT_VINFO_IN_PATTERN_P.
2516 (vect_recog_vector_vector_shift_pattern): Likewise.
2517 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
2518 the pattern statement rather than the original statement; check
2519 directly for a WIDEN_MULT_EXPR here.
2520 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
2521 vect_is_simple_use to return the pattern statement rather
2522 than the original statement; use is_pattern_stmt_p to check
2523 for such a pattern statement.
2524 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
2525 to return the pattern statement rather than the original statement;
2526 don't do the same transformation here.
2527 (vect_is_simple_use): If the defining statement has been replaced
2528 by a pattern statement, return the pattern statement instead.
2529 Remove the corresponding (local) transformation from the vectype
2532 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2534 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
2535 end and default to null.
2536 * tree-vect-loop.c (vect_create_epilog_for_reduction)
2537 (vectorizable_reduction): Update calls accordingly, dropping the
2538 gimple ** argument if the passed-back statement isn't needed.
2539 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
2540 (vect_recog_rotate_pattern): Likewise.
2541 (vect_recog_mask_conversion_pattern): Likewise.
2542 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2543 (vect_mask_constant_operand_p): Likewise.
2544 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
2545 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
2546 (get_group_load_store_type, get_load_store_type): Likewise.
2547 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
2548 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
2549 (vectorizable_conversion, vectorizable_assignment): Likewise.
2550 (vectorizable_shift, vectorizable_operation): Likewise.
2551 (vectorizable_store, vect_is_simple_cond): Likewise.
2552 (vectorizable_condition, vectorizable_comparison): Likewise.
2553 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
2554 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
2555 and move it to the end. Cope with null def_stmt_outs.
2557 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2559 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
2561 2018-06-29 Jeff Law <law@redhat.com>
2563 * config/v850/v850.c (v850_legitimate_address_p): Handle large
2564 displacements for TARGET_V850E2V3 and newer.
2565 (TARGET_LRA_P): Remove. Defaults to LRA now.
2566 * config/v850/v850.md (sign23byte_load): Remove.
2567 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
2568 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
2570 2018-06-29 Martin Liska <mliska@suse.cz>
2573 * coverage.c (coverage_init): Mangle full path name.
2574 * doc/invoke.texi: Document the change.
2575 * gcov-io.c (mangle_path): New.
2576 * gcov-io.h (mangle_path): Likewise.
2577 * gcov.c (mangle_name): Use mangle_path for path mangling.
2579 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2581 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
2582 if starting source register is not even.
2584 2018-06-29 Martin Liska <mliska@suse.cz>
2586 PR tree-optimization/86263
2587 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
2588 Make edge redirection.
2590 2018-06-29 David Malcolm <dmalcolm@redhat.com>
2592 * dumpfile.c (dump_loc): Add indentation based on scope depth.
2593 (dump_scope_depth): New variable.
2594 (get_dump_scope_depth): New function.
2595 (dump_begin_scope): New function.
2596 (dump_end_scope): New function.
2597 * dumpfile.h (get_dump_scope_depth): New declaration.
2598 (dump_begin_scope): New declaration.
2599 (dump_end_scope): New declaration.
2600 (class auto_dump_scope): New class.
2601 (AUTO_DUMP_SCOPE): New macro.
2602 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
2605 2018-06-29 Richard Biener <rguenther@suse.de>
2607 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
2608 compute_all_dependences succeeds.
2609 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
2610 exceed --param loop-max-datarefs-for-datadeps.
2612 2018-06-29 Jakub Jelinek <jakub@redhat.com>
2614 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
2616 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
2619 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
2620 alternative 0 in preferred_for_speed attribute.
2622 2018-06-28 Paul Koning <ni1d@arrl.net>
2624 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
2625 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
2626 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
2627 * config/pdp11/pdp11.md: Correct "length" attribute calculation
2628 for shift insn patterns.
2630 2018-06-28 David Malcolm <dmalcolm@redhat.com>
2632 * cgraph.c (cgraph_node::get_body): Replace assignments to
2633 "dump_file" with calls to set_dump_file.
2634 * dumpfile.c (alt_dump_file): Make static, and group with...
2635 (alt_flags): ...this definition.
2636 (dumps_are_enabled): New variable.
2637 (refresh_dumps_are_enabled): New function.
2638 (set_dump_file): New function.
2639 (set_alt_dump_file): New function.
2640 (gcc::dump_manager::dump_start): Replace assignments to
2641 "dump_file" and "alt_dump_file" with calls to set_dump_file and
2643 (gcc::dump_manager::dump_finish): Likewise.
2644 * dumpfile.h (alt_dump_file): Delete decl.
2645 (dumps_are_enabled): New variable decl.
2646 (set_dump_file): New function decl.
2647 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
2649 * tree-nested.c (lower_nested_functions): Replace assignments to
2650 "dump_file" with calls to set_dump_file.
2652 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
2654 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
2655 goto_locus of each outgoing edge of each basic block.
2657 2018-06-28 Richard Biener <rguenther@suse.de>
2659 * dwarf2out.c (decl_scope_table): Remove.
2660 (push_decl_scope): Likewise.
2661 (pop_decl_scope): Likewise.
2662 (gen_type_die_for_member): Do not call push/pop_decl_scope.
2663 (gen_struct_or_union_type_die): Likewise.
2664 (gen_tagged_type_die): Likewise.
2665 (dwarf2out_init): Do not initialize decl_scope_table.
2666 (dwarf2out_c_finalize): Do not free it.
2668 2018-06-28 Richard Biener <rguenther@suse.de>
2670 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
2671 deciding whether to not re-use a DIE.
2673 2018-06-28 Richard Biener <rguenther@suse.de>
2675 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
2676 DW_AT_abstract_origin attribute.
2678 2018-06-28 Martin Liska <mliska@suse.cz>
2680 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2681 Use newly introduced constants.
2682 * tree-switch-conversion.h (struct jump_table_cluster):
2683 Define max_ratio_for_size and max_ratio_for_speed.
2685 2018-06-28 Martin Liska <mliska@suse.cz>
2687 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2688 Add new checking assert to catch invalid state.
2689 (jump_table_cluster::can_be_handled): Handle single case
2691 (jump_table_cluster::is_beneficial): Bail out for such case.
2692 (bit_test_cluster::find_bit_tests):
2693 Add new checking assert to catch invalid state.
2694 (bit_test_cluster::can_be_handled): Handle single case
2696 (bit_test_cluster::is_beneficial): Bail out for such case.
2697 (switch_decision_tree::analyze_switch_statement):
2700 2018-06-28 Martin Liska <mliska@suse.cz>
2702 * common.opt: Introduce -completion option.
2703 * gcc.c (driver_handle_option): Handle it.
2704 (driver::main): Print completions if completion
2706 * opt-suggestions.c (option_proposer::get_completions):
2708 (option_proposer::suggest_completion): Likewise.
2709 (option_proposer::find_param_completions): Likewise.
2710 (verify_autocompletions): Likewise.
2711 (test_completion_valid_options): Likewise.
2712 (test_completion_valid_params): Likewise.
2713 (in_completion_p): Likewise.
2714 (empty_completion_p): Likewise.
2715 (test_completion_partial_match): Likewise.
2716 (test_completion_garbage): Likewise.
2717 (opt_proposer_c_tests): Likewise.
2718 * opt-suggestions.h: Declare new functions.
2719 * opts.c (common_handle_option): Handle OPT__completion_.
2720 * selftest-run-tests.c (selftest::run_tests): Add
2721 opt_proposer_c_tests.
2722 * selftest.c (assert_str_startswith): New.
2723 * selftest.h (assert_str_startswith): Likewise.
2724 (opt_proposer_c_tests): New.
2725 (ASSERT_STR_STARTSWITH): Likewise.
2727 2018-06-28 Martin Liska <mliska@suse.cz>
2729 * Makefile.in: Add opt-suggestions.o.
2730 * gcc-main.c: Include opt-suggestions.h.
2731 * gcc.c (driver::driver): Likewise.
2732 (driver::~driver): Remove m_option_suggestions.
2733 (driver::build_option_suggestions): Moved to option_proposer.
2734 (driver::suggest_option): Likewise.
2735 (driver::handle_unrecognized_options): Use option_proposer.
2736 * gcc.h (class driver): Add new memver m_option_proposer.
2737 * opt-suggestions.c: New file.
2738 * opt-suggestions.h: New file.
2740 2018-06-28 Martin Liska <mliska@suse.cz>
2742 * vec.h (class auto_string_vec): New (moved from auto_argvec).
2743 (auto_string_vec::~auto_string_vec): Likewise.
2745 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
2747 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
2748 prevent_decl_creation_for_types fields up and add reset_location field.
2749 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
2750 statement if id->reset_location is true.
2751 (copy_edges_for_bb): Do not set goto_locus on the new edges if
2752 id->reset_location is true.
2753 (copy_phis_for_bb): Force input_location on the arguments if
2754 id->reset_location is true.
2755 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
2756 is set on the function to be inlined.
2758 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
2760 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
2762 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
2764 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
2765 registers for Pmode.
2766 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
2767 hard registers for the clobbered pseudo.
2769 2018-06-27 Paul Koning <ni1d@arrl.net>
2771 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
2772 mutually exclusive options.
2773 * config/pdp11/constraints.md (h): New constraint.
2774 (O): Update definition to match shift code generation.
2775 (D): New constraint.
2776 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
2778 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
2780 (output_jump): Change arguments.
2781 (pdp11_fixed_cc_regs): New function.
2782 (pdp11_cc_mode): Ditto.
2783 (pdp11_expand_shift): Ditto.
2784 (pdp11_assemble_shift): Ditto.
2785 (pdp11_small_shift): Ditto.
2786 (pdp11_branch_cost): Remove.
2787 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
2789 (pdp11_register_move_cost): Update for CC registers.
2790 (pdp11_rtx_costs): Add case for LSHIFTRT.
2791 (pdp11_output_jump): Add CCNZ mode conditional branches.
2792 (notice_update_cc_on_set): Remove.
2793 (pdp11_cc_mode): New function.
2794 (simple_memory_operand): Correct pre/post decrement case.
2795 (no_side_effect_operand): New function.
2796 (pdp11_regno_reg_class): Add CC_REGS class.
2797 (pdp11_fixed_cc_regs): New function.
2798 (pdp11_small_shift): New function.
2799 (pdp11_expand_shift): New function to expand shift insns.
2800 (pdp11_assemble_shift): New function to output shifts.
2801 (pdp11_branch_cost): Remove.
2802 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
2803 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
2804 (WCHAR_TYPE): Ditto.
2805 (PTRDIFF_TYPE): Ditto.
2806 (ADJUST_INSN_LENGTH): New macro.
2807 (FIXED_REGISTERS): Add CC registers.
2808 (CALL_USED_REGISTERS): Ditto.
2810 (REG_CLASS_NAMES): Ditto.
2811 (REG_CLASS_CONTENTS): Ditto.
2812 (SELECT_CC_MODE): Use new function.
2813 (TARGET_FLAGS_REGNUM): New macro.
2814 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
2815 (cc0_reg_rtx): Remove.
2816 (CC_STATUS_MDEP): Remove.
2817 (CC_STATUS_MDEFP_INIT): Remove.
2818 (CC_IN_FPU): Remove.
2819 (NOTICE_UPDATE_CC): Remove.
2820 (REGISTER_NAMES): Add CC registers.
2821 (BRANCH_COST): Change to constant 1.
2822 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
2824 * config/pdp11/pdp11.opt (mbcopy): Remove.
2825 (mbcopy-builtin): Remove.
2826 (mbranch-cheap): Remove.
2827 (mbranch-expensive): Remove.
2828 * config/pdp11/predicates.md (expand_shift_operand): Update to
2829 match shift code generation.
2830 (ccnz_operator): New predicate.
2831 * doc/invoke.texi (PDP-11 Options): Remove deleted options
2832 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
2833 Remove non-existent option -mabshi, -mno-abshi. Document mutually
2835 * doc/md.texi (PDP-11): Document new D and h constraints. Update
2836 description of O constraint.
2838 2018-06-27 Jeff Law <law@redhat.com>
2839 Austin Law <austinklaw@gmail.com>
2841 * config/v850/v850.md (addsi3_set_flags): New pattern.
2842 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
2843 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
2844 (zero_extendhisi2_v850_set_flags): Likewise.
2845 (zero_extendqisi2_v850_set_flags): Likewise.
2846 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
2847 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
2848 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
2850 * config/v850/v850-protos.h (notice_update_cc): Remove.
2851 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
2852 (v850_print_operand): Handle 'D' and "d".
2853 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
2854 Add handling of arithmetic/logical operations compared against zero.
2855 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
2856 Do not look at v850_compare_op, instead get mode from last argument.
2857 (v850_gen_compare): Remove
2858 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
2859 after reload for prologue insns.
2860 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
2862 (construct_save_jarl): Likewise.
2863 (TARGET_FLAGS_REGNUM): Define.
2864 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
2865 (NOTICE_UPDATE_CC): Remove.
2866 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
2867 than cc0. Conditionalize on reload_completed.
2868 (cmpsi_insn, setfcc_insn): Likewise.
2869 (tst1 splitter): Turn into define_and_split which sets the flags
2871 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
2872 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
2873 (cstoresf4, cstoredf4): Clobber the flags.
2874 (cmpsi, cmpsf, cmpdf): Remove expanders.
2875 (setf_insn): Remove pattern.
2876 (addsi3): Turn into define_and_split which clobbers the flags after
2877 reload and a suitable pattern (addsi3_clobber_flags) for use after
2879 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
2880 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
2881 (ashrsi3, ashrsi3_v850e2): Likewise.
2882 (bins): Clobber the flags.
2883 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
2884 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
2885 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
2886 (call_value_internal_short, call_value_internal_long): Likewise.
2887 (callt_save_interrupt, callt_return_interrupt): Likewise.
2888 (save_interrupt, return_interrupt): Likewise.
2889 (callt_save_all_interrupt, save_all_interrupt): Likewise.
2890 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
2891 (restore_all_interrupt, _restore_all_interrupt): Likewise.
2892 (All FP comparisons): Only allow after reload has completed.
2894 (divh, divhu): Tweak output template.
2895 (branch_z_normal, branch_z_invert): Remove
2896 (branch_nz_normal, branch_nz_invert): Likewise.
2897 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
2899 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
2900 * config/v850/v850.c (notice_update_cc): Remove.
2901 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
2902 (CC_NO_CARRY): Likewise.
2903 (NOTICE_UPDATE_CC): Define to nothing.
2904 * config/v850/v850.md: Remove block comment on cc0 handling
2905 Remove "cc" attribute from all patterns. Remove cc_status handling
2906 from all patterns. Minor formatting fixes.
2908 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2910 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
2911 (cortex-a76.cortex-a55): Likewise.
2912 * config/aarch64/aarch64-tune.md: Regenerate.
2913 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
2914 cortex-a76.cortex-a55.
2916 2018-06-27 Jeff Law <law@redhat.com>
2918 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
2919 (MULTILIB_DIRNAMES): Similarly.
2921 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
2923 * gimple.h (gimple_return_retbnd): Delete.
2924 (gimple_return_set_retbnd): Likewise.
2925 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
2926 gimple_return_set_retbnd.
2927 * gimple-pretty-print.c (dump_gimple_return): Remove call to
2928 gimple_return_retbnd and adjust.
2929 * tree-inline.h (struct copy_body_data): Remove retbnd field.
2930 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
2931 Explicitly return NULL in a couple more cases. Move assertion
2932 on debug statements and remove unreachable code.
2933 (reset_debug_binding): Do not test id->retbnd.
2934 (expand_call_inline): Do not set it.
2936 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2938 * configure.ac: Add --disable-gcov option.
2939 * configure: Regenerate.
2940 * Makefile.in: Honour @enable_gcov@.
2941 * doc/install.texi: Document --disable-gcov.
2943 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2945 * config/arm/arm-cpus.in (cortex-a76): New entry.
2946 (cortex-a76.cortex-a55): Likewise.
2947 * config/arm/arm-tables.opt: Regenerate.
2948 * config/arm/arm-tune.md: Likewise.
2949 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
2950 * doc/invoke.texi (ARM Options): Document cortex-a76 and
2951 cortex-a76.cortex-a55.
2953 2018-06-27 Tamar Christina <tamar.christina@arm.com>
2956 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
2958 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
2960 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
2962 (EPILOGUE_USES): Likewise.
2964 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
2966 * tree-inline.c (remap_location): New function extracted from...
2967 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
2968 (copy_phis_for_bb): ...here. Call remap_location.
2969 (copy_cfg_body): Adjust call to copy_edges_for_bb.
2971 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
2973 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
2974 unaligned vsx for 16B memset.
2976 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
2979 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
2980 ieee128_float_type_node to long_double_type_node unless
2981 TARGET_LONG_DOUBLE_128 is set.
2983 2018-06-26 David Malcolm <dmalcolm@redhat.com>
2985 * cfgloop.c (get_loop_location): Convert return type from
2986 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
2987 by implicit construction from rtx_insn *, and using
2988 dump_user_location_t::from_function_decl for the fallback case.
2989 * cfgloop.h (get_loop_location): Convert return type from
2990 location_t to dump_user_location_t.
2991 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
2992 dump_printf_loc to pass in a dump_location_t rather than a
2993 location_t, via the gimple stmt.
2994 * coverage.c (get_coverage_counts): Update calls to
2995 dump_printf_loc to pass in dump_location_t rather than a
2997 * doc/optinfo.texi (Dump types): Convert example of
2998 dump_printf_loc from taking "locus" to taking "insn". Update
2999 description of the "_loc" calls to cover dump_location_t.
3000 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
3002 (dump_user_location_t::dump_user_location_t): New constructors,
3003 from gimple *stmt and rtx_insn *.
3004 (dump_user_location_t::from_function_decl): New function.
3005 (dump_loc): Make static.
3006 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
3007 const dump_location_t &.
3008 (dump_generic_expr_loc): Delete.
3009 (dump_printf_loc): Convert param "loc" from location_t to
3010 const dump_location_t &.
3011 (selftest::test_impl_location): New function.
3012 (selftest::dumpfile_c_tests): New function.
3013 * dumpfile.h: Include "profile-count.h".
3014 (class dump_user_location_t): New class.
3015 (struct dump_impl_location_t): New struct.
3016 (class dump_location_t): New class.
3017 (dump_printf_loc): Convert 2nd param from source_location to
3018 const dump_location_t &.
3019 (dump_generic_expr_loc): Delete.
3020 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
3021 const dump_location_t &.
3022 * gimple-fold.c (fold_gimple_assign): Update call to
3023 dump_printf_loc to pass in a dump_location_t rather than a
3024 location_t, via the gimple stmt.
3025 (gimple_fold_call): Likewise.
3026 * gimple-loop-interchange.cc
3027 (loop_cand::analyze_iloop_reduction_var): Update for change to
3028 check_reduction_path.
3029 (tree_loop_interchange::interchange): Update for change to
3031 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
3032 change in return-type of find_loop_location.
3033 (graphite_regenerate_ast_isl): Likewise.
3034 * graphite-optimize-isl.c (optimize_isl): Likewise.
3035 * graphite.c (graphite_transform_loops): Likewise.
3036 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
3037 pass in a dump_location_t rather than a location_t, via the
3039 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3040 * ipa.c (walk_polymorphic_call_targets): Likewise.
3041 * loop-unroll.c (report_unroll): Convert "locus" param from
3042 location_t to dump_location_t.
3043 (decide_unrolling): Update for change to get_loop_location's
3045 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
3046 location_t to dump_user_location_t.
3047 (grid_find_single_omp_among_assignments_1): Updates calls to
3048 dump_printf_loc to pass in a dump_location_t rather than a
3049 location_t, via the gimple stmt.
3050 (grid_parallel_clauses_gridifiable): Convert "tloc" from
3051 location_t to dump_location_t. Updates calls to dump_printf_loc
3052 to pass in a dump_location_t rather than a location_t, via the
3054 (grid_inner_loop_gridifiable_p): Likewise.
3055 (grid_dist_follows_simple_pattern): Likewise.
3056 (grid_gfor_follows_tiling_pattern): Likewise.
3057 (grid_target_follows_gridifiable_pattern): Likewise.
3058 (grid_attempt_target_gridification): Convert initialization
3059 of local "grid" from memset to zero-initialization; FIXME: does
3060 this require C++11? Update call to dump_printf_loc to pass in a
3061 optinfo_location rather than a location_t, via the gimple stmt.
3062 * profile.c (read_profile_edge_counts): Updates call to
3063 dump_printf_loc to pass in a dump_location_t rather than a
3065 (compute_branch_probabilities): Likewise.
3066 * selftest-run-tests.c (selftest::run_tests): Call
3068 * selftest.h (dumpfile_c_tests): New decl.
3069 * tree-loop-distribution.c (pass_loop_distribution::execute):
3070 Update for change in return type of find_loop_location.
3071 * tree-parloops.c (parallelize_loops): Likewise.
3072 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
3073 "locus" from location_t to dump_user_location_t.
3074 (canonicalize_loop_induction_variables): Likewise.
3075 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
3076 for change in return type of find_loop_location.
3077 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
3078 to dump_printf_loc to pass in a dump_location_t rather than a
3079 location_t, via the stmt.
3080 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
3082 * tree-vect-loop-manip.c (find_loop_location): Convert return
3083 type from source_location to dump_user_location_t.
3084 (vect_do_peeling): Update for above change.
3085 (vect_loop_versioning): Update for change in type of
3087 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
3088 from location_t to dump_user_location_t.
3089 (vect_estimate_min_profitable_iters): Update for change in type
3091 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
3092 location_t to dump_location_t.
3093 (vect_slp_bb): Update for change in type of vect_location.
3094 * tree-vectorizer.c (vect_location): Convert from source_location
3095 to dump_user_location_t.
3096 (try_vectorize_loop_1): Update for change in vect_location's type.
3097 (vectorize_loops): Likewise.
3098 (increase_alignment): Likewise.
3099 * tree-vectorizer.h (vect_location): Convert from source_location
3100 to dump_user_location_t.
3101 (find_loop_location): Convert return type from source_location to
3102 dump_user_location_t.
3103 (check_reduction_path): Convert 1st param from location_t to
3104 dump_user_location_t.
3105 * value-prof.c (check_counter): Update call to dump_printf_loc to
3106 pass in a dump_user_location_t rather than a location_t; update
3107 call to error_at for change in type of "locus".
3108 (check_ic_target): Update call to dump_printf_loc to
3109 pass in a dump_user_location_t rather than a location_t, via the
3112 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
3114 * config/s390/s390.h (enum processor_flags): Do not use
3115 default tune parameter when -march was specified.
3117 2018-06-26 Jakub Jelinek <jakub@redhat.com>
3120 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
3121 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
3124 2018-06-26 Richard Biener <rguenther@suse.de>
3126 PR tree-optimization/86287
3128 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
3129 (vect_analyze_loop): Initialize n_stmts.
3131 2018-06-26 Richard Biener <rguenther@suse.de>
3134 * fold-const.c (fold_convertible_p): Pointer extension
3137 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
3140 * dwarf2out.c (loc_list_has_views): Adjust comments.
3141 (dw_loc_list): Split single cross-partition range with
3144 2018-06-25 Jeff Law <law@redhat.com>
3146 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
3147 on -mbig-switch by default.
3149 * config/v850/predicates.md (const_float_1_operand): Fix match_code
3151 (const_float_0_operand): Remove unused predicate.
3152 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
3153 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
3154 (recipsf2): New expander. Original pattern now called
3156 (recipdf2, recipdf2_insn): Similarly.
3157 (rsqrtsf2, rsqrtsf2_insn): Similarly
3158 (rsqrtdf2, rsqrtdf2_insn): Similarly
3160 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
3162 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
3163 Simplify logic for FreeBSD (twice).
3165 2018-06-25 Martin Sebor <msebor@redhat.com>
3167 PR tree-optimization/86204
3168 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
3169 a strnlen result if it's less than the length of the string.
3171 2018-06-25 Martin Sebor <msebor@redhat.com>
3173 PR tree-optimization/85700
3174 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
3175 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
3176 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
3178 2018-06-25 Martin Sebor <msebor@redhat.com>
3180 * doc/extend.texi (Zero-length arrays): Update and clarify.
3182 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
3184 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
3185 added IEEE/IBM long double multilib support on PowerPC little
3186 endian Linux systems.
3187 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
3188 (MULTILIB_DEFAULTS): Likewise.
3189 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3191 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
3192 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
3193 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
3195 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
3198 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
3199 (REORDER_45): Likewise.
3201 2018-06-25 Jeff Law <law@redhat.com>
3203 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
3204 dividend to 32 bits. Adjust length.
3205 (udivmodhi4): Cleanup output template. Fix length.
3207 2018-06-25 Carl Love <cel@us.ibm.com>
3209 * config/rs6000/vsx.md: Change word selector to prefered location.
3211 2018-06-25 Richard Biener <rguenther@suse.de>
3213 PR tree-optimization/86304
3214 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
3215 epilogue-if-converted loops as well.
3217 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
3219 * lto-section-out.c (lto_begin_section): Do not print section
3220 name for noaddr and unnumbered dumps.
3222 2018-06-25 Richard Biener <rguenther@suse.de>
3224 * tree-vectorizer.h (struct vec_info_shared): New structure
3225 with parts split out from struct vec_info and loop_nest from
3226 struct _loop_vec_info.
3227 (struct vec_info): Adjust accordingly.
3228 (struct _loop_vec_info): Likewise.
3229 (LOOP_VINFO_LOOP_NEST): Adjust.
3230 (LOOP_VINFO_DATAREFS): Likewise.
3231 (LOOP_VINFO_DDRS): Likewise.
3232 (struct _bb_vec_info): Likewise.
3233 (BB_VINFO_DATAREFS): Likewise.
3234 (BB_VINFO_DDRS): Likewise.
3235 (struct _stmt_vec_info): Add dr_aux member.
3236 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
3237 (DR_MISALIGNMENT_UNINITIALIZED): New.
3238 (set_dr_misalignment): Adjust.
3239 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
3240 (vect_analyze_loop): Adjust prototype.
3241 (vect_analyze_loop_form): Likewise.
3242 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
3243 Compute dependences lazily.
3244 (vect_record_base_alignments): Use shared datarefs/ddrs.
3245 (vect_verify_datarefs_alignment): Likewise.
3246 (vect_analyze_data_refs_alignment): Likewise.
3247 (vect_analyze_data_ref_accesses): Likewise.
3248 (vect_analyze_data_refs): Likewise.
3249 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
3250 constructor parameter for shared part.
3251 (vect_analyze_loop_form): Pass in shared part and adjust.
3252 (vect_analyze_loop_2): Pass in storage for the number of
3253 stmts. Move loop nest finding to the caller. Compute
3255 (vect_analyze_loop): Pass in shared part.
3256 (vect_transform_loop): Verify shared datarefs are unchanged.
3257 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
3258 constructor parameter for shared part.
3259 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
3260 (vect_slp_bb): Verify shared datarefs are unchanged before
3262 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
3264 (new_stmt_vec_info): Initialize DR_AUX misalignment to
3265 DR_MISALIGNMENT_UNINITIALIZED.
3266 * tree-vectorizer.c (vec_info::vec_info): Add constructor
3267 parameter for shared part.
3268 (vec_info::~vec_info): Adjust.
3269 (vec_info_shared::vec_info_shared): New.
3270 (vec_info_shared::~vec_info_shared): Likewise.
3271 (vec_info_shared::save_datarefs): Likewise.
3272 (vec_info_shared::check_datarefs): Likewise.
3273 (try_vectorize_loop_1): Construct shared part live for analyses
3274 of a single loop for multiple vector sizes.
3275 * tree-parloops.c (gather_scalar_reductions): Adjust.
3277 2018-06-25 Richard Biener <rguenther@suse.de>
3279 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
3280 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
3281 (vect_analyze_data_refs): Remove similar code from here and
3282 simplify accordingly.
3284 2018-06-25 Richard Biener <rguenther@suse.de>
3286 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
3287 for reverse storage order accesses rather than asserting
3288 they cannot happen here.
3290 2018-06-25 Tom de Vries <tdevries@suse.de>
3293 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
3294 Use data16 instead of .byte for insn prefix.
3296 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
3299 * parser.c (make_char_string_pack): Pass this literal chars
3300 through cpp_interpret_string.
3301 (cp_parser_userdef_numeric_literal): Check the result of
3302 make_char_string_pack.
3304 2018-06-24 Maya Rashish <coypu@sdf.org>
3306 * ginclude/stddef.h: Simplify conditions around avoiding
3307 re-definition of __size_t.
3309 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3311 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
3312 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
3314 2018-06-22 Maya Rashish <coypu@sdf.org>
3316 * doc/invoke.texi (mno-fancy-math-387): Update for changes
3317 made to OpenBSD and NetBSD through the years.
3319 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3321 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3322 behavior of vec_pack (vector double, vector double) to match
3323 behavior of vec_float2 (vector double, vector double).
3325 2018-06-22 Olivier Hainque <hainque@adacore.com>
3327 * gimplify.c (gimplify_function_tree): Prevent creation
3328 of a trampoline for the address of the current function
3329 passed to entry/exit instrumentation hooks.
3331 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
3334 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
3337 2018-06-22 Martin Liska <mliska@suse.cz>
3339 PR tree-optimization/86263
3340 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3341 Bail out if is_enabled is false.
3342 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
3344 (jump_table_cluster::is_enabled): New function.
3346 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3348 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
3349 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
3350 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
3351 (lto_input_ts_binfo_tree_pointers): Likewise.
3352 * tree-streamer-out.c (streamer_write_tree_bitfields,
3353 write_ts_binfo_tree_pointers): Likewise.
3354 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
3356 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3358 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
3360 2018-06-22 Martin Liska <mliska@suse.cz>
3362 * symbol-summary.h (get): Make it pure and inline move
3363 functionality from ::get function.
3364 (get): Remove and inline into ::get and ::get_create.
3365 (get_create): Move code from ::get function.
3367 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3370 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
3371 -x assembler-with-cpp.
3373 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3375 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
3376 _FILE_OFFSET_BITS=64 for C++.
3378 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
3380 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
3381 conversion insn that shows up when pr85657-3.c is compiled using
3382 IEEE 128-bit long double.
3383 (neg<mode>2_internal): Use the correct mode to check whether the
3384 mode is IBM extended.
3385 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
3386 multiply and divide external functions from being created more
3389 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3391 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
3393 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
3394 the edge can be forwarded.
3395 (cfg_layout_merge_blocks): Likewise.
3397 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3399 * except.c (finish_eh_generation): Commit edge insertions only after
3400 the EH edges have been redirected from post-landing to landing pads.
3402 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3404 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
3405 create_tmp_var_for to create the FRAME decl.
3406 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
3408 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3410 * tree-inline.c (copy_edges_for_bb): Minor tweak.
3411 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
3412 debug statement when resetting its value.
3413 (expand_call_inline): Copy the locus of the call onto the assignment
3414 of the return value, if any. Use local variable in more cases.
3416 2018-06-21 Martin Liska <mliska@suse.cz>
3418 * ipa-pure-const.c (propagate_nothrow): Use
3419 funct_state_summaries->get.
3420 (dump_malloc_lattice): Likewise.
3421 (propagate_malloc): Likewise.
3423 2018-06-21 Richard Biener <rguenther@suse.de>
3425 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
3426 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
3427 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
3429 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
3430 BLOCK_ABSTRACT_ORIGIN unconditionally.
3432 2018-06-21 David Malcolm <dmalcolm@redhat.com>
3434 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
3436 * ipa-reference.c (ipa_reference_c_finalize): Delete
3437 ipa_ref_opt_sum_summaries and set it to NULL.
3439 2018-06-21 Tom de Vries <tdevries@suse.de>
3441 PR tree-optimization/85859
3442 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
3443 test with comment from bb_no_side_effects_p.
3445 2018-06-21 Richard Biener <rguenther@suse.de>
3447 PR tree-optimization/86232
3448 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
3449 max for constant niter.
3451 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
3453 * config/aarch64/aarch64-simd.md
3454 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
3456 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
3458 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
3459 Make opernads of the unspec commutative.
3461 2018-06-21 Richard Biener <rguenther@suse.de>
3463 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
3464 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
3465 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
3466 (vect_analyze_data_ref_dependence): Re-order checks to deal with
3468 (vect_record_base_alignments): Do not record base alignment
3469 for gathers or scatters.
3470 (vect_compute_data_ref_alignment): Drop return value that is always
3471 true. Bail out early for gathers or scatters.
3472 (vect_enhance_data_refs_alignment): Bail out early for gathers
3474 (vect_find_same_alignment_drs): Likewise.
3475 (vect_analyze_data_refs_alignment): Remove dead code.
3476 (vect_slp_analyze_and_verify_node_alignment): Likewise.
3477 (vect_analyze_data_refs): For possible gathers or scatters do
3478 not create an alternate DR, just check their possible validity
3479 and mark them. Adjust DECL_NONALIASED handling to not rely
3481 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
3482 update inits of gathers or scatters.
3483 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
3484 Also copy gather/scatter flag to pattern vinfo.
3486 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
3488 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3489 behavior of vec_packsu (vector unsigned long long, vector unsigned
3490 long long) to match behavior of vec_packs with same signature.
3492 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
3493 Thomas Schwinge <thomas@codesourcery.com>
3494 Cesar Philippidis <cesar@codesourcery.com>
3496 * gimplify.c (gimplify_scan_omp_clauses): Add support for
3497 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
3498 (gimplify_adjust_omp_clauses): Likewise.
3499 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
3500 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
3501 (gimplify_omp_target_update): Update handling of acc update and
3503 * omp-low.c (install_var_field): Remove unused parameter
3504 base_pointers_restrict.
3505 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
3506 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
3508 (omp_target_base_pointers_restrict_p): Delete.
3509 (scan_omp_target): Update call to scan_sharing_clauses.
3510 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
3512 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
3513 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
3514 (convert_local_omp_clauses): Likewise.
3515 * tree-pretty-print.c (dump_omp_clause): Likewise.
3516 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
3518 (omp_clause_code_name): Likewise.
3520 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3523 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
3526 PR tree-optimization/86231
3527 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
3528 anti-range don't overwrite *vr0min before using it to compute *vr0max.
3530 2018-06-20 Tom de Vries <tdevries@suse.de>
3532 PR tree-optimization/86097
3533 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
3534 iv type if signedness of iv type is not the same as that of *nit.
3536 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3538 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
3539 EDGE_EH edges, verify they are all EDGE_EH.
3541 2018-06-20 Maya Rashish <coypu@sdf.org>
3543 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
3545 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3547 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
3548 * config/aarch64/aarch64.c (xgene1_tunings): Add
3549 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
3550 (aarch64_mode_valid_for_sched_fusion_p):
3551 Allow 16-byte modes.
3552 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
3553 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
3555 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
3557 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
3558 * config/aarch64/iterators.md (VQ2): New mode iterator.
3560 2018-06-20 Martin Liska <mliska@suse.cz>
3562 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
3563 Change default ratio from 10 to 8.
3565 2018-06-20 Martin Liska <mliska@suse.cz>
3567 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3569 (bit_test_cluster::find_bit_tests): Likewise.
3570 (switch_decision_tree::analyze_switch_statement): Find clusters.
3571 * tree-switch-conversion.h (struct jump_table_cluster): Document
3574 2018-06-20 Martin Liska <mliska@suse.cz>
3576 * tree-switch-conversion.c (switch_conversion::collect):
3577 Record m_uniq property.
3578 (switch_conversion::expand): Bail out for special conditions.
3579 (group_cluster::~group_cluster): New.
3580 (group_cluster::group_cluster): Likewise.
3581 (group_cluster::dump): Likewise.
3582 (jump_table_cluster::emit): New.
3583 (switch_decision_tree::fix_phi_operands_for_edges): New.
3584 (struct case_node): Remove struct.
3585 (jump_table_cluster::can_be_handled): New.
3586 (case_values_threshold): Moved to header.
3587 (reset_out_edges_aux): Likewise.
3588 (jump_table_cluster::is_beneficial): New.
3589 (bit_test_cluster::can_be_handled): Likewise.
3590 (add_case_node): Remove.
3591 (bit_test_cluster::is_beneficial): New.
3592 (case_bit_test::cmp): New.
3593 (bit_test_cluster::emit): New.
3594 (expand_switch_as_decision_tree_p): Remove.
3595 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
3596 (fix_phi_operands_for_edge): Likewise.
3597 (switch_decision_tree::analyze_switch_statement): New.
3598 (compute_cases_per_edge): Move ...
3599 (switch_decision_tree::compute_cases_per_edge): ... here.
3600 (try_switch_expansion): Likewise.
3601 (switch_decision_tree::try_switch_expansion): Likewise.
3602 (record_phi_operand_mapping): Likewise.
3603 (switch_decision_tree::record_phi_operand_mapping): Likewise.
3604 (emit_case_decision_tree): Likewise.
3605 (switch_decision_tree::emit): Likewise.
3606 (balance_case_nodes): Likewise.
3607 (switch_decision_tree::balance_case_nodes): Likewise.
3608 (dump_case_nodes): Likewise.
3609 (switch_decision_tree::dump_case_nodes): Likewise.
3610 (emit_jump): Likewise.
3611 (switch_decision_tree::emit_jump): Likewise.
3612 (emit_cmp_and_jump_insns): Likewise.
3613 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
3614 (emit_case_nodes): Likewise.
3615 (switch_decision_tree::emit_case_nodes): Likewise.
3616 (conditional_probability): Remove.
3617 * tree-switch-conversion.h (enum cluster_type): New.
3619 (struct cluster): Likewise.
3620 (cluster::cluster): Likewise.
3621 (struct simple_cluster): Likewise.
3622 (simple_cluster::simple_cluster): Likewise.
3623 (struct group_cluster): Likewise.
3624 (struct jump_table_cluster): Likewise.
3625 (struct bit_test_cluster): Likewise.
3626 (struct min_cluster_item): Likewise.
3627 (struct case_tree_node): Likewise.
3628 (case_tree_node::case_tree_node): Likewise.
3629 (jump_table_cluster::case_values_threshold): Likewise.
3630 (struct case_bit_test): Likewise.
3631 (struct switch_decision_tree): Likewise.
3632 (struct switch_conversion): Likewise.
3633 (switch_decision_tree::reset_out_edges_aux): Likewise.
3635 2018-06-20 Martin Liska <mliska@suse.cz>
3637 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
3638 (hoist_edge_and_branch_if_true): Likewise.
3639 (expand_switch_using_bit_tests_p): Likewise.
3640 (struct case_bit_test): Likewise.
3641 (case_bit_test_cmp): Likewise.
3642 (emit_case_bit_tests): Likewise.
3643 (switch_conversion::switch_conversion): New class.
3644 (struct switch_conv_info): Remove old struct.
3645 (collect_switch_conv_info): More to ...
3646 (switch_conversion::collect): ... this.
3647 (check_range): Likewise.
3648 (switch_conversion::check_range): Likewise.
3649 (check_all_empty_except_final): Likewise.
3650 (switch_conversion::check_all_empty_except_final): Likewise.
3651 (check_final_bb): Likewise.
3652 (switch_conversion::check_final_bb): Likewise.
3653 (create_temp_arrays): Likewise.
3654 (switch_conversion::create_temp_arrays): Likewise.
3655 (free_temp_arrays): Likewise.
3656 (gather_default_values): Likewise.
3657 (switch_conversion::gather_default_values): Likewise.
3658 (build_constructors): Likewise.
3659 (switch_conversion::build_constructors): Likewise.
3660 (constructor_contains_same_values_p): Likewise.
3661 (switch_conversion::contains_same_values_p): Likewise.
3662 (array_value_type): Likewise.
3663 (switch_conversion::array_value_type): Likewise.
3664 (build_one_array): Likewise.
3665 (switch_conversion::build_one_array): Likewise.
3666 (build_arrays): Likewise.
3667 (switch_conversion::build_arrays): Likewise.
3668 (gen_def_assigns): Likewise.
3669 (switch_conversion::gen_def_assigns): Likewise.
3670 (prune_bbs): Likewise.
3671 (switch_conversion::prune_bbs): Likewise.
3672 (fix_phi_nodes): Likewise.
3673 (switch_conversion::fix_phi_nodes): Likewise.
3674 (gen_inbound_check): Likewise.
3675 (switch_conversion::gen_inbound_check): Likewise.
3676 (process_switch): Use the newly created class.
3677 (switch_conversion::expand): New.
3678 (switch_conversion::~switch_conversion): New.
3679 * tree-switch-conversion.h: New file.
3681 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3683 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
3684 tree-vect-patterns.c.
3685 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
3686 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
3687 (vect_recog_sad_pattern): Likewise.
3688 (vect_recog_widen_sum_pattern): Likewise.
3689 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
3690 (vect_recog_widen_shift_pattern): Remove the type_in argument.
3691 (vect_recog_rotate_pattern): Likewise.
3692 (vect_recog_mult_pattern): Likewise.
3693 (vect_recog_vector_vector_shift_pattern): Likewise.
3694 (vect_recog_divmod_pattern): Likewise.
3695 (vect_recog_mixed_size_cond_pattern): Likewise.
3696 (vect_recog_bool_pattern): Likewise.
3697 (vect_recog_mask_conversion_pattern): Likewise.
3698 (vect_try_gather_scatter_pattern): Likewise.
3699 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
3700 (vect_recog_over_widening_pattern): Likewise.
3701 (vect_recog_gather_scatter_pattern): Likewise.
3702 (vect_recog_func_ptr): Move from tree-vectorizer.h
3703 (vect_vect_recog_func_ptrs): Move further down the file.
3704 (vect_recog_func): Likewise. Remove the third argument.
3705 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
3706 (vect_pattern_recog_1): Expect the pattern function to do any
3707 necessary target tests. Also expect it to provide a vector type.
3708 Remove the type_in handling.
3710 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3712 * tree-vect-patterns.c (vect_pattern_detected): New function.
3713 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
3714 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
3715 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
3716 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
3717 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
3718 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
3719 (vect_recog_mask_conversion_pattern)
3720 (vect_try_gather_scatter_pattern): Likewise.
3722 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3724 * tree-vect-patterns.c (vect_get_internal_def): New function.
3725 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
3726 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
3727 (search_type_for_mask_1): Use it.
3729 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3731 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
3732 redundant WIDEN_SUM_EXPR handling.
3733 (vect_recog_sad_pattern): Likewise.
3735 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3737 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
3738 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
3739 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
3740 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
3741 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
3743 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3745 * tree-vect-stmts.c (vectorizable_call): Make sure that we
3746 use the stmt_vec_info of the original bb statement for the
3747 new zero assignment, even if the call is part of a pattern.
3749 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3751 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
3752 that the sequence is attached to the original statement rather
3753 than the pattern statement.
3754 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
3755 PATTERN_DEF_SEQ from the original statement rather than
3756 the main pattern statement.
3757 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
3758 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
3759 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
3761 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3763 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
3764 definition statements before the early exit for statements that aren't
3766 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
3768 (vect_transform_loop): ...here. Process pattern definition
3769 statements without first checking whether the main pattern
3770 statement is live or relevant.
3772 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
3774 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
3775 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
3777 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
3779 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
3780 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
3781 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
3782 (expand_block_compare): Change select_block_compare_mode call.
3783 (expand_strncmp_align_check): Use new functions, fix comment.
3784 (emit_final_str_compare_gpr): New function.
3785 (expand_strn_compare): Refactor and clean up code.
3786 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
3788 2018-06-19 Tony Reix <tony.reix@atos.com>
3789 Damien Bergamini <damien.bergamini@atos.com>
3790 David Edelsohn <dje.gcc@gmail.com>
3792 * collect2.c (static_obj): New variable.
3793 (static_libs): New variable.
3794 (is_in_list): Uncomment declaration.
3795 (main): Track AIX libraries linked statically.
3796 (is_in_list): Uncomment definition.
3797 (scan_prog_file): Don't add AIX shared libraries initializer
3798 to constructor list if linking statically.
3800 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
3802 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
3804 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
3806 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
3808 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
3811 2018-06-19 Martin Liska <mliska@suse.cz>
3813 * config/i386/i386.c (ix86_can_inline_p): Do not use
3814 ipa_fn_summaries::get_create.
3815 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
3817 (devirtualization_time_bonus): Likewise.
3818 (ipcp_propagate_stage): Likewise.
3819 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3820 (edge_set_predicate): Likewise.
3821 (evaluate_conditions_for_known_args): Likewise.
3822 (evaluate_properties_for_edge): Likewise.
3823 (ipa_call_summary::reset): Tranform to ...
3824 (ipa_call_summary::~ipa_call_summary): ... this.
3825 (ipa_fn_summary::reset): Transform to ...
3826 (ipa_fn_summary::~ipa_fn_summary): ... this.
3827 (ipa_fn_summary_t::remove): Rename to ...
3828 (ipa_fn_summary_t::remove_callees): ... this.
3829 (ipa_fn_summary_t::duplicate): Use placement new
3830 instead of memory copy.
3831 (ipa_call_summary_t::duplicate): Likewise.
3832 (ipa_call_summary_t::remove): Remove.
3833 (dump_ipa_call_summary): Change get_create to get.
3834 (ipa_dump_fn_summary): Dump only when summary exists.
3835 (analyze_function_body): Use symbol_summary::get instead
3837 (compute_fn_summary): Likewise.
3838 (estimate_edge_devirt_benefit): Likewise.
3839 (estimate_edge_size_and_time): Likewise.
3840 (inline_update_callee_summaries): Likewise.
3841 (remap_edge_change_prob): Likewise.
3842 (remap_edge_summaries): Likewise.
3843 (ipa_merge_fn_summary_after_inlining): Likewise.
3844 (write_ipa_call_summary): Likewise.
3845 (ipa_fn_summary_write): Likewise.
3846 (ipa_free_fn_summary): Likewise.
3847 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
3848 (struct ipa_call_summary): Likewise.
3849 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
3851 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
3852 (estimate_size_after_inlining): Likewise.
3853 (estimate_growth): Likewise.
3854 (growth_likely_positive): Likewise.
3855 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3856 (inline_call): Likewise.
3857 * ipa-inline.c (caller_growth_limits): Likewise.
3858 (can_inline_edge_p): Likewise.
3859 (can_inline_edge_by_limits_p): Likewise.
3860 (compute_uninlined_call_time): Likewise.
3861 (compute_inlined_call_time): Likewise.
3862 (want_inline_small_function_p): Likewise.
3863 (edge_badness): Likewise.
3864 (update_caller_keys): Likewise.
3865 (update_callee_keys): Likewise.
3866 (inline_small_functions): Likewise.
3867 (inline_to_all_callers_1): Likewise.
3868 (dump_overall_stats): Likewise.
3869 (early_inline_small_functions): Likewise.
3870 (early_inliner): Likewise.
3871 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3872 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3873 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3874 * ipa-split.c (execute_split_functions): Likewise.
3875 * symbol-summary.h: Likewise.
3876 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3878 2018-06-19 Richard Biener <rguenther@suse.de>
3880 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
3881 (vectorize_loops): ... here. Fix dbgcnt handling.
3882 (try_vectorize_loop): Wrap try_vectorize_loop_1.
3884 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
3887 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
3888 ieee128 argument takes up only one (vector) register, not two (floating
3891 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
3893 * gimplify.c (gimplify_init_constructor): Really never clear for an
3894 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
3896 2018-06-19 Richard Biener <rguenther@suse.de>
3898 PR tree-optimization/86179
3899 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
3900 after failed recognition.
3902 2018-06-18 Martin Sebor <msebor@redhat.com>
3905 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
3906 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
3907 Handle integer subtraction.
3908 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
3909 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
3911 2018-06-18 David Malcolm <dmalcolm@redhat.com>
3913 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
3914 param from rtx to rtx_insn *.
3915 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
3917 (frv_ifcvt_modify_insn): Likwise.
3918 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
3919 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
3920 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
3921 as_a <rtx_insn *> cast to local "unprotected_region" once
3922 it's been established that it's not NULL or pc_rtx.
3923 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
3924 param "sethi" from rtx to rtx_insn *.
3925 (nds32_group_float_insns): Likewise for param "insn".
3926 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
3928 (vax_output_int_subtract): Likewise.
3929 * config/vax/vax.c (vax_output_int_add): Likewise for param
3931 (vax_output_int_subtract): Likewise.
3932 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
3933 (emit_pattern_after): Likewise for param "after".
3934 (emit_insn_after): Likewise.
3935 (emit_jump_insn_after): Likewise.
3936 (emit_call_insn_after): Likewise.
3937 (emit_debug_insn_after): Likewise.
3938 (emit_pattern_before): Likewise for param "before".
3939 (emit_insn_before): Likewise.
3940 (emit_jump_insn_before): Likewise.
3941 * final.c (get_insn_template): Likewise for param "insn", removing
3943 * output.h (get_insn_template): Likewise for 2nd param.
3944 * rtl.h (emit_insn_before): Likewise.
3945 (emit_jump_insn_before): Likewise.
3946 (emit_debug_insn_before_noloc): Likewise.
3947 (emit_insn_after): Likewise.
3948 (emit_jump_insn_after): Likewise.
3949 (emit_call_insn_after): Likewise.
3950 (emit_debug_insn_after): Likewise.
3951 (set_insn_deleted): Likewise for param.
3953 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
3956 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
3957 floating point modes, so that IFmode is numerically greater than
3958 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
3959 to declare the ordering. This prevents IFmode from being
3960 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
3961 machine. Include rs6000-modes.h to share the fractional values
3962 between genmodes* and the rest of the compiler.
3966 * config/rs6000/rs6000-modes.h: New file.
3967 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
3968 meaning of rs6000_long_double_size so that 126..128 selects an
3969 appropriate 128-bit floating point type.
3970 (rs6000_option_override_internal): Likewise.
3971 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
3972 (TARGET_LONG_DOUBLE_128): Change the meaning of
3973 rs6000_long_double_size so that 126..128 selects an appropriate
3974 128-bit floating point type.
3975 (LONG_DOUBLE_TYPE_SIZE): Update comment.
3976 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
3977 source and destination to match the standard usage.
3978 (truncifkf2): Likewise.
3979 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
3980 ISA 2.07 to use an explicit clobber, instead of passing in a
3982 (copysign<mode>3_soft): Likewise.
3984 2018-06-18 David Malcolm <dmalcolm@redhat.com>
3986 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
3987 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
3988 (vect_slp_analyze_instance_dependence): Likewise.
3989 (vect_enhance_data_refs_alignment): Likewise.
3990 (vect_analyze_data_refs_alignment): Likewise.
3991 (vect_slp_analyze_and_verify_instance_alignment
3992 (vect_analyze_data_ref_accesses): Likewise.
3993 (vect_prune_runtime_alias_test_list): Likewise.
3994 (vect_analyze_data_refs): Likewise.
3995 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
3996 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
3997 (vect_analyze_scalar_cycles_1): Likewise.
3998 (vect_get_loop_niters): Likewise.
3999 (vect_analyze_loop_form_1): Likewise.
4000 (vect_update_vf_for_slp): Likewise.
4001 (vect_analyze_loop_operations): Likewise.
4002 (vect_analyze_loop): Likewise.
4003 (vectorizable_induction): Likewise.
4004 (vect_transform_loop): Likewise.
4005 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
4006 * tree-vect-slp.c (vect_analyze_slp): Likewise.
4007 (vect_make_slp_decision): Likewise.
4008 (vect_detect_hybrid_slp): Likewise.
4009 (vect_slp_analyze_operations): Likewise.
4010 (vect_slp_bb): Likewise.
4011 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
4012 (vectorizable_bswap): Likewise.
4013 (vectorizable_call): Likewise.
4014 (vectorizable_simd_clone_call): Likewise.
4015 (vectorizable_conversion): Likewise.
4016 (vectorizable_assignment): Likewise.
4017 (vectorizable_shift): Likewise.
4018 (vectorizable_operation): Likewise.
4019 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
4021 2018-06-18 Martin Sebor <msebor@redhat.com>
4023 PR tree-optimization/81384
4024 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
4025 * builtins.c (expand_builtin_strnlen): New function.
4026 (expand_builtin): Call it.
4027 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
4028 * builtins.def (BUILT_IN_STRNLEN): New.
4029 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
4030 Warn for bounds in excess of maximum object size.
4031 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
4032 single-value ranges. Handle strnlen.
4033 (handle_builtin_strlen): Handle strnlen.
4034 (strlen_check_and_optimize_stmt): Same.
4035 * doc/extend.texi (Other Builtins): Document strnlen.
4037 2018-06-18 Maya Rashish <coypu@sdf.org>
4039 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
4040 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
4041 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
4043 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
4045 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
4047 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4049 * tree.c (escaped_string::escape): Replace cast to char * by
4050 const_cast<char *> (unescaped).
4052 2018-06-18 Nick Clifton <nickc@redhat.com>
4055 * tree.c (escaped_string): New class. Converts an unescaped
4056 string into its escaped equivalent.
4057 (warn_deprecated_use): Use the new class to convert the
4058 deprecation message, if present.
4059 (test_escaped_strings): New self test.
4060 (test_c_tests): Add test_escaped_strings.
4061 * doc/extend.texi (deprecated): Add a note that the
4062 deprecation message is affected by the -fmessage-length
4063 option, and that control characters will be escaped.
4064 (#pragma GCC error): Document this pragma.
4065 (#pragma GCC warning): Likewise.
4066 * doc/invoke.texi (-fmessage-length): Document this option's
4067 effect on the #warning and #error preprocessor directives and
4068 the deprecated attribute.
4070 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
4072 * tree.c (decl_value_expr_lookup): Revert latest change.
4073 (decl_value_expr_insert): Likewise.
4075 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
4077 * gimplify.c (nonlocal_vlas): Delete.
4078 (nonlocal_vla_vars): Likewise.
4079 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
4081 (gimplify_body): Do not create and destroy nonlocal_vlas.
4082 * tree-nested.c: Include diagnostic.h.
4083 (use_pointer_in_frame): Tweak.
4084 (lookup_field_for_decl): Add assertion and declare the transformation.
4085 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
4086 internal error when the reference is in a wrong context. Do not
4087 create a debug decl by default.
4088 (note_nonlocal_block_vlas): Delete.
4089 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
4090 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
4091 create a debug decl by default.
4092 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
4093 call is in a wrong context.
4094 (fixup_vla_decls): New function.
4095 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
4096 debug variables were created.
4097 * tree.c (decl_value_expr_lookup): Add checking assertion.
4098 (decl_value_expr_insert): Likewise.
4100 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
4103 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
4104 * tree-scalar-evolution.c (interpret_expr): Likewise.
4105 (expression_expensive_p): Likewise.
4106 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
4107 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
4108 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
4109 (ssa_defined_by_minus_one_stmt_p): New.
4111 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
4114 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
4115 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
4116 * dojump.c (do_jump): Likewise.
4117 * expr.c (expand_expr_real_2): Check operand type's sign.
4118 * fold-const.c (const_unop): Handle ABSU_EXPR.
4119 (fold_abs_const): Likewise.
4120 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
4121 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
4122 (strip_sign_op_1): Likesise.
4123 * match.pd: Add new pattern to generate ABSU_EXPR.
4124 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
4125 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4126 * tree-eh.c (operation_could_trap_helper_p): Likewise.
4127 * tree-inline.c (estimate_operator_cost): Likewise.
4128 * tree-pretty-print.c (dump_generic_node): Likewise.
4129 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
4130 * tree.def (ABSU_EXPR): New.
4132 2018-06-16 Jakub Jelinek <jakub@redhat.com>
4135 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
4136 documented as preserved for backward compatibility only.
4137 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
4139 PR rtl-optimization/86108
4140 * bb-reorder.c (create_forwarder_block): Renamed to ...
4141 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
4142 jump from new landing pad to the second part.
4143 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
4146 2018-06-15 Jakub Jelinek <jakub@redhat.com>
4149 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
4150 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
4151 Only call store_expr for halves if the mode is the same.
4154 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
4155 Fix up comment formatting.
4157 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
4159 * typed-splay-tree.h (typed_splay_tree::remove): New function.
4160 (typed_splay_tree::closure,
4161 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
4162 (typed_splay_tree::typed_splay_tree,
4163 typed_splay_tree::operator =): Declared private.
4164 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
4165 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
4166 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
4167 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
4168 typed_splay_tree::splay_tree_splay,
4169 typed_splay_tree::splay_tree_foreach_helper,
4170 typed_splay_tree::splay_tree_insert,
4171 typed_splay_tree::splay_tree_remove,
4172 typed_splay_tree::splay_tree_lookup,
4173 typed_splay_tree::splay_tree_predecessor,
4174 typed_splay_tree::splay_tree_successor,
4175 typed_splay_tree::splay_tree_min,
4176 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
4177 (typed_splay_tree::root, typed_splay_tree::comp,
4178 typed_splay_tree::delete_key,
4179 typed_splay_tree::delete_value): New data members.
4180 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
4181 typed_splay_tree::remove.
4183 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
4185 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
4186 -mginv and -mno-ginv to the assembler.
4187 * config/mips/mips.opt (-mcrc): New option.
4189 * doc/invoke.text (-mcrc): Document.
4192 2018-06-15 Nick Clifton <nickc@redhat.com>
4195 * tree.c (escaped_string): New class. Converts an unescaped
4196 string into its escaped equivalent.
4197 (warn_deprecated_use): Use the new class to convert the
4198 deprecation message, if present.
4199 (test_escaped_strings): New self test.
4200 (test_c_tests): Add test_escaped_strings.
4201 * doc/extend.texi (deprecated): Add a note that the
4202 deprecation message is affected by the -fmessage-length
4203 option, and that control characters will be escaped.
4204 (#pragma GCC error): Document this pragma.
4205 (#pragma GCC warning): Likewise.
4206 * doc/invoke.texi (-fmessage-length): Document this option's
4207 effect on the #warning and #error preprocessor directives and
4208 the deprecated attribute.
4210 2018-06-15 Richard Biener <rguenther@suse.de>
4212 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
4213 here, also noting vector size used.
4214 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
4215 size used in MSG_OPTIMIZED_LOCATIONS dump.
4216 (pass_slp_vectorize::execute): Adjust.
4218 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
4221 * config/arc/arc.c (arc_return_address_register): Fix
4224 2018-06-15 Richard Biener <rguenther@suse.de>
4227 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
4228 leave useless conversion stripping to force_gimple_operand_gsi.
4229 (gimplify_build2): Likewise.
4230 (gimplify_build1): Likewise.
4232 2018-06-15 Richard Biener <rguenther@suse.de>
4235 * tree-cfg.c (move_stmt_op): unshare invariant addresses
4236 before adjusting their block.
4238 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
4240 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
4241 multilibs for *-*-rtems*.
4242 * config/riscv/t-rtems: New file.
4244 2018-06-14 Jakub Jelinek <jakub@redhat.com>
4247 * match.pd ((A +- CST1) +- CST2): Punt if last resort
4248 unsigned_type_for returns NULL.
4251 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
4252 subregs of multi-word pseudos unless the float mode has word size.
4254 2018-06-14 Richard Biener <rguenther@suse.de>
4257 * tree-vect-generic.c (build_word_mode_vector_type): Remove
4258 duplicate and harmful type_hash_canon.
4259 * tree.c (type_hash_canon): Assert we didn't find ourselves.
4261 2018-06-14 Richard Biener <rguenther@suse.de>
4264 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
4267 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
4269 * config/rtems.h (STDINT_LONG32): Define.
4271 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
4272 Prachi Godbole <prachi.godbole@imgtec.com>
4274 * config/mips/mips-cpus.def: Define P6600.
4275 * config/mips/mips-tables.opt: Regenerate.
4276 * config/mips/mips.c (mips_ucbranch_type): New enum.
4277 (mips_rtx_cost_data): Add support for P6600.
4278 (mips_issue_rate): Likewise.
4279 (mips_multipass_dfa_lookahead): Likewise.
4280 (mips_avoid_hazard): Likewise.
4281 (mips_reorg_process_insns): Likewise.
4282 (mips_classify_branch_p6600): New function.
4283 * config/mips/mips.h (TUNE_P6600): New define.
4284 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
4285 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
4286 * config/mips/mips.md: Include p6600.md.
4287 (processor): Add p6600.
4288 * config/mips/p6600.md: New file.
4289 * doc/invoke.texi: Add p6600 to supported architectures.
4291 2018-06-13 Martin Sebor <msebor@redhat.com>
4293 PR tree-optimization/86114
4294 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
4296 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
4298 2018-06-13 Richard Biener <rguenther@suse.de>
4300 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
4301 Properly set vector type of the intermediate stmt.
4302 * tree-vect-stmts.c (vectorizable_operation): The destination
4303 var always has vectype_out type.
4305 2018-06-13 Jeff Law <law@redhat.com>
4307 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
4308 integer 0 for argument to print_rtl_with_bb.
4309 (rl78_reorg): Likewise.
4311 2018-06-13 David Malcolm <dmalcolm@redhat.com>
4313 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
4314 from rtx to rtx_insn *.
4315 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
4317 (add_sched_insns_for_speculation): Likewise for local "target",
4318 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
4319 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
4320 from rtx_insn ** to rtx_code_label **.
4321 (reorg_emit_nops): Likewise.
4322 (c6x_reorg): Likewise for local "call_labels".
4323 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
4325 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
4326 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
4327 the loops over LABEL_REFS.
4328 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
4330 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
4331 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
4332 (split_branches): Strengthen local "olabel" from rtx to
4333 rtx_insn *, adding a safe_as_a cast.
4334 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
4336 (add_insn_after): Likewise for first two params.
4337 (add_insn_before): Likewise.
4338 (remove_insn): Likewise for param.
4339 (emit_pattern_before_noloc): Likewise for second and third params.
4340 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
4341 (emit_call_insn_before_noloc): Likewise.
4342 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
4344 (emit_barrier_before): Likewise.
4345 (emit_label_before): Strengthen "label" param from "rtx" to
4346 "rtx_code_label *". Strengthen "before" param from "rtx" to
4348 (emit_insn_after_1): Strengthen "after" param from "rtx" to
4350 (emit_pattern_after_noloc): Likewise.
4351 (emit_insn_after_noloc): Likewise.
4352 (emit_jump_insn_after_noloc): Likewise.
4353 (emit_call_insn_after_noloc): Likewise.
4354 (emit_debug_insn_after_noloc): Likewise.
4355 (emit_barrier_after): Likewise.
4356 (emit_label_after): Likewise for both params.
4357 (emit_pattern_after_setloc): Likewise for "after" param. Convert
4358 "loc" param from "int" to "location_t".
4359 (emit_insn_after_setloc): Likewise.
4360 (emit_jump_insn_after_setloc): Likewise.
4361 (emit_call_insn_after_setloc): Likewise.
4362 (emit_debug_insn_after_setloc): Likewise.
4363 (emit_pattern_before_setloc): Likewise for "before" param. Convert
4364 "loc" param from "int" to "location_t".
4365 (emit_pattern_before): Convert NULL_RTX to NULL.
4366 (emit_insn_before_setloc): Convert "loc" param from "int" to
4368 (emit_jump_insn_before_setloc): Likewise.
4369 (emit_call_insn_before_setloc): Likewise.
4370 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
4371 rtx_insn *. Convert "loc" param from "int" to "location_t".
4372 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
4373 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
4374 Convert 3rd param from "int" to "location_t".
4375 (emit_barrier_before, emit_barrier_after, next_real_insn):
4376 Strengthen param from rtx to rtx_insn *.
4377 (emit_label_before): Strengthen 1st param from "rtx" to
4378 "rtx_code_label *". Strengthen 2nd param from "rtx" to
4380 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
4381 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
4382 Strengthen 2nd param from "rtx" to "rtx_insn *".
4383 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
4384 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
4385 Likewise. Convert 3rd param from "int" to "location_t".
4386 (emit_label_after): Strengthen 1st param from "rtx" to
4388 (next_real_insn, remove_insn): Strengthen param from "rtx" to
4390 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
4391 from "rtx" to "rtx_insn *".
4393 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
4395 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
4396 bodies streamed in with -Q.
4397 * dumpfile.c (dump_files): Add lto-stream-out dump file.
4398 * dumpfile.h (tree_dump_index): Add lto_stream_out.
4399 * gimple-streamer-out.c: Include gimple-pretty-print.h
4400 (output_bb): Dump stmts streamed.
4401 * lto-section-out.c: Include print-tree.h
4402 (lto_begin_section): Dump sections created.
4403 (lto_output_decl_index): Dump decl encoded.
4404 * lto-streamer-out.c: Include print-tree.h
4405 (create_output_block): Dump output block created.
4406 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
4407 (output_function): Dump function output.
4408 (output_constructor): Dump constructor streamed.
4409 (write_global_stream): Output indexes encoded.
4410 (produce_asm_for_decls): Dump streams encoded.
4411 * lto-streamer.c (streamer_dump_file): New global var.
4412 * lto-streamer.h (streamer_dump_file): Declare.
4413 * passes.c (ipa_write_summaries): Initialize streamer dump.
4414 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
4417 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
4420 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
4421 offsets for register save directives. Emit a second batch of save
4422 directives, if need be, when the function accesses prior frames.
4424 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4426 * config/arc/fpu.md (fmasf4): Force operand to register.
4427 (fnmasf4): Likewise.
4429 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4431 * config/arc/arc-protos.h (arc_pad_return): Remove.
4432 * config/arc/arc.c (machine_function): Remove force_short_suffix
4434 (arc_print_operand): Adjust printing of '&'.
4435 (arc_verify_short): Remove conditional printing of short suffix.
4436 (arc_final_prescan_insn): Remove reference to size_reason.
4437 (pad_return): New function.
4438 (arc_reorg): Call pad_return.
4439 (arc_pad_return): Remove.
4440 (arc_init_machine_status): Remove reference to force_short_suffix.
4441 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
4442 (attr length): When attribute iscompact is true force to 2
4443 regardless; in the case of maybe check if we want to force the
4444 instruction to have 4 bytes length.
4445 (nopv): Change it to generate 4 byte long nop as well.
4446 (blockage): New pattern.
4447 (simple_return): Remove call to arc_pad_return.
4448 (p_return_i): Likewise.
4450 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4452 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
4454 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4456 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
4459 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4461 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
4462 for ARC700 and ARCv2.
4464 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
4467 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
4468 operands[2] instead of operands[1].
4471 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
4473 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
4474 case, check whether the outer register overlaps an unallocatable
4475 register, not just whether it fits the required class.
4477 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
4479 * poly-int.h (can_div_trunc_p): Add new overload in which all values
4481 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
4482 (memrefs_conflict_p): Likewise.
4483 (init_alias_analysis): Likewise.
4484 * cfgexpand.c (expand_debug_expr): Likewise.
4485 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
4486 * cse.c (fold_rtx): Likewise.
4487 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
4488 * expr.c (emit_block_move_hints): Likewise.
4489 (clear_storage_hints, push_block, emit_push_insn): Likewise.
4490 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
4491 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
4492 (emit_group_store): Likewise.
4493 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
4494 to read the PRE/POST_MODIFY increment.
4495 * calls.c (store_one_arg): Use strip_offset.
4496 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
4498 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
4500 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
4501 (simplify_binary_operation_1): Extend CONST_INT handling to
4503 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
4504 than a HOST_WIDE_INT.
4505 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
4507 (adjust_mems, add_stores): Update accodingly.
4508 (vt_canonicalize_addr): Track polynomial offsets.
4509 (emit_note_insn_var_location): Likewise.
4510 (vt_add_function_parameter): Likewise.
4511 (vt_initialize): Likewise.
4513 2018-06-12 Jeff Law <law@redhat.com>
4515 * config.gcc (alpha*-*-freebsd*): Remove.
4516 * config/alpha/freebsd.h: Remove.
4518 2018-06-12 David Malcolm <dmalcolm@redhat.com>
4521 * spellcheck-tree.c (levenshtein_distance): Rename to...
4522 (get_edit_distance): ...this, and update for underlying renaming.
4523 * spellcheck-tree.h (levenshtein_distance): Rename to...
4524 (get_edit_distance): ...this.
4525 * spellcheck.c (levenshtein_distance): Rename to...
4526 (get_edit_distance): ...this. Convert from Levenshtein distance
4527 to Damerau-Levenshtein distance by supporting transpositions of
4528 adjacent characters. Rename "v1" to "v_next" and "v0" to
4530 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
4531 (selftest::test_edit_distance_unit_test_oneway): ...this, and
4532 update for underlying renaming.
4533 (selftest::levenshtein_distance_unit_test): Rename to...
4534 (selftest::test_get_edit_distance_unit): ...this, and update for
4535 underlying renaming.
4536 (selftest::test_find_closest_string): Add example from PR 69968
4537 where transposition helps
4538 (selftest::test_metric_conditions): Update for renaming.
4539 (selftest::test_metric_conditions): Likewise.
4540 (selftest::spellcheck_c_tests): Likewise.
4541 * spellcheck.h (levenshtein_distance): Rename both overloads to...
4542 (get_edit_distance): ...this.
4543 (best_match::consider): Update for renaming.
4545 2018-06-12 Martin Sebor <msebor@redhat.com>
4547 PR tree-optimization/85259
4548 * builtins.c (compute_objsize): Handle constant offsets.
4549 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
4550 true iff a warning has been issued.
4551 * gimple.h (gimple_nonartificial_location): New function.
4552 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
4553 gimple_nonartificial_location and handle -Wno-system-headers.
4554 (handle_builtin_stxncpy): Same.
4556 2018-06-12 Martin Sebor <msebor@redhat.com>
4559 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
4561 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
4563 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
4564 BUILTIN_VEC_XST entries for pointer to double and long long.
4566 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
4569 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
4571 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
4574 2018-06-12 Martin Liska <mliska@suse.cz>
4576 * doc/options.texi: Document IntegerRange.
4578 2018-06-12 Martin Liska <mliska@suse.cz>
4580 * config/i386/i386.opt: Make MPX-related options as Deprecated.
4581 * opt-functions.awk: Handle Deprecated flag.
4582 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
4584 (read_cmdline_option): Report warning for a deprecated option.
4585 * opts.h (struct cl_option): Add new field cl_deprecated.
4586 (CL_ERR_DEPRECATED): New.
4588 2018-06-12 Martin Liska <mliska@suse.cz>
4590 * doc/options.texi: Document Deprecated option flag.
4592 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4594 * config/arc/arc-arch.h (arc_extras): New enum.
4595 (arc_cpu_t):Add field extra.
4596 (arc_cpu_types): Consider the extras.
4597 * config/arc/arc-cpus.def: Add extras info.
4598 * config/arc/arc-opts.h (processor_type): Consider extra field.
4599 * config/arc/arc.c (arc_override_options): Handle extra field.
4601 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4603 * config/arc/arc-arch.h: Update ARC_OPTX macro.
4604 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
4606 * config/arc/arc.c (arc_init): Update pic warning.
4607 (irq_range): Update irq range parsing warnings.
4608 (arc_override_options): Update various warning messages.
4609 (arc_handle_aux_attribute): Likewise.
4611 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
4613 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
4615 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4617 * doc/sourcebuild.texi: Document usage of line number 0 in verify
4618 compiler messages directives.
4620 2018-06-12 Matthew Fortune <mfortune@gmail.com>
4622 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
4623 * config/mips/mips-tables.opt: Regenerate.
4624 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
4626 * doc/invoke.texi: Document -march=i6500.
4628 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
4630 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
4631 (i6400_gpmul): Add cpu_unit.
4632 (i6400_gpdiv): Likewise.
4633 (i6400_msa_add_d): Update reservations.
4634 (i6400_msa_int_add) Likewise.
4635 (i6400_msa_short_logic3) Likewise.
4636 (i6400_msa_short_logic2) Likewise.
4637 (i6400_msa_short_logic) Likewise.
4638 (i6400_msa_move) Likewise.
4639 (i6400_msa_cmp) Likewise.
4640 (i6400_msa_short_float2) Likewise.
4641 (i6400_msa_div_d) Likewise.
4642 (i6400_msa_long_logic1) Likewise.
4643 (i6400_msa_long_logic2) Likewise.
4644 (i6400_msa_mult) Likewise.
4645 (i6400_msa_long_float2) Likewise.
4646 (i6400_msa_long_float4) Likewise.
4647 (i6400_msa_long_float5) Likewise.
4648 (i6400_msa_long_float8) Likewise.
4649 (i6400_fpu_fadd): Include frint type.
4650 (i6400_fpu_store): New define_insn_reservation.
4651 (i6400_fpu_load): Likewise.
4652 (i6400_fpu_move): Likewise.
4653 (i6400_fpu_fcmp): Likewise.
4654 (i6400_fpu_fmadd): Likewise.
4655 (i6400_int_mult): Include imul3nc type and update reservation.
4656 (i6400_int_div): Include idiv3 type and update reservation.
4657 (i6400_int_load): Update to check type not move_type.
4658 (i6400_int_store): Likewise.
4659 (i6400_int_prefetch): Set zero latency.
4661 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
4663 * gcc.c: Document new %@{...} sequence.
4664 (LINK_COMMAND_SPEC): Use it for the -L switches.
4665 (cpp_unique_options): Use it for the -I switches.
4666 (at_file_argbuf): New global variable.
4667 (in_at_file): Likewise.
4668 (alloc_args): Create at_file_argbuf.
4669 (clear_args): Truncate at_file_argbuf.
4670 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
4671 (open_at_file): New function.
4672 (close_at_file): Likewise.
4673 (create_at_file): Delete.
4674 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
4677 (validate_switches_from_spec): Deal with %@{...} sequence.
4678 (validate_switches): Likewise.
4679 (driver::finalize): Call clear_args.
4681 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
4683 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
4685 2018-06-11 Martin Sebor <msebor@redhat.com>
4687 * doc/invoke.texi (-Wall): List -Wc++17-compat.
4688 (Wno-class-memaccess): Add @opindex.
4689 (Wno-templates, Wno-multiple-inheritance): Same.
4690 (Wno-virtual-inheritance, Wno-namespaces): Same.
4691 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
4692 (Wno-format-overflow, Wno-format-truncation): Same.
4693 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
4694 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
4695 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
4696 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
4697 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
4698 (Wno-misspelled-isr): Same.
4700 2018-06-11 Martin Sebor <msebor@redhat.com>
4702 * PR tree-optimization/86083
4703 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
4705 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
4707 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
4709 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
4712 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
4713 on the correct operand.
4714 (*movdi_internal64): Ditto.
4716 2018-06-11 Martin Liska <mliska@suse.cz>
4718 PR tree-optimization/86089
4719 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
4721 2018-06-11 Julia Koval <julia.koval@intel.com>
4723 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
4724 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
4725 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
4727 2018-06-11 Olivier Hainque <hainque@adacore.com>
4729 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
4730 for Ada with strict dwarf2.
4732 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
4735 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
4738 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
4740 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
4742 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
4744 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
4747 2018-06-08 Martin Liska <mliska@suse.cz>
4749 * tree-cfg.h (debug_function): Fix argument type to match
4752 2018-06-08 Martin Liska <mliska@suse.cz>
4754 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
4755 Remove usage of MPX-related (and removed) fields.
4756 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
4758 2018-06-08 David Malcolm <dmalcolm@redhat.com>
4760 * cfg.c (debug): Use TDF_NONE rather than 0.
4761 * cfghooks.c (debug): Likewise.
4762 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
4763 (struct dump_option_value_info): Convert to...
4764 (struct kv_pair): ...this template type.
4765 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
4767 (optinfo_verbosity_options): Likewise.
4768 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
4770 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
4771 than int for "optgroup_flags" param.
4772 (dump_generic_expr_loc): Use dump_flags_t rather than int for
4774 (dump_dec): Likewise.
4775 (dump_finish): Use TDF_NONE rather than 0.
4776 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
4777 rather than int for "optgroup_flags" param. Use TDF_NONE rather
4778 than 0. Update for change to option_ptr.
4779 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
4780 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
4781 0. Update for changes to optinfo_verbosity_options and
4783 (opt_info_switch_p): Convert optgroup_flags from int to
4785 (dump_basic_block): Use dump_flags_t rather than int
4786 for "dump_kind" param.
4787 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
4788 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
4789 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
4790 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
4791 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
4792 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
4793 TDF_NONE): Convert from macros to...
4794 (enum dump_flag): ...this new enum.
4795 (dump_flags_t): Update to use enum.
4796 (operator|, operator&, operator~, operator|=, operator&=):
4797 Implement for dump_flags_t.
4798 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
4799 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
4800 Convert from macros to...
4801 (enum optgroup_flag): ...this new enum.
4802 (optgroup_flags_t): New typedef.
4803 (operator|, operator|=): Implement for optgroup_flags_t.
4804 (struct dump_file_info): Convert field "alt_flags" to
4805 dump_flags_t. Convert field "optgroup_flags" to
4807 (dump_basic_block): Use dump_flags_t rather than int for param.
4808 (dump_generic_expr_loc): Likewise.
4809 (dump_dec): Likewise.
4810 (dump_register): Convert param "optgroup_flags" to
4812 (opt_info_enable_passes): Likewise.
4813 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
4815 * gimple-pretty-print.c (debug): Likewise.
4816 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
4817 (merged_store_group::apply_stores): Likewise.
4818 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
4819 * gimple.c (verify_gimple_pp): Likewise.
4820 * graphite-poly.c (print_pbb_body): Likewise.
4821 * passes.c (pass_manager::register_one_dump_file): Convert
4822 local "optgroup_flags" to optgroup_flags_t.
4823 * print-tree.c (print_node): Use TDF_NONE rather than 0.
4825 (debug_body): Likewise.
4826 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
4827 to optgroup_flags_t.
4828 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
4830 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
4831 (convert_mult_to_fma): Likewise.
4832 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
4833 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
4834 * tree-vect-data-refs.c (dump_lower_bound): Convert param
4835 "dump_kind" to dump_flags_t.
4837 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
4839 * config/rs6000/rs6000.c (min, max): Delete.
4841 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
4843 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
4844 -mabi=spe and -mabi=no-spe.
4846 2018-06-08 Martin Liska <mliska@suse.cz>
4848 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
4849 where we expect an existing summary.
4851 2018-06-08 Martin Liska <mliska@suse.cz>
4853 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
4854 * ipa-inline.h (estimate_edge_growth): Likewise.
4856 2018-06-08 Martin Liska <mliska@suse.cz>
4858 * cgraph.c (function_version_hasher::hash): Use
4859 cgraph_node::get_uid ().
4860 (function_version_hasher::equal):
4861 * cgraph.h (cgraph_node::get_uid): New method.
4862 * ipa-inline.c (update_caller_keys): Use
4863 cgraph_node::get_uid ().
4864 (update_callee_keys): Likewise.
4865 * ipa-utils.c (searchc): Likewise.
4866 (ipa_reduced_postorder): Likewise.
4867 * lto-cgraph.c (input_node): Likewise.
4868 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
4869 * symbol-summary.h (symtab_insertion): Likewise.
4870 (symtab_removal): Likewise.
4871 (symtab_duplication): Likewise.
4872 * tree-pretty-print.c (dump_function_header): Likewise.
4873 * tree-sra.c (convert_callers_for_node): Likewise.
4875 2018-06-08 Martin Liska <mliska@suse.cz>
4877 * cgraph.c (symbol_table::create_edge): Always assign a new
4879 (symbol_table::free_edge): Do not recycle numbers.
4880 * cgraph.h (cgraph_edge::get): New method.
4881 * symbol-summary.h (symtab_removal): Use it.
4882 (symtab_duplication): Likewise.
4883 (call_summary::hashable_uid): Remove.
4885 2018-06-08 Martin Liska <mliska@suse.cz>
4887 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
4888 (initialize_growth_caches): Remove.
4889 (free_growth_caches): Likewise.
4890 (do_estimate_edge_time): Use edge_growth_cache.
4891 (do_estimate_edge_size): Likewise.
4892 (do_estimate_edge_hints): Likewise.
4893 * ipa-inline.c (reset_edge_caches): Likewise.
4894 (recursive_inlining): Likewise.
4895 (inline_small_functions): Likewise.
4896 * ipa-inline.h (initialize_growth_caches): Remove.
4897 (estimate_edge_size): Likewise.
4898 (estimate_edge_time): Likewise.
4899 (estimate_edge_hints): Likewise.
4900 (reset_edge_growth_cache): Likewise.
4901 * symbol-summary.h (call_summary::remove): New method.
4903 2018-06-08 Martin Liska <mliska@suse.cz>
4905 * ipa-cp.c (class edge_clone_summary): New summary.
4906 (grow_edge_clone_vectors): Remove.
4907 (ipcp_edge_duplication_hook): Remove.
4908 (class edge_clone_summary_t): New call_summary class.
4909 (ipcp_edge_removal_hook): Remove.
4910 (edge_clone_summary_t::duplicate): New function.
4911 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
4912 (create_specialized_node): Likewise.
4913 (ipcp_driver): Initialize edge_clone_summaries and do not
4916 2018-06-08 Martin Liska <mliska@suse.cz>
4918 * symbol-summary.h (get): New function.
4919 (call_summary::m_initialize_when_cloning): New class member.
4921 2018-06-08 Martin Liska <mliska@suse.cz>
4923 * cgraph.c (cgraph_node::remove): Do not recycle uid.
4924 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
4925 (symbol_table::allocate_cgraph_symbol): Do not set uid.
4926 * passes.c (uid_hash_t): Record removed_nodes by their uids.
4927 (remove_cgraph_node_from_order): Use the removed_nodes set.
4928 (do_per_function_toporder): Likwise.
4929 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
4930 instead of summary_uid.
4931 (symtab_removal): Likewise.
4932 (symtab_duplication): Likewise.
4934 2018-06-08 Martin Liska <mliska@suse.cz>
4936 * ipa-cp.c (ipcp_store_bits_results): Use
4937 ipcp_transformation_sum.
4938 (ipcp_store_vr_results): Likewise.
4939 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
4941 (ipcp_transformation_initialize): ... this.
4942 (ipa_set_node_agg_value_chain):
4943 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
4944 (write_ipcp_transformation_info): Likewise.
4945 (read_ipcp_transformation_info): Likewise.
4946 (ipcp_update_bits): Likewise.
4947 (ipcp_update_vr): Likewise.
4948 (ipcp_transform_function): Likewise.
4949 * ipa-prop.h: Rename ipcp_transformation_summary to
4950 ipcp_transformation.
4951 (class ipcp_transformation_t): New function summary.
4952 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
4953 (ipa_get_agg_replacements_for_node): Likewise.
4955 2018-06-08 Martin Liska <mliska@suse.cz>
4957 * ipa-pure-const.c (struct funct_state_d): Do it class instead
4959 (class funct_state_summary_t): New function_summary class.
4960 (has_function_state): Remove.
4961 (get_function_state): Likewise.
4962 (set_function_state): Likewise.
4963 (add_new_function): Likewise.
4964 (funct_state_summary_t::insert): New function.
4965 (duplicate_node_data): Remove.
4966 (remove_node_data): Remove.
4967 (funct_state_summary_t::duplicate): New function.
4968 (register_hooks): Create new funct_state_summaries.
4969 (pure_const_generate_summary): Use it.
4970 (pure_const_write_summary): Likewise.
4971 (pure_const_read_summary): Likewise.
4972 (propagate_pure_const): Likewise.
4973 (propagate_nothrow): Likewise.
4974 (dump_malloc_lattice): Likewise.
4975 (propagate_malloc): Likewise.
4976 (execute): Do not register hooks, just remove summary
4978 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
4981 2018-06-08 Martin Liska <mliska@suse.cz>
4983 * ipa-reference.c (remove_node_data): Remove.
4984 (duplicate_node_data): Likewise.
4985 (class ipa_ref_var_info_summary_t): New class.
4986 (class ipa_ref_opt_summary_t): Likewise.
4987 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
4988 (get_reference_optimization_summary): Use
4989 ipa_ref_opt_sum_summaries.
4990 (set_reference_vars_info): Remove.
4991 (set_reference_optimization_summary): Likewise.
4992 (ipa_init): Create summaries.
4993 (init_function_info): Use function summary.
4994 (ipa_ref_opt_summary_t::duplicate): New function.
4995 (ipa_ref_opt_summary_t::remove): New function.
4996 (get_read_write_all_from_node): Fix GNU coding style.
4997 (propagate): Use function summary.
4998 (write_node_summary_p): Fix GNU coding style.
4999 (stream_out_bitmap): Likewise.
5000 (ipa_reference_read_optimization_summary): Use function summary.
5001 (ipa_reference_c_finalize): Do not release hooks.
5003 2018-06-08 Martin Liska <mliska@suse.cz>
5005 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
5006 (analyze_function_body): Extract multiple calls of get_create.
5007 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
5008 * ipa-inline.c (recursive_inlining): Use ::get method.
5009 * ipa-inline.h (estimate_edge_growth): Likewise.
5011 2018-06-08 Martin Liska <mliska@suse.cz>
5013 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
5015 (hsa_function_summary::hsa_function_summary): Use the new enum
5017 (hsa_gpu_implementation_p): Use hsa_summaries::get.
5018 * hsa-gen.c (hsa_get_host_function): Likewise.
5019 (get_brig_function_name): Likewise.
5020 * ipa-hsa.c (process_hsa_functions): Likewise.
5021 (ipa_hsa_write_summary): Likewise.
5022 * symbol-summary.h (symtab_duplication): Use ::get function/
5023 (get): New function.
5025 2018-06-08 Martin Liska <mliska@suse.cz>
5027 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
5029 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
5030 (hsa_register_kernel): Likewise.
5031 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
5032 * hsa-gen.c (hsa_get_host_function): Likewise.
5033 (get_brig_function_name): Likewise.
5034 (generate_hsa): Likewise.
5035 (pass_gen_hsail::execute): Likewise.
5036 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
5037 (devirtualization_time_bonus): Likewise.
5038 (ipcp_propagate_stage): Likewise.
5039 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
5040 (edge_set_predicate): Likewise.
5041 (evaluate_conditions_for_known_args): Likewise.
5042 (evaluate_properties_for_edge): Likewise.
5043 (ipa_fn_summary::reset): Likewise.
5044 (ipa_fn_summary_t::duplicate): Likewise.
5045 (dump_ipa_call_summary): Likewise.
5046 (ipa_dump_fn_summary): Likewise.
5047 (analyze_function_body): Likewise.
5048 (compute_fn_summary): Likewise.
5049 (estimate_edge_devirt_benefit): Likewise.
5050 (estimate_edge_size_and_time): Likewise.
5051 (estimate_calls_size_and_time): Likewise.
5052 (estimate_node_size_and_time): Likewise.
5053 (inline_update_callee_summaries): Likewise.
5054 (remap_edge_change_prob): Likewise.
5055 (remap_edge_summaries): Likewise.
5056 (ipa_merge_fn_summary_after_inlining): Likewise.
5057 (ipa_update_overall_fn_summary): Likewise.
5058 (read_ipa_call_summary): Likewise.
5059 (inline_read_section): Likewise.
5060 (write_ipa_call_summary): Likewise.
5061 (ipa_fn_summary_write): Likewise.
5062 (ipa_free_fn_summary): Likewise.
5063 * ipa-hsa.c (process_hsa_functions): Likewise.
5064 (ipa_hsa_write_summary): Likewise.
5065 (ipa_hsa_read_section): Likewise.
5066 * ipa-icf.c (sem_function::merge): Likewise.
5067 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
5068 (do_estimate_edge_time): Likewise.
5069 (estimate_size_after_inlining): Likewise.
5070 (estimate_growth): Likewise.
5071 (growth_likely_positive): Likewise.
5072 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
5073 (inline_call): Likewise.
5074 * ipa-inline.c (caller_growth_limits): Likewise.
5075 (can_inline_edge_p): Likewise.
5076 (can_inline_edge_by_limits_p): Likewise.
5077 (compute_uninlined_call_time): Likewise.
5078 (compute_inlined_call_time): Likewise.
5079 (want_inline_small_function_p): Likewise.
5080 (edge_badness): Likewise.
5081 (update_caller_keys): Likewise.
5082 (update_callee_keys): Likewise.
5083 (recursive_inlining): Likewise.
5084 (inline_small_functions): Likewise.
5085 (inline_to_all_callers_1): Likewise.
5086 (dump_overall_stats): Likewise.
5087 (early_inline_small_functions): Likewise.
5088 (early_inliner): Likewise.
5089 * ipa-inline.h (estimate_edge_growth): Likewise.
5090 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
5091 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5092 * ipa-prop.h (IPA_NODE_REF): Likewise.
5093 (IPA_EDGE_REF): Likewise.
5094 * ipa-pure-const.c (malloc_candidate_p): Likewise.
5095 (propagate_malloc): Likewise.
5096 * ipa-split.c (execute_split_functions): Likewise.
5097 * symbol-summary.h: Rename get to get_create.
5099 (get_create): Likewise.
5100 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
5102 2018-06-08 Martin Liska <mliska@suse.cz>
5104 * symbol-summary.h (release): Move definition out of class
5106 (symtab_removal): Likewise.
5107 (symtab_duplication): Likewise.
5109 2018-06-08 Martin Liska <mliska@suse.cz>
5111 * symbol-summary.h (function_summary): Move constructor
5112 implementation out of class declaration.
5113 (release): Likewise.
5114 (symtab_insertion): Likewise.
5115 (symtab_removal): Likewise.
5116 (symtab_duplication): Likewise.
5119 2018-06-08 Martin Liska <mliska@suse.cz>
5121 * Makefile.in: Remove support for MPX (macros, related functions,
5122 fields in cgraph_node, ...).
5123 * builtin-types.def (BT_BND): Likewise.
5124 (BT_FN_BND_CONST_PTR): Likewise.
5125 (BT_FN_CONST_PTR_BND): Likewise.
5126 (BT_FN_VOID_PTR_BND): Likewise.
5127 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
5128 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
5129 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
5130 (expand_builtin_mempcpy_with_bounds): Likewise.
5131 (expand_builtin_memset_with_bounds): Likewise.
5132 (expand_builtin_memset_args): Likewise.
5133 (std_expand_builtin_va_start): Likewise.
5134 (expand_builtin): Likewise.
5135 (expand_builtin_with_bounds): Likewise.
5136 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
5137 (DEF_LIB_BUILTIN_CHKP): Likewise.
5138 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
5139 (DEF_CHKP_BUILTIN): Likewise.
5140 (BUILT_IN_MEMCPY): Likewise.
5141 (BUILT_IN_MEMMOVE): Likewise.
5142 (BUILT_IN_MEMPCPY): Likewise.
5143 (BUILT_IN_MEMSET): Likewise.
5144 (BUILT_IN_STPCPY): Likewise.
5145 (BUILT_IN_STRCAT): Likewise.
5146 (BUILT_IN_STRCHR): Likewise.
5147 (BUILT_IN_STRCPY): Likewise.
5148 (BUILT_IN_STRLEN): Likewise.
5149 (BUILT_IN_MEMCPY_CHK): Likewise.
5150 (BUILT_IN_MEMMOVE_CHK): Likewise.
5151 (BUILT_IN_MEMPCPY_CHK): Likewise.
5152 (BUILT_IN_MEMSET_CHK): Likewise.
5153 (BUILT_IN_STPCPY_CHK): Likewise.
5154 (BUILT_IN_STRCAT_CHK): Likewise.
5155 (BUILT_IN_STRCPY_CHK): Likewise.
5156 * calls.c (store_bounds): Likewise.
5157 (emit_call_1): Likewise.
5158 (special_function_p): Likewise.
5159 (maybe_warn_nonstring_arg): Likewise.
5160 (initialize_argument_information): Likewise.
5161 (finalize_must_preallocate): Likewise.
5162 (compute_argument_addresses): Likewise.
5163 (expand_call): Likewise.
5164 * cfgexpand.c (expand_call_stmt): Likewise.
5165 (expand_return): Likewise.
5166 (expand_gimple_stmt_1): Likewise.
5167 (pass_expand::execute): Likewise.
5168 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
5169 (cgraph_node::remove): Likewise.
5170 (cgraph_node::dump): Likewise.
5171 (cgraph_node::verify_node): Likewise.
5172 * cgraph.h (chkp_function_instrumented_p): Likewise.
5173 (symtab_node::get_alias_target): Likewise.
5174 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
5175 (cgraph_local_p): Likewise.
5176 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
5177 (cgraph_edge::rebuild_references): Likewise.
5178 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
5179 (walk_polymorphic_call_targets): Likewise.
5180 (cgraph_node::expand_thunk): Likewise.
5181 (symbol_table::output_weakrefs): Likewise.
5182 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
5183 (ix86_handle_option): Likewise.
5184 * config/i386/constraints.md: Likewise.
5185 * config/i386/i386-builtin-types.def (BND): Likewise.
5189 * config/i386/i386-builtin.def (BDESC_END): Likewise.
5190 (BDESC_FIRST): Likewise.
5192 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
5193 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
5194 * config/i386/i386.c (enum reg_class): Likewise.
5195 (ix86_target_string): Likewise.
5196 (ix86_option_override_internal): Likewise.
5197 (ix86_conditional_register_usage): Likewise.
5198 (ix86_valid_target_attribute_inner_p): Likewise.
5199 (ix86_set_indirect_branch_type): Likewise.
5200 (ix86_set_current_function): Likewise.
5201 (ix86_function_arg_regno_p): Likewise.
5202 (init_cumulative_args): Likewise.
5203 (ix86_function_arg_advance): Likewise.
5204 (ix86_function_arg): Likewise.
5205 (ix86_pass_by_reference): Likewise.
5206 (ix86_function_value_regno_p): Likewise.
5207 (ix86_function_value_1): Likewise.
5208 (ix86_function_value_bounds): Likewise.
5209 (ix86_return_in_memory): Likewise.
5210 (ix86_setup_incoming_vararg_bounds): Likewise.
5211 (ix86_va_start): Likewise.
5212 (indirect_thunk_need_prefix): Likewise.
5213 (print_reg): Likewise.
5214 (ix86_print_operand): Likewise.
5215 (ix86_expand_call): Likewise.
5216 (ix86_output_function_return): Likewise.
5217 (reg_encoded_number): Likewise.
5218 (BDESC_VERIFYS): Likewise.
5219 (ix86_init_mpx_builtins): Likewise.
5220 (ix86_init_builtins): Likewise.
5221 (ix86_emit_cmove): Likewise.
5222 (ix86_emit_move_max): Likewise.
5223 (ix86_expand_builtin): Likewise.
5224 (ix86_builtin_mpx_function): Likewise.
5225 (ix86_get_arg_address_for_bt): Likewise.
5226 (ix86_load_bounds): Likewise.
5227 (ix86_store_bounds): Likewise.
5228 (ix86_load_returned_bounds): Likewise.
5229 (ix86_store_returned_bounds): Likewise.
5230 (ix86_class_likely_spilled_p): Likewise.
5231 (ix86_hard_regno_mode_ok): Likewise.
5232 (x86_order_regs_for_local_alloc): Likewise.
5233 (ix86_mitigate_rop): Likewise.
5234 (ix86_bnd_prefixed_insn_p): Likewise.
5235 (ix86_mpx_bound_mode): Likewise.
5236 (ix86_make_bounds_constant): Likewise.
5237 (ix86_initialize_bounds): Likewise.
5238 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
5239 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
5240 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
5241 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
5242 (TARGET_CHKP_BOUND_MODE): Likewise.
5243 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
5244 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
5245 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
5246 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
5247 * config/i386/i386.h (TARGET_MPX): Likewise.
5248 (TARGET_MPX_P): Likewise.
5249 (VALID_BND_REG_MODE): Likewise.
5250 (FIRST_BND_REG): Likewise.
5251 (LAST_BND_REG): Likewise.
5252 (enum reg_class): Likewise.
5253 (BND_REG_P): Likewise.
5254 (BND_REGNO_P): Likewise.
5255 (BNDmode): Likewise.
5256 (ADJUST_INSN_LENGTH): Likewise.
5257 * config/i386/i386.md: Likewise.
5258 * config/i386/i386.opt: Likewise.
5259 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
5260 (defined): Likewise.
5261 (LINK_MPX): Likewise.
5262 (MPX_SPEC): Likewise.
5263 (LIBMPX_SPEC): Likewise.
5264 (LIBMPXWRAPPERS_SPEC): Likewise.
5265 (CHKP_SPEC): Likewise.
5266 * config/i386/predicates.md: Likewise.
5267 * dbxout.c (dbxout_type): Likewise.
5268 * doc/extend.texi: Likewise.
5269 * doc/invoke.texi: Likewise.
5270 * doc/md.texi: Likewise.
5271 * doc/tm.texi: Likewise.
5272 * doc/tm.texi.in: Likewise.
5273 * dwarf2out.c (is_base_type): Likewise.
5274 (gen_formal_types_die): Likewise.
5275 (gen_subprogram_die): Likewise.
5276 (gen_type_die_with_usage): Likewise.
5277 (gen_decl_die): Likewise.
5278 (dwarf2out_late_global_decl): Likewise.
5279 * expr.c (expand_assignment): Likewise.
5280 (emit_storent_insn): Likewise.
5281 (store_expr_with_bounds): Likewise.
5282 (store_expr): Likewise.
5283 (expand_expr_real_1): Likewise.
5284 * expr.h (store_expr_with_bounds): Likewise.
5285 * function.c (use_register_for_decl): Likewise.
5286 (struct bounds_parm_data): Likewise.
5287 (assign_parms_augmented_arg_list): Likewise.
5288 (assign_parm_find_entry_rtl): Likewise.
5289 (assign_parm_is_stack_parm): Likewise.
5290 (assign_parm_load_bounds): Likewise.
5291 (assign_bounds): Likewise.
5292 (assign_parms): Likewise.
5293 (expand_function_start): Likewise.
5294 * gcc.c (CHKP_SPEC): Likewise.
5295 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
5296 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
5297 (wrestrict_dom_walker::check_call): Likewise.
5298 * gimple.c (gimple_build_call_from_tree): Likewise.
5299 * gimple.h (enum gf_mask): Likewise.
5300 (gimple_call_with_bounds_p): Likewise.
5301 (gimple_call_set_with_bounds): Likewise.
5302 * gimplify.c (gimplify_init_constructor): Likewise.
5303 * ipa-cp.c (initialize_node_lattices): Likewise.
5304 (propagate_constants_across_call): Likewise.
5305 (find_more_scalar_values_for_callers_subset): Likewise.
5306 * ipa-hsa.c (process_hsa_functions): Likewise.
5307 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
5308 * ipa-icf.c (sem_function::merge): Likewise.
5309 * ipa-inline.c (early_inliner): Likewise.
5310 * ipa-pure-const.c (warn_function_noreturn): Likewise.
5311 (warn_function_cold): Likewise.
5312 (propagate_pure_const): Likewise.
5313 * ipa-ref.h (enum GTY): Likewise.
5314 * ipa-split.c (find_retbnd): Likewise.
5315 (consider_split): Likewise.
5316 (split_function): Likewise.
5317 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
5318 * ipa.c (walk_polymorphic_call_targets): Likewise.
5319 (symbol_table::remove_unreachable_nodes): Likewise.
5320 (process_references): Likewise.
5321 (cgraph_build_static_cdtor_1): Likewise.
5322 * lto-cgraph.c (lto_output_node): Likewise.
5323 (output_refs): Likewise.
5324 (compute_ltrans_boundary): Likewise.
5325 (input_overwrite_node): Likewise.
5326 (input_node): Likewise.
5327 (input_cgraph_1): Likewise.
5328 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
5329 * passes.c (pass_manager::execute_early_local_passes): Likewise.
5330 (class pass_chkp_instrumentation_passes): Likewise.
5331 (make_pass_chkp_instrumentation_passes): Likewise.
5332 * passes.def: Likewise.
5333 * rtl.h (struct GTY): Likewise.
5334 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
5335 * stor-layout.c (layout_type): Likewise.
5336 * symtab.c: Likewise.
5337 * target.def: Likewise.
5338 * targhooks.c (default_chkp_bound_type): Likewise.
5339 (default_chkp_bound_mode): Likewise.
5340 (default_builtin_chkp_function): Likewise.
5341 (default_chkp_function_value_bounds): Likewise.
5342 (default_chkp_make_bounds_constant): Likewise.
5343 (default_chkp_initialize_bounds): Likewise.
5344 * targhooks.h (default_chkp_bound_type): Likewise.
5345 (default_chkp_bound_mode): Likewise.
5346 (default_builtin_chkp_function): Likewise.
5347 (default_chkp_function_value_bounds): Likewise.
5348 (default_chkp_make_bounds_constant): Likewise.
5349 (default_chkp_initialize_bounds): Likewise.
5350 * toplev.c (compile_file): Likewise.
5351 (process_options): Likewise.
5352 * tree-core.h (DEF_BUILTIN): Likewise.
5353 (DEF_BUILTIN_CHKP): Likewise.
5354 * tree-inline.c (declare_return_variable): Likewise.
5355 (remap_gimple_stmt): Likewise.
5356 (copy_bb): Likewise.
5357 (initialize_inlined_parameters): Likewise.
5358 (expand_call_inline): Likewise.
5359 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
5360 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
5361 (make_pass_ipa_chkp_produce_thunks): Likewise.
5362 (make_pass_chkp): Likewise.
5363 (make_pass_chkp_opt): Likewise.
5364 (make_pass_chkp_instrumentation_passes): Likewise.
5365 * tree-pretty-print.c (dump_generic_node): Likewise.
5366 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
5367 * tree-ssa-dce.c (propagate_necessity): Likewise.
5368 (eliminate_unnecessary_stmts): Likewise.
5369 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
5370 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
5371 * tree-ssa-sccvn.h: Likewise.
5372 * tree-ssa-strlen.c (get_string_length): Likewise.
5373 (valid_builtin_call): Likewise.
5374 (adjust_last_stmt): Likewise.
5375 (handle_builtin_strchr): Likewise.
5376 (handle_builtin_strcpy): Likewise.
5377 (handle_builtin_stxncpy): Likewise.
5378 (handle_builtin_memcpy): Likewise.
5379 (handle_builtin_strcat): Likewise.
5380 (strlen_check_and_optimize_stmt): Likewise.
5381 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
5382 * tree-streamer-in.c: Likewise.
5383 * tree-streamer.c (record_common_node): Likewise.
5384 * tree.c (tree_code_size): Likewise.
5385 (wide_int_to_tree_1): Likewise.
5386 (type_contains_placeholder_1): Likewise.
5387 (build_common_tree_nodes): Likewise.
5388 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
5389 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
5390 (POINTER_BOUNDS_P): Likewise.
5391 (BOUNDED_TYPE_P): Likewise.
5392 (BOUNDED_P): Likewise.
5393 (CALL_WITH_BOUNDS_P): Likewise.
5394 (pointer_bounds_type_node): Likewise.
5395 * value-prof.c (gimple_ic): Likewise.
5396 * var-tracking.c (vt_add_function_parameters): Likewise.
5397 * varasm.c (make_decl_rtl): Likewise.
5398 (assemble_start_function): Likewise.
5399 (output_constant): Likewise.
5400 (maybe_assemble_visibility): Likewise.
5401 * varpool.c (ctor_for_folding): Likewise.
5402 * chkp-builtins.def: Remove.
5403 * ipa-chkp.c: Remove.
5404 * ipa-chkp.h: Remove.
5405 * rtl-chkp.c: Remove.
5406 * rtl-chkp.h: Remove.
5407 * tree-chkp-opt.c: Remove.
5408 * tree-chkp.c: Remove.
5409 * tree-chkp.h: Remove.
5411 2018-06-07 Carl Love <cel@us.ibm.com>
5413 * config/rs6000/vsx.md (vextract_fp_from_shorth,
5414 vextract_fp_from_shortl): Add BE support.
5416 2018-06-07 Paul Koning <ni1d@arrl.net>
5418 * compare-elim.c (try_merge_compare): Don't merge compare if
5419 address contains a side effect.
5420 (try_eliminate_compare): Likewise.
5422 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
5424 * config.gcc: Support "tremont".
5425 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
5426 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5428 * config/i386/i386.c (m_TREMONT): Define.
5429 (processor_target_table): Add "tremont".
5430 (PTA_TREMONT): Define.
5431 (ix86_lea_outperforms): Add TARGET_TREMONT.
5432 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
5433 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
5434 and M_INTEL_GOLDMONT_PLUS.
5435 (fold_builtin_cpu): Add "tremont".
5436 (ix86_add_stmt_cost): Add TARGET_TREMONT.
5437 (ix86_option_override_internal): Add "tremont".
5438 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
5439 (processor_type): Add PROCESSOR_TREMONT.
5440 * config/i386/x86-tune.def: Add m_TREMONT.
5441 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
5443 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5445 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
5446 symbol defined for msp430i* devices to be lower case.
5448 2018-06-07 Richard Biener <rguenther@suse.de>
5450 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
5451 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
5452 Properly wrap signed arithmetic if overflow wraps.
5454 2018-06-07 Jakub Jelinek <jakub@redhat.com>
5456 PR tree-optimization/69615
5457 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
5458 of a cast from a same precision integral SSA_NAME in a bb dominated
5459 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
5460 cast to utype if rhs2 has already a compatible type.
5462 2018-06-07 Richard Biener <rguenther@suse.de>
5464 PR tree-optimization/85935
5465 * graphite-scop-detection.c (find_params_in_bb): Analyze
5466 condition operands with respect to the correct loop. Assert
5467 the analysis doesn't fail.
5469 2018-06-04 Carl Love <cel@us.ibm.com>
5471 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
5472 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
5473 as it is slightly cheaper.
5474 (first_match_or_eos_index_<mode>):
5475 Calculate index using natural element order.
5476 (first_match_index_<mode>):
5477 Calculate index using natural element order.
5478 (first_match_or_eos_index_<mode>):
5479 Calculate index using natural order.
5480 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
5481 for BE and LE modes.
5482 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
5483 P9V_BUILTIN_VCLZLSBB_V16QI.
5484 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
5487 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
5489 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
5490 indentation and line wrap for many prototypes. Add missing
5491 @smallexample directives around block of prototypes for vec_xl and
5494 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
5496 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
5497 track if we pass or return IEEE 128-bit floating point.
5498 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
5499 C++ mangling that is compatible with GCC 8.1.
5500 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
5501 (init_cumulative_args): Note if we pass or return IEEE 128-bit
5502 floating point types.
5503 (rs6000_function_arg_advance_1): Likewise.
5504 (rs6000_mangle_type): Optionally generate mangled names that match
5505 what GCC 8.1 generated for IEEE 128-bit floating point types.
5506 (rs6000_globalize_decl_name): If we have an external function that
5507 passes or returns IEEE 128-bit types, generate a weak reference
5508 from the mangled name used in GCC 8.1 to the current mangled
5510 (rs6000_init_builtins): Make __ibm128 use the long double type if
5511 long double is IBM extended double. Make __float128 use the long
5512 double type if long double is IEEE 128-bit.
5515 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
5516 macro for __ibm128 built-in functions.
5517 (PACK_IF): Add __ibm128 pack/unpack functions.
5518 (UNPACK_IF): Likewise.
5519 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
5520 enable long double built-in functions if long double is IEEE
5521 128-bit floating point.
5522 (rs6000_invalid_builtin): Update long double built-in function
5524 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
5525 functions, adjust the built-in function to use the long double
5526 built-in function if __ibm128 and long double are the same type.
5527 * doc/extend.texi (PowerPC builtins): Update documention for
5528 __builtin_{,un}pack_longdouble. Add documentation for
5529 __builtin_{,un}pack_ibm128.
5531 2018-06-06 Jim Wilson <jimw@sifive.com>
5533 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
5534 (struct machine_function): New field interrupt_mode.
5535 (riscv_handle_type_attribute): New function. Add forward declaration.
5536 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
5537 (riscv_expand_epilogue): Check interrupt_mode field.
5538 (riscv_set_current_function): Check interrupt attribute args and
5539 set interrupt_mode field.
5540 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
5541 (riscv_sret, riscv_uret): New.
5542 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
5543 new arguments to interrupt attribute.
5545 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
5548 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
5549 Don't handle -mcpu=power8 if -mpower9-vector is also used.
5551 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
5553 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
5554 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
5555 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5556 several redundant entries.
5558 2018-06-06 David Malcolm <dmalcolm@redhat.com>
5560 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
5561 type from "rtx" to "rtx_insn *".
5562 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
5563 for local "call_insn", removing cast.
5564 (ix86_expand_call): Likewise, introducing a "call_insn" local.
5566 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
5568 PR tree-optimization/86066
5569 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
5570 for BIT_INSERT_EXPR stores.
5572 2018-06-06 Richard Biener <rguenther@suse.de>
5574 PR tree-optimization/86062
5575 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5576 component refs ontop
5577 of to be offsetted base.
5579 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5581 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
5582 to be static and remove check on interrupt attribute name.
5584 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
5586 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5587 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
5589 2018-06-05 Steve Ellcey <sellcey@cavium.com>
5592 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
5594 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
5595 Remove second argument, change how error is called.
5596 (aarch64_layout_arg): Remove second argument from
5597 aarch64_err_no_fpadvsimd call.
5598 (aarch64_init_cumulative_args): Ditto.
5599 (aarch64_gimplify_va_arg_expr): Ditto.
5600 * config/aarch64/aarch64.md (mov<mode>): Ditto.
5602 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
5604 * config/i386/i386.md (simple_return_indirect_internal): New expander.
5605 (*simple_return_indirect_internal<mode>): Rename from
5606 simple_return_indirect_internal. Use W mode iterator.
5607 (rstorssp): New expander.
5608 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
5609 (clrssbsy): New expander.
5610 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
5612 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
5614 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
5616 (cmse_check_pointed_object): Likewise.
5618 2018-06-05 Martin Liska <mliska@suse.cz>
5620 PR gcov-profile/47618
5621 * doc/invoke.texi: Document how -fprofile-dir format
5624 2018-06-05 Richard Biener <rguenther@suse.de>
5626 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
5627 removal pretend DOM info isn't available so we do not update
5628 it and only remove edges, not dominated blocks. Actually free
5629 DOM info in case we removed something. Remove unreachable blocks.
5630 (mfb_keep_latches): Work with either DOM info or marked backedges.
5631 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
5632 first. Mark backedges if DOM info isn't available.
5633 (Re-)compute DOM info after cleanup_control_flow_pre.
5635 2018-06-05 Richard Biener <rguenther@suse.de>
5637 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
5638 (locus_discrim_hasher::hash): Adjust.
5639 (locus_discrim_hasher::equal): Likewise.
5640 (next_discriminator_for_locus): Work on line directly.
5641 (same_line_p): Pass in expanded locus1 as well.
5642 (assign_discriminators): Avoid redundant location expansions.
5644 2018-06-05 Richard Biener <rguenther@suse.de>
5646 PR tree-optimization/86046
5647 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
5648 if required after clearing TREE_ADDRESSABLE.
5650 2018-06-05 Richard Biener <rguenther@suse.de>
5652 PR tree-optimization/86047
5653 * tree-ssa-loop.c (for_each_index): Glob handling of all
5654 decls and constants and really handle all of them.
5656 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5659 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
5660 qualifier_void_pointer and qualifier_const_void_pointer.
5661 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
5662 (arm_init_builtins): Handle the above.
5663 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
5664 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
5665 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
5668 2018-06-05 Martin Liska <mliska@suse.cz>
5670 * auto-profile.c (read_autofdo_file): Do not use
5671 gcov_ctr_summary struct.
5672 (afdo_callsite_hot_enough_for_early_inline): Likewise.
5673 * coverage.c (struct counts_entry): Likewise.
5674 (read_counts_file): Read just single summary entry.
5675 (get_coverage_counts): Use gcov_summary struct.
5676 * coverage.h (get_coverage_counts): Likewise.
5677 * gcov-dump.c (dump_working_sets): Likewise.
5678 (tag_summary): Dump just single summary.
5679 * gcov-io.c (gcov_write_summary): Write just histogram
5681 (gcov_read_summary): Read just single summary.
5682 (compute_working_sets): Use gcov_summary struct.
5683 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
5684 of GCOV_COUNTERS_SUMMABLE.
5685 (GCOV_COUNTERS_SUMMABLE): Remove.
5686 (GCOV_FIRST_VALUE_COUNTER): Replace with
5687 GCOV_COUNTER_V_INTERVAL.
5688 (struct gcov_ctr_summary): Remove.
5689 (struct gcov_summary): Directly use fields of former
5691 (compute_working_sets): Use gcov_summary struct.
5692 * gcov.c (read_count_file): Do not use ctrs fields.
5693 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
5695 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
5697 * profile.c: Likewise.
5698 * profile.h: Likewise.
5700 2018-06-05 Martin Liska <mliska@suse.cz>
5702 PR gcov-profile/84846
5703 * gcov.c (output_lines): Print working directory only
5704 in intermediate format.
5706 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
5708 * config/s390/s390-builtin-types.def: Add void function type.
5709 * config/s390/s390-builtins.def: Use the function type for the
5712 2018-06-04 Jim Wilson <jimw@sifive.com>
5714 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
5716 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
5717 and maybe_eh_return. Change regno to unsigned int. Use new args to
5718 handle EH_RETURN_DATA_REGNO registers properly.
5719 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
5720 (riscv_expand_epilogue): Update comment. Change argument name and
5721 type. Update code to use new name and type. Pass new args to
5722 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
5724 * config/riscv/riscv.md (NORMAL_RETURN): New.
5725 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
5726 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
5727 (eh_return): Call gen_eh_return_internal and emit barrier.
5728 (eh_return_internal): Call riscv_expand_epilogue.
5730 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
5732 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
5733 bit_insertion field and declare can_be_merged_into method.
5734 (merged_store_group::can_be_merged_into): New method.
5735 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
5736 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
5737 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
5739 2018-06-04 Richard Biener <rguenther@suse.de>
5741 PR tree-optimization/85955
5742 * builtins.c (fold_builtin_sincos): Convert pointers to
5743 destination to appropriate type before dereferencing.
5745 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
5747 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
5749 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
5751 * expr.c (expand_expr_real_1): Force the operand into memory if
5752 its TYPE_MODE is BLKmode and if there is no integer mode for
5753 the number of bits being extracted.
5755 2018-06-04 Jakub Jelinek <jakub@redhat.com>
5759 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5760 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
5762 2018-06-04 Richard Biener <rguenther@suse.de>
5764 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
5765 (cleanup_tree_cfg_noloop): ... single caller. Do
5766 start_recording_case_labels later.
5768 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
5770 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
5771 to _IMMINTRIN_H_INCLUDED.
5772 * config/i386/pconfigintrin.h: Ditto.
5773 * config/i386/waitpkgintrin.h: Ditto.
5774 * config/i386/immintrin.h: Add includes for sgxintrin.h,
5775 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
5776 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
5777 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
5778 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
5779 waitpkgintrin.h and cldemoteintrin.h.
5781 2018-06-04 Richard Biener <rguenther@suse.de>
5783 PR tree-optimization/86038
5784 * tracer.c (find_best_successor): Check probability for
5785 being initialized, bail out if not.
5787 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
5790 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
5791 of bits to ignore when comparing architectures.
5793 2018-06-04 Jakub Jelinek <jakub@redhat.com>
5795 PR tree-optimization/69615
5796 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
5797 maximum or minimum of the type, try to merge it also as if
5798 range1 is + [-, x - 1] or + [x + 1, -].
5801 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
5803 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
5805 PR tree-optimization/86034
5806 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
5807 the unsigned bitfield type in a bit insertion sequence if it does not
5808 have a larger precision than the bitfield size.
5809 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
5811 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
5813 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
5815 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
5817 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
5818 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
5819 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
5820 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
5822 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
5824 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
5825 Disable -fdelete-null-pointer-checks for ELF toolchain.
5827 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
5828 Kito Cheng <kito.cheng@gmail.com>
5830 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
5831 (nds32le-*-*, nds32be-*-*): Integrate checking process.
5832 (nds32*-*-*): Add glibc and uclibc conditions.
5833 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
5834 (TARGET_EXCEPT_UNWIND_INFO): Define.
5835 * config/nds32/elf.h: New file.
5836 * config/nds32/linux.h: New file.
5837 * config/nds32/nds32-elf.opt: New file.
5838 * config/nds32/nds32-linux.opt: New file.
5839 * config/nds32/nds32-fp-as-gp.c
5840 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
5841 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
5843 (nds32_asm_file_end): Ditto.
5844 (nds32_print_operand): Ditto.
5845 (nds32_insert_attributes): Ditto.
5846 (nds32_init_libfuncs): New function.
5847 (TARGET_HAVE_TLS): Define.
5848 (TARGET_INIT_LIBFUNCS): Define.
5849 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
5851 (TARGET_ELF): Apply different mcmodel setting.
5852 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
5853 been migrated into elf.h and linux.h files.
5854 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
5855 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
5856 (mcmodel): The content has been migrated into nds32-elf.opt and
5857 nds32-linux.opt files.
5858 * config/nds32/t-elf: New file.
5859 * config/nds32/t-linux: New file.
5861 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
5862 Shiva Chen <shiva0217@gmail.com>
5864 * config/nds32/constants.md (unspec_volatile_element): Add
5865 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
5866 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
5868 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
5869 (make_pass_nds32_fp_as_gp): Declare.
5870 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
5872 (nds32_asm_function_end_prologue): Remove unused asm output.
5873 (nds32_asm_function_begin_epilogue): Remove unused asm output.
5874 (nds32_asm_file_start): Output necessary fp_as_gp information.
5875 (nds32_option_override): Adjust register usage.
5876 (nds32_expand_prologue): Consider fp_as_gp situation.
5877 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
5878 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
5881 (simple_return): Ditto.
5882 (omit_fp_begin): Output special directive for fp_as_gp.
5883 (omit_fp_end): Output special directive for fp_as_gp.
5884 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
5885 mforbid-fp-as-gp): New options.
5887 2018-06-01 Mark Wielaard <mark@klomp.org>
5889 * dwarf2out.c (dwarf2out_finish): Remove generation of
5890 DW_AT_loclists_base.
5892 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
5894 * gimple-ssa-store-merging.c: Include gimple-fold.h.
5895 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
5896 (struct merged_store_group): Add bit_insertion field.
5897 (dump_char_array): Use standard hexadecimal format.
5898 (merged_store_group::merged_store_group): Set bit_insertion to false.
5899 (merged_store_group::apply_stores): Use optimal buffer size. Deal
5900 with BIT_INSERT_EXPR stores. Move up code updating the mask and
5901 also print the mask in the dump file.
5902 (pass_store_merging::gate): Minor tweak.
5903 (imm_store_chain_info::coalesce_immediate): Fix wrong association
5904 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
5905 stores with INTEGER_CST stores.
5906 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
5907 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
5908 and use it throughout. Generate bit insertion sequences if need be.
5909 (pass_store_merging::process_store): Remove redundant condition.
5910 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
5912 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
5914 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
5915 the 128-bit floating point types. Fix function comment.
5917 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5919 * config/aarch64/aarch64-simd.md
5920 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
5922 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
5925 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
5927 PR tree-optimization/85989
5928 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
5930 (backprop::intersect_uses): Check it when deciding whether this
5931 is a backedge reference.
5932 (backprop::process_block): Add each phi to m_visited_phis
5933 after visiting it, then clear it at the end.
5935 2018-06-01 Richard Biener <rguenther@suse.de>
5937 * tree-vectorizer.h (vect_dr_stmt): New function.
5938 (vect_get_load_cost): Adjust.
5939 (vect_get_store_cost): Likewise.
5940 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
5941 Use vect_dr_stmt instead of DR_SMTT.
5942 (vect_record_base_alignments): Likewise.
5943 (vect_calculate_target_alignment): Likewise.
5944 (vect_compute_data_ref_alignment): Likewise and make static.
5945 (vect_update_misalignment_for_peel): Likewise.
5946 (vect_verify_datarefs_alignment): Likewise.
5947 (vector_alignment_reachable_p): Likewise.
5948 (vect_get_data_access_cost): Likewise. Pass down
5949 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
5950 (vect_get_peeling_costs_all_drs): Likewise.
5951 (vect_peeling_hash_get_lowest_cost): Likewise.
5952 (vect_enhance_data_refs_alignment): Likewise.
5953 (vect_find_same_alignment_drs): Likewise.
5954 (vect_analyze_data_refs_alignment): Likewise.
5955 (vect_analyze_group_access_1): Likewise.
5956 (vect_analyze_group_access): Likewise.
5957 (vect_analyze_data_ref_access): Likewise.
5958 (vect_analyze_data_ref_accesses): Likewise.
5959 (vect_vfa_segment_size): Likewise.
5960 (vect_small_gap_p): Likewise.
5961 (vectorizable_with_step_bound_p): Likewise.
5962 (vect_prune_runtime_alias_test_list): Likewise.
5963 (vect_analyze_data_refs): Likewise.
5964 (vect_supportable_dr_alignment): Likewise.
5965 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
5966 (vect_gen_prolog_loop_niters): Likewise.
5967 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5968 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
5970 (vect_recog_mask_conversion_pattern): Likewise.
5971 (vect_try_gather_scatter_pattern): Likewise.
5972 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
5973 to vect_get_store_cost.
5974 (vect_get_store_cost): Get stmt_info instead of DR.
5975 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
5976 (vect_get_load_cost): Get stmt_info instead of DR.
5978 2018-06-01 Richard Biener <rguenther@suse.de>
5981 * gimple-fold.c (var_decl_component_p): Also allow offsetted
5982 vars wrapped in MEM_REFs.
5984 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
5986 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
5987 Fix subreg tests so that we only return a choice between
5988 GENERAL_REGS and FP_REGS if the original classes included both.
5990 2018-06-01 Richard Biener <rguenther@suse.de>
5993 * tree-ssa-structalias.c (get_function_part_constraint):
5994 Handle NULL fi->decl.
5995 (find_func_aliases_for_call): Properly handle indirect
5996 fi from direct call.
5997 (find_func_clobbers): Likewise.
5998 (ipa_pta_execute): Likewise.
5999 (create_variable_info_for): For functions that are ifunc_resolver
6000 resolve to a varinfo that contains the result of the resolver call.
6001 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
6004 2018-05-31 Michael Collison <michael.collison@arm.com>
6006 * config/aarch64/aarch64.md:
6007 (*fix_to_zero_extenddfdi2): New pattern.
6008 * gcc.target/aarch64/fix_extend1.c: New testcase.
6010 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
6014 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
6015 and BUILT_IN_STRNCMP_EQ.
6016 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
6017 BUILT_IN_STRNCMP_EQ.
6018 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
6019 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
6020 (gimple_fold_builtin): Likewise.
6021 * tree-ssa-strlen.c (compute_string_length): New function.
6022 (determine_min_obsize): New function.
6023 (handle_builtin_string_cmp): New function to handle calls to
6024 string compare functions.
6025 (strlen_optimize_stmt): Add handling to builtin string compare
6027 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
6028 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
6029 * tree.c (build_common_builtin_nodes): Add new defines of
6030 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
6032 2018-05-31 Jakub Jelinek <jakub@redhat.com>
6035 * bb-reorder.c (pass_partition_blocks::gate): Return false for
6036 functions with naked attribute.
6038 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
6040 * config/i386/sse.md (avx_vec_concat<mode>):
6041 Substitute concat_tg_mode mode attribute with xtg_mode.
6042 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
6043 (concat_tg_mode): Remove mode attribute.
6045 2018-05-31 Martin Sebor <msebor@redhat.com>
6048 * calls.c (alloc_max_size): Correct a logic error/typo.
6049 Treat excessive arguments as infinite. Warn for invalid arguments.
6050 * doc/invoke.texi (-Walloc-size-larger-than): Update.
6052 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
6055 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
6056 and movx for Haswell.
6058 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
6059 Cesar Philippidis <cesar@codesourcery.com>
6062 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
6063 when emitting error on private/firstprivate reductions.
6064 * omp-low.c (lower_omp_target): Avoid reference-type processing
6065 on pointers for firstprivate clause.
6067 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
6069 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
6072 * config/aarch64/aarch64-simd.md
6073 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
6074 (aarch64_ld1_x3_<mode>): Likewise
6075 (aarch64_st1x2<VALLDIF:mode>): Likewise
6076 (aarch64_st1_x2_<mode>): Likewise
6077 (aarch64_st1x3<VALLDIF:mode>): Likewise
6078 (aarch64_st1_x3_<mode>): Likewise
6079 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
6080 (vld1_s8_x3): Likewise.
6081 (vld1_u16_x3): Likewise.
6082 (vld1_s16_x3): Likewise.
6083 (vld1_u32_x3): Likewise.
6084 (vld1_s32_x3): Likewise.
6085 (vld1_u64_x3): Likewise.
6086 (vld1_s64_x3): Likewise.
6087 (vld1_f16_x3): Likewise.
6088 (vld1_f32_x3): Likewise.
6089 (vld1_f64_x3): Likewise.
6090 (vld1_p8_x3): Likewise.
6091 (vld1_p16_x3): Likewise.
6092 (vld1_p64_x3): Likewise.
6093 (vld1q_u8_x3): Likewise.
6094 (vld1q_s8_x3): Likewise.
6095 (vld1q_u16_x3): Likewise.
6096 (vld1q_s16_x3): Likewise.
6097 (vld1q_u32_x3): Likewise.
6098 (vld1q_s32_x3): Likewise.
6099 (vld1q_u64_x3): Likewise.
6100 (vld1q_s64_x3): Likewise.
6101 (vld1q_f16_x3): Likewise.
6102 (vld1q_f32_x3): Likewise.
6103 (vld1q_f64_x3): Likewise.
6104 (vld1q_p8_x3): Likewise.
6105 (vld1q_p16_x3): Likewise.
6106 (vld1q_p64_x3): Likewise.
6107 (vst1_s64_x2): Likewise.
6108 (vst1_u64_x2): Likewise.
6109 (vst1_f64_x2): Likewise.
6110 (vst1_s8_x2): Likewise.
6111 (vst1_p8_x2): Likewise.
6112 (vst1_s16_x2): Likewise.
6113 (vst1_p16_x2): Likewise.
6114 (vst1_s32_x2): Likewise.
6115 (vst1_u8_x2): Likewise.
6116 (vst1_u16_x2): Likewise.
6117 (vst1_u32_x2): Likewise.
6118 (vst1_f16_x2): Likewise.
6119 (vst1_f32_x2): Likewise.
6120 (vst1_p64_x2): Likewise.
6121 (vst1q_s8_x2): Likewise.
6122 (vst1q_p8_x2): Likewise.
6123 (vst1q_s16_x2): Likewise.
6124 (vst1q_p16_x2): Likewise.
6125 (vst1q_s32_x2): Likewise.
6126 (vst1q_s64_x2): Likewise.
6127 (vst1q_u8_x2): Likewise.
6128 (vst1q_u16_x2): Likewise.
6129 (vst1q_u32_x2): Likewise.
6130 (vst1q_u64_x2): Likewise.
6131 (vst1q_f16_x2): Likewise.
6132 (vst1q_f32_x2): Likewise.
6133 (vst1q_f64_x2): Likewise.
6134 (vst1q_p64_x2): Likewise.
6135 (vst1_s64_x3): Likewise.
6136 (vst1_u64_x3): Likewise.
6137 (vst1_f64_x3): Likewise.
6138 (vst1_s8_x3): Likewise.
6139 (vst1_p8_x3): Likewise.
6140 (vst1_s16_x3): Likewise.
6141 (vst1_p16_x3): Likewise.
6142 (vst1_s32_x3): Likewise.
6143 (vst1_u8_x3): Likewise.
6144 (vst1_u16_x3): Likewise.
6145 (vst1_u32_x3): Likewise.
6146 (vst1_f16_x3): Likewise.
6147 (vst1_f32_x3): Likewise.
6148 (vst1_p64_x3): Likewise.
6149 (vst1q_s8_x3): Likewise.
6150 (vst1q_p8_x3): Likewise.
6151 (vst1q_s16_x3): Likewise.
6152 (vst1q_p16_x3): Likewise.
6153 (vst1q_s32_x3): Likewise.
6154 (vst1q_s64_x3): Likewise.
6155 (vst1q_u8_x3): Likewise.
6156 (vst1q_u16_x3): Likewise.
6157 (vst1q_u32_x3): Likewise.
6158 (vst1q_u64_x3): Likewise.
6159 (vst1q_f16_x3): Likewise.
6160 (vst1q_f32_x3): Likewise.
6161 (vst1q_f64_x3): Likewise.
6162 (vst1q_p64_x3): Likewise.
6164 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6166 * config/msp430/msp430.c (msp430_output_labelref): Prepend
6167 user_label_prefix to name.
6169 * tree-core.h: Update comment about the format of NAME string
6170 passed to handler in attribute_spec.
6172 * config/msp430/msp430.md: Remove erroneous subreg expression from
6173 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
6174 zero_extend{q,h}isi2.
6176 2018-05-30 Borislav Petkov <bp@suse.de>
6178 * doc/extend.texi: Document some architecture specific
6179 constraints and sort entries.
6181 2018-05-30 Martin Sebor <msebor@redhat.com>
6184 * builtins.c (expand_builtin_stpcpy_1): New function.
6185 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
6186 only if the former succeeds.
6188 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
6190 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
6193 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6195 * doc/invoke.texi (-flinker-output): Document
6197 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6199 * passes.c (ipa_write_summaries): Only modify statements if body
6201 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
6202 incrementally linking.
6203 (ipa_passes): Likewise.
6204 * lto-cgraph.c (lto_output_node): When incrementally linking do not
6205 pass down resolution info.
6206 * common.opt (flag_incremental_link): Update info.
6207 * gcc.c (plugin specs): Turn flinker-output=* to
6208 -plugin-opt=-linker-output-known
6209 * toplev.c (compile_file): Also cut compilation when doing incremental
6211 * flag-types. (enum lto_partition_model): Add
6212 LTO_LINKER_OUTPUT_NOLTOREL.
6213 (invoke.texi): Add -flinker-output docs.
6214 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
6215 link same way as WPA; do not stream in dead initializers.
6217 * dwarf2out.c (dwarf2out_die_ref_for_decl,
6218 darf2out_register_external_decl): Support incremental link.
6220 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6222 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
6224 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6226 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
6227 it down to simple_object_copy_lto_debug_sections.
6228 (run_gcc): Determine incremental LTO link time and configure
6229 lto1 into non-wpa mode, disable renaming of debug sections.
6231 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
6233 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
6234 descriptions of various incorrectly documented functions.
6236 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
6239 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6240 address check not strict.
6242 2018-05-30 Richard Biener <rguenther@suse.de>
6244 PR tree-optimization/85964
6245 * tracer.c (better_p): Drop initialized count check, we only
6246 call the function with initialized counts now.
6247 (find_best_successor): Do find a best edge if one
6248 has uninitialized count.
6249 (find_best_predecessor): Likewise. Do BB frequency check only
6250 if count is initialized.
6252 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
6254 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
6255 (aarch64_ldrstr_offset_compare): New.
6256 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
6257 load/store orderings.
6258 (aarch64_gen_adjusted_ldpstp): Likewise.
6260 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
6262 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
6263 Check for subset of GENERAL_REGS and FP_REGS.
6264 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
6267 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
6269 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
6270 and wi::to_poly_offset. Add the current offset and then check
6271 whether the sum fits, rather than using an unchecked addition of
6272 a checked term. Check for a shwi rather than a uhwi.
6273 * expr.c (get_bit_range): Use tree_to_poly_uint64.
6274 (store_constructor): Use poly_int_tree_p.
6275 (expand_expr_real_1): Likewise.
6276 * function.c (assign_temp): Likewise.
6277 * fold-const.c (const_binop): Use poly_int_tree_p and
6279 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
6281 * ipa-icf-gimple.c (func_checker::compare_operand): Use
6282 to_poly_offset for MEM offsets.
6283 * ipa-icf.c (sem_variable::equals): Likewise.
6284 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
6285 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
6286 wi::to_poly_offset for BIT_FIELD_REF offsets.
6287 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
6289 * var-tracking.c (emit_note_insn_var_location): Use
6290 tree_to_poly_uint64.
6292 2018-05-29 Jim Wilson <jimw@sifive.com>
6294 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
6296 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
6299 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
6300 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
6302 (sse4_1_round<mode>2): Use nonimmediate_operand
6303 for operand 1 predicate.
6305 2018-05-29 Martin Sebor <msebor@redhat.com>
6306 Richard Biener <rguenther@suse.de>
6309 * calls.c (get_size_range): Call determine_value_range instead
6310 of get_value_range..
6311 * tree-vrp.h (determine_value_range): Declared new function.
6312 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
6314 2018-05-29 Richard Biener <rguenther@suse.de>
6316 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
6317 sure to use non-pattern stmts for get_earlier_stmt arguments.
6318 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
6319 called on pattern stmts.
6320 (get_later_stmt): Likewise.
6322 2018-05-29 Martin Liska <mliska@suse.cz>
6324 PR gcov-profile/85759
6325 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
6328 2018-05-29 Jakub Jelinek <jakub@redhat.com>
6330 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
6332 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
6336 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
6337 VEC_PACK_FLOAT_EXPR): New tree codes.
6338 * tree-pretty-print.c (op_code_prio): Handle
6339 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
6340 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
6341 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
6342 * tree-inline.c (estimate_operator_cost): Likewise.
6343 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
6344 * fold-const.c (const_binop): Likewise.
6345 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
6346 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
6347 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6348 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
6349 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
6350 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
6351 * expr.c (expand_expr_real_2): Likewise.
6352 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
6353 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
6354 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
6356 * optabs.c (expand_widen_pattern_expr): For
6357 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
6358 sign from result type rather than operand's type.
6359 (expand_binop_directly): For vec_packu_float_optab and
6360 vec_packs_float_optab allow result type to be different from operand's
6362 * optabs-tree.c (optab_for_tree_code): Handle
6363 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
6364 VEC_PACK_FLOAT_EXPR. Formatting fixes.
6365 * tree-vect-generic.c (expand_vector_operations_1): Handle
6366 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
6367 VEC_PACK_FLOAT_EXPR.
6368 * tree-vect-stmts.c (supportable_widening_operation): Handle
6370 (supportable_narrowing_operation): Handle FLOAT_EXPR.
6371 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
6372 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
6373 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
6374 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
6376 (vec_pack<floatprefix>_float_<mode>): New expander.
6377 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
6379 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
6380 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
6381 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
6382 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
6383 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
6385 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
6386 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
6387 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
6388 VEC_PACK_FLOAT_EXPR): Document.
6390 2018-05-29 Richard Biener <rguenther@suse.de>
6392 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
6394 (stmt_vec_info_vec): Make pointer.
6395 (init_stmt_vec_info_vec): Remove.
6396 (free_stmt_vec_info_vec): Likewise.
6397 (set_stmt_vec_info_vec): New function.
6398 (free_stmt_vec_infos): Likewise.
6399 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
6400 (set_vinfo_for_stmt): Likewise.
6401 (get_earlier_stmt): Likewise.
6402 (get_later_stmt): Likewise.
6403 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
6404 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
6405 (vec_info::~vec_info): Free stmt_vec_infos.
6406 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
6407 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
6408 (pass_slp_vectorize::execute): Likewise.
6409 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
6410 (free_stmt_vec_info_vec): Likewise.
6411 (set_stmt_vec_info_vec): New function.
6412 (free_stmt_vec_infos): Likewise.
6413 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
6414 the global stmt_vec_info_vec.
6415 * tree-parloops.c (gather_scalar_reductions): Use
6416 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
6419 2018-05-29 Richard Biener <rguenther@suse.de>
6421 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
6423 2018-05-29 Martin Liska <mliska@suse.cz>
6424 David Malcolm <dmalcolm@redhat.com>
6426 * vec.c (test_reverse): New.
6427 (vec_c_tests): Add new test.
6428 * vec.h (vl_ptr>::reverse): New function.
6430 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
6432 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
6434 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
6437 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
6439 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
6441 2018-05-28 Richard Biener <rguenther@suse.de>
6443 PR tree-optimization/85933
6444 * tree-vect-data-refs.c (vect_record_base_alignments): Only
6445 look at stmts marked as vectorizable.
6447 2018-05-28 Richard Biener <rguenther@suse.de>
6449 PR tree-optimization/85934
6450 * tree-vect-generic.c (expand_vector_operations_1): Hoist
6451 vector boolean check before scalar optimization.
6453 2018-05-28 Jakub Jelinek <jakub@redhat.com>
6455 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
6458 2018-05-28 Mark Wielaard <mark@klomp.org>
6460 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
6461 if it is an expression containing a minus sign.
6463 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
6465 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
6467 2018-05-27 Paul Koning <ni1d@arrl.net>
6469 * config/pdp11/pdp11.md (truncsihi2): Remove.
6471 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
6472 Chung-Ju Wu <jasonwucj@gmail.com>
6474 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
6476 (unaligned_store_dw): Ditto.
6477 * config/nds32/nds32-memory-manipulation.c
6478 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
6479 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
6480 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
6481 (emit_setmem_word_loop): Rename to ...
6482 (emit_setmem_doubleword_loop): ... this.
6483 (nds32_gen_dup_4_byte_to_word_value): New function.
6484 (nds32_gen_dup_8_byte_to_double_word_value): New function.
6485 (nds32_expand_setmem_loop): Refine implementation.
6486 (nds32_expand_setmem_loop_v3m): Ditto.
6487 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
6490 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
6492 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
6494 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
6496 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
6497 (nds32_init_machine_status): Initialize machine->attr_naked_p and
6498 machine->attr_no_prologue_p.
6499 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
6500 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
6501 (nds32_expand_epilogue): Consider attr_naked_p.
6502 (nds32_expand_epilogue_v3pop): Likewise.
6503 (nds32_can_use_return_insn): Likewise.
6504 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
6505 attr_no_prologue_p fields.
6506 * config/nds32/nds32.opt (mret-in-naked-func): New option.
6508 2018-05-27 Jakub Jelinek <jakub@redhat.com>
6511 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
6513 * config/i386/sse.md
6514 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
6516 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
6518 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
6520 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
6522 (*<floatsuffix>floatv2div2sf2): Rename to ...
6523 (*float<floatunssuffix>v2div2sf2): ... this.
6524 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
6525 (float<floatunssuffix>v2div2sf2_mask): ... this.
6526 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
6527 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
6528 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
6530 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
6532 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
6534 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
6536 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
6538 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
6540 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
6541 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
6542 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
6543 gen_ufix_truncv8dfv8si2.
6544 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
6545 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
6546 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
6547 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
6548 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
6549 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
6550 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
6551 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
6553 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
6557 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
6559 2018-05-25 Jim Wilson <jimw@sifive.com>
6561 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
6562 * config/riscv/riscv.c (struct machine_function): Add
6563 interrupt_handler_p and attribute_checked_p fields.
6564 (riscv_attribute_table): Add interrupt.
6565 (riscv_interrupt_type_p): New.
6566 (riscv_save_reg_p): Save extra regs for interrupt handler.
6567 (riscv_use_save_libcall): Return false for interrupt handler.
6568 (riscv_first_stack_step): Add forward declaration.
6569 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
6570 for interrupt handler with large frame. Use it for saved reg list.
6571 (riscv_expand_prologue): Move flag_stack_usage_info support to
6572 eliminate duplication.
6573 (riscv_expand_epilogue): Generate mret for interrupt handler.
6574 (riscv_epilogue_uses): New.
6575 (riscv_can_use_return_insn): Return false for interrupt handler.
6576 (riscv_function_ok_for_sibcall): Likewise.
6577 (riscv_set_current_function): Add interrupt handler support.
6578 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
6579 * config/riscv/riscv.md (UNSPECV_MRET): New.
6581 (riscv_frflags, riscv_fsflags): Use tab after opcode.
6583 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
6585 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
6587 PR tree-optimization/85712
6588 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
6589 this candidate has already been replaced in-situ by a copy.
6591 2018-05-25 Jason Merrill <jason@redhat.com>
6593 PR c++/80485 - inline function non-zero address.
6594 * symtab.c (nonzero_address): Check DECL_COMDAT.
6596 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
6599 * config/alpha/alpha.md (ashlsi3): New insn pattern.
6600 (*ashlsi_se): Rename from *ashldi_se. Define as sign
6601 extension of SImode operation. Use const123_operand predicate.
6602 (*saddsi_1): Remove.
6603 (*saddl_se_1): Ditto.
6605 (*ssubl_se_1): Ditto.
6606 * config/alpha/predicates.md (const123_operand): New predicate.
6607 * config/alpha/constraints.md (P): Use IN_RANGE.
6609 2018-05-25 Richard Biener <rguenther@suse.de>
6611 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
6613 (ref_maybe_used_by_stmt_p): Likewise.
6614 (stmt_may_clobber_ref_p): Likewise.
6615 (stmt_may_clobber_ref_p_1): Likewise.
6616 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
6618 (ref_maybe_used_by_stmt_p): Likewise.
6619 (stmt_may_clobber_ref_p): Likewise.
6620 (stmt_may_clobber_ref_p_1): Likewise.
6621 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
6622 the alias oracle to disambiguate DRs with stmts DR analysis
6624 (vect_analyze_data_refs): Do not give up on not analyzable
6625 DRs for BB vectorization. Remove code truncating the dataref
6628 2018-05-25 Jakub Jelinek <jakub@redhat.com>
6631 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
6632 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
6633 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
6635 2018-05-25 Richard Biener <rguenther@suse.de>
6637 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
6638 function, combining stmt data ref gathering and fatal analysis
6640 (vect_analyze_data_refs): Remove now redudnant code and simplify.
6641 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
6642 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
6643 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
6644 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
6646 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6648 PR tree-optimization/85720
6649 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
6650 SCC if all partitions are builtins.
6651 (version_loop_by_alias_check): New parameter. Generate cancelable
6652 runtime alias check if all partitions are builtins.
6653 (distribute_loop): Update call to above function.
6655 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6657 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
6658 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
6659 (parm_default_def_partition_arg): Ditto.
6660 (set_parm_default_def_partition): Ditto.
6661 (get_parm_default_def_partitions): Ditto and make it static.
6662 (get_undefined_value_partitions): Ditto and make it static.
6663 (remove_ssa_form): Refactor call to init_var_map here.
6664 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
6665 computation for loop region.
6666 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
6667 (register_default_def): Delete.
6668 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
6669 (parm_default_def_partition_arg): Ditto.
6670 (set_parm_default_def_partition): Ditto.
6671 (get_parm_default_def_partitions): Ditto and make it static.
6672 (get_undefined_value_partitions): Ditto and make it static.
6673 (coalesce_with_default, coalesce_with_default): Update comment.
6674 (create_coalesce_list_for_region): New func factored out from
6675 create_outofssa_var_map.
6676 (populate_coalesce_list_for_outofssa): New func factored out from
6677 create_outofssa_var_map and coalesce_ssa_name.
6678 (create_outofssa_var_map): Delete.
6679 (coalesce_ssa_name): Refactor to support live range computation.
6680 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
6681 (get_parm_default_def_partitions): Delete.
6682 (get_undefined_value_partitions): Ditto.
6683 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
6684 computation for loop region.
6685 (new_tree_live_info, loe_visit_block): Ditto.
6686 (live_worklist, set_var_live_on_entry): Ditto.
6687 (calculate_live_on_exit, verify_live_on_entry): Ditto.
6688 * tree-ssa-live.h (struct _var_map): New fields.
6689 (init_var_map): Change decl.
6690 (region_contains_p): New.
6692 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6694 * tree-ssa-live.h (live_merge_and_clear): Delete.
6696 2018-05-25 Richard Biener <rguenther@suse.de>
6699 * tree-dump.c (dequeue_and_dump): Remove access to removed
6700 operand 2 of a SWITCH_EXPR.
6702 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6704 * doc/sourcebuild.texi (vect_double_cond_arith): Include
6705 multiplication and division.
6706 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
6707 (cond_udiv@var{m}, cond_umod@var{m}): Document.
6708 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
6709 (cond_udiv_optab, cond_umod_optab): New optabs.
6710 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
6711 (IFN_COND_RDIV): New internal functions.
6712 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
6713 TRUNC_MOD_EXPR and RDIV_EXPR.
6714 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
6715 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
6717 (SVE_INT_BINARY): Include mult.
6718 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
6719 (optab, sve_int_op): Handle mult.
6720 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
6722 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
6723 for SVE_INT_BINARY_SD.
6725 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6727 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
6728 (optab, sve_int_op): Handle div and udiv.
6729 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
6730 for SVE_INT_BINARY_SD.
6731 (*<optab><mode>3): New insn for the same.
6733 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6735 * tree-vect-patterns.c: Include predict.h.
6736 (vect_recog_divmod_pattern): Restrict check for division support
6737 to when optimizing for size.
6739 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6741 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
6742 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
6743 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
6744 (gimple_match_op::set_op): Likewise.
6745 (gimple_resimplify4): Declare.
6746 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
6747 (expr::gen_transform): Likewise.
6748 (decision_tree::gen): Generate a simplification routine for 4 operands.
6749 * gimple-match-head.c (gimple_simplify): Add an overload for
6750 4 operands. In the top-level function, handle up to 4 call
6751 arguments and call gimple_resimplify4.
6752 (gimple_resimplify4): New function.
6753 (build_call_internal): Pass a fourth operand.
6754 (maybe_push_to_seq): Likewise.
6755 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
6756 Fold VEC_COND_EXPRs of an operation and a default value into
6757 an IFN_COND_* function if possible.
6758 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
6760 (SVE_COND_FP_BINARY): Include them.
6761 (optab, sve_fp_op): Handle them.
6762 (SVE_INT_BINARY_REV): New code iterator.
6763 (SVE_COND_FP_BINARY_REV): New int iterator.
6764 (commutative): New int attribute.
6765 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
6767 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
6769 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
6770 (*cond_<optab><mode>): New patterns for reversed operands.
6772 2018-05-25 Richard Biener <rguenther@suse.de>
6774 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
6775 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
6776 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
6777 (STMT_VINFO_GROUPED_ACCESS): Adjust.
6778 * tree-vect-data-refs.c (everywhere): Adjust users.
6779 * tree-vect-loop.c (everywhere): Likewise.
6780 * tree-vect-slp.c (everywhere): Likewise.
6781 * tree-vect-stmts.c (everywhere): Likewise.
6782 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
6784 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6786 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
6788 (gcc_cv_as_section_exclude): ... this.
6789 Try Solaris as #exclude syntax.
6790 * configure: Regenerate.
6791 * config.in: Regenerate.
6792 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
6794 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
6795 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
6797 * varasm.c (default_elf_asm_named_section): Don't check if
6798 HAVE_GAS_SECTION_EXCLUDE is defined.
6800 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6802 * doc/md.texi: Update the documentation of the cond_* optabs
6803 to mention the new final operand. Fix GET_MODE_NUNITS call.
6804 Describe the scalar case too.
6805 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
6806 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
6808 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
6809 (get_conditional_internal_fn): Update comment.
6810 * tree-vect-loop.c (vectorizable_reduction): Pass the original
6811 accumulator value as a final argument to conditional functions.
6812 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
6813 a define_expand and add an "else" operand. Assert for now that
6814 the else operand is equal to operand 2. Use SVE_INT_BINARY and
6815 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
6816 (*cond_<optab><mode>): New patterns.
6817 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
6818 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6819 (UNSPEC_COND_EOR): Delete.
6820 (optab): Remove associated mappings.
6821 (SVE_INT_BINARY): New code iterator.
6822 (sve_int_op): Remove int attribute and add "minus" to the code
6824 (SVE_COND_INT_OP): Delete.
6825 (SVE_COND_FP_OP): Rename to...
6826 (SVE_COND_FP_BINARY): ...this.
6828 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6830 * optabs.c (can_reuse_operands_p): New function.
6831 (maybe_legitimize_operands): Try to reuse the results for
6834 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
6836 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
6837 Add {q} suffix to insn mnemonic.
6839 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6841 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
6842 (msp430_warn_func_return): New.
6844 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
6846 * fold-const.c (tree_nonzero_bits): New function.
6847 * fold-const.h (tree_nonzero_bits): Likewise.
6848 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
6849 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
6851 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
6855 * varasm.c (assemble_alias): Check ifunc_resolver only on
6858 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
6861 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
6862 when memory input operand is handled.
6864 2018-05-24 Luis Machado <luis.machado@linaro.org>
6866 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
6868 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
6870 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
6872 * match.pd: Delay FMA folds until after vectorization.
6874 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
6877 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6878 address check not strict.
6880 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
6882 * gimple-match.h (gimple_match_op): New class.
6883 (mprts_hook): Replace parameters with a gimple_match_op *.
6884 (maybe_build_generic_op): Likewise.
6885 (gimple_simplified_result_is_gimple_val): Replace parameters with
6886 a const gimple_match_op *.
6887 (gimple_simplify): Replace code_helper * and tree * parameters with
6888 a gimple_match_op * parameter.
6889 (gimple_resimplify1): Replace code_helper *, tree and tree *
6890 parameters with a gimple_match_op * parameter.
6891 (gimple_resimplify2): Likewise.
6892 (gimple_resimplify3): Likewise.
6893 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
6894 parameters with a gimple_match_op * parameter.
6895 * gimple-match-head.c (gimple_simplify): Change prototypes of
6896 auto-generated functions to take a gimple_match_op * instead of
6897 separate code_helper * and tree * parameters. Make the same
6898 change in the top-level overload and update calls to the
6899 gimple_resimplify routines. Update calls to the auto-generated
6900 functions and to maybe_push_res_to_seq in the publicly-facing
6901 operation-specific gimple_simplify overloads.
6902 (gimple_match_op::MAX_NUM_OPS): Define.
6903 (gimple_resimplify1): Replace rcode and ops with a single res_op
6904 parameter. Update call to gimple_simplify.
6905 (gimple_resimplify2): Likewise.
6906 (gimple_resimplify3): Likewise.
6907 (mprts_hook): Replace parameters with a gimple_match_op *.
6908 (maybe_build_generic_op): Likewise.
6909 (build_call_internal): Replace type, nargs and ops with
6910 a gimple_match_op *.
6911 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
6912 with a single gimple_match_op *. Update calls to mprts_hook,
6913 build_call_internal and gimple_simplified_result_is_gimple_val.
6914 Factor out code that is common to the tree_code and combined_fn cases.
6915 * genmatch.c (expr::gen_transform): Replace tem_code and
6916 tem_ops with a gimple_match_op called tem_op. Update calls
6917 to the gimple_resimplify functions and maybe_push_res_to_seq.
6918 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
6919 res_ops. Update call to the gimple_resimplify functions.
6920 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
6921 (decision_tree::gen): Make the functions take a gimple_match_op *
6922 called res_op instead of separate res_code and res_ops parameters.
6923 Update call accordingly.
6924 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
6925 and ops with a single res_op parameter. Update calls to
6926 maybe_build_generic_op and maybe_push_res_to_seq.
6927 (fold_stmt_1): Update calls to gimple_simplify and
6928 replace_stmt_with_simplification.
6929 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
6930 and gimple_simplified_result_is_gimple_val.
6931 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
6933 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
6934 with a gimple_match_op *.
6935 (vn_nary_build_or_lookup): Likewise. Update call to
6936 vn_nary_build_or_lookup_1.
6937 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
6938 gimple_match_op *. Update calls to the gimple_resimplify routines
6939 and to gimple_simplified_result_is_gimple_val.
6940 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
6941 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
6942 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
6943 (visit_nary_op): Likewise.
6944 (visit_reference_op_load): Likewise.
6946 2018-05-23 Luis Machado <luis.machado@linaro.org>
6948 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
6949 modifier for printing the step amount.
6951 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
6954 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
6957 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
6959 * doc/sourcebuild.texi (Endianness): New subsubsection.
6961 2018-05-23 Luis Machado <luis.machado@linaro.org>
6963 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6964 <prefetch_dynamic_strides>: New const bool field.
6965 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6966 prefetch_dynamic_strides.
6967 (exynosm1_prefetch_tune): Likewise.
6968 (thunderxt88_prefetch_tune): Likewise.
6969 (thunderx_prefetch_tune): Likewise.
6970 (thunderx2t99_prefetch_tune): Likewise.
6971 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
6973 (aarch64_override_options_internal): Update to set
6974 PARAM_PREFETCH_DYNAMIC_STRIDES.
6975 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6976 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6977 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6978 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6979 prefetch-dynamic-strides setting.
6981 2018-05-23 Luis Machado <luis.machado@linaro.org>
6983 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6984 <minimum_stride>: New const int field.
6985 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6986 minimum_stride field defaulting to -1.
6987 (exynosm1_prefetch_tune): Likewise.
6988 (thunderxt88_prefetch_tune): Likewise.
6989 (thunderx_prefetch_tune): Likewise.
6990 (thunderx2t99_prefetch_tune): Likewise.
6991 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
6992 <default_opt_level>: Set to 3.
6993 (aarch64_override_options_internal): Update to set
6994 PARAM_PREFETCH_MINIMUM_STRIDE.
6995 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6996 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6997 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6998 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6999 stride is constant and is below the minimum stride threshold.
7001 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7003 * config/arm/arm-cpus.in (mode26): Delete.
7004 (armv4): Delete mode26 reference.
7005 * config/arm/arm.c (arm_configure_build_target): Delete use of
7008 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
7010 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
7012 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
7013 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
7015 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
7016 (floatunsdidf2): Ditto.
7018 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
7020 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
7021 (fixuns_trunc<mode>si2_avx512f): Ditto.
7022 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
7023 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
7024 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
7026 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
7028 PR rtl-optimization/79985
7029 * df-scan.c (df_insn_refs_collect): Remove special case for
7030 global registers and asm statements.
7032 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
7034 * extend.texi (Global Register Variables): Rewrite the bullet list.
7035 Note that the register is available for allocation. Note that access
7036 via inline asm must use constraints. Add note about async-signal
7037 handlers. Remove paragraph about automagic register selection.
7039 2018-05-23 Richard Biener <rguenther@suse.de>
7041 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
7042 of fixed offset from memset VN.
7044 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
7046 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
7048 (alloc_cand_and_find_basis): Initialize first_interp field.
7049 (slsr_process_mul): Modify first_interp field.
7050 (slsr_process_add): Likewise.
7051 (slsr_process_cast): Modify first_interp field for each new
7053 (slsr_process_copy): Likewise.
7054 (dump_candidate): Dump first_interp field.
7055 (replace_mult_candidate): Process all interpretations, not just
7057 (replace_rhs_if_not_dup): Likewise.
7058 (replace_one_candidate): Likewise.
7060 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
7062 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
7064 (aarch64_needs_frame_chain): New function.
7065 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
7067 2018-05-23 Sudakshina Das <sudi.das@arm.com>
7070 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
7071 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
7072 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
7073 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
7074 as true for strict-align.
7075 (aarch64_can_inline_p): Perform checks even when callee has no
7076 attributes to check for strict alignment.
7077 * doc/extend.texi (AArch64 Function Attributes): Document
7079 * doc/invoke.texi: (AArch64 Options): Likewise.
7081 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
7083 PR tree-optimization/85853
7084 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
7085 the handling of the root of the node to...
7086 (vect_slp_analyze_node_operations_1): ...this new function,
7087 and run the whole thing with the child nodes' def types
7088 set according to their SLP node's def type.
7090 2018-05-23 Richard Biener <rguenther@suse.de>
7093 * tree-data-ref.c (create_runtime_alias_checks): Defer
7094 and ignore overflow warnings.
7096 2018-05-23 Yury Gribov <tetra2005@gmail.com>
7098 PR tree-optimization/85822
7099 * tree-vrp.c (is_masked_range_test): Fix handling of negative
7102 2018-05-23 Richard Biener <rguenther@suse.de>
7104 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
7105 memset constants via native_interpret_expr.
7107 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
7110 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
7112 (cgraph_node::create_alias): Likewise.
7113 (cgraph_node::get_availability): Check ifunc_resolver instead
7114 of looking up ifunc attribute.
7115 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
7116 * varasm.c (do_assemble_alias): Likewise.
7117 (assemble_alias): Likewise.
7118 (default_binds_local_p_3): Likewise.
7119 * cgraph.h (cgraph_node): Add ifunc_resolver.
7120 (cgraph_node::only_called_directly_or_aliased_p): Return false
7122 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
7124 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
7125 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
7126 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
7127 instead of looking up ifunc attribute.
7129 2018-05-22 Luis Machado <luis.machado@linaro.org>
7131 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
7133 2018-05-22 Martin Sebor <msebor@redhat.com>
7136 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
7137 only when expasion succeeds.
7138 (expand_builtin_strcmp): Same.
7139 (expand_builtin_strncmp): Same.
7141 2018-05-22 Martin Sebor <msebor@redhat.com>
7143 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
7145 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
7146 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7148 * config/aarch64/aarch64-ldpstp.md: Replace uses of
7149 aarch64_mem_pair_operand with memory_operand and delete operand swapping
7151 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
7152 Add check for legitimate_address.
7153 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
7154 (aarch64_swap_ldrstr_operands): New.
7155 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
7158 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
7159 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7161 * config/aarch64/aarch64.md: New patterns to generate stp
7163 (store_pair_sw, store_pair_dw): New patterns to generate stp for
7164 single words and double words.
7165 (load_pair_sw, load_pair_dw): Likewise.
7166 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
7168 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
7170 * config/aarch64/aarch64-ldpstp.md: Modify peephole
7171 for different mode ldpstp and add peephole for merged zero stores.
7173 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
7175 (aarch64_gen_store_pair): Rename calls to match new patterns.
7176 (aarch64_gen_load_pair): Rename calls to match new patterns.
7177 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
7178 (load_pair<DREG:mode><DREG2:mode>): ... This.
7179 (store_pair<mode>): Rename to...
7180 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
7181 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
7183 (V_INT_EQUIV): Handle SImode.
7184 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
7187 2018-05-22 Martin Sebor <msebor@redhat.com>
7190 * calls.c (maybe_warn_nonstring_arg): Use string length to set
7191 or ajust the presumed bound on an operation to avoid unnecessary
7194 2018-05-22 Martin Sebor <msebor@redhat.com>
7196 PR tree-optimization/85826
7197 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
7198 assuming that a DECL necesarily has a constant size.
7200 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
7203 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
7205 2018-05-22 Richard Biener <rguenther@suse.de>
7207 PR tree-optimization/85834
7208 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
7209 non-constant and non-zero memset arguments.
7211 2018-05-22 Martin Liska <mliska@suse.cz>
7214 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
7216 2018-05-22 Richard Biener <rguenther@suse.de>
7218 PR tree-optimization/85863
7219 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
7220 comparisons when vectype is specified.
7221 (vectorizable_condition): Do not specify vectype for
7222 vect_is_simple_cond when SLP vectorizing.
7224 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
7227 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
7228 define __ibm128 as long double.
7229 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
7230 as a distinct type when IEEE 128-bit support is enabled.
7231 (init_float128_ieee): Fix up conversions between IFmode and IEEE
7232 128-bit types to use the correct functions.
7233 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
7234 convert between 128-bit floating point types that have different
7235 modes but the same representation, instead of using gen_lowpart to
7237 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
7239 (IFKF_reg): New attributes to give the register constraints for
7241 (extend<mode>tf2_internal): New insns to mark an explicit
7242 conversion between 128-bit floating point types that have a
7243 different mode but share the same representation.
7245 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
7247 PR tree-optimization/85814
7248 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
7249 a null return from get_strinfo when unsharing the next
7250 strinfo in the chain.
7252 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7255 * varasm.c (weak_finish): Clean up weak_decls.
7257 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7259 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
7260 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
7261 UNSPEC_UADALP values.
7262 * config/aarch64/iterators.md (ABAL): New int iterator.
7265 (sur): Add mappings for the above.
7266 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
7268 (aarch64_<sur>abal<mode>_4): Likewise.
7269 (aarch64_<sur>adalp<mode>_3): Likewise.
7270 (<sur>sadv16qi): New define_expand.
7272 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
7274 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
7275 (*movdf_internal): Ditto.
7276 (*rcpsf2_sse): Ditto.
7277 (*rsqrtsf2_sse): Ditto.
7278 (*sqrt<mode>2_sse): Ditto.
7280 2018-05-21 Tamar Christina <tamar.christina@arm.com>
7282 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
7284 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
7285 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
7286 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
7287 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
7289 * config/aarch64/arm_neon.h: Likewise.
7290 * config/aarch64/iterators.md (VQ_I): New.
7292 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
7294 * config.gcc: Add arc/t-multilib-linux to tmake_file for
7296 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
7299 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
7301 * config/nds32/constraints.md (S): New constraint.
7302 * config/nds32/nds32.md (call_internal): Use constraint S.
7303 (call_value_internal): Likewise.
7304 (sibcall_internal): Likewise.
7305 (sibcall_value_internal): Likewise.
7307 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
7308 Chung-Ju Wu <jasonwucj@gmail.com>
7310 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
7313 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
7314 Chung-Ju Wu <jasonwucj@gmail.com>
7316 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
7317 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
7318 (nds32_rtx_costs_impl): Simplify.
7319 (nds32_address_cost_impl): Simplify.
7320 (nds32_init_rtx_costs): New function.
7321 (nds32_rtx_costs_speed_prefer): Likewise.
7322 (nds32_rtx_costs_size_prefer): Likewise.
7323 (nds32_address_cost_speed_prefer): Likewise.
7324 (nds32_address_cost_speed_fwprop): Likewise.
7325 (nds32_address_cost_size_prefer): Likewise.
7326 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
7327 * config/nds32/nds32.c (nds32_option_override): Use
7328 nds32_init_rtx_costs function.
7330 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
7332 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
7333 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
7334 (TARGET_PIPELINE_N8): Likewise.
7335 (TARGET_PIPELINE_N10): Likewise.
7336 (TARGET_PIPELINE_N13): Likewise.
7337 (TARGET_PIPELINE_GRAYWOLF): Likewise.
7339 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
7341 * config/nds32/nds32-fpu.md: Update copyright year.
7343 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7345 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
7347 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7349 * config/nds32/nds32.c
7350 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
7351 * config/nds32/nds32.opt (minline-asm-r15): New option.
7353 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7355 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
7357 * config/nds32/nds32.md (abssi2): New pattern.
7359 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
7361 * config/i386/i386.md (rex64namesuffix): New mode attribute.
7362 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
7363 Merge insn pattern from sse_cvtsi2ss<round_name> and
7364 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
7365 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
7366 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
7367 using SWI48 mode iterator.
7368 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
7369 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
7370 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
7371 pattern from sse_cvttss2si<round_saeonly_name>
7372 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
7373 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
7374 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
7375 using SWI48 mode iterator.
7376 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
7377 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
7378 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
7379 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
7380 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
7381 using SWI48 mode iterator.
7382 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
7383 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
7384 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
7385 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
7386 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
7387 SWI48 mode iterator.
7388 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
7389 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
7390 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
7391 pattern from sse_cvttsd2si<round_saeonly_name>
7392 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
7394 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7396 * config/nds32/nds32-md-auxiliary.c
7397 (nds32_valid_smw_lwm_base_p): Refine.
7398 (nds32_output_smw_single_word): Refine.
7399 (nds32_output_smw_double_word): New.
7400 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
7402 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7404 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
7405 (nds32_output_stack_pop): Refine.
7406 (nds32_expand_unaligned_load): Refine.
7407 (nds32_expand_unaligned_store): Refine.
7409 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7410 Chung-Ju Wu <jasonwucj@gmail.com>
7412 * config/nds32/constants.md: Add TP_REGNUM constant.
7413 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
7414 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
7416 * config/nds32/nds32-doubleword.md: Consider flag_pic.
7417 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
7418 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
7419 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
7420 and PIC code generation.
7421 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
7423 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
7425 * config/nds32/nds32.md: Support TLS and PIC.
7426 * config/nds32/nds32.c: Support TLS and PIC.
7427 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
7428 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
7431 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7433 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
7434 mode with E_ prefix.
7436 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7437 Chung-Ju Wu <jasonwucj@gmail.com>
7439 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
7440 * config/nds32/nds32-md-auxiliary.c
7441 (symbolic_reference_mentioned_p): New.
7442 (nds32_legitimize_ict_address): New.
7443 (nds32_expand_ict_move): New.
7444 (nds32_indirect_call_referenced_p): New.
7445 (nds32_symbol_binds_local_p): Delete.
7446 (nds32_long_call_p): Modify.
7447 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
7448 * config/nds32/nds32-protos.h
7449 (symbolic_reference_mentioned_p): Declare.
7450 (nds32_legitimize_ict_address): Declare.
7451 (nds32_expand_ict_move): Declare.
7452 (nds32_indirect_call_referenced_p): Declare.
7453 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
7454 (nds32_relax_group): Use nds32_ict_const_p as condition.
7455 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
7456 (nds32_asm_file_start): Output ict_model directive in asm code.
7457 (nds32_legitimate_address_p): Consider indirect call.
7458 (nds32_print_operand): Consider indirect call.
7459 (nds32_print_operand_address): Consider indirect call.
7460 (nds32_insert_attributes): Handle "indirect_call" attribute.
7461 (TARGET_LEGITIMATE_ADDRESS_P): Define.
7462 (TARGET_LEGITIMATE_CONSTANT_P): Define.
7463 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
7464 (TARGET_DELEGITIMIZE_ADDRESS): Define.
7465 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
7466 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
7467 (TARGET_ICT_MODEL_SMALL): Define.
7468 (TARGET_ICT_MODEL_LARGE): Define.
7469 * config/nds32/nds32.md (movsi): Consider ict model.
7470 (call, call_value): Consider ict model.
7471 (sibcall, sibcall_value): Consider ict model.
7472 * config/nds32/nds32.opt (mict-model): New option.
7473 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
7476 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
7477 Monk Chiang <sh.chiang04@gmail.com>
7478 Jim Wilson <jimw@sifive.com>
7480 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
7481 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
7482 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
7483 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
7484 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
7485 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
7486 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
7487 compute save_libcall_adjustment properly.
7488 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
7489 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
7490 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
7491 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
7492 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
7493 (ABI_SPEC): Handle mabi=ilp32e.
7494 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
7495 (RVE): Add RVE mask.
7496 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
7497 <-march>: Add rv32e as an example.
7499 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
7502 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
7503 (intra_create_variable_infos): Handle C++ constructors.
7505 2018-05-18 Martin Liska <mliska@suse.cz>
7507 * passes.def: Remove a redundant pass.
7509 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
7512 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
7514 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7516 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
7518 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
7520 (armv2, armv2a, armv3, armv3m): Delete architectures.
7521 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
7522 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
7523 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
7525 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
7526 (*mulsidi3adddi): Likewise.
7527 (mulsidi3): Likewise.
7528 (*mulsidi3_nov6): Likewise.
7529 (umulsidi3): Likewise.
7530 (umulsidi3_nov6): Likewise.
7531 (umaddsidi4): Likewise.
7532 (*umulsidi3adddi): Likewise.
7533 (smulsi3_highpart): Likewise.
7534 (*smulsi3_highpart_nov6): Likewise.
7535 (umulsi3_highpart): Likewise.
7536 (*umulsi3_highpart_nov6): Likewise.
7537 * config/arm/arm.h (arm_arch3m): Delete.
7538 * config/arm/arm.c (arm_arch3m): Delete.
7539 (arm_option_override_internal): Update armv3-related comment.
7540 (arm_configure_build_target): Delete use of isa_bit_mode32.
7541 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
7542 (arm_rtx_costs_internal): Delete check of arm_arch3m.
7543 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
7545 (mulusa3): Likewise.
7546 * config/arm/arm-protos.h (arm_arch3m): Delete.
7547 * config/arm/arm-tables.opt: Regenerate.
7548 * config/arm/arm-tune.md: Likewise.
7549 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
7550 deleted architectures.
7552 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7554 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
7555 (armv5t, armv5te): New features.
7556 (ARMv5, ARMv5e): Delete fgroups.
7557 (ARMv5t, ARMv5te): Adjust for above changes.
7559 (armv5, armv5e): Delete arches.
7560 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
7562 (*call_reg_arm): Likewise.
7563 (*call_value_reg_armv5): Likewise.
7564 (*call_value_reg_arm): Likewise.
7565 (*call_symbol): Likewise.
7566 (*call_value_symbol): Likewise.
7567 (*sibcall_insn): Likewise.
7568 (*sibcall_value_insn): Likewise.
7570 (prefetch): Likewise.
7571 (define_split and define_peephole2 dependent on arm_arch5):
7573 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
7575 (TARGET_ARM_QBIT): Likewise.
7576 (TARGET_DSP_MULTIPLY): Likewise.
7577 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
7578 (arm_arch5, arm_arch5e): Delete.
7579 (arm_arch5t, arm_arch5te): Declare.
7580 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
7581 (arm_arch5t): Declare.
7582 (arm_option_reconfigure_globals): Update for the above.
7583 (arm_options_perform_arch_sanity_checks): Update comment, replace
7584 use of arm_arch5 with arm_arch5t.
7585 (use_return_insn): Likewise.
7586 (arm_emit_call_insn): Likewise.
7587 (output_return_instruction): Likewise.
7588 (arm_final_prescan_insn): Likewise.
7589 (arm_coproc_builtin_available): Likewise.
7590 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
7591 arm_arch5e with arm_arch5t and arm_arch5te.
7592 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
7593 (arm_arch5t, arm_arch5te): Declare.
7594 * config/arm/arm-tables.opt: Regenerate.
7595 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
7596 * config/arm/t-multilib: Likewise.
7597 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
7598 instead of arm_arch5.
7599 (*call_reg_thumb1): Likewise.
7600 (*call_value_reg_thumb1_v5): Likewise.
7601 (*call_value_reg_thumb1): Likewise.
7602 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
7604 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
7606 2018-05-18 Martin Liska <mliska@suse.cz>
7608 PR gcov-profile/84846
7609 * doc/gcov.texi: Document -t option of gcov tool.
7611 2018-05-18 Martin Liska <mliska@suse.cz>
7613 PR gcov-profile/84846
7614 * gcov.c (print_usage): Add new -t option.
7615 (process_args): Handle the option.
7616 (generate_results): Use stdout as output when requested by
7619 2018-05-18 Martin Liska <mliska@suse.cz>
7621 PR gcov-profile/84846
7622 * coverage.c (coverage_init): Write PWD to .gcno file.
7623 * doc/gcov.texi: Document how working directory is printed.
7624 * gcov-dump.c (dump_gcov_file): Print PWD.
7625 * gcov.c (output_intermediate_file): Likewise.
7626 (read_graph_file): Read PWD string.
7627 (output_lines): Print PWD.
7629 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7632 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
7633 for retval and return false if all args to phi are zero.
7635 2018-05-18 Richard Biener <rguenther@suse.de>
7637 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
7639 (evrp_dom_walker::before_dom_children): Call it.
7641 2018-05-18 Richard Biener <rguenther@suse.de>
7643 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
7644 results when processing array refs with variable index.
7646 2018-05-18 Toon Moene <toon@moene.org>
7648 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
7649 directly after that of -floop-interchange. Indicate that both
7650 options are enabled by default when specifying -O3.
7652 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7654 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
7655 iterator. Delete separate integer-mode vec_set<mode> expander.
7656 (aarch64_simd_vec_setv2di): Delete.
7657 (vec_setv2di): Delete.
7658 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
7659 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
7660 the "w, r" alternative.
7662 2018-05-18 Martin Liska <mliska@suse.cz>
7664 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
7665 * tree-pass.h (make_pass_lower_switch_O0): New function.
7666 * tree-switch-conversion.c (node_has_low_bound): Remove.
7667 (node_has_high_bound): Likewise.
7668 (node_is_bounded): Likewise.
7669 (class pass_lower_switch): Make it a template type and create
7671 (pass_lower_switch::execute): Add template argument.
7672 (make_pass_lower_switch): New function.
7673 (make_pass_lower_switch_O0): New function.
7674 (do_jump_if_equal): Remove.
7675 (emit_case_nodes): Simplify to just handle all 3 cases and leave
7676 all the hard work to tree optimization passes.
7678 2018-05-18 Martin Liska <mliska@suse.cz>
7680 * dbgcnt.c (limit_low): Renamed from limit.
7681 (limit_high): New variable.
7682 (dbg_cnt_is_enabled): Check for upper limit.
7683 (dbg_cnt): Adjust dumping.
7684 (dbg_cnt_set_limit_by_index): Add new argument for high
7686 (dbg_cnt_set_limit_by_name): Likewise.
7687 (dbg_cnt_process_single_pair): Parse new format.
7688 (dbg_cnt_process_opt): Use strtok.
7689 (dbg_cnt_list_all_counters): Remove 'value' and add
7691 * doc/invoke.texi: Document changes.
7693 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
7695 * doc/sourcebuild.texi (scalar_all_fma): Document.
7696 * tree.def (FMA_EXPR): Delete.
7697 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
7698 * internal-fn.c (ternary_direct): New macro.
7699 (expand_ternary_optab_fn): Likewise.
7700 (direct_ternary_optab_supported_p): Likewise.
7701 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
7702 * builtins.c (fold_builtin_fma): Delete.
7703 (fold_builtin_3): Don't call it.
7704 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
7705 * expr.c (expand_expr_real_2): Likewise.
7706 * fold-const.c (operand_equal_p): Likewise.
7707 (fold_ternary_loc): Likewise.
7708 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
7709 * gimple.c (DEFTREECODE): Likewise.
7710 * gimplify.c (gimplify_expr): Likewise.
7711 * optabs-tree.c (optab_for_tree_code): Likewise.
7712 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
7713 * tree-eh.c (operation_could_trap_p): Likewise.
7714 (stmt_could_throw_1_p): Likewise.
7715 * tree-inline.c (estimate_operator_cost): Likewise.
7716 * tree-pretty-print.c (dump_generic_node): Likewise.
7717 (op_code_prio): Likewise.
7718 * tree-ssa-loop-im.c (stmt_cost): Likewise.
7719 * tree-ssa-operands.c (get_expr_operands): Likewise.
7720 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
7721 * fold-const-call.h (fold_fma): Delete.
7722 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
7723 CFN_FNMA and CFN_FNMS.
7725 * genmatch.c (combined_fn): New enum.
7726 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
7727 (commutative_op): New function.
7728 (commutate): Use it. Handle more than 2 operands.
7729 (dt_operand::gen_gimple_expr): Use commutative_op.
7730 (parser::parse_expr): Allow :c to be used with non-binary
7731 operators if the commutative operand is known.
7732 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
7733 CFN_FMS, CFN_FNMA and CFN_FNMS.
7734 (backprop::process_assign_use): Remove FMA_EXPR handling.
7735 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
7736 (gen_hsa_fma): New function.
7737 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
7738 IFN_FNMA and IFN_FNMS.
7739 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
7740 * gimple-fold.h (follow_all_ssa_edges): Declare.
7741 * gimple-fold.c (follow_all_ssa_edges): New function.
7742 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
7743 gimple_build interface and use follow_all_ssa_edges to fold the result.
7744 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
7745 instead of checking for optabs directly.
7746 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
7747 rather than FMA_EXPRs.
7748 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
7749 call to IFN_FMA instead of an FMA_EXPR.
7751 2018-05-17 Jim Wilson <jimw@sifive.com>
7753 * expr.c (do_tablejump): When converting index to Pmode, if we have a
7754 sign extended promoted subreg, and the range does not have the sign bit
7755 set, then do a sign extend.
7757 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
7758 test, check for sign extended subreg and/or constant operands, and
7759 do a sign extend in that case.
7761 2018-05-17 Steve Ellcey <sellcey@cavium.com>
7763 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
7764 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
7766 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
7767 Change logics_shift_reg to logics_shift_imm.
7768 (thunderx2t99_fp_loadpair_basic): Delete.
7769 (thunderx2t99_fp_storepair_basic): Delete.
7770 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
7771 (thunderx2t99_asimd_polynomial): Delete.
7772 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
7773 and neon_fp_mul_d_scalar_q.
7774 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
7775 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
7776 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
7777 (thunderx2t99_asimd_lut): Add missing tbl types.
7778 (thunderx2t99_asimd_ext): Delete.
7779 (thunderx2t99_asimd_load1_1_mult): Delete.
7780 (thunderx2t99_asimd_load1_2_mult): Delete.
7781 (thunderx2t99_asimd_load1_ldp): New.
7782 (thunderx2t99_asimd_load1): New.
7783 (thunderx2t99_asimd_load2): Add missing *load2* types.
7784 (thunderx2t99_asimd_load3): New.
7785 (thunderx2t99_asimd_load4): New.
7786 (thunderx2t99_asimd_store1_1_mult): Delete.
7787 (thunderx2t99_asimd_store1_2_mult): Delete.
7788 (thunderx2t99_asimd_store2_mult): Delete.
7789 (thunderx2t99_asimd_store2_onelane): Delete.
7790 (thunderx2t99_asimd_store_stp): New.
7791 (thunderx2t99_asimd_store1): New.
7792 (thunderx2t99_asimd_store2): New.
7793 (thunderx2t99_asimd_store3): New.
7794 (thunderx2t99_asimd_store4): New.
7796 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
7798 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
7799 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
7801 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
7802 Segher Boessenkool <segher@kernel.crashing.org>
7805 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
7808 2018-05-17 Richard Biener <rguenther@suse.de>
7810 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
7811 for pruning loop and prune defs feeding only already visited PHIs.
7813 2018-05-17 Richard Biener <rguenther@suse.de>
7815 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
7817 2018-05-17 Bin Cheng <bin.cheng@arm.com>
7818 Richard Biener <rguenther@suse.de>
7820 PR tree-optimization/85793
7821 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
7822 for VMAT_ELEMENTWISE.
7824 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
7826 * internal-fn.h (lookup_internal_fn): Declare
7827 * internal-fn.c (lookup_internal_fn): New function.
7828 * gimple.c (gimple_build_call_from_tree): Handle calls to
7830 * gimple-pretty-print.c (dump_gimple_call): Print "." before
7831 internal function names.
7832 * tree-pretty-print.c (dump_generic_node): Likewise.
7833 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
7835 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
7837 * gimple-fold.h (gimple_build): Make the function forms take
7838 combined_fn rather than built_in_function.
7839 (gimple_simplify): Likewise.
7840 * gimple-match-head.c (gimple_simplify): Likewise.
7841 * gimple-fold.c (gimple_build): Likewise.
7842 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
7843 rather than gimple_build_call_internal.
7844 (get_initial_defs_for_reduction): Likewise.
7845 (vect_create_epilog_for_reduction): Likewise.
7846 (vectorizable_live_operation): Likewise.
7848 2018-05-17 Martin Liska <mliska@suse.cz>
7850 * gimple-ssa-sprintf.c (format_directive): Do not use
7851 space in between 'G_' and '('.
7853 2018-05-17 Jakub Jelinek <jakub@redhat.com>
7856 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
7857 even if the mask is not all ones.
7860 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
7862 (ix86_gimple_fold_builtin): Likewise.
7865 * config/i386/i386.c: Include tree-vector-builder.h.
7866 (ix86_vector_shift_count): New function.
7867 (ix86_fold_builtin): Fold shift builtins by scalar count.
7868 (ix86_gimple_fold_builtin): Likewise.
7870 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
7871 _mm512_setzero): New intrinsics.
7873 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
7874 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7876 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
7877 code generation for cases where splatting a value is not useful.
7878 * simplify-rtx.c (simplify_ternary_operation): Simplify
7879 vec_merge across a vec_duplicate and a paradoxical subreg forming
7880 a vector mode to a vec_concat.
7882 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
7884 * config.gcc: Support "goldmont-plus".
7885 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
7887 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7888 PROCESSOR_GOLDMONT_PLUS.
7889 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
7890 (processor_target_table): Add "goldmont-plus".
7891 (PTA_GOLDMONT_PLUS): Define.
7892 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
7893 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
7894 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
7895 (fold_builtin_cpu): Add "goldmont-plus".
7896 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
7897 (ix86_option_override_internal): Add "goldmont-plus".
7898 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
7899 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
7900 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
7901 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
7903 2018-05-17 Richard Biener <rguenther@suse.de>
7905 PR tree-optimization/85757
7906 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
7907 remove defs that only feed that PHI from further processing.
7909 2018-05-16 Jim Wilson <jimw@sifive.com>
7911 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
7913 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
7915 2018-05-16 Mark Wielaard <mark@klomp.org>
7917 * dwarf2out.c (count_index_strings): New function.
7918 (output_indirect_strings): Call count_index_strings and generate
7919 header for dwarf_version >= 5.
7921 2018-05-16 Mark Wielaard <mark@klomp.org>
7923 * dwarf2out.c (dwarf_FORM): New function.
7924 (set_indirect_string): Use dwarf_FORM.
7925 (reset_indirect_string): Likewise.
7926 (size_of_die): Likewise.
7927 (value_format): Likewise.
7928 (output_die): Likewise.
7929 (add_skeleton_AT_string): Likewise.
7930 (output_macinfo_op): Likewise.
7931 (index_string): Likewise.
7932 (output_index_string_offset): Likewise.
7933 (output_index_string): Likewise.
7934 (count_index_strings): Likewise.
7936 2018-05-16 Carl Love <cel@us.ibm.com>
7938 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7939 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
7941 2018-05-16 Martin Jambor <mjambor@suse.cz>
7943 * ipa-prop.c (ipa_free_all_edge_args): Remove.
7944 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
7946 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
7948 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
7949 (fnma<mode>4): Likewise.
7950 (fms<mode>4): Likewise.
7951 (fnms<mode>4): Likewise.
7952 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
7953 (aarch64_fnma<mode>4): Likewise.
7954 (aarch64_fms<mode>4): Likewise.
7955 (aarch64_fnms<mode>4): Likewise.
7956 (aarch64_fnmadd<mode>4): Likewise.
7958 2018-05-16 Jason Merrill <jason@redhat.com>
7960 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
7962 2018-05-16 Richard Biener <rguenther@suse.de>
7964 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
7965 (dump_stmt_cost): Declare.
7966 (add_stmt_cost): Dump cost we add.
7967 (add_stmt_costs): New function.
7968 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
7970 (vect_analyze_stmt): Adjust prototype.
7971 (vectorizable_condition): Likewise.
7972 (vectorizable_live_operation): Likewise.
7973 (vectorizable_reduction): Likewise.
7974 (vectorizable_induction): Likewise.
7975 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
7976 cost vector to pass to vectorizable_ and record afterwards.
7977 (vect_model_reduction_cost): Take cost vector argument and adjust.
7978 (vect_model_induction_cost): Likewise.
7979 (vectorizable_reduction): Likewise.
7980 (vectorizable_induction): Likewise.
7981 (vectorizable_live_operation): Likewise.
7982 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
7983 SLP_TREE_NUMBER_OF_VEC_STMTS.
7984 (vect_analyze_slp_cost_1): Remove.
7985 (vect_analyze_slp_cost): Likewise.
7986 (vect_slp_analyze_node_operations): Take visited args and
7987 a target cost vector. Avoid processing already visited stmt sets.
7988 (vect_slp_analyze_operations): Use a local cost vector to gather
7989 costs and register those of non-discarded instances.
7990 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
7991 (vect_schedule_slp_instance): Remove copying of
7992 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
7994 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
7995 adding cost. Record cost entry location.
7996 (vect_prologue_cost_for_slp_op): Function to compute cost of
7997 a constant or invariant generated for SLP vect in the prologue,
7998 split out from vect_analyze_slp_cost_1.
7999 (vect_model_simple_cost): Make static. Adjust for SLP costing.
8000 (vect_model_promotion_demotion_cost): Likewise.
8001 (vect_model_store_cost): Likewise, make static.
8002 (vect_model_load_cost): Likewise.
8003 (vectorizable_bswap): Add cost vector arg and adjust.
8004 (vectorizable_call): Likewise.
8005 (vectorizable_simd_clone_call): Likewise.
8006 (vectorizable_conversion): Likewise.
8007 (vectorizable_assignment): Likewise.
8008 (vectorizable_shift): Likewise.
8009 (vectorizable_operation): Likewise.
8010 (vectorizable_store): Likewise.
8011 (vectorizable_load): Likewise.
8012 (vectorizable_condition): Likewise.
8013 (vectorizable_comparison): Likewise.
8014 (can_vectorize_live_stmts): Likewise.
8015 (vect_analyze_stmt): Likewise.
8016 (vect_transform_stmt): Adjust calls to vectorizable_*.
8017 * tree-vectorizer.c: Include gimple-pretty-print.h.
8018 (dump_stmt_cost): New function.
8020 2018-05-16 Richard Biener <rguenther@suse.de>
8022 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
8023 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
8024 * tree-ssa-dse.c: Include tree-ssa-loop.h.
8025 (check_name): New callback.
8026 (dse_classify_store): Track cycles via a visited bitmap of PHI
8027 defs and simplify handling of in-loop and across loop dead stores
8028 and properly fail for loop-variant refs. Handle byte-tracking with
8029 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
8032 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
8034 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
8035 (vect_get_mask_type_for_stmt): Likewise.
8036 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
8038 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
8039 to determine the statement's vector type and the vector type that
8040 should be used for calculating nunits. Deal with cases in which
8041 the type has to be deferred.
8042 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
8043 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
8044 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
8045 (vect_determine_vf_for_stmt): New functions, split out from...
8046 (vect_determine_vectorization_factor): ...here.
8047 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
8048 (vect_get_mask_type_for_stmt): New functions, split out from
8049 vect_determine_vectorization_factor.
8051 2018-05-16 Richard Biener <rguenther@suse.de>
8053 * tree-cfg.c (verify_gimple_assign_ternary): Properly
8054 verify the [VEC_]COND_EXPR embedded comparison.
8056 2018-05-15 Martin Sebor <msebor@redhat.com>
8058 PR tree-optimization/85753
8059 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
8060 RECORD_TYPE in addition to ARRAY_TYPE.
8062 2018-05-15 Martin Sebor <msebor@redhat.com>
8065 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
8067 2018-05-15 Richard Biener <rguenther@suse.de>
8069 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
8070 add by_clobber_p one. Change algorithm to collect all defs
8071 representing uses we need to walk and try reducing them to
8072 a single one before failing.
8073 (dse_dom_walker::dse_optimize_stmt): Adjust.
8075 2018-05-13 Mark Wielaard <mark@klomp.org>
8077 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
8078 (size_of_loc_descr): Likewise.
8079 (output_loc_operands): Likewise.
8080 (output_loc_operands_raw): Likewise.
8081 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
8082 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
8083 (hash_loc_operands): Likewise.
8084 (compare_loc_operands): Likewise.
8086 2018-05-14 Mark Wielaard <mark@klomp.org>
8088 * dwarf2out.c (count_index_addrs): New function.
8089 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
8091 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8093 PR tree-optimization/83648
8094 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
8095 return value as malloc candidate.
8097 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8100 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
8101 param as true in call to suggest_attribute.
8103 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
8105 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
8106 -mreadonly-in-sdata.
8108 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8110 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
8112 (aarch64_crypto_aesd_fused): Likewise.
8114 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
8116 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
8117 (movsi_aarch64): Likewise.
8118 (load_pairsi): Likewise.
8119 (load_pairdi): Likewise.
8120 (store_pairsi): Likewise.
8121 (store_pairdi): Likewise.
8122 (load_pairsf): Likewise.
8123 (load_pairdf): Likewise.
8124 (store_pairsf): Likewise.
8125 (store_pairdf): Likewise.
8126 (zero_extend): Likewise.
8127 (trunc): Swap alternatives.
8128 (fcvt_target): Add '?' to prefer w over r.
8130 2018-05-14 Jakub Jelinek <jakub@redhat.com>
8133 * config/i386/i386.md: Disallow non-commutative arithmetics in
8134 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
8135 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
8136 in the peephole2 before it.
8138 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
8140 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
8141 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
8142 (ix86_handle_option): Handle -mcldemote.
8143 * config.gcc: New header.
8144 * config/i386/cldemoteintrin.h: New file.
8145 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
8146 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
8148 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8149 OPTION_MASK_ISA_CLDEMOTE.
8150 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
8151 (ix86_valid_target_attribute_inner_p): Ditto.
8152 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
8153 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
8154 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
8155 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
8156 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
8158 * config/i386/i386.opt: Add -mcldemote.
8159 * config/i386/x86intrin.h: New header.
8160 * doc/invoke.texi: Add -mcldemote.
8162 2018-05-14 Richard Biener <rguenther@suse.de>
8164 * doc/match-and-simplify.texi: Adjust :s documentation.
8166 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
8168 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
8169 intended memcpy size.
8170 (REORDER_45): Likewise.
8172 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
8174 * sort.cc: New file.
8175 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
8176 * vec.c (qsort_chk): Use gcc_qsort.
8177 * Makefile.in (OBJS-libcommon): Add sort.o.
8178 (build/sort.o): New target. Use it...
8179 (BUILD_RTL): ... here, and...
8180 (build/gencfn-macros): ... here, and...
8181 (build/genmatch): ... here.
8183 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8184 Chung-Ju Wu <jasonwucj@gmail.com>
8186 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
8187 * config/nds32/nds32-graywolf.md: New file.
8188 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
8189 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
8191 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
8192 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
8193 * config/nds32/nds32.md (pipeline_model): Add graywolf.
8194 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
8195 * config/nds32/pipelines.md: Include n15 settings.
8197 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8198 Chung-Ju Wu <jasonwucj@gmail.com>
8200 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
8201 * config/nds32/nds32-n13.md: New file.
8202 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
8203 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
8205 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
8206 * config/nds32/nds32.md (pipeline_model): Add n13.
8207 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
8208 * config/nds32/pipelines.md: Include n13 settings.
8210 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8211 Chung-Ju Wu <jasonwucj@gmail.com>
8213 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
8214 * config/nds32/nds32-n10.md: New file.
8215 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
8216 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
8218 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
8219 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
8220 * config/nds32/nds32.md (pipeline_model): Add n10.
8221 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
8222 * config/nds32/pipelines.md: Include n10 settings.
8224 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
8225 Kito Cheng <kito.cheng@gmail.com>
8226 Chung-Ju Wu <jasonwucj@gmail.com>
8228 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
8229 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8230 Add enum values for DSP extension instructions.
8231 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
8233 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
8234 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
8236 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
8237 * config/nds32/nds32-dspext.md: New file for DSP implementation.
8238 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
8239 * config/nds32/nds32-intrinsic.md: Likewise.
8240 * config/nds32/nds32_intrinsic.h: Likewise.
8241 * config/nds32/nds32-md-auxiliary.c: Likewise.
8242 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
8243 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
8244 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
8245 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
8246 * config/nds32/nds32-protos.h: New declarations for DSP extension.
8247 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
8248 TYPE_DMAC in switch statement.
8249 * config/nds32/nds32.c: New checking and implementation for DSP
8250 extension instructions.
8251 * config/nds32/nds32.h: Likewise.
8252 * config/nds32/nds32.md: Likewise.
8253 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
8254 * config/nds32/predicates.md: Implement new predicates for DSP
8257 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8259 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
8260 Reformat alternatives and attributes so it is easier to identify
8261 which constraints/attributes go with which instruction.
8262 (mov<mode>_hardfloat32, FMOVE64): Likewise.
8263 (mov<mode>_softfloat32, FMOVE64): Likewise.
8264 (mov<mode>_hardfloat64, FMOVE64): Likewise.
8265 (mov<mode>_softfloat64, FMOVE64): Likewise.
8267 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
8269 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8271 (Basic PowerPC Built-in Functions): The new name of the
8272 subsection previously known as "PowerPC Built-in Functions".
8273 (Basic PowerPC Built-in Functions Available on all Configurations):
8275 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
8276 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
8277 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
8278 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
8280 2018-05-11 Martin Jambor <mjambor@suse.cz>
8283 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
8286 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
8289 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
8291 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
8293 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
8294 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
8295 (ix86_handle_option): Handle -mwaitpkg.
8296 * config.gcc: New header.
8297 * config/i386/cpuid.h (bit_WAITPKG): New bit.
8298 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
8299 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
8301 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8302 OPTION_MASK_ISA_WAITPKG.
8303 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
8304 (ix86_option_override_internal): Add PTA_WAITPKG.
8305 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
8306 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
8307 IX86_BUILTIN_TPAUSE.
8308 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
8309 __builtin_ia32_umwait and __builtin_ia32_tpause.
8310 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
8311 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
8312 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
8313 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
8314 UNSPECV_TPAUSE): New.
8315 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
8316 * config/i386/i386.opt: Add -mwaitpkg.
8317 * config/i386/waitpkgintrin.h: New file.
8318 * config/i386/x86intrin.h: New header.
8319 * doc/invoke.texi: Add -mwaitpkg.
8321 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
8324 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
8326 (cortex-m0): Use armv6s-m isa.
8327 (cortex-m0plus): Likewise.
8328 (cortex-m1): Likewise.
8329 (cortex-m0.small-multiply): Likewise.
8330 (cortex-m0plus.small-multiply): Likewise.
8331 (cortex-m1.small-multiply): Likewise.
8333 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
8334 Jakub Jelinek <jakub@redhat.com>
8336 PR tree-optimization/85692
8337 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
8338 source permute as well.
8340 2018-05-11 Martin Liska <mliska@suse.cz>
8343 * doc/extend.texi: Document LLVM style format for no_sanitize
8346 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8348 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
8349 mode_supports_vsx_dform_quad to mode_supports_dq_form.
8350 (mode_supports_vsx_dform_quad): Likewise.
8351 (mode_supports_vmx_dform): Move these functions to be next to the
8352 other mode_supports functions.
8353 (mode_supports_dq_form): Likewise.
8354 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
8355 mode_supports_dq_form.
8356 (reg_offset_addressing_ok_p): Likewise.
8357 (offsettable_ok_by_alignment): Likewise.
8358 (rs6000_legitimate_offset_address_p): Likewise.
8359 (legitimate_lo_sum_address_p): Likewise.
8360 (rs6000_legitimize_address): Likewise.
8361 (rs6000_legitimize_reload_address): Likewise.
8362 (rs6000_secondary_reload_inner): Likewise.
8363 (rs6000_preferred_reload_class): Likewise.
8364 (rs6000_output_move_128bit): Likewise.
8366 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
8368 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
8369 Generate SImode target register for null target.
8370 <case IX86_BUILTIN_XGETBV>: Ditto.
8371 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
8372 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
8374 2018-05-10 Carl Love <cel@us.ibm.com>
8376 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
8377 dcbtt and dcbtstt if operands[2] is 0.
8379 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
8382 * config/i386/sse.md (usadv64qi): New expander.
8384 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
8386 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
8387 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
8388 -maltivec=be support.
8389 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
8390 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
8391 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
8392 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
8393 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
8394 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
8395 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
8396 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
8397 altivec_vsumsws): Adjust.
8398 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
8399 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
8400 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
8402 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
8403 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
8404 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
8405 (altivec_lve<VI_char>x): Delete expand.
8406 (*altivec_lve<VI_char>x_internal): Rename to...
8407 (altivec_lve<VI_char>x): ... this.
8408 (altivec_lvxl_<mode>): Delete expand.
8409 (*altivec_lvxl_<mode>_internal): Rename to ...
8410 (altivec_lvxl_<mode>): ... this.
8411 (altivec_stvxl_<mode>): Delete expand.
8412 (*altivec_stvxl_<mode>_internal): Rename to ...
8413 (altivec_stvxl_<mode>): ... this.
8414 (altivec_stve<VI_char>x): Delete expand.
8415 (*altivec_stve<VI_char>x_internal): Rename to ...
8416 (altivec_stve<VI_char>x): ... this.
8417 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
8418 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
8419 reduc_plus_scal_<mode>): Adjust.
8420 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
8422 (rs6000_cpu_cpp_builtins): Adjust.
8423 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
8424 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
8425 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
8426 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
8427 -maltivec=be support.
8428 (rs6000_split_vec_extract_var): Adjust.
8429 (rs6000_split_v4si_init): Adjust.
8430 (swap_selector_for_mode): Delete.
8431 (altivec_expand_lvx_be, altivec_expand_stvx_be,
8432 altivec_expand_stvex_be): Delete.
8433 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
8434 -maltivec=be support.
8435 (rs6000_gimple_fold_builtin): Ditto.
8436 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
8438 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
8439 (TARGET_DIRECT_MOVE_64BIT): Adjust.
8440 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
8441 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
8442 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
8443 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
8444 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
8445 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
8446 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
8447 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
8448 anonymous split): Adjust.
8449 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
8450 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
8452 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
8454 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
8455 when --with-gxx-include-dir is also specified.
8456 * configure: Regenerate.
8458 2018-05-09 Jim Wilson <jimw@sifive.com>
8461 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
8462 * config/riscv/t-withmultilib: New.
8463 * config/riscv/withmultilib.h: New.
8464 * doc/install.texi: Document RISC-V --with-multilib-list support.
8466 2018-05-09 Richard Biener <rguenther@suse.de>
8468 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
8470 (vect_bb_vectorization_profitable_p): Adjust. Compute
8471 actual scalar cost using the cost vector and the add_stmt_cost
8474 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8476 PR rtl-optimization/85645
8477 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
8478 in the REG_CFA_REGISTER note for LR, don't leave it empty.
8480 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8482 PR rtl-optimization/85645
8483 * shrink-wrap.c (spread_components): Return a boolean saying if
8484 anything was changed.
8485 (try_shrink_wrapping_separate): Iterate spread_components until
8486 nothing changes anymore.
8488 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8490 PR rtl-optimization/85645
8491 * regrename.c (build_def_use): Also kill the chains that include the
8492 destination of a REG_CFA_REGISTER note.
8494 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8496 PR rtl-optimization/85645
8497 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
8498 insn that has a REG_CFA_REGISTER note.
8500 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
8502 * cfgexpand.c (expand_clobber): New function.
8503 (expand_gimple_stmt_1): Use it.
8504 * tree-vect-stmts.c (vect_clobber_variable): New function,
8506 (vectorizable_simd_clone_call): ...here.
8507 (vectorizable_store): Emit a clobber either side of an
8508 IFN_STORE_LANES sequence.
8509 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
8511 2018-05-09 Tom de Vries <tom@codesourcery.com>
8514 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
8515 (define_insn "trap_if_false"): Add exit after trap.
8517 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
8519 PR rtl-optimization/85638
8520 * bb-reorder.c: Include common/common-target.h.
8521 (create_forwarder_block): New function extracted from...
8522 (fix_up_crossing_landing_pad): ...here. Rename into...
8523 (dw2_fix_up_crossing_landing_pad): ...this.
8524 (sjlj_fix_up_crossing_landing_pad): New function.
8525 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
8526 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
8527 from both partitions and exit the loop after one iteration.
8529 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
8532 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8534 (Basic PowerPC Built-in Functions): The new name of the
8535 subsection previously known as "PowerPC Built-in Functions".
8536 (Basic PowerPC Built-in Functions Available on all Configurations):
8538 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
8540 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
8542 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
8544 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
8547 2018-05-08 Jim Wilson <jimw@sifive.com>
8549 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
8550 (LD_EMUL_SUFFIX): New.
8551 (LINK_SPEC): Use it.
8553 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
8555 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8557 (Basic PowerPC Built-in Functions): The new name of the
8558 subsection previously known as "PowerPC Built-in Functions".
8559 (Basic PowerPC Built-in Functions Available on all Configurations):
8561 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
8563 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
8565 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
8567 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
8570 2018-05-08 Jakub Jelinek <jakub@redhat.com>
8573 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
8574 after cmpelim optimization.
8576 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
8578 * config.gcc: Support "goldmont".
8579 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
8580 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8582 * config/i386/i386.c (m_GOLDMONT): Define.
8583 (processor_target_table): Add "goldmont".
8584 (PTA_GOLDMONT): Define.
8585 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
8586 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
8587 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
8588 (fold_builtin_cpu): Add "goldmont".
8589 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
8590 (ix86_option_override_internal): Add "goldmont".
8591 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
8592 (processor_type): Add PROCESSOR_GOLDMONT.
8593 * config/i386/i386.md: Add CPU "glm".
8594 * config/i386/glm.md: New file.
8595 * config/i386/x86-tune.def: Add m_GOLDMONT.
8596 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
8598 2018-05-08 Jakub Jelinek <jakub@redhat.com>
8601 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
8603 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
8604 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
8605 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
8608 * config/i386/i386.c (ix86_fold_builtin): Handle
8609 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
8612 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
8613 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
8615 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
8618 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
8619 (check_arch): Likewise.
8620 (check_fpu): Return the result rather than printing it.
8621 (end arch): Fix operator precedence.
8622 (end cpu): Likewise.
8623 (END): Print the result from check_fpu.
8625 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
8626 Alan Hayward <alan.hayward@arm.com>
8627 David Sherwood <david.sherwood@arm.com>
8629 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
8630 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
8631 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
8632 (*fcmuo<mode>_and): New patterns.
8634 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
8636 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
8637 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
8638 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
8639 (cmp_op, sve_imm_con): New code attributes.
8640 (SVE_COND_INT_CMP, imm_con): Delete.
8641 (cmp_op): Remove above unspecs from int attribute.
8642 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
8644 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
8645 comparison-specific unspecs.
8646 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
8647 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
8648 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
8649 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
8650 (*vec_fcm<cmp_op><mode>): Rename to...
8651 (*fcm<cmp_op><mode>): ...this and adjust likewise.
8652 (*vec_fcmuo<mode>): Rename to...
8653 (*fcmuo<mode>): ...this and adjust likewise.
8654 (*pred_fcm<cmp_op><mode>): New pattern.
8655 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
8656 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
8658 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
8660 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
8661 (aarch64_emit_sve_predicated_cond): New function.
8662 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
8663 (aarch64_emit_unspec_cond_or): Replace with...
8664 (aarch64_emit_sve_or_conds): ...this new function. Use
8665 aarch64_emit_sve_ptrue_op for the individual comparisons and
8666 aarch64_emit_binop to OR them together.
8667 (aarch64_emit_inverted_unspec_cond): Replace with...
8668 (aarch64_emit_sve_inverted_cond): ...this new function. Use
8669 aarch64_emit_sve_ptrue_op for the comparison and
8670 aarch64_emit_unop to invert the result.
8671 (aarch64_expand_sve_vec_cmp_float): Update after the above
8672 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
8674 2018-05-07 Nathan Sidwell <nathan@acm.org>
8676 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
8677 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
8678 (Backwards Compatibility): Likewise.
8680 2018-05-07 Luis Machado <luis.machado@linaro.org>
8684 2018-05-07 Luis Machado <luis.machado@linaro.org>
8686 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8687 <prefetch_dynamic_strides>: New const bool field.
8688 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8689 prefetch_dynamic_strides.
8690 (exynosm1_prefetch_tune): Likewise.
8691 (thunderxt88_prefetch_tune): Likewise.
8692 (thunderx_prefetch_tune): Likewise.
8693 (thunderx2t99_prefetch_tune): Likewise.
8694 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
8696 (aarch64_override_options_internal): Update to set
8697 PARAM_PREFETCH_DYNAMIC_STRIDES.
8698 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
8699 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
8700 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
8701 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
8702 prefetch-dynamic-strides setting.
8704 2018-05-07 Luis Machado <luis.machado@linaro.org>
8706 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8707 <minimum_stride>: New const int field.
8708 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8709 minimum_stride field.
8710 (exynosm1_prefetch_tune): Likewise.
8711 (thunderxt88_prefetch_tune): Likewise.
8712 (thunderx_prefetch_tune): Likewise.
8713 (thunderx2t99_prefetch_tune): Likewise.
8714 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
8715 (aarch64_override_options_internal): Update to set
8716 PARAM_PREFETCH_MINIMUM_STRIDE.
8717 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
8718 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
8719 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
8720 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
8721 stride is constant and is below the minimum stride threshold.
8723 2018-05-07 Luis Machado <luis.machado@linaro.org>
8725 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
8728 2018-05-07 Luis Machado <luis.machado@linaro.org>
8730 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8731 <prefetch_dynamic_strides>: New const bool field.
8732 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8733 prefetch_dynamic_strides.
8734 (exynosm1_prefetch_tune): Likewise.
8735 (thunderxt88_prefetch_tune): Likewise.
8736 (thunderx_prefetch_tune): Likewise.
8737 (thunderx2t99_prefetch_tune): Likewise.
8738 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
8740 (aarch64_override_options_internal): Update to set
8741 PARAM_PREFETCH_DYNAMIC_STRIDES.
8742 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
8743 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
8744 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
8745 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
8746 prefetch-dynamic-strides setting.
8748 2018-05-07 Luis Machado <luis.machado@linaro.org>
8750 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8751 <minimum_stride>: New const int field.
8752 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8753 minimum_stride field.
8754 (exynosm1_prefetch_tune): Likewise.
8755 (thunderxt88_prefetch_tune): Likewise.
8756 (thunderx_prefetch_tune): Likewise.
8757 (thunderx2t99_prefetch_tune): Likewise.
8758 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
8759 (aarch64_override_options_internal): Update to set
8760 PARAM_PREFETCH_MINIMUM_STRIDE.
8761 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
8762 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
8763 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
8764 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
8765 stride is constant and is below the minimum stride threshold.
8767 2018-05-06 Jakub Jelinek <jakub@redhat.com>
8770 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
8771 the type is addressable. Don't force op into register if it has
8774 2018-05-05 Roland McGrath <mcgrathr@google.com>
8777 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
8778 any section for which we don't know a specific type it should have,
8779 regardless of name. Previously this was done only for the exact
8780 names ".init_array", ".fini_array", and ".preinit_array".
8781 (default_elf_asm_named_section): Add comment about
8782 relationship with default_section_type_flags and SECTION_NOTYPE.
8783 (get_section): Don't consider it a type conflict if one side has
8784 SECTION_NOTYPE and the other doesn't, as long as neither has the
8785 SECTION_BSS et al used in the default_section_type_flags logic.
8787 2018-05-05 Tom de Vries <tom@codesourcery.com>
8790 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
8791 (workaround_barsyncs): New function.
8792 (nvptx_reorg): Use workaround_barsyncs.
8793 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
8794 (define_expand "nvptx_membar_cta"): New define_expand.
8795 (define_insn "*nvptx_membar_cta"): New insn.
8797 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
8799 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
8800 To improve optimization opportunities.
8801 * builtin-types.def: The new needed builtin types for the above.
8803 2018-05-04 Richard Biener <rguenther@suse.de>
8805 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
8806 * gimple-ssa-store-merging.c
8807 (imm_store_chain_info::output_merged_store): Remove redundant create,
8808 release split_store vector contents on failure.
8809 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
8810 scalar stmt vector on cache hit.
8812 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
8814 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
8816 * config.gcc (powerpc-xilinx-eabi*): Remove.
8817 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
8819 (fusion_addis_mem_combo_load): Ditto.
8820 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
8822 (rs6000_cpu_cpp_builtins): Ditto.
8823 * config/rs6000/rs6000-linux.c
8824 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
8825 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
8826 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
8828 (rs6000_setup_reg_addr_masks): Ditto.
8829 (rs6000_init_hard_regno_mode_ok): Ditto.
8830 (rs6000_option_override_internal): Ditto.
8831 (legitimate_lo_sum_address_p): Ditto.
8832 (rs6000_legitimize_address): Ditto.
8833 (rs6000_legitimize_reload_address): Ditto.
8834 (rs6000_legitimate_address_p): Ditto.
8835 (abi_v4_pass_in_fpr): Ditto.
8836 (setup_incoming_varargs): Ditto.
8837 (rs6000_gimplify_va_arg): Ditto.
8838 (rs6000_split_multireg_move): Ditto.
8839 (rs6000_savres_strategy): Ditto.
8840 (rs6000_emit_prologue_components): Ditto.
8841 (rs6000_emit_epilogue_components): Ditto.
8842 (rs6000_emit_prologue): Ditto.
8843 (rs6000_emit_epilogue): Ditto.
8844 (rs6000_elf_file_end): Ditto.
8845 (rs6000_function_value): Ditto.
8846 (rs6000_libcall_value): Ditto.
8847 * config/rs6000/rs6000.h: Ditto.
8848 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
8849 (TARGET_MINMAX): ... this. New.
8850 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
8851 * config/rs6000/rs6000.md: Remove Xilinx FP support.
8852 (*movsi_internal1_single): Delete.
8853 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
8854 mfpu=, mxilinx-fpu): Delete.
8855 * config/rs6000/singlefp.h: Delete.
8856 * config/rs6000/sysv4.h: Remove Xilinx FP support.
8857 * config/rs6000/t-rs6000: Ditto.
8858 * config/rs6000/t-xilinx: Delete.
8859 * config/rs6000/titan.md: Adjust for fp_type removal.
8860 * config/rs6000/vsx.md: Remove Xilinx FP support.
8861 (VStype_simple): Delete.
8862 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
8863 * config/rs6000/xfpu.h: Delete.
8864 * config/rs6000/xfpu.md: Delete.
8865 * config/rs6000/xilinx.h: Delete.
8866 * config/rs6000/xilinx.opt: Delete.
8867 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
8868 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
8870 2018-05-04 Tom de Vries <tom@codesourcery.com>
8873 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
8876 2018-05-04 Richard Biener <rguenther@suse.de>
8879 * tree-complex.c (update_complex_assignment): We are always in SSA form.
8880 (expand_complex_div_wide): Likewise.
8881 (expand_complex_operations_1): Likewise.
8882 (expand_complex_libcall): Preserve EH info of the original stmt.
8883 (tree_lower_complex): Handle removed blocks.
8884 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
8885 on complex multiplication and division libcall builtins.
8887 2018-05-04 Richard Biener <rguenther@suse.de>
8890 * fold-const.c (negate_expr_p): Restrict negation of operand
8891 zero of a division to when we know that can happen without
8893 (fold_negate_expr_1): Likewise.
8895 2018-05-04 Jakub Jelinek <jakub@redhat.com>
8898 * real.h (real_nextafter): Declare.
8899 * real.c (real_nextafter): New function.
8900 * fold-const-call.c (fold_const_nextafter): New function.
8901 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
8902 CASE_CFN_NEXTTOWARD.
8903 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
8904 even when arg1_mode is different from arg0_mode.
8906 2018-05-03 Nathan Sidwell <nathan@acm.org>
8908 * doc/extend.texi (Deprecated Features): Remove
8910 (Backwards Compatibility): Likewise.
8911 * doc/invoke.texi (C++ Language Options): Likewise.
8912 (C++ Dialect Options): Likewise.
8914 2018-05-03 Jakub Jelinek <jakub@redhat.com>
8917 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
8918 _mm512_mask_mullox_epi64): New intrinsics.
8920 2018-05-03 Tom de Vries <tom@codesourcery.com>
8923 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8924 dump files): Add offload-tree.
8926 2018-05-03 Richard Biener <rguenther@suse.de>
8928 PR tree-optimization/85615
8929 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
8930 to loops not nested in BBs loop father to avoid creating multi-entry
8933 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8935 PR tree-optimization/70291
8936 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
8937 arguments. Change return type to tree. Emit libcall as a new
8938 statement rather than replacing existing one when inplace_p is true.
8939 (expand_complex_multiplication_components): New function.
8940 (expand_complex_multiplication): Expand floating-point complex
8941 multiplication using the above.
8942 (expand_complex_division): Rename inner_type parameter to type.
8943 Update expand_complex_libcall call-site.
8944 (expand_complex_operations_1): Update expand_complex_multiplication
8945 and expand_complex_division call-sites.
8947 2018-05-02 Jakub Jelinek <jakub@redhat.com>
8950 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
8951 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
8952 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
8953 the highest significant bit of the shift count mask is clear. In
8954 check whether and[sq]i3 is needed verify that all significant bits
8955 of the shift count other than the highest are set.
8957 2018-05-02 Tom de Vries <tom@codesourcery.com>
8960 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
8961 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
8962 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
8963 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
8964 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
8965 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
8966 __builtin_goacc_parlevel_size.
8968 2018-05-02 Richard Biener <rguenther@suse.de>
8970 PR tree-optimization/85597
8971 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
8972 do not use split vect_get_vec_defs call but call vect_get_slp_defs
8975 2018-05-02 Tom de Vries <tom@codesourcery.com>
8978 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8979 dump files): Add ltrans-tree.
8981 2018-05-02 Tom de Vries <tom@codesourcery.com>
8984 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8985 dump files): Add wpa-ipa.
8987 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
8989 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
8990 powerpc*-*-linux*paired* target.
8991 * config/rs6000/750cl.h: Delete.
8992 * config/rs6000/paired.h: Delete.
8993 * config/rs6000/paired.md: Delete.
8994 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
8996 * config/rs6000/rs6000-builtin.def: Remove paired float support.
8997 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
8998 comment. Remove paired float support.
8999 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
9000 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
9002 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
9003 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
9005 * config/rs6000/rs6000.c: Remove paired float support.
9006 (paired_expand_vector_init, paired_expand_vector_move,
9007 paired_emit_vector_compare, paired_emit_vector_cond_expr,
9008 (paired_expand_lv_builtin, paired_expand_stv_builtin,
9009 paired_expand_builtin, paired_expand_predicate_builtin,
9010 paired_init_builtins): Delete.
9011 * config/rs6000/rs6000.h: Remove paired float support.
9012 * config/rs6000/rs6000.md: Remove paired float support.
9013 (move_from_CR_ov_bit): Delete.
9014 * config/rs6000/rs6000.opt (mpaired): Delete.
9015 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
9016 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
9018 2018-05-02 Richard Biener <rguenther@suse.de>
9021 * gimplify.c (gimplify_save_expr): When in SSA form allow
9022 SAVE_EXPRs to compute to SSA vars.
9024 2018-05-02 Jakub Jelinek <jakub@redhat.com>
9027 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
9028 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
9029 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
9030 clobber operands[2], instead use a new pseudo. Formatting fixes.
9032 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
9034 PR tree-optimization/85586
9035 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
9036 exit early for statements in the same group if the accesses are
9039 2018-05-02 Tom de Vries <tom@codesourcery.com>
9042 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
9045 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
9047 PR tree-optimization/85143
9048 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
9050 2018-05-01 Tom de Vries <tom@codesourcery.com>
9053 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
9054 not found" error message.
9056 2018-05-01 Tom de Vries <tom@codesourcery.com>
9059 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
9060 * vec.c (test_ordered_remove_if): New function.
9061 (vec_c_tests): Call test_ordered_remove_if.
9062 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
9063 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
9064 * tree-vect-patterns.c (vect_pattern_recog_1): Use
9065 VEC_ORDERED_REMOVE_IF.
9067 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9069 PR tree-optimization/82665
9070 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
9071 pointer subtraction where arguments come from a memchr call.
9073 2018-05-01 Jakub Jelinek <jakub@redhat.com>
9075 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
9076 --push-state --as-needed and --pop-state instead of --as-needed and
9077 --no-as-needed if ld supports it.
9078 * configure: Regenerated.
9081 * doc/install.texi2html: Replace _002d with - and _002a with * in
9082 generated html files using sed.
9084 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9087 * gcc-rich-location.c (blank_line_before_p): New function.
9088 (use_new_line): New function.
9089 (gcc_rich_location::add_fixit_insert_formatted): New function.
9090 * gcc-rich-location.h
9091 (gcc_rich_location::add_fixit_insert_formatted): New function.
9093 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9095 * selftest.c (assert_streq): Rename "expected" and "actual" to
9096 "val1" and "val2". Extend NULL-handling to cover both inputs
9097 symmetrically, while still requiring both to be non-NULL for a pass.
9098 * selftest.h (assert_streq): Rename "expected" and "actual" to
9100 (ASSERT_EQ): Likewise.
9101 (ASSERT_EQ_AT): Likewise.
9102 (ASSERT_KNOWN_EQ): Likewise.
9103 (ASSERT_KNOWN_EQ_AT): Likewise.
9104 (ASSERT_NE): Likewise.
9105 (ASSERT_MAYBE_NE): Likewise.
9106 (ASSERT_MAYBE_NE_AT): Likewise.
9107 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
9108 the assertion to pass.
9109 (ASSERT_STREQ_AT): Likewise.
9111 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
9113 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
9114 interaction with -pie.
9116 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9118 * selftest.h: Fix alphabetization of per-source-file selftest
9121 2018-04-30 Jason Merrill <jason@redhat.com>
9123 PR c++/61982 - dead stores to destroyed objects.
9124 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
9127 2018-04-30 Jason Merrill <jason@redhat.com>
9129 * tree.c (build_clobber): New.
9130 * tree.h: Declare it.
9131 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
9133 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9135 * diagnostic-show-locus.c (layout::layout): Update for
9136 location_get_source_line returning a char_span.
9137 (struct char_span): Move to input.h.
9138 (struct correction): Update for fields in char_span becoming
9140 (struct source_line): Update for location_get_source_line
9141 returning a char_span.
9142 (layout::print_line): Likewise.
9143 * edit-context.c (edited_file::print_content): Likewise.
9144 (edited_file::print_diff_hunk): Likewise.
9145 (edited_file::print_run_of_changed_lines): Likewise.
9146 (edited_file::get_num_lines): Likewise.
9147 (edited_line::edited_line): Likewise.
9148 * final.c (asm_show_source): Likewise.
9149 * input.c (location_get_source_line): Convert return type
9150 from const char * to char_span, losing the final "line_len"
9152 (dump_location_info): Update for the above.
9153 (get_substring_ranges_for_loc): Likewise. Use a char_span
9154 when handling the literal within the line.
9155 (test_reading_source_line): Update for location_get_source_line
9156 returning a char_span.
9157 * input.h (class char_span): Move here from
9158 diagnostic-show-locus.c, converting from a struct to a class.
9159 Make data members private.
9160 (char_span::operator bool): New.
9161 (char_span::length): New.
9162 (char_span::get_buffer): New.
9163 (char_span::operator[]): New.
9164 (char_span::subspan): Make const.
9165 (char_span::xstrdup): New.
9166 (location_get_source_line): Convert return type from const char *
9167 to char_span, losing the final "line_size" param.
9169 2018-04-30 Jan Hubicka <jh@suse.cz>
9171 * lto-wrapper.c (ltrans_priorities): New static var.
9172 (cmp_priority): New.
9173 (run_gcc): Read priorities and if doing parallel build order
9174 the Makefile by them.
9176 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9178 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
9180 2018-04-30 Richard Biener <rguenther@suse.de>
9182 * tree-cfg.c (verify_address): Remove base argument, add
9183 flag whether to check TREE_ADDRESSABLE and do that.
9184 (verify_expr): Remove.
9185 (verify_types_in_gimple_reference): Add pieces from verify_expr.
9186 (verify_gimple_assign_single): Likewise.
9187 (verify_gimple_switch): Likewise.
9188 (verify_expr_location_1): Dereference tp once. Add (disabled)
9189 piece from verify_expr.
9190 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
9192 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9194 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
9196 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9198 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
9199 (small_data_pattern): Likewise.
9200 (arc_rewrite_small_data): Likewise.
9201 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
9202 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
9203 (get_symbol_alignment): New function.
9204 (legitimate_small_data_address_p): Likewise.
9205 (legitimate_scaled_address): Update, call
9206 legitimate_small_data_address_p.
9207 (output_sdata): New static variable.
9208 (arc_print_operand): Update how we handle small data operands.
9209 (arc_print_operand_address): Likewise.
9210 (arc_legitimate_address_p): Update, use
9211 legitimate_small_data_address_p.
9212 (arc_rewrite_small_data_p): Remove.
9213 (arc_rewrite_small_data_1): Likewise.
9214 (arc_rewrite_small_data): Likewise.
9215 (small_data_pattern): Likewise.
9216 (compact_sda_memory_operand): Update to use
9217 legitimate_small_data_address_p and get_symbol_alignment.
9218 (prepare_move_operands): Don't rewite sdata pattern.
9219 (prepare_extend_operands): Remove.
9220 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
9222 (zero_extendqisi2): Likewise.
9223 (zero_extendhisi2): Likewise.
9224 (extendqihi2): Likewise.
9225 (extendqisi2): Likewise.
9226 (extendhisi2): Likewise.
9230 * config/arc/constraints.md (Usd): Change it to memory constraint.
9232 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9234 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
9235 as source of std instructions.
9236 * config/arc/arc.md (movsi_insn): Update pattern predicate to
9237 allow 6-bit constants as source for store instructions.
9238 (movdi_insn): Update instruction pattern to allow 6-bit constants
9239 as source for store instructions.
9241 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
9243 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
9245 2018-04-30 Nathan Sidwell <nathan@acm.org>
9246 Sandra Loosemore <sandra@codesourcery.com>
9248 * dumpfile.c (dump_open): Allow '-' for stdout.
9249 * doc/invoke.texi (Developer Options): Document dump filename
9250 determination early. Document stdin/stdout selection.
9252 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
9254 Microblaze Target: PIC data text relative
9256 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
9257 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
9259 * config/microblaze/microblaze.h (microblaze_constant_address_p):
9260 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
9261 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
9262 New addressing mode for data-text relative position indepenedent code.
9263 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
9264 'ADDRESS_SYMBOLIC_TXT_REL'.
9265 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
9266 (microblaze_legitimate_pic_operand): Exclude function calls from
9267 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
9268 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
9270 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
9271 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
9272 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
9273 for 'address + offset'.
9274 (microblaze_expand_prologue): Add new function prologue call for
9276 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
9277 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
9278 table in case of TARGET_PIC_DATA_TEXT_REL.
9279 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
9280 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
9281 Add new macros 'UNSPEC_TEXT',
9282 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
9283 + exclude function calls from 'UNSPEC_PLT' in case of data text
9285 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
9286 new target hook for generating address diff vector tables in case of
9288 * doc/tm.texi : Regenerate.
9289 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
9290 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
9291 of addr diff vector generation.
9292 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
9293 target hook definition.
9294 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
9295 Add default function for generate_pic_addr_diff_vec -> flag_pic.
9296 * doc/invoke.texi (Add new pic option): Add new microblaze pic
9297 option for data text relative.
9299 2018-04-30 Richard Biener <rguenther@suse.de>
9301 * tree-chrec.h (evolution_function_is_constant_p): Remove
9303 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
9305 2018-04-30 Richard Biener <rguenther@suse.de>
9308 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
9310 2018-04-30 Richard Biener <rguenther@suse.de>
9312 PR tree-optimization/28364
9313 PR tree-optimization/85275
9314 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
9315 copying first exit test.
9317 2018-04-28 Mark Wielaard <mark@klomp.org>
9319 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
9321 (dwarf_AT): Handle DW_AT_addr_base.
9322 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
9324 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
9327 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
9328 (*ashl<dwi>3_doubleword_mask_1): Ditto.
9329 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
9330 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
9332 2018-04-28 Richard Biener <rguenther@suse.de>
9334 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
9335 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
9336 to reflect use. Only add interesting stmts.
9338 2018-04-27 Martin Jambor <mjambor@suse.cz>
9341 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
9342 the jump function allows for passing through aggregate values.
9344 2018-04-27 David Malcolm <dmalcolm@redhat.com>
9346 * input.h (in_system_header_at): Convert from macro to inline
9348 (from_macro_expansion_at): Likewise.
9349 (from_macro_definition_at): Likewise.
9351 2018-04-27 Jeff Law <law@redhat.com>
9353 * config.gcc: Mark tile* targets as deprecated/obsolete.
9355 2018-04-27 Richard Biener <rguenther@suse.de>
9357 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
9360 2018-04-27 Richard Biener <rguenther@suse.de>
9362 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
9364 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
9366 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
9367 with Yd constraint. Set "preferred_for_speed" attribute from
9368 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
9370 (*movdi_internal): Ditto.
9371 (movti_interunit splitters): Remove
9372 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
9373 (movdi_interunit splitters): Ditto.
9374 * config/i386/constraints.md (Ye): Remove.
9375 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
9377 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9380 * config/aarch64/constraints.md (Usg): Limit to 31.
9383 2018-04-27 Jakub Jelinek <jakub@redhat.com>
9385 PR tree-optimization/85529
9386 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
9387 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
9388 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
9389 zero extension or masking of the MSB bit.
9390 (optimize_range_tests): Add FIRST_BB argument, pass it through
9391 to optimize_range_tests_var_bound.
9392 (maybe_optimize_range_tests, reassociate_bb): Adjust
9393 optimize_range_tests callers.
9395 2018-04-26 Richard Biener <rguenther@suse.de>
9396 Jakub Jelinek <jakub@redhat.com>
9398 * cgraph.h (symbol_table): Just declare debug method here.
9399 * symtab.c (symbol_table::debug): Define.
9401 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
9403 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
9405 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
9407 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
9408 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
9409 (*movdi_internal): Substitute Yi and Yj constraint with x
9410 and Ym and Yn constraint with y constraint. Update "isa"
9411 attribute and set "preferred_for_speed" attribute from
9412 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
9413 (*movsi_internal): Ditto.
9414 (*movdf_internal): Ditto.
9415 (*movsf_internal): Ditto.
9416 (*zero_extendsidi2): Ditto.
9417 * config/i386/sse.md (vec_set<mode>_0): Ditto.
9418 (sse2_loadld): Ditto.
9419 (*vec_extract<ssevecmodelower>_0): Ditto.
9420 (*vec_extractv4si_0_zext_sse4): Ditto.
9421 (vec_concatv2di): Ditto.
9422 (*vec_dup<mode>): Ditto.
9423 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
9424 * config/i386/constraints.md (Yi): Remove.
9429 2018-04-26 Nathan Sidwell <nathan@acm.org>
9431 * dumpfile.c (dump_open): New.
9432 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
9433 (dump_finish): Detect stdio/stderr by value not name.
9435 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
9437 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
9439 2018-04-26 Tom de Vries <tom@codesourcery.com>
9442 * config/nvptx/nvptx.c (verify_neutering_jumps)
9443 (verify_neutering_labels): New function
9444 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
9446 2018-04-26 Tom de Vries <tom@codesourcery.com>
9449 * config/nvptx/nvptx.c (needs_neutering_p): New function.
9450 (nvptx_single): Use needs_neutering_p to skip over insns that do not
9453 2018-04-26 Richard Biener <rguenther@suse.de>
9454 Tom de Vries <tom@codesourcery.com>
9457 * lto-streamer-out.c (output_function): Fixup loops if required to match
9458 discovery done in the reader.
9460 2018-04-26 Richard Biener <rguenther@suse.de>
9462 PR tree-optimization/85116
9463 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
9464 have a loop exit from the single latch predecessor. Remove
9465 case of header with just condition.
9466 (ch_base::copy_headers): Exclude infinite loops from any
9468 (pass_ch::execute): Record exits.
9470 2018-04-26 Richard Biener <rguenther@suse.de>
9472 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
9473 prologue cost vector and pass it to vect_get_load_cost.
9474 (vect_get_peeling_costs_all_drs): Likewise.
9475 (vect_peeling_hash_get_lowest_cost): Likewise.
9476 (vect_enhance_data_refs_alignment): Likewise.
9478 2018-04-26 Richard Biener <rguenther@suse.de>
9481 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
9482 checking of integer<->pointer conversions.
9483 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
9484 sign-/zero-extending pointer types.
9485 (expand_omp_for_static_chunk): Likewise.
9487 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
9488 Jean Lee <xiaoyur347@gmail.com>
9490 * config/mips/mips.c (mips_asan_shadow_offset): New function.
9491 (TARGET_ASAN_SHADOW_OFFSET): Define.
9492 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
9493 true for -fsanitize=address.
9495 2018-04-25 Mark Wielaard <mark@klomp.org>
9497 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
9500 2018-04-25 Jakub Jelinek <jakub@redhat.com>
9502 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
9503 than "alu", remove explicit "memory" and "imm_disp" attributes.
9504 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
9507 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
9508 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
9509 gen_lowpart_no_emit.
9511 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
9514 * config/i386/i386.c (ix86_expand_builtin): Change memory
9515 operand to XI, extend p0 to Pmode.
9516 * config/i386/i386.md: Change unspec volatile and operand
9517 1 mode to XI, change operand 0 mode to P.
9519 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9521 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
9522 GET_MODE_MASK before any checking.
9523 (nds32_can_use_bset_p): Likewise.
9524 (nds32_can_use_btgl_p): Likewise.
9526 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9528 * config/nds32/nds32-doubleword.md: New define_split pattern for
9529 illegal register number.
9531 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9533 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
9535 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9537 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
9539 2018-04-25 Richard Biener <rguenther@suse.de>
9541 * lto-streamer.h (LTO_major_version): Bump to 8.
9543 2018-04-25 Jakub Jelinek <jakub@redhat.com>
9545 * BASE-VER: Set to 9.0.0.
9547 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
9549 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
9550 in __abskf2 and __powikf2.
9552 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9555 * config/aarch64/constraints.md (Usg, Usj): New constraints.
9556 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
9557 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
9558 Use the above on operand 2. Reindent.
9559 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
9561 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
9564 * common/config/i386/i386-common.c (ix86_handle_option): Don't
9566 * config/i386/i386.opt (mcet): Removed.
9567 * doc/install.texi: Remove -mcet documentation.
9568 * doc/invoke.texi: Likewise.
9570 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
9573 * doc/install.texi: Remove -mcet from bootstrap-cet.
9575 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9578 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
9579 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
9583 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
9584 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
9585 containing a CONST_VECTOR.
9587 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
9589 * doc/install.texi: Update newlib dependency for nvptx.
9591 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9594 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
9595 instead of INTVAL when shifting x left.
9597 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
9599 PR tree-optimization/85478
9600 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
9601 vect_grouped_store_supported for single element vectors.
9603 2018-04-24 Richard Biener <rguenther@suse.de>
9606 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
9607 load cost increase to the case of non-constant step.
9609 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9612 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
9613 destination if any_malformed_asm.
9615 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
9618 * expr.c (store_field): In the bitfield case, if the value comes from
9619 a function call and is returned in registers by means of a PARALLEL,
9620 do not change the mode of the temporary unless BLKmode and VOIDmode.
9622 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
9624 PR rtl-optimization/85423
9625 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
9626 dependencies to debug insns when the previous insn is non-debug.
9628 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
9630 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
9631 enums into a single definition.
9632 (fls): Fix predicates and printing.
9635 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
9637 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
9638 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
9639 and short u6 immediate.
9640 (check_if_valid_sleep_operand): Remove.
9641 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
9643 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9645 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
9646 flag_always_save_lp condition.
9647 * config/nds32/nds32.opt (malways-save-lp): New option.
9649 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9651 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
9652 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
9653 * config/nds32/nds32.h
9654 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
9655 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
9657 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9659 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
9660 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
9662 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9663 Chung-Ju Wu <jasonwucj@gmail.com>
9665 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
9667 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
9668 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
9670 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9672 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
9674 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9676 * config/nds32/nds32-protos.h (nds32_data_alignment,
9677 nds32_local_alignment): Declare.
9678 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
9679 nds32_local_alignment): New functions.
9680 (TARGET_CONSTANT_ALIGNMENT): Define.
9681 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
9683 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9685 * config/nds32/nds32.c
9686 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
9687 (TARGET_MODES_TIEABLE_P): Likewise.
9689 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9691 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
9694 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
9695 Chung-Ju Wu <jasonwucj@gmail.com>
9697 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9698 for unaligned access.
9699 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9701 * config/nds32/nds32-intrinsic.md: Likewise.
9702 * config/nds32/nds32_intrinsic.h: Likewise.
9703 * config/nds32/nds32.h (nds32_builtins): Likewise.
9704 * config/nds32/nds32.opt (munaligned-access): New option.
9705 * config/nds32/nds32.c (nds32_asm_file_start): Display
9706 flag_unaligned_access status.
9708 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
9710 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
9711 -mno-relax is present.
9712 * config/riscv/linux.h (LINK_SPEC): Ditto.
9714 2018-04-20 Martin Sebor <msebor@redhat.com>
9717 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
9719 (gimple_fold_builtin_stxcpy_chk): Same.
9720 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
9722 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
9725 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
9726 __powikf2 when long double is IEEE 128-bit.
9728 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
9730 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
9731 step to make sure stack always aligned.
9733 2018-04-20 Carl Love <cel@us.ibm.com>
9736 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
9737 size check for arg0.
9739 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
9740 Tom de Vries <tom@codesourcery.com>
9743 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
9744 Emit insns for calls too.
9745 (nvptx_find_par): Always look for worker-level predecessor insn.
9746 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
9748 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
9749 (nvptx_process_pars): Propagate frames for calls.
9751 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
9754 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
9756 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
9757 (ix86_handle_option): Don't handle OPT_mibt.
9758 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
9760 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
9762 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
9763 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
9764 (ix86_target_macros): Define __CET__ with flag_cf_protection
9765 for -fcf-protection.
9766 * config/i386/i386.c (isa2_opts): Remove -mibt.
9767 * config/i386/i386.h (TARGET_IBT): Removed.
9768 (TARGET_IBT_P): Likewise.
9769 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
9770 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
9771 * config/i386/i386.opt (mcet): Update help message.
9774 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
9775 -mcet as an alias for -mshstk.
9777 2018-04-20 Richard Biener <rguenther@suse.de>
9780 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
9781 complexity by forcing a single use of the multiply operand.
9783 2018-04-20 Martin Jambor <mjambor@suse.cz>
9786 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
9787 recursion dependency to only apply to non-clones.
9789 2018-04-20 Martin Jambor <mjambor@suse.cz>
9792 * ipa-cp.c (create_specialized_node): Check that clones of
9793 self-recursive edges exist during IPA-CP.
9795 2018-04-19 Toon Moene <toon@moene.org>
9797 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
9800 2018-04-19 Jakub Jelinek <jakub@redhat.com>
9802 PR tree-optimization/85467
9803 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
9804 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
9805 VECTOR_CST element to type.
9807 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9810 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
9811 * config/i386/i386.md (builtin_setjmp_setup): Removed.
9812 (builtin_longjmp): Likewise.
9813 (save_stack_nonlocal): New pattern.
9814 (restore_stack_nonlocal): Likewise.
9816 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9819 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
9820 Replace ASM_OUTPUT_LABEL with fprintf.
9822 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9825 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
9826 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
9827 * config/i386/i386-c.c (ix86_target_macros_internal): Also
9828 define __IBT__ and __SHSTK__ for -fcf-protection.
9829 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
9831 (ix86_trampoline_init): Likewise.
9832 (x86_output_mi_thunk): Likewise.
9833 (ix86_notrack_prefixed_insn_p): Likewise.
9834 (ix86_option_override_internal): Don't disallow -fcf-protection.
9835 * config/i386/i386.md (rdssp<mode>): Also enable for
9837 (incssp<mode>): Likewise.
9838 (nop_endbr): Likewise.
9839 * config/i386/i386.opt (mcet): Change help message to built-in
9843 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
9844 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
9845 enable CET built-in functions.
9847 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
9849 * common/config/i386/i386-common.c
9850 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
9851 OPTION_MASK_ISA_MOVDIRI_UNSET,
9852 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
9853 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
9854 * config.gcc (movdirintrin.h): New header.
9855 * config/i386/cpuid.h (bit_MOVDIRI,
9856 bit_MOVDIR64B): New bits.
9857 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
9859 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
9860 (VOID, PVOID, PCVOID)): New function types.
9861 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
9862 __builtin_ia32_directstoreu_u64,
9863 __builtin_ia32_movdir64b): New builtins.
9864 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
9865 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
9867 (ix86_valid_target_attribute_inner_p): Ditto.
9868 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
9869 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
9870 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
9871 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
9872 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
9873 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
9874 (movdiri<mode>, movdir64b_<mode>): New.
9875 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
9876 * config/i386/immintrin.h: Include movdirintrin.h.
9877 * config/i386/movdirintrin.h: New file.
9878 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
9880 2018-04-19 Richard Biener <rguenther@suse.de>
9883 * cfg.c (clear_bb_flags): When loop state says we have
9884 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
9886 2018-04-19 Richard Biener <rguenther@suse.de>
9888 PR tree-optimization/84737
9889 * tree-vect-data-refs.c (vect_copy_ref_info): New function
9890 copying restrict info.
9891 (vect_setup_realignment): Use it.
9892 * tree-vectorizer.h (vect_copy_ref_info): Declare.
9893 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
9894 the first DR to all generated stores.
9895 (vectorizable_load): Likewise for loads.
9897 2018-04-19 Jakub Jelinek <jakub@redhat.com>
9899 PR tree-optimization/85446
9900 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
9901 the integral and pointer types to have the same precision.
9903 * doc/install.texi: Document --disable-cet being the default and
9906 2018-04-18 Martin Liska <mliska@suse.cz>
9908 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
9911 2018-04-18 Martin Liska <mliska@suse.cz>
9914 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9917 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9918 arguments if they are comparable.
9920 2018-04-18 Martin Liska <mliska@suse.cz>
9923 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9926 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9929 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
9932 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
9933 ENDBR after calling __morestack.
9935 2018-04-18 David Malcolm <dmalcolm@redhat.com>
9938 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
9939 by using gcc_base_ver to generate a gcc_driver_version, and use
9940 it when generating GCC_DRIVER_NAME.
9941 * configure: Regenerate.
9943 2018-04-18 Jakub Jelinek <jakub@redhat.com>
9946 * config.gcc: Obsolete powerpc*-*-*spe*.
9948 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9951 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
9952 (stabstr_D): Change type of unum from unsigned int to
9953 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
9956 2018-04-17 Jim Wilson <jimw@sifive.com>
9959 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
9960 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
9961 Set arg_pointer_offset after using pretend_args_size.
9963 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9965 PR rtl-optimization/85431
9966 * dse.c (record_store): Ignore zero width stores.
9969 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
9970 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
9971 __builtin_stack_restore rather than after it.
9972 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
9973 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
9974 argument instead of virtual_dynamic_stack_rtx.
9976 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
9978 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
9980 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
9981 Add note to error message to explain internal mapping of overloaded
9982 built-in function name to non-overloaded built-in function name.
9983 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
9986 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9989 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
9990 where the inputs overlap with the output.
9992 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9994 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
9995 (=v, v) alternative and explicit "memory" attribute.
9996 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
9997 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9999 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
10000 "sselog1" type instead of "sselog".
10001 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
10002 "sselog". Remove explicit "memory" attribute.
10003 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
10004 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10006 (vec_extract_hi_v32hi): Merge all alternatives into one, use
10007 "sselog1" type instead of "sselog". Remove explicit "memory"
10009 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
10010 use "sselog1" type instead of "sselog". Remove explicit "memory"
10012 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
10013 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10015 (vec_extract_hi_v64qi): Merge all alternatives into one, use
10016 "sselog1" type instead of "sselog". Remove explicit "memory"
10018 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
10019 use "sselog1" type instead of "sselog". Remove explicit "memory"
10023 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
10025 PR middle-end/85414
10026 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
10029 2018-04-17 Martin Jambor <mjambor@suse.cz>
10032 * ipa-cp.c (create_specialized_node): Call
10033 expand_all_artificial_thunks if necessary.
10035 2018-04-17 Martin Liska <mliska@suse.cz>
10038 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
10039 in message, remote space in between '_G' and '('.
10041 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10044 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
10045 avx512f_vmcmp<mode>3<round_saeonly_name>,
10046 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
10047 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
10048 avx512f_rndscale<mode><round_saeonly_name>,
10049 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
10050 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
10051 Use %<iptr>2 instead of %2 for -masm=intel.
10052 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
10053 avx512f_vcvttss2usi<round_saeonly_name>,
10054 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
10056 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
10057 avx512f_vcvttsd2usi<round_saeonly_name>,
10058 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
10059 Use %q1 instead of %1 for -masm=intel.
10060 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
10061 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
10062 of %3 for -masm=intel.
10063 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
10065 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
10067 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
10069 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
10070 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
10072 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
10074 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
10075 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
10076 %g1 and one with %0 and %1.
10077 (avx512er_vmrcp28<mode><round_saeonly_name>,
10078 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
10079 %1 for -masm=intel.
10080 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
10081 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
10082 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
10083 of %0 and %{%4%} for -masm=intel.
10084 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
10085 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
10086 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
10087 order of %0 and %{%5%}%{z%} for -masm=intel.
10089 2018-04-17 Jan Hubicka <jh@suse.cz>
10092 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
10094 2018-04-17 Martin Liska <mliska@suse.cz>
10097 * multiple_target.c (create_dispatcher_calls): Set apostrophes
10098 for target_clone error message. Make default implementation
10099 clone to be a local declaration.
10100 (separate_attrs): Add new argument and check for an empty
10102 (expand_target_clones): Handle it.
10103 (ipa_target_clone): Make redirection just for target_clones
10106 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
10107 Tom de Vries <tom@codesourcery.com>
10109 PR middle-end/84955
10110 * omp-expand.c (expand_oacc_for): Add dummy false branch for
10111 tiled basic blocks without omp continue statements.
10113 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
10116 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
10117 vec_extract expression as having side effects to make sure it gets
10120 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
10123 * config/i386/i386.c (get_builtin_code_for_version): Check
10126 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
10129 * config.gcc: Support "skylake".
10130 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10132 * config/i386/i386.c (m_SKYLAKE): Define.
10133 (processor_target_table): Add "skylake".
10134 (ix86_option_override_internal): Add "skylake".
10135 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
10136 PROCESSOR_CANNONLAKE.
10137 (get_builtin_code_for_version): Fix priority for
10138 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
10139 PROCESSOR_SKYLAKE-AVX512.
10140 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
10141 (processor_type): Add PROCESSOR_SKYLAKE.
10143 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
10144 Jason Merrill <jason@redhat.com>
10147 * convert.c (convert_to_integer_1): Use direct recursion for
10148 enumeral types and types with a precision less than the number
10149 of bits in their mode.
10151 2018-04-16 Julia Koval <julia.koval@intel.com>
10154 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
10155 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
10157 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
10160 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
10161 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
10162 and -mno-direct-move.
10164 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
10167 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
10168 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
10169 Ensure negative shifts result in {0}.
10171 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
10173 PR rtl-optimization/79916
10174 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
10175 regs (if any) to define how to gnerate SD moves when LRA is in
10178 2018-04-13 Jakub Jelinek <jakub@redhat.com>
10180 PR rtl-optimization/85393
10181 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
10182 * except.c (expand_dw2_landing_pad_for_region): Make static.
10183 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
10184 a label and unconditional jump to old_bb, rather than
10185 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
10188 PR rtl-optimization/85376
10189 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
10190 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
10191 instead of a specific value.
10193 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
10194 Bin Cheng <bin.cheng@arm.com>
10196 PR tree-optimization/82965
10197 PR tree-optimization/83991
10198 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
10199 by_profile_only parameter.
10200 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
10201 information if the loop was predicted to iterate too many times.
10202 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
10204 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
10207 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
10210 2018-04-13 Martin Liska <mliska@suse.cz>
10211 Jakub Jelinek <jakub@redhat.com>
10213 PR middle-end/81657
10214 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
10215 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
10216 * builtins.c (expand_builtin_memory_copy_args): Use
10217 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
10218 handle dest_addr == pc_rtx.
10220 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
10223 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
10224 asked to not generate direct moves.
10225 (fix_trunc<mode>si2_stfiwx): Similar.
10226 (fix_trunc<mode>si2_internal): Similar.
10228 2018-04-12 Jakub Jelinek <jakub@redhat.com>
10231 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
10232 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
10233 lookup if dest in some wider mode is known to be const0_rtx and
10234 if so, record permanent equivalence for it to be ZERO_EXTEND of
10235 the narrower mode destination.
10237 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
10239 * lto-streamer-out.c (output_function): Revert 259346.
10240 * omp-expand.c (expand_oacc_for): Likewise.
10242 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
10244 PR rtl-optimization/85354
10245 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
10246 * sel-sched.c (sel_global_init): ... here.
10248 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
10251 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
10252 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
10253 mode for PE-COFF targets.
10254 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
10255 (i386_pe_asm_lto_end): Likewise.
10256 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
10257 (TARGET_ASM_LTO_END): Likewise.
10258 * config/i386/winnt.c (saved_debug_info_level): New static variable.
10259 (i386_pe_asm_lto_start): New function.
10260 (i386_pe_asm_lto_end): Likewise.
10262 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
10263 Richard Biener <rguenther@suse.de>
10265 PR middle-end/84955
10266 * lto-streamer-out.c (output_function): Fix CFG loop state before
10268 * omp-expand.c (expand_oacc_for): Handle calls to internal
10269 functions like regular functions.
10271 2018-04-12 Richard Biener <rguenther@suse.de>
10274 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
10275 for the early LTO debug to properly generate references to it
10276 during DIE emission. Do not re-use that for the skeleton for
10278 (dwarf2out_early_finish): Likewise.
10280 2018-04-12 Jakub Jelinek <jakub@redhat.com>
10283 * config/i386/sse.md
10284 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
10285 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
10286 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
10287 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
10288 and output is a reg, avoid creating invalid lowpart subreg, but
10289 instead split into a 512-bit move. Don't split if not AVX512VL,
10290 input is xmm16+ reg and output is a mem.
10291 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
10292 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
10293 xmm16+ reg and output is a mem.
10295 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10297 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
10298 also for flag_dwarf2_cfi_asm.
10300 2018-04-12 Jakub Jelinek <jakub@redhat.com>
10302 PR rtl-optimization/85342
10303 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
10304 a bool scalar var inside of the loop instead. Don't try to update
10305 recog_data.operand after failed apply_change_group.
10307 2018-04-12 Tom de Vries <tom@codesourcery.com>
10310 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
10311 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
10312 array with flexible array member as array without given dimension.
10313 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
10314 argument for undefined param to true.
10316 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
10319 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
10320 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
10321 from PowerPC section.
10322 * config/rs6000/sysv4.opt (mcall-): Improve help text.
10323 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
10324 help text that is too long.
10325 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
10326 help text that is too long.
10327 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
10328 help text that is too long.
10330 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
10332 * config/alpha/alpha.md (stack_probe_internal): Rename
10333 from "probe_stack". Update all callers.
10335 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10337 PR rtl-optimization/84566
10338 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
10339 sched_macro_fuse_insns.
10341 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10344 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
10345 (compute_block_dependences): ... from here.
10347 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10349 PR tree-optimization/85331
10350 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
10351 from int to HOST_WIDE_INT.
10353 2018-04-11 Martin Jambor <mjambor@suse.cz>
10356 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
10357 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
10358 not the same as the source val.
10359 (cgraph_edge_brings_value_p): New parameter.
10360 (gather_edges_for_value): Pass destination value to
10361 cgraph_edge_brings_value_p.
10362 (perhaps_add_new_callers): Likewise.
10363 (get_info_about_necessary_edges): Likewise and exclude values brought
10364 only by self-recursive edges.
10365 (create_specialized_node): Redirect only clones of self-calling edges.
10366 (+self_recursive_pass_through_p): New function.
10367 (find_more_scalar_values_for_callers_subset): Use it.
10368 (find_aggregate_values_for_callers_subset): Likewise.
10369 (known_aggs_to_agg_replacement_list): Removed.
10370 (decide_whether_version_node): Re-calculate known constants for all
10371 remaining context clones.
10373 2018-04-11 Richard Biener <rguenther@suse.de>
10376 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
10377 from early DWARF output.
10378 (dwarf2out_early_finish): Output line info unconditionally into
10379 early DWARF and add reference to it.
10381 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10384 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
10385 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
10386 other than V2DFmode using iptr mode attribute.
10387 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
10389 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10391 PR rtl-optimization/84659
10392 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
10394 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10397 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
10399 (output_loc_list): Pass address of a dummy size variable even in the
10400 locview handling loop.
10401 (index_location_lists): Add comment on why skip_loc_list_entry can't
10404 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
10407 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
10410 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
10413 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
10414 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
10415 and -mstring-compare-inline-limit.
10417 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10420 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
10421 for stack clash protection in a register whenever we need it to be in
10424 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10426 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
10427 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
10429 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10432 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
10434 (mlong-double-): Ditto.
10435 * config/rs6000/sysv4.opt (msdata=): Ditto.
10436 (mtls-size=): Ditto.
10438 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10440 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
10441 erroneous entries for
10442 "vector int vec_ldl (int, long int *)", and
10443 "vector unsigned int vec_ldl (int, unsigned long int *)".
10444 Add comments and entries for
10445 "vector bool char vec_ldl (int, bool char *)",
10446 "vector bool short vec_ldl (int, bool short *)",
10447 "vector bool int vec_ldl (int, bool int *)",
10448 "vector bool long long vec_ldl (int, bool long long *)",
10449 "vector pixel vec_ldl (int, pixel *)",
10450 "vector long long vec_ldl (int, long long *)",
10451 "vector unsigned long long vec_ldl (int, unsigned long long *)".
10452 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
10453 type tree bool_long_long_type_node and correct definition of
10454 bool_V2DI_type_node to make reference to this new type tree.
10455 (rs6000_mangle_type): Replace erroneous reference to
10456 bool_long_type_node with bool_long_long_type_node.
10457 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
10458 comments to emphasize sign distinctions for char and int types and
10459 replace RS6000_BTI_bool_long constant with
10460 RS6000_BTI_bool_long_long constant. Also add comment to restrict
10461 use of RS6000_BTI_pixel.
10462 (bool_long_type_node): Remove this macro definition.
10463 (bool_long_long_type_node): New macro definition
10465 2018-04-10 Jakub Jelinek <jakub@redhat.com>
10467 PR rtl-optimization/85300
10468 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
10469 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
10470 simplify_unary_operation fails.
10472 2018-04-10 Martin Liska <mliska@suse.cz>
10474 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
10475 cgraph_edge and ipa_ref.
10477 2018-04-10 Jakub Jelinek <jakub@redhat.com>
10481 * config/i386/sse.md
10482 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
10483 computation of the VEC_MERGE selector from mask.
10484 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
10485 Fix decoding of the VEC_MERGE selector into mask.
10487 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
10489 PR tree-optimization/85286
10490 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
10492 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
10494 * final.c (final_1): Set insn_last_address as well as
10495 insn_current_address.
10497 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10500 * explow.c (emit_stack_probe): Call validize_mem on memory location
10501 before passing it to gen_probe_stack. Create address operand and
10502 legitimize it for the probe_stack_address case.
10504 2018-04-09 Jan Hubicka <jh@suse.cz>
10507 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
10508 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
10509 * tree.c (free_lang_data_in_type): Fix handling of binfos;
10511 (free_lang_data): Rebuild type inheritance graph.
10513 2018-04-09 Martin Sebor <msebor@redhat.com>
10515 * invoke.texi (-finline-small-functions): Mention other optimization
10517 (-findirect-inlining, -fpartial-inlining): Same.
10518 (-finline-functions-called-once): Same.
10519 (-freorder-blocks-and-partition): Same.
10521 2018-04-09 Jan Hubicka <jh@suse.cz>
10524 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
10525 jumps; choose last target that matches the criteria (i.e.
10526 no partition changes for non-crossing jumps).
10527 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
10528 support for redirecting crossing jumps to non-crossing.
10530 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
10532 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
10533 also for naked functions.
10535 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
10537 * config/arc/arc.md (add_shift): New pattern.
10538 (add_shift2): Likewise.
10539 (sub_shift): Likewise.
10540 (sub_shift_cmp0_noout): Likewise.
10541 (compare_si_ashiftsi): Likewise.
10542 (xbfu_cmp0_noout): New combine pattern.
10543 (xbfu_cmp0"): Likewise.
10544 (movsi_set_cc_insn): Place the predicable variant first.
10545 (commutative_binary_cmp0_noout): Remove clobber.
10546 (commutative_binary_cmp0): New pattern.
10547 (noncommutative_binary_cmp0): Likewise.
10548 (noncommutative_binary_cmp0_noout): Likewise.
10549 (noncommutative_binary_comparison_result_used): Removed.
10550 (rsub_cmp0): New pattern.
10551 (rsub_cmp0_noout): Likewise.
10552 (extzvsi): Changed, keep only meaningful variants.
10553 (SQH, SEZ): New iterators.
10554 (SQH_postfix): New mode attribute.
10555 (SEZ_prefix): New code attribute.
10556 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
10557 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
10558 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
10559 of numerical value.
10560 (noncommutative_operator): Check the availability of barrel
10563 2018-04-09 Richard Biener <rguenther@suse.de>
10565 PR tree-optimization/85284
10566 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
10567 Only use the niter constraining form of simple_iv when the exit
10568 is always executed.
10570 2018-04-09 Tom de Vries <tom@codesourcery.com>
10573 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
10574 (define_expand "*memory_barrier"): New define_expand.
10575 (define_insn "memory_barrier"): New insn.
10577 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10579 PR rtl-optimization/80463
10580 PR rtl-optimization/83972
10581 PR rtl-optimization/83480
10583 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
10584 correct producer for the insn.
10585 (tidy_control_flow): Fixup seqnos in case of debug insns.
10587 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10589 PR rtl-optimization/83913
10591 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
10592 different sched-times when merging exprs.
10594 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10596 PR rtl-optimization/83962
10598 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
10599 tidy_fallthru_edge and tidy_control_flow.
10601 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10603 PR rtl-optimization/83530
10605 * sel-sched.c (force_next_insn): New global variable.
10606 (remove_insn_for_debug): When force_next_insn is true, also leave only
10607 next insn in the ready list.
10608 (sel_sched_region): When the region wasn't scheduled, make another pass
10609 over it with force_next_insn set to 1.
10611 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
10613 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
10615 * config/nds32/constants.md (unspec_volatile_element): Add enum values
10616 for interrupt control.
10617 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
10618 functions for interrupt control.
10619 * config/nds32/nds32-intrinsic.md: Likewise.
10620 * config/nds32/nds32_intrinsic.h: Likewise.
10621 * config/nds32/nds32.h (nds32_builtins): Likewise.
10623 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
10625 * config/nds32/nds32.c (nds32_init_machine_status,
10626 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
10627 strict_aligned_p field.
10628 (nds32_expand_to_rtl_hook): New function.
10629 (TARGET_EXPAND_TO_RTL_HOOK): Define.
10630 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
10632 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10633 Chung-Ju Wu <jasonwucj@gmail.com>
10635 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
10636 * config/nds32/nds32-n7.md: New file.
10637 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
10638 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
10640 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
10641 * config/nds32/nds32.md (pipeline_model): Add n7.
10642 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
10643 * config/nds32/pipelines.md: Include n7 settings.
10645 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10646 Chung-Ju Wu <jasonwucj@gmail.com>
10648 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
10649 * config/nds32/nds32-e8.md: New file.
10650 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
10651 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
10653 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
10654 * config/nds32/nds32.md (pipeline_model): Add e8.
10655 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
10656 * config/nds32/pipelines.md: Include e8 settings.
10658 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10659 Chung-Ju Wu <jasonwucj@gmail.com>
10661 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
10662 * config/nds32/nds32-n8.md: New file.
10663 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
10664 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
10666 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
10667 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
10668 * config/nds32/nds32.md (pipeline_model): Add n8.
10669 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
10670 * config/nds32/pipelines.md: Include n8 settings.
10672 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10673 Chung-Ju Wu <jasonwucj@gmail.com>
10675 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
10676 * config/nds32/nds32-n9-2r1w.md: New file.
10677 * config/nds32/nds32-n9-3r2w.md: New file.
10678 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
10679 nds32_register_ports): New or modify for cpu n9.
10680 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
10682 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
10683 * config/nds32/nds32-utils.c: New file.
10684 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
10685 TARGET_MUL_SLOW): Define.
10686 * config/nds32/nds32.md (pipeline_model): New attribute.
10687 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
10688 New options that support cpu n9.
10689 * config/nds32/pipelines.md: Include n9 settings.
10690 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
10692 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
10694 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
10695 information if necessary.
10696 (output_cond_branch_compare_zero): Likewise.
10697 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
10698 (nds32_target_alignment): Refine for alignment.
10699 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
10700 (FUNCTION_BOUNDARY): Modify.
10701 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
10703 * config/nds32/nds32.opt (malways-align, malign-functions): New.
10705 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
10707 * config/nds32/constants.md (unspec_volatile_element): Add values for
10708 TLB operation and data prefetch.
10709 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
10710 functions for TLB operation and data prefetch.
10711 * config/nds32/nds32-intrinsic.md: Likewise.
10712 * config/nds32/nds32_intrinsic.h: Likewise.
10713 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
10714 (nds32_print_operand): Likewise.
10715 * config/nds32/nds32.h (nds32_builtins): Likewise.
10717 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
10718 Andrew Pinski <pinsika@gcc.gnu.org>
10720 PR middle-end/82976
10721 * match.pd: Use constant_boolean_node of correct type instead of
10722 boolean_true_node or boolean_false_node for simplifying
10723 pointer comparisons to zero.
10725 2018-04-07 Jakub Jelinek <jakub@redhat.com>
10727 PR tree-optimization/80021
10728 * tree.c (verify_type_variant): Make error call in verify_variant_match
10729 translatable and remove final full stop.
10731 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10733 * config/nds32/constants.md (unspec_volatile_element): Add
10734 UNSPEC_VOLATILE_EH_RETURN.
10735 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
10736 nds32_output_stack_pop): Support dwarf exception handling process.
10737 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
10738 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
10739 exception handling process.
10740 (nds32_compute_stack_frame): Likewise.
10741 (nds32_return_addr_rtx): Likewise.
10742 (nds32_initial_elimination_offset): Likewise.
10743 (nds32_expand_prologue): Likewise.
10744 (nds32_expand_epilogue): Likewise.
10745 (nds32_dynamic_chain_address): New function.
10746 * config/nds32/nds32.h (machine_function): Add fields for dwarf
10747 exception handling.
10748 (DYNAMIC_CHAIN_ADDRESS): Define.
10749 (EH_RETURN_DATA_REGNO): Define.
10750 (EH_RETURN_STACKADJ_RTX): Define.
10751 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
10752 patterns for dwarf exception handling.
10754 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10756 * config/nds32/nds32.h: Clean up obsolete macros.
10758 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10760 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10761 Add enum values for particular instructions.
10762 * config/nds32/nds32-intrinsic.c: Implementation of expanding
10763 particular intrinsic functions.
10764 * config/nds32/nds32-intrinsic.md: Likewise.
10765 * config/nds32/nds32_intrinsic.h: Likewise.
10766 * config/nds32/nds32.h (nds32_builtins): Likewise.
10767 * config/nds32/nds32.md (type): Add pbsad and pbsada.
10768 (btst, ave): New patterns for particular instructions.
10770 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10772 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10773 Add enum values for atomic load/store and memory sync.
10774 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
10776 * config/nds32/nds32-intrinsic.md: Likewise.
10777 * config/nds32/nds32_intrinsic.h: Likewise.
10778 * config/nds32/nds32.h (nds32_builtins): Likewise.
10780 2018-04-07 Jakub Jelinek <jakub@redhat.com>
10782 PR tree-optimization/85257
10783 * fold-const.c (native_encode_vector): If not all elts could fit
10784 and off is -1, return 0 rather than offset.
10785 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
10786 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
10787 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
10788 adjust buffer in native_interpret_expr call.
10790 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10792 * config/nds32/constants.md (unspec_volatile_element): Add cache
10793 control enum values.
10794 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
10795 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
10796 * config/nds32/nds32.c (nds32_cctl_names): New.
10797 (nds32_print_operand): Handle cache control register names.
10798 * config/nds32/nds32.h (nds32_builtins): New enum values.
10799 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
10801 * config/nds32/nds32.md (type): Add mmu.
10802 * config/nds32/pipelines.md (simple_insn): Add mmu.
10804 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10806 * config/nds32/nds32.md (type): Remove call.
10807 * config/nds32/pipelines.md (simple_insn): Likewise.
10809 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10811 * config/nds32/constants.md (unspec_volatile_element): Add
10812 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
10813 UNSPEC_VOLATILE_FMFCFG.
10814 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
10815 description for fmfcfg and fmfcsr.
10816 (bdesc_1arg): Add fmtcsr.
10817 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
10818 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
10819 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
10820 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
10821 unspec_fmfcfg): New patterns.
10822 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
10823 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
10824 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
10825 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
10826 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
10827 __nds32__fmfcfg): Define.
10829 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10831 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
10832 intrinsic register names.
10833 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
10834 intrinsic register enum values and macros.
10836 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10838 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
10839 for load/store addressing form.
10840 (nds32_print_operand_address): Likewise.
10842 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
10845 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
10846 based on LABEL_REF. Remove useless assertion.
10847 (pic_address_needs_scratch): Fix formatting.
10848 (sparc_legitimize_pic_address): Minor tweaks.
10849 (sparc_delegitimize_address): Adjust assertion accordingly.
10850 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
10851 into symbolic_operand.
10852 (movsi_high_pic_label_ref): Likewise.
10853 (movsi_lo_sum_pic_label_ref): Likewise.
10854 (movdi_pic_label_ref): Likewise.
10855 (movdi_high_pic_label_ref): Likewise.
10856 (movdi_lo_sum_pic_label_ref): Likewise.
10858 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
10860 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
10861 custom LIB_SPEC setup.
10863 2018-04-06 Ruslan Bukin <br@bsdpad.com>
10864 Kito Cheng <kito.cheng@gmail.com>
10866 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
10867 * config/riscv/freebsd.h: New.
10869 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10871 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
10872 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
10875 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10876 Kito Cheng <kito.cheng@gmail.com>
10878 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
10879 nds32_output_call, nds32_symbol_binds_local_p): New functions.
10880 * config/nds32/nds32-protos.h (nds32_output_call,
10881 nds32_output_return): Declare.
10882 * config/nds32/nds32.md: Refine all the call and return patterns.
10884 2018-04-06 Jakub Jelinek <jakub@redhat.com>
10887 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
10888 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
10890 PR rtl-optimization/84872
10891 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
10892 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
10893 EDGE_CROSSING edge.
10895 2018-04-06 Tamar Christina <tamar.christina@arm.com>
10897 * expr.c (copy_blkmode_to_reg): Revert 254862.
10898 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
10900 2018-04-06 Richard Biener <rguenther@suse.de>
10902 PR middle-end/85244
10903 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
10904 after seeing a component reference with an adjacent field. Treat
10905 refs to arrays at struct end of external decls similar to
10906 refs to unconstrained commons.
10908 2018-04-06 Jakub Jelinek <jakub@redhat.com>
10911 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
10912 look through SAVE_EXPRs with non-side-effects argument. Adjust
10914 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
10917 2018-04-06 Richard Biener <rguenther@suse.de>
10919 PR middle-end/85180
10920 * alias.c (find_base_term): New wrapper around find_base_term
10921 unwinding CSELIB_VAL_PTR changes.
10922 (find_base_term): Do not restore CSELIB_VAL_PTR during the
10925 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10927 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
10929 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
10930 constant definitions.
10931 ("nop"): lr 0,0 -> nopr r0
10932 ("nop_lr0", "nop_lr1"): New insn definitions.
10934 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10936 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
10937 NDS32_V3PUSH_AVAILABLE_P macro.
10939 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
10940 Chung-Ju Wu <jasonwucj@gmail.com>
10942 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
10943 (nds32*-*-*): Add float and fpu_config into supported_defaults.
10944 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
10945 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
10946 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
10947 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
10948 * config/nds32/constraints.md: New constraints and checking for hard
10949 float configuration.
10950 * config/nds32/iterators.md: New mode iterator and attribute for hard
10951 float configuration.
10952 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
10954 * config/nds32/nds32-fpu.md: New file.
10955 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
10956 deal with hard float code generation.
10957 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
10959 (abi_type, float_reg_number): New enum type.
10960 * config/nds32/nds32-predicates.c: New predicates for hard float.
10961 * config/nds32/nds32-protos.h: Declare functions for hard float.
10962 * config/nds32/nds32.c: Implementation for hard float configuration.
10963 * config/nds32/nds32.h: Definitions for hard float configuration.
10964 * config/nds32/nds32.md: Include hard float machine description and
10965 modify patterns for hard float configuration.
10966 * config/nds32/nds32.opt: New options for hard float configuration.
10967 * config/nds32/predicates.md: New predicates for hard float
10970 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10972 * common/config/nds32/nds32-common.c
10973 (nds32_option_optimization_table): Enable -mreleax-hint by default.
10975 2018-04-05 Jakub Jelinek <jakub@redhat.com>
10977 PR middle-end/85195
10978 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
10979 CONSTRUCTOR_ELT (ctor, ...)->value.
10981 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
10984 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
10986 2018-04-05 Tom de Vries <tom@codesourcery.com>
10989 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
10992 2018-04-05 Shiva Chen <shiva0217@gmail.com>
10993 Kito Cheng <kito.cheng@gmail.com>
10995 * config/nds32/constraints.md (U33): Fine-tune checking condition.
10996 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
10997 * config/nds32/nds32.h (nds32_16bit_address_type): Add
10998 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
11000 2018-04-05 Shiva Chen <shiva0217@gmail.com>
11001 Kito Cheng <kito.cheng@gmail.com>
11003 * config/nds32/constraints.md (Ufe): New memory constraint.
11004 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
11005 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
11006 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
11008 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
11009 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
11011 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11013 * config/nds32/nds32.md: Use optimize_size in the condition for
11014 alu-shift instructions.
11016 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11018 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
11020 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11022 * config/nds32/nds32.md (negsi2): Refine pattern.
11024 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
11025 Chung-Ju Wu <jasonwucj@gmail.com>
11027 * config/nds32/iterators.md (shift_rotate): New code iterator.
11028 (shift): New code attribute.
11029 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
11030 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
11031 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
11032 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
11033 bit-wise operations.
11034 (andsi3, *andsi3): Ditto.
11035 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
11036 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
11037 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
11038 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
11039 nds32_ior_operand, nds32_xor_operand): New predicates.
11041 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11043 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
11044 (addsi3, subsi3): ... this.
11046 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11048 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
11050 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11052 * config/nds32/nds32.md: Adjust indention.
11054 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
11056 * config/nds32/nds32.md (feature): New attribute.
11058 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11060 * config/nds32/nds32.md (subtype): New attribute.
11062 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
11065 * config/arm/arm-builtins.c (arm_expand_builtin): Change
11066 expansion to perform a bitwise AND of the argument followed by a
11067 boolean negation of the result.
11069 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
11071 PR rtl-optimization/84878
11072 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
11073 the basic block. Assert the use reference is not artificial and that
11074 it has an associated insn.
11076 2018-04-04 Michael Matz <matz@suse.de>
11078 * builtins.c (compute_objsize): Pass correct operand
11079 to array_at_struct_end_p.
11081 2018-04-04 Richard Biener <rguenther@suse.de>
11084 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
11085 from contexts for DINFO_LEVEL_TERSE and below.
11087 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11089 * config/nds32/nds32-doubleword.md (move_<mode>): Require
11090 resiter_operand condition.
11091 * config/nds32/nds32.md (*move<mode>): Ditto.
11093 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11094 Monk Chiang <sh.chiang04@gmail.com>
11096 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
11098 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11100 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
11102 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11103 Kito Cheng <kito.cheng@gmail.com>
11105 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
11106 nds32_cond_code_str, output_cond_branch,
11107 output_cond_branch_compare_zero, nds32_expand_cbranch,
11108 nds32_expand_cstore, nds32_expand_movcc,
11109 nds32_output_cbranchsi4_equality_zero,
11110 nds32_output_cbranchsi4_equality_reg,
11111 nds32_output_cbranchsi4_equality_reg_or_const_int,
11112 nds32_output_cbranchsi4_greater_less_zero: New functions.
11113 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
11114 nds32_expand_cstore, nds32_expand_movcc,
11115 nds32_output_cbranchsi4_equality_zero,
11116 nds32_output_cbranchsi4_equality_reg,
11117 nds32_output_cbranchsi4_equality_reg_or_const_int,
11118 nds32_output_cbranchsi4_greater_less_zero): Declare.
11119 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
11120 nds32_rimm11s_operand): New predicates.
11121 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
11122 * config/nds32/nds32.md: Rewrite all the branch and conditional move
11125 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11127 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
11128 * config/nds32/nds32.md: Ditto.
11129 * config/nds32/pipelines.md: Ditto.
11131 2018-04-04 Richard Biener <rguenther@suse.de>
11133 PR tree-optimization/85168
11134 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
11135 propagating abnormals.
11137 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11139 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
11141 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11142 Kito Cheng <kito.cheng@gmail.com>
11144 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
11145 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
11146 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
11147 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
11148 * config/nds32/nds32.md (sibcall_internal): New.
11149 (sibcall_register): Remove.
11150 (sibcall_immediate): Remove.
11151 (sibcall_value_internal): New.
11152 (sibcall_value_register): Remove.
11153 (sibcall_value_immediate): Remove.
11154 * config/nds32/predicates.md (nds32_general_register_operand): New.
11155 (nds32_call_address_operand): New.
11157 2018-04-03 Jakub Jelinek <jakub@redhat.com>
11159 PR rtl-optimization/85167
11160 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
11161 bb_defs if *split_p, instead preinitialize it to NULL.
11163 PR tree-optimization/85156
11164 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
11165 evaluating the argument multiple times.
11167 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
11169 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
11171 (_mm_cvtpd_ps): Likewise.
11172 (_mm_cvttpd_epi32): Likewise.
11173 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
11174 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
11175 vector, pixel, and bool following altivec.h include.
11177 2018-04-03 Martin Sebor <msebor@redhat.com>
11179 * doc/extend.texi (Common Function Attributes): Clarify.
11180 (const attribute): Likewise.
11181 (pure attribute): Likewise.
11183 2018-04-03 Jakub Jelinek <jakub@redhat.com>
11186 * config/i386/i386.c (ix86_expand_vector_set): Use
11187 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
11189 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
11191 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
11192 instructions when changing rounding bits to preserve precision bits
11193 in the x87 control word.
11195 2018-04-03 Martin Liska <mliska@suse.cz>
11197 PR tree-optimization/82491
11198 * rtl.h (strip_offset_and_add): Replace += suboffset with
11199 poly_uint64 () + suboffset.
11201 2018-03-29 Martin Liska <mliska@suse.cz>
11202 Martin Jambor <mjambor@suse.cz>
11205 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
11206 param_type is not an integral or pointer type.
11208 2018-04-03 Richard Biener <rguenther@suse.de>
11210 * sese.h (recompute_all_dominators): Remove.
11212 2018-04-02 Martin Sebor <msebor@redhat.com>
11214 * doc/invoke.texi (-Wrestrict): Fix typos.
11216 2018-04-02 Jim Wilson <jimw@sifive.com>
11218 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
11219 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
11220 (<optab>di3, <optab>si3_extend): Likewise.
11221 (<optab>si3_mask, <optab>si3_mask_1): New.
11222 (<optab>di3_mask, <optab>di3_mask_1): New.
11223 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
11224 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
11225 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
11227 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
11229 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
11232 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
11234 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
11235 (nds32_canonicalize_comparison): New function.
11237 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
11238 Kito Cheng <kito.cheng@gmail.com>
11239 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11241 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
11242 * config/nds32/constants.md (unspec_volatile_element): Add
11243 UNSPEC_VOLATILE_RELAX_GROUP.
11244 * config/nds32/nds32-relax-opt.c: New file.
11245 * config/nds32/nds32-predicates.c
11246 (nds32_symbol_load_store_p): New function.
11247 * config/nds32/nds32-protos.h
11248 (nds32_symbol_load_store_p): Declare function.
11249 (make_pass_nds32_relax_opt): Declare new rtl pass function.
11250 * config/nds32/nds32.c
11251 (nds32_register_pass): New function to register pass.
11252 (nds32_register_passes): New function to register passes.
11253 * config/nds32/nds32.md (relax_group): New pattern.
11254 * config/nds32/nds32.opt (mrelax-hint): New option.
11255 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
11257 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
11259 * config/nds32/t-nds32: Modify files dependency.
11261 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
11263 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
11264 (PROFILE_HOOK): Define its implementation.
11266 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
11268 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
11269 type and 32-bit size.
11271 2018-04-01 Jakub Jelinek <jakub@redhat.com>
11273 PR middle-end/85090
11274 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
11275 (V_128_256): New mode iterator.
11276 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
11277 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
11278 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
11280 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
11281 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
11283 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
11286 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
11287 NaN inputs correctly.
11289 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
11292 * config/rs6000/vsx.md (??r): New mode attribute.
11293 (*vsx_mov<mode>_64bit): Use it.
11294 (*vsx_mov<mode>_32bit): Likewise.
11296 2018-03-30 Martin Sebor <msebor@redhat.com>
11298 PR tree-optimization/84818
11299 * builtins.c (check_access): Use warning_n.
11301 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11304 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
11306 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
11309 2018-03-30 Julia Koval <julia.koval@intel.com>
11312 * x86-tune.def (movx, partial_reg_dependency): Enable for
11315 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
11317 PR inline-asm/84985
11318 * lra-constraints.c (process_alt_operands): Move setting
11319 this_alternative_matches below.
11321 2018-03-29 Martin Liska <mliska@suse.cz>
11324 * doc/invoke.texi: Document how LTO works with debug info.
11325 Describe auto-load support of binutils. Mention 'x86-64'
11326 as valid option value of -march option.
11328 2018-03-29 Jakub Jelinek <jakub@redhat.com>
11330 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
11333 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
11334 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
11335 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
11338 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
11341 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
11342 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
11343 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
11344 for RS6000_BTM_POWERPC64.
11345 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
11346 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
11347 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
11350 (DIVDEU): Likewise.
11352 2018-03-28 Carl Love <cel@us.ibm.com>
11355 2017-09-27 Carl Love <cel@us.ibm.com>
11357 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
11358 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
11359 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
11362 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11364 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
11365 instead of __vector bool.
11366 (_mm_max_pu8): Likewise.
11367 (_mm_min_pi16): Likewise.
11369 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
11372 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
11373 (DIVWEUO): Likewise.
11374 (DIVDEO): Likewise.
11375 (DIVDEUO): Likewise.
11376 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
11377 DIVWEUO and DIVDEUO.
11378 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
11379 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
11380 (div_extend): Likewise.
11381 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
11383 (__builtin_divweuo): Likewise.
11384 (__builtin_divdeo): Likewise.
11385 (__builtin_divdeuo): Likewise.
11387 2018-03-28 Jakub Jelinek <jakub@redhat.com>
11390 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
11391 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
11393 PR tree-optimization/82004
11394 * gimple-match-head.c (optimize_pow_to_exp): New function.
11395 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
11396 Don't fold to exp if optimize_pow_to_exp is false.
11398 2018-03-28 Martin Liska <mliska@suse.cz>
11401 * calls.c (initialize_argument_information): Fix trailing space.
11402 * common.opt: Fix typo and provide better explanation for
11403 -fsanitize-coverage option.
11404 * config/i386/i386.opt: Fix typo.
11406 2018-03-28 Jakub Jelinek <jakub@redhat.com>
11407 Martin Liska <mliska@suse.cz>
11410 * gimplify.c (asan_poison_variable): Don't do the check for
11411 gimplify_omp_ctxp here.
11412 (gimplify_decl_expr): Do it here.
11413 (gimplify_target_expr): Likewise.
11415 2018-03-28 Martin Liska <mliska@suse.cz>
11418 * config/i386/i386.c (ix86_function_arg_advance): Do not call
11419 chkp_type_bounds_count if MPX is not enabled.
11421 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
11423 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
11425 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
11428 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
11429 function to create the function decl for complex long double
11430 multiply and divide for -mabi=ieeelongdouble.
11431 (init_float128_ieee): Call it.
11433 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
11436 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
11437 -fcf-protection=branch -mibt.
11438 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
11440 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11443 * config/arm/arm.c (arm_valid_symbolic_address): Handle
11444 arm_word_relocations.
11446 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
11449 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
11450 extern array declarations.
11452 2018-03-27 Richard Biener <rguenther@suse.de>
11454 PR middle-end/84067
11455 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
11456 explicit single_use checks.
11458 2018-03-27 Richard Biener <rguenther@suse.de>
11460 PR tree-optimization/85082
11461 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
11464 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11466 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
11467 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
11468 Turn on fasynchronous-unwind-tables and funwind-tables.
11470 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
11473 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
11474 (*bmi_blsr_<mode>_ccz): Ditto.
11476 2018-03-26 Tom de Vries <tom@codesourcery.com>
11478 PR tree-optimization/85063
11479 * omp-general.c (offloading_function_p): New function. Factor out
11481 * omp-offload.c (pass_omp_target_link::gate): ... here.
11482 * omp-general.h (offloading_function_p): Declare.
11483 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
11484 with attribute omp declare target for offloading functions.
11486 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
11488 PR tree-optimization/84005
11489 * tree-data-ref.h (get_base_for_alignment): Declare.
11490 * tree-data-ref.c (get_base_for_alignment_1): New function.
11491 (get_base_for_alignment): Likewise.
11492 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
11493 get_base_for_alignment to find a suitable base object, instead
11494 of always using drb->base_address.
11496 2018-03-23 Jakub Jelinek <jakub@redhat.com>
11498 PR inline-asm/85022
11499 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
11500 known size by default.
11502 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
11504 PR inline-asm/85030
11505 * lra-constraints.c (process_alt_operands): Don't match BLKmode
11506 and non BLKmode operands.
11508 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11511 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
11512 Clean up attributes.
11514 2018-03-23 Richard Biener <rguenther@suse.de>
11517 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
11518 we are going to emit early debug for LTO.
11520 2018-03-23 Jakub Jelinek <jakub@redhat.com>
11522 PR inline-asm/85034
11523 * function.c (match_asm_constraints_1): Don't optimize if input
11524 doesn't satisfy general_operand predicate for output's mode.
11526 PR inline-asm/85022
11527 * alias.c (write_dependence_p): Don't require for x_canonicalized
11528 non-VOIDmode if x has VOIDmode.
11531 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
11532 just don't try to optimize it rather than assert it never happens.
11534 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
11536 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
11537 macro expansions for definition of ST_INTERNAL_<mode> and
11538 LD_INTERNAL_<mode> builtins.
11539 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
11541 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
11543 (altivec_expand_st_builtin): Likewise.
11544 (altivec_expand_builtin): Remove calls to deleted functions.
11545 (rs6000_address_for_altivec): Delete this function.
11546 * config/rs6000/vector.md: Remove expands for
11547 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
11549 2018-03-22 Sudakshina Das <sudi.das@arm.com>
11552 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
11553 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
11554 re-computing once computed.
11555 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
11556 (arm_init_machine_status): Initialize
11557 machine->static_chain_stack_bytes.
11559 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
11562 * doc/extend.texi: Add four new prototypes for vec_ld.
11563 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
11564 definitions for more logical presentation.
11565 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
11566 entries for V1TI variants of __builtin_altivec_ld builtin.
11567 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
11568 handling of V1TI variant of LVX icode pattern.
11569 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
11570 (rs6000_gimple_fold_builtin): Likewise.
11571 (altivec_init_builtins): Add code to define
11572 __builtin_altivec_lvx_v1ti function.
11574 2018-03-22 Jakub Jelinek <jakub@redhat.com>
11576 PR inline-asm/84941
11577 * function.c (match_asm_constraints_1): Don't do the optimization
11578 if input isn't a REG, SUBREG, MEM or constant.
11580 2018-03-22 Tom de Vries <tom@codesourcery.com>
11582 PR tree-optimization/84956
11583 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
11584 bb_has_abnormal_pred.
11586 2018-03-22 Jakub Jelinek <jakub@redhat.com>
11589 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
11590 DECL_INITIAL (decl) to decl at the end.
11591 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
11592 adjust the comment.
11594 2018-03-21 Joseph Myers <joseph@codesourcery.com>
11596 * doc/extend.texi (__builtin_tgmath): Document when complex
11597 integer types are treated as _Complex _Float64.
11599 2018-03-21 Tom de Vries <tom@codesourcery.com>
11601 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
11603 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11605 PR tree-optimization/84960
11606 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
11607 if it is ENTRY block, move them into single succ of ENTRY in that case.
11609 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
11611 PR tree-optimization/84811
11612 * poly-int.h (poly_span_traits): Remove the T3 parameter and
11613 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
11614 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
11615 (known_subrange_p): Update accordingly. Cast each value involved
11616 in the size comparison, rather than casting the result of the
11619 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11621 PR tree-optimization/84982
11622 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
11623 by flipping the least significant bit rather than all bits from
11624 bitpos to bitpos + bitsize - 1.
11626 2018-03-21 Nathan Sidwell <nathan@acm.org>
11628 * doc/extend.texi (Deprecated Features): Remove mention of
11629 long-deleted deprecations.
11631 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11634 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
11635 * configure: Regenerate.
11637 2018-03-21 Tom de Vries <tom@codesourcery.com>
11639 PR tree-optimization/83126
11640 * tree-parloops.c (num_phis): New function.
11641 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
11643 2018-03-21 Nathan Sidwell <nathan@acm.org>
11645 * doc/extend.texi (Deprecated Features): Update deprecated flags,
11646 mention anon-struct/union members and trailing attributes.
11648 2018-03-21 Bin Cheng <bin.cheng@arm.com>
11650 PR tree-optimization/84969
11651 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
11652 builtin memset partitions if they set different rhs values.
11654 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11656 PR rtl-optimization/84989
11657 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
11658 VEC_DUPLICATE with scalar result mode.
11660 2018-03-21 Martin Liska <mliska@suse.cz>
11663 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
11664 not intended return statement.
11666 2018-03-21 Martin Liska <mliska@suse.cz>
11669 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
11670 (chkp_find_bound_slots_1): Limit number of iterations.
11672 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
11675 * Minor grammar fixes for x86 options.
11677 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11680 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
11681 holding REG_CFA_RESTORE notes, instead turn them into a USE.
11683 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
11686 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
11687 (altivec_lvx_<mode>_1op): Likewise.
11688 (altivec_stvx_<mode>_2op): Likewise.
11689 (altivec_stvx_<mode>_1op): Likewise.
11690 (altivec_lvx_<VM2:mode>): New define_expand.
11691 (altivec_stvx_<VM2:mode>): Likewise.
11692 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
11693 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
11694 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
11695 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
11696 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
11697 (rs6000_gen_lvx): Likewise.
11698 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
11699 (altivec_expand_stv_builtin): Likewise.
11700 (altivec_expand_builtin): Likewise.
11701 * config/rs6000/vector.md: Likewise.
11703 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11706 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
11709 2018-03-20 Richard Biener <rguenther@suse.de>
11712 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
11713 sign-conversions as zero, fall back to standard scalar_stmt
11716 2018-03-20 Martin Liska <mliska@suse.cz>
11719 * predict.c (rebuild_frequencies): Handle case when we have
11720 PROFILE_ABSENT, but flag_guess_branch_prob is false.
11722 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11725 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
11726 flag_section_anchors.
11727 * varasm.c (use_blocks_for_decl_p): Remove hack for
11728 dw2_force_const_mem.
11731 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
11733 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
11734 be created, use lowpart_subreg of operands[0] rather than operands[0]
11736 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
11737 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
11738 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
11739 and n constraint instead of aarch64_shift_imm_di and Usd.
11740 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
11741 (*aarch64_<optab>_reg_minus<mode>3): ... this.
11743 2018-03-20 Sudakshina Das <sudi.das@arm.com>
11746 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
11747 to favor GPR over NEON registers.
11748 (<shift>di3_neon): Likewise.
11750 2018-03-20 Tom de Vries <tom@codesourcery.com>
11753 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
11754 (nvptx_process_pars): Emit bar.sync asap and alap.
11756 2018-03-20 Tom de Vries <tom@codesourcery.com>
11759 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
11760 seen_label if seen_label is already set.
11762 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11765 * config/i386/i386.c (fold_builtin_cpu): For features above 31
11766 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
11767 Use 1U instead of 1. Formatting fixes.
11770 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
11771 instead of TREE_TYPE (s1) for the return value.
11773 2018-03-19 Jakub Jelinek <jakub@redhat.com>
11775 PR tree-optimization/84946
11776 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
11777 bitsize + bitsize in poly_uint64 rather than poly_int64.
11780 * dwarf2asm.c: Include fold-const.c.
11781 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
11782 of decl rather than decl itself.
11784 PR rtl-optimization/84643
11785 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
11787 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
11790 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
11791 calling assemble_variable.
11793 2018-03-19 Sudakshina Das <sudi.das@arm.com>
11796 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
11797 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
11799 2018-03-19 Jim Wilson <jimw@sifive.com>
11802 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
11803 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
11804 (riscv_first_stack_step): Likewise.
11805 (riscv_option_override): Use STACK_BOUNDARY instead of
11806 MIN_STACK_BOUNDARY.
11807 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
11808 MIN_STACK_BOUNDARY.
11809 (BIGGEST_ALIGNMENT): Set to 128.
11810 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
11811 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
11814 2018-03-19 Richard Biener <rguenther@suse.de>
11816 PR tree-optimization/84933
11817 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
11818 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
11820 2018-03-19 Richard Biener <rguenther@suse.de>
11822 PR tree-optimization/84859
11823 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
11824 (cond_if_else_store_replacement): Perform sinking operation on
11825 single-store BBs regardless of MAX_STORES_TO_SINK setting.
11826 Generalize what a BB with a single eligible store is.
11828 2018-03-19 Richard Biener <rguenther@suse.de>
11830 PR tree-optimization/84929
11831 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
11832 chrec_is_positive against non-chrec arg.
11834 2018-03-19 Tamar Christina <tamar.christina@arm.com>
11837 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
11839 2018-03-18 Martin Liska <mliska@suse.cz>
11841 PR rtl-optimization/84635
11842 * regrename.c (build_def_use): Use matches_mode only when
11845 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
11847 PR tree-optimization/84913
11848 * tree-vect-loop.c (vectorizable_reduction): Don't try to
11849 vectorize chains of COND_EXPRs.
11851 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11853 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
11855 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11857 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
11859 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11861 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
11863 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
11864 Kito Cheng <kito.cheng@gmail.com>
11866 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
11867 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
11868 (nds32_adjust_reg_alloc_order): New function.
11869 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
11871 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
11873 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
11874 nds32_print_operand, nds32_print_operand_address): Use
11875 HOST_WIDE_INT_PRINT_DEC instead.
11877 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
11879 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
11881 2018-03-17 Jakub Jelinek <jakub@redhat.com>
11884 * config/i386/i386.c (initial_ix86_tune_features,
11885 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
11886 unsigned long long.
11887 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
11888 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
11889 rather than 1u << ix86_tune. Formatting fix.
11890 (ix86_option_override_internal): Change ix86_arch_mask from
11891 unsigned int to unsigned HOST_WIDE_INT, initialize to
11892 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
11893 (ix86_function_specific_restore): Likewise.
11895 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11898 * postreload.c (reload_combine_recognize_pattern): Perform
11899 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
11900 truncate_int_for_mode the result for the destination's mode.
11903 * hsa-gen.c (mem_type_for_type): Fix comment typo.
11904 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
11906 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11909 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
11912 * lra-assigns.c (lra_split_hard_reg_for): Don't use
11913 regno_allocno_class_array and sorted_pseudos.
11914 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
11915 insns where regno is used.
11917 2018-03-16 Martin Liska <mliska@suse.cz>
11920 * multiple_target.c (create_dispatcher_calls): Redirect
11921 reference in the symbol table.
11923 2018-03-16 Martin Liska <mliska@suse.cz>
11926 * multiple_target.c (create_dispatcher_calls): Redirect also
11929 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11933 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
11934 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
11935 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
11937 2018-03-16 Julia Koval <julia.koval@intel.com>
11939 * doc/invoke.texi (Skylake Server): Add CLWB.
11940 Cannonlake): Remove CLWB.
11942 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11944 PR tree-optimization/84841
11945 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
11947 (FLOAT_ONE_CONST_TYPE): Define.
11948 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
11949 (sort_by_operand_rank): Put entries with higher constant_type last
11950 rather than first to match comments.
11952 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
11954 * config/nios2/nios2.md (movsi_internal): Fix thinko in
11957 2018-03-15 Jakub Jelinek <jakub@redhat.com>
11960 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
11961 check and use address of target always.
11963 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
11966 * config/i386/i386.c (indirect_thunk_needed): Update comments.
11967 (indirect_thunk_bnd_needed): Likewise.
11968 (indirect_thunks_used): Likewise.
11969 (indirect_thunks_bnd_used): Likewise.
11970 (indirect_return_needed): New.
11971 (indirect_return_bnd_needed): Likewise.
11972 (output_indirect_thunk_function): Add a bool argument for
11974 (output_indirect_thunk_function): Don't generate alias for
11975 function return thunk.
11976 (ix86_code_end): Call output_indirect_thunk_function to generate
11977 function return thunks.
11978 (ix86_output_function_return): Set indirect_return_bnd_needed
11979 and indirect_return_needed instead of indirect_thunk_bnd_needed
11980 and indirect_thunk_needed.
11982 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
11984 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
11985 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
11986 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
11988 2018-03-15 David Malcolm <dmalcolm@redhat.com>
11989 Paul Hua <paul.hua.gm@gmail.com>
11992 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
11994 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
11996 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
11997 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
11998 resp. SFmode cases.
12000 2018-03-15 Tamar Christina <tamar.christina@arm.com>
12003 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
12004 instead of GET_MODE_SIZE when comparing Units.
12006 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
12009 * varasm.c (hash_section): Return an unchangeble hash value
12010 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
12011 Return !aarch64_can_use_per_function_literal_pools_p ().
12013 2018-03-15 Jakub Jelinek <jakub@redhat.com>
12016 * optabs.c (emit_conditional_move): Pass address of cmode's copy
12017 rather than address of cmode as last argument to prepare_cmp_insn.
12019 2018-03-15 Julia Koval <julia.koval@intel.com>
12021 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
12022 F_AVX512VNNI, F_AVX512BITALG): New.
12024 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
12027 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
12028 insn for floating-point loads and stores.
12030 2018-03-14 Carl Love <cel@us.ibm.com>
12032 * config/rs6000/rs6000-c.c: Add macro definitions for
12033 ALTIVEC_BUILTIN_VEC_PERMXOR.
12034 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
12035 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
12036 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
12037 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
12039 * config/doc/extend.texi: Add prototypes for vec_permxor.
12041 2018-03-14 David Malcolm <dmalcolm@redhat.com>
12044 * diagnostic-show-locus.c (class layout_point): Convert m_line
12045 from int to linenum_type.
12046 (line_span::comparator): Use linenum "compare" function when
12047 comparing line numbers.
12048 (test_line_span): New function.
12049 (layout_range::contains_point): Convert param "row" from int to
12051 (layout_range::intersects_line_p): Likewise.
12052 (layout::will_show_line_p): Likewise.
12053 (layout::print_source_line): Likewise.
12054 (layout::should_print_annotation_line_p): Likewise.
12055 (layout::print_annotation_line): Likewise.
12056 (layout::print_leading_fixits): Likewise.
12057 (layout::annotation_line_showed_range_p): Likewise.
12058 (struct line_corrections): Likewise for field m_row.
12059 (line_corrections::line_corrections): Likewise for param "row".
12060 (layout::print_trailing_fixits): Likewise.
12061 (layout::get_state_at_point): Likewise.
12062 (layout::get_x_bound_for_row): Likewise.
12063 (layout::print_line): Likewise.
12064 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
12065 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
12066 * input.c (selftest::test_linenum_comparisons): New function.
12067 (selftest::input_c_tests): Call it.
12068 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
12069 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
12070 * selftest.h (ASSERT_GT): New macro.
12071 (ASSERT_GT_AT): New macro.
12072 (ASSERT_LT): New macro.
12073 (ASSERT_LT_AT): New macro.
12075 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
12077 PR rtl-optimization/84780
12078 * combine.c (distribute_links): Don't make a link based on pc_rtx.
12080 2018-03-14 Martin Liska <mliska@suse.cz>
12082 * tree.c (record_node_allocation_statistics): Use
12083 get_stats_node_kind.
12084 (get_stats_node_kind): New function extracted from
12085 record_node_allocation_statistics.
12086 (free_node): Use get_stats_node_kind.
12088 2018-03-14 Richard Biener <rguenther@suse.de>
12090 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
12091 that the value-set of ANTIC_IN doesn't grow.
12094 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
12096 (BB_VISITED_WITH_VISITED_SUCCS): New define.
12097 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
12099 2018-03-14 Julia Koval <julia.koval@intel.com>
12101 * config.gcc (icelake-client, icelake-server): New.
12103 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
12104 (initial_ix86_arch_features): Ditto.
12105 (PTA_SKYLAKE): Add SGX.
12106 (PTA_ICELAKE): Remove.
12107 (PTA_ICELAKE_CLIENT): New.
12108 (PTA_ICELAKE_SERVER): New.
12109 (ix86_option_override_internal): Split up icelake on icelake client and
12111 (get_builtin_code_for_version): Ditto.
12112 (fold_builtin_cpu): Ditto.
12113 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
12114 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
12115 * config/i386/i386.h (processor_type): Ditto.
12116 * doc/invoke.texi: Ditto.
12118 2018-03-14 Jakub Jelinek <jakub@redhat.com>
12121 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
12122 INTEGER_CST offset, add it together with bitpos / 8 and
12123 sign extend based on POINTER_SIZE.
12127 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
12130 * config/i386/constraints.md (Yc): New register constraint.
12131 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
12132 Use Yc constraint for alternative 2 of operand 0. Remove
12133 preferred_for_speed attribute.
12135 2018-03-14 Richard Biener <rguenther@suse.de>
12137 PR tree-optimization/84830
12138 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
12139 with the old one to avoid oscillations.
12141 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
12144 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
12146 (assign_by_spills): Return a flag of reload assignment failure.
12147 Do not process the reload assignment failures. Do not spill other
12148 reload pseudos if they has the same reg class. Update n if
12150 (lra_assign): Add a return arg. Set up from the result of
12151 assign_by_spills call.
12152 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12153 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12154 usage_insns if it is not NULL.
12155 (spill_hard_reg_in_range): New function.
12156 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12157 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12158 function prototypes.
12159 (lra_assign): Change prototype.
12160 * lra.c (lra): Add code to deal with fails by splitting hard reg
12163 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
12165 * config/riscv/riscv.opt (mrelax): New option.
12166 * config/riscv/riscv.c (riscv_file_start): Emit ".option
12167 "norelax" when riscv_mrelax is disabled.
12168 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
12170 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12173 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
12174 reassociation for int modes.
12176 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12178 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
12179 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
12181 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
12182 * config/aarch64/aarch64-sve.md
12183 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
12184 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
12185 (*extend<mode><Vwide>2): Rename to...
12186 (aarch64_sve_extend<mode><Vwide>2): ...this.
12187 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
12188 renaming the old pattern to...
12189 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
12191 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
12192 define_expand, renaming the old pattern to...
12193 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
12194 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
12195 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
12196 account when deciding which SVE instruction the optab should use.
12197 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
12199 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12201 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
12202 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
12203 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
12204 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
12205 (tlsdesc_small_<mode>): Turn a define_expand and use
12206 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
12207 (tlsdesc_small_advsimd_<mode>): ...this.
12208 (tlsdesc_small_sve_<mode>): New pattern.
12210 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12212 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
12213 (UNSPEC_UMUL_HIGHPART): New constants.
12214 (MUL_HIGHPART): New int iteraor.
12215 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
12216 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
12218 (*<su>mul<mode>3_highpart): New define_insn.
12220 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
12223 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
12226 2018-03-13 Martin Liska <mliska@suse.cz>
12229 * (sem_item_optimizer::sem_item_optimizer): Initialize new
12231 (sem_item_optimizer::~sem_item_optimizer): Release it.
12232 (sem_item_optimizer::merge_classes): Register variable aliases.
12233 (sem_item_optimizer::fixup_pt_set): New function.
12234 (sem_item_optimizer::fixup_points_to_sets): Likewise.
12235 * ipa-icf.h: Declare new variables and functions.
12237 2018-03-13 Jakub Jelinek <jakub@redhat.com>
12239 PR middle-end/84834
12240 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
12241 integer_pow2p@2 and test integer_pow2p in condition.
12242 (A < 0 ? C : 0): Similarly for @1.
12244 PR middle-end/84831
12245 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
12246 characters starting at p contain '\0' character, don't look beyond
12250 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
12251 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
12254 * reg-stack.c (change_stack): Change update_end var from int to
12255 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
12256 also call set_block_for_insn on the newly added insns and rescan.
12259 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
12260 on the last operand.
12263 * tree.c (stabilize_reference_1): Return save_expr (e) for
12264 STATEMENT_LIST even if it doesn't have side-effects.
12266 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
12268 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
12270 2018-03-12 Renlin Li <renlin.li@arm.com>
12272 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
12273 aarch64_output_scalar_simd_mov_immediate.
12275 2018-03-12 Martin Sebor <msebor@redhat.com>
12277 PR tree-optimization/83456
12278 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
12279 for perfectly overlapping calls to memcpy.
12280 (gimple_fold_builtin_memory_chk): Same.
12281 (gimple_fold_builtin_strcpy): Handle no-warning.
12282 (gimple_fold_builtin_stxcpy_chk): Same.
12283 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
12285 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
12287 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
12288 parameter. Use it for SFmode.
12289 (rs6000_function_arg_advance_1): Adjust.
12290 (rs6000_function_arg): Adjust.
12291 (rs6000_gimplify_va_arg): Pass false for that new parameter.
12293 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
12295 PR rtl-optimization/84169
12296 PR rtl-optimization/84780
12297 * combine.c (can_combine_p): Check for a 2-insn combination whether
12298 the destination register is used between the two insns, too.
12300 2018-03-12 Richard Biener <rguenther@suse.de>
12302 PR tree-optimization/84803
12303 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
12304 for refs DR analysis didn't process.
12306 2018-03-12 Richard Biener <rguenther@suse.de>
12308 PR tree-optimization/84777
12309 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
12310 force-vectorize loops ignore whether we are optimizing for size.
12312 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
12314 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
12315 (TARGET_MD_ASM_ADJUST): Define.
12317 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
12318 Kito Cheng <kito.cheng@gmail.com>
12319 Chung-Ju Wu <jasonwucj@gmail.com>
12321 * config/nds32/nds32.c (nds32_compute_stack_frame,
12322 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
12323 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
12324 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
12325 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
12326 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
12327 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
12328 * config/nds32/nds32.md (prologue, epilogue): Use macro
12329 NDS32_V3PUSH_AVAILABLE_P to do checking.
12331 2018-03-11 Jakub Jelinek <jakub@redhat.com>
12334 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
12335 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
12336 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
12337 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
12338 addition of most attributes on !orig_type_die or the attribute not
12339 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
12341 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12342 Chung-Ju Wu <jasonwucj@gmail.com>
12344 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
12345 __NDS32_VH__ macro.
12346 * config/nds32/nds32.opt (mvh): New option.
12348 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12349 Chung-Ju Wu <jasonwucj@gmail.com>
12351 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
12353 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
12354 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
12357 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12358 Chung-Ju Wu <jasonwucj@gmail.com>
12360 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
12362 * config/nds32/nds32-multiple.md (strlensi): New pattern.
12363 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
12365 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
12366 Kito Cheng <kito.cheng@gmail.com>
12367 Chung-Ju Wu <jasonwucj@gmail.com>
12369 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
12370 UNSPEC_FFMISM and UNSPEC_FLMISM.
12371 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
12372 for ffb, ffmism and flmism.
12373 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
12374 (unspec_ffmism): Ditto.
12375 (unspec_flmism): Ditto.
12376 (nds32_expand_builtin_impl): Check if string extension is available.
12377 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
12378 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
12380 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
12383 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
12386 * lra-assigns.c (assign_by_spills): Return a flag of reload
12387 assignment failure. Do not process the reload assignment
12388 failures. Do not spill other reload pseudos if they has the same
12390 (lra_assign): Add a return arg. Set up from the result of
12391 assign_by_spills call.
12392 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12393 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12394 usage_insns if it is not NULL.
12395 (spill_hard_reg_in_range): New function.
12396 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12397 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12398 function prototypes.
12399 (lra_assign): Change prototype.
12400 * lra.c (lra): Add code to deal with fails by splitting hard reg
12403 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
12406 * config/i386/i386.opt: Replace Enforcment with Enforcement.
12408 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
12411 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
12412 (dw_val_node): Add val_symbolic_view.
12413 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
12414 (symview_upper_bound): New.
12415 (new_line_info_table): Initialize symviews_since_reset.
12416 (dwarf2out_source_line): Count symviews_since_reset and set
12417 symview_upper_bound.
12418 (dw_val_equal_p): Handle symview.
12419 (add_AT_symview): New.
12420 (print_dw_val): Handle symview.
12421 (attr_checksum, attr_checksum_ordered): Likewise.
12422 (same_dw_val_p, size_of_die): Likewise.
12423 (value_format, output_die): Likewise.
12424 (add_high_low_attributes): Use add_AT_symview for entry_view.
12425 (dwarf2out_finish): Reset symview_upper_bound, clear
12428 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
12431 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
12432 Add strict argument and use it.
12433 (rs6000_split_multireg_move): Update for new strict argument.
12434 (mem_operand_gpr): Disallow all non-offsettable addresses.
12435 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
12437 2018-03-09 Jakub Jelinek <jakub@redhat.com>
12440 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
12441 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
12442 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
12445 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
12446 decl, use remap_type if we want to use the type.
12448 2018-03-09 Martin Sebor <msebor@redhat.com>
12450 PR tree-optimization/84526
12451 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
12453 (builtin_access::generic_overlap): Be prepared to handle non-array
12456 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
12458 PR rtl-optimization/84682
12459 * lra-constraints.c (process_address_1): Check is_address flag
12460 for address constraints.
12461 (process_alt_operands): Likewise.
12462 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
12463 preprocess_constraints.
12464 * recog.h (preprocess_constraints): Add oploc parameter.
12466 * recog.c (preprocess_constraints): Test address_operand for
12467 CT_ADDRESS constraints.
12469 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
12472 * lra-assigns.c (assign_by_spills): Return a flag of reload
12473 assignment failure. Do not process the reload assignment
12474 failures. Do not spill other reload pseudos if they has the same
12476 (lra_assign): Add a return arg. Set up from the result of
12477 assign_by_spills call.
12478 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12479 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12480 usage_insns if it is not NULL.
12481 (spill_hard_reg_in_range): New function.
12482 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12483 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12484 function prototypes.
12485 (lra_assign): Change prototype.
12486 * lra.c (lra): Add code to deal with fails by splitting hard reg
12489 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12492 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
12493 Accept complain bool parameter. Only emit errors if it is true.
12494 (arm_parse_cpu_option_name): Likewise.
12495 (arm_target_thumb_only): Adjust callers of the above.
12496 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
12497 prototype to take a default true bool parameter.
12498 (arm_parse_arch_option_name): Likewise.
12500 2018-03-09 David Malcolm <dmalcolm@redhat.com>
12501 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
12505 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
12506 * configure: Regenerate.
12507 * configure.ac ("linker --version-script option"): New.
12508 ("linker soname option"): New.
12510 2018-03-09 Richard Biener <rguenther@suse.de>
12512 PR tree-optimization/84775
12513 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
12514 immediate uses of predicate stmts and mark them modified.
12517 PR tree-optimization/84178
12518 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
12520 (version_loop_for_if_conversion): Delay update_ssa call.
12521 (tree_if_conversion): Delay update_ssa until after predicate
12524 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
12527 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
12528 when the function accesses prior frames.
12530 2018-03-08 Jakub Jelinek <jakub@redhat.com>
12533 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
12534 gen_llsym, otherwise call maybe_gen_llsym.
12536 PR inline-asm/84742
12537 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
12538 has ',' character inside of it.
12540 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12543 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
12544 as clobbering CC_REGNUM.
12546 2018-03-08 Richard Biener <rguenther@suse.de>
12548 PR middle-end/84552
12549 * tree-scalar-evolution.c: Include tree-into-ssa.h.
12550 (follow_copies_to_constant): Do not follow SSA names registered
12553 2018-03-08 Richard Biener <rguenther@suse.de>
12555 PR tree-optimization/84178
12556 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
12558 (version_loop_for_if_conversion): Delay update_ssa call.
12559 (tree_if_conversion): Delay update_ssa until after predicate
12562 2018-03-08 David Malcolm <dmalcolm@redhat.com>
12564 PR tree-optimization/84178
12565 * tree-if-conv.c (release_bb_predicate): Remove the
12566 the assertion that the stmts have NULL use_ops.
12567 Discard the statements, asserting that they haven't
12568 yet been added to a BB.
12570 2018-03-08 Richard Biener <rguenther@suse.de>
12572 PR tree-optimization/84746
12573 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
12574 (phi_translate): Pass in destination ANTIC_OUT set.
12575 (phi_translate_1): Likewise. For a simplified result lookup
12576 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
12577 (phi_translate_set): Adjust.
12578 (do_pre_regular_insertion): Likewise.
12579 (do_pre_partial_partial_insertion): Likewise.
12581 2018-03-08 Martin Liska <mliska@suse.cz>
12583 PR gcov-profile/84735
12584 * doc/gcov.texi: Document usage of profile files.
12585 * gcov-io.h: Document changes in the format.
12587 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
12591 * dwarf2out.c (struct dw_line_info_table): Update comments for
12593 (FORCE_RESET_NEXT_VIEW): New.
12594 (FORCE_RESETTING_VIEW_P): New.
12595 (RESETTING_VIEW_P): Check for -1 too.
12596 (ZERO_VIEW_P): Likewise.
12597 (new_line_info_table): Force-reset next view.
12598 (dwarf2out_begin_function): Likewise.
12599 (dwarf2out_source_line): Simplify zero_view_p initialization.
12600 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
12601 view directly. Omit view when omitting .loc at line 0.
12603 2018-03-08 Jakub Jelinek <jakub@redhat.com>
12605 PR tree-optimization/84740
12606 * tree-switch-conversion.c (process_switch): Call build_constructors
12607 only if info.phi_count is non-zero.
12609 PR tree-optimization/84739
12610 * tree-tailcall.c (find_tail_calls): Check call arguments against
12611 DECL_ARGUMENTS (current_function_decl) rather than
12612 DECL_ARGUMENTS (func) when checking for tail recursion.
12614 2018-03-07 Jakub Jelinek <jakub@redhat.com>
12616 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
12617 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
12618 Volker Reichelt's entry and add entries for people that perform
12619 GCC fuzzy testing and report numerous bugs.
12621 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
12624 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
12625 readonly data in sdata, if that is disabled.
12626 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
12627 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
12628 -mreadonly-in-sdata option.
12630 2018-03-07 Martin Sebor <msebor@redhat.com>
12632 PR tree-optimization/84468
12633 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
12634 basic block when looking for nul assignment.
12636 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
12639 * except.h (output_function_exception_table): Adjust prototype.
12640 * except.c (output_function_exception_table): Remove FNNAME parameter
12641 and add SECTION parameter. Ouput one part of the table at a time.
12642 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
12643 the first part of the exception table and emit unwind directives.
12644 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
12645 (i386_pe_seh_cold_init): Likewise.
12646 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
12647 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
12648 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
12649 (ix86_output_call_insn): Emit a nop in one more case for SEH.
12650 * config/i386/winnt.c: Include except.h.
12651 (struct seh_frame_state): Add reg_offset, after_prologue and
12652 in_cold_section fields.
12653 (i386_pe_seh_end_prologue): Set seh->after_prologue.
12654 (i386_pe_seh_cold_init): New function.
12655 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
12656 to seh->in_cold_section.
12657 (seh_emit_push): Record the offset of the push.
12658 (seh_emit_save): Record the offet of the save.
12659 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
12660 Test seh->after_prologue to disregard the epilogue.
12661 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
12662 (i386_pe_end_cold_function): New function.
12664 2018-03-07 Jakub Jelinek <jakub@redhat.com>
12667 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
12668 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
12671 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
12673 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
12674 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
12676 PR middle-end/84723
12677 * multiple_target.c: Include tree-inline.h and intl.h.
12678 (expand_target_clones): Diagnose and fail if node->definition and
12679 !tree_versionable_function_p (node->decl).
12681 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
12683 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
12685 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
12686 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
12687 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
12689 2018-03-06 Jakub Jelinek <jakub@redhat.com>
12692 * combine.c (try_combine): Use reg_or_subregno instead of handling
12693 just paradoxical SUBREGs and REGs.
12695 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
12697 * config/arc/arc.c (arc_finalize_pic): Remove function.
12698 (arc_must_save_register): We use single base PIC register, remove
12699 checks to save/restore the PIC register.
12700 (arc_expand_prologue): Likewise.
12701 * config/arc/arc-protos.h (arc_set_default_type_attributes):
12703 (arc_verify_short): Likewise.
12704 (arc_attr_type): Likewise.
12705 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
12706 (walk_stores): Likewise.
12707 (arc_address_cost): Make it static.
12708 (arc_verify_short): Likewise.
12709 (branch_dest): Likewise.
12710 (arc_attr_type): Likewise.
12711 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
12712 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
12713 (arc_final_prescan_insn): Remove inserting the nops due to
12714 hardware hazards. It is done in reorg step.
12715 (insn_length_variant_t): Remove.
12716 (insn_length_parameters_t): Likewise.
12717 (arc_insn_length_parameters): Likewise.
12718 (arc_get_insn_variants): Likewise.
12719 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
12721 2018-03-06 Jakub Jelinek <jakub@redhat.com>
12723 PR inline-asm/84683
12724 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
12727 PR tree-optimization/84687
12728 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
12730 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
12732 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12734 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
12735 Rename to ppc_speculation_barrier.
12736 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
12737 __builtin_ppc_speculation_barrier.
12739 2018-03-05 Jakub Jelinek <jakub@redhat.com>
12742 * combine.c (combine_simplify_rtx): Don't try to simplify if
12743 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
12746 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
12748 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
12749 to 32 bytes when compiling for POWER9.
12751 2018-03-05 Jakub Jelinek <jakub@redhat.com>
12754 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
12755 regparm >= 3 with no arg reg available also for calls with
12756 flag_force_indirect_call. Pass decl to ix86_function_regparm.
12759 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
12761 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
12763 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
12766 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
12768 2018-03-05 Richard Biener <rguenther@suse.de>
12770 PR tree-optimization/84486
12771 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
12772 When inserting a __builtin_assume_aligned call set the LHS
12773 SSA name alignment info accordingly.
12775 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
12777 PR tree-optimization/84114
12778 * config/aarch64/aarch64.c (aarch64_reassociation_width)
12779 Avoid reassociation of FLOAT_MODE addition.
12781 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
12783 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
12784 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
12785 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
12786 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
12787 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
12788 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
12789 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
12791 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
12792 __builtin_ia32_wbinvd): New builtins.
12793 (SPECIAL_ARGS2): New.
12794 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
12795 (SPECIAL_ARGS2): New.
12796 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
12797 (ix86_valid_target_attribute_inner_p): Ditto.
12798 (ix86_init_mmx_sse_builtins): Add special_args2.
12799 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
12800 TARGET_WBNOINVD_P): New.
12801 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
12802 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
12803 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
12804 * config/i386/immintrin.h (_wbinvd): New intrinsic.
12805 * config/i386/pconfigintrin.h: New file.
12806 * config/i386/wbnoinvdintrin.h: Ditto.
12807 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
12809 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
12811 2018-03-05 Richard Biener <rguenther@suse.de>
12813 PR tree-optimization/84670
12814 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
12816 (BB_VISITED_WITH_VISITED_SUCCS): New define.
12817 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
12818 (compute_antic_aux): Only assert the number of values in ANTIC_IN
12819 doesn't grow if all successors (recursively) were visited at least
12822 2018-03-05 Richard Biener <rguenther@suse.de>
12824 PR tree-optimization/84650
12825 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
12826 if executed in the loop pipeline.
12828 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
12830 * doc/configfiles.texi (Configuration Files): Move info about
12831 conditionalizing $target-protos.h to...
12832 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
12833 differs from $target-protos.h.
12835 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
12836 Chung-Ju Wu <jasonwucj@gmail.com>
12838 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
12839 * config/nds32/nds32-multiple.md (setmemsi): Define.
12840 * config/nds32/nds32-memory-manipulation.c
12841 (nds32_gen_dup_4_byte_to_word_value): New.
12842 (emit_setmem_word_loop): New.
12843 (emit_setmem_byte_loop): New.
12844 (nds32_expand_setmem_loop): New.
12845 (nds32_expand_setmem_loop_v3m): New.
12846 (nds32_expand_setmem_unroll): New.
12847 (nds32_expand_setmem): New.
12849 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12850 Chung-Ju Wu <jasonwucj@gmail.com>
12852 * config/nds32/nds32-memory-manipulation.c
12853 (nds32_emit_load_store): New.
12854 (nds32_emit_post_inc_load_store): New.
12855 (nds32_emit_mem_move): New.
12856 (nds32_emit_mem_move_block): New.
12857 (nds32_expand_movmemsi_loop_unknown_size): New.
12858 (nds32_expand_movmemsi_loop_known_size): New.
12859 (nds32_expand_movmemsi_loop): New.
12860 (nds32_expand_movmemsi_unroll): New.
12861 (nds32_expand_movmemqi): Rename ...
12862 (nds32_expand_movmemsi): ... to this.
12863 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
12864 (movmemsi): ... to this.
12865 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
12866 (nds32_expand_movmemsi): ... to this.
12868 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12869 Monk Chiang <sh.chiang04@gmail.com>
12870 Chung-Ju Wu <jasonwucj@gmail.com>
12872 * config/nds32/nds32-protos.h
12873 (nds32_expand_load_multiple): New arguments.
12874 (nds32_expand_store_multiple): Ditto.
12875 (nds32_valid_multiple_load_store): Rename ...
12876 (nds32_valid_multiple_load_store_p): ... to this.
12877 * config/nds32/nds32-memory-manipulation.c
12878 (nds32_expand_load_multiple): Refine implementation.
12879 (nds32_expand_store_multiple): Ditto.
12880 * config/nds32/nds32-multiple.md
12881 (load_multiple): Update nds32_expand_load_multiple interface.
12882 (store_multiple): Update nds32_expand_store_multiple interface.
12883 * config/nds32/nds32-predicates.c
12884 (nds32_valid_multiple_load_store): Rename ...
12885 (nds32_valid_multiple_load_store_p): ... to this and refine
12887 * config/nds32/predicates.md
12888 (nds32_load_multiple_and_update_address_operation): New predicate.
12889 (nds32_store_multiple_and_update_address_operation): New predicate.
12891 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12892 Chung-Ju Wu <jasonwucj@gmail.com>
12894 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
12895 (combo): New attribute.
12896 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
12898 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
12900 * config/nds32/nds32.opt: Change -mcmodel= default value.
12902 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
12903 Monk Chiang <sh.chiang04@gmail.com>
12904 Chung-Ju Wu <jasonwucj@gmail.com>
12906 * config/nds32/constants.md (unspec_element): New enum.
12907 * config/nds32/constraints.md (Umw): New constraint.
12908 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
12909 * config/nds32/nds32-intrinsic.md: Likewise.
12910 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
12911 (nds32_valid_smw_lwm_base_p): New.
12912 (nds32_output_smw_single_word): New.
12913 (nds32_output_lmw_single_word): New.
12914 (nds32_expand_unaligned_load): New.
12915 (nds32_expand_unaligned_store): New.
12916 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
12917 (nds32_output_smw_single_word): Declare.
12918 (nds32_output_lmw_single_word): Declare.
12919 (nds32_expand_unaligned_load): Declare.
12920 (nds32_expand_unaligned_store): Declare.
12921 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
12922 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
12923 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
12924 NDS32_BUILTIN_UASTORE_DW.
12925 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
12928 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
12929 Kito Cheng <kito.cheng@gmail.com>
12930 Chung-Ju Wu <jasonwucj@gmail.com>
12932 * config/nds32/nds32-intrinsic.c
12933 (nds32_expand_builtin_null_ftype_reg): Delete.
12934 (nds32_expand_builtin_reg_ftype_imm): Ditto.
12935 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
12936 (nds32_read_argument): New.
12937 (nds32_legitimize_target): Ditto.
12938 (nds32_legitimize_argument): Ditto.
12939 (nds32_check_constant_argument): Ditto.
12940 (nds32_expand_unop_builtin): Ditto.
12941 (nds32_expand_unopimm_builtin): Ditto.
12942 (nds32_expand_binop_builtin): Ditto.
12943 (nds32_builtin_decl_impl): Ditto.
12944 (builtin_description): Ditto.
12945 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
12946 (nds32_init_builtins_impl): Ditto.
12947 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
12948 (nds32_builtin_decl): New.
12949 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
12950 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
12952 2018-03-02 Jeff Law <law@redhat.com>
12954 * reorg.c (stop_search_p): Handle DEBUG_INSN.
12955 (redundant_insn, fill_simple_delay_slots): Likewise.
12956 (fill_slots_from_thread): Likewise.
12957 * resource.c (mark_referenced_resources): Likewise.
12958 (mark_set_resources, find_dead_or_set_registers): Likewise.
12960 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12962 * substring-locations.h (format_warning_va): Formatting fix for
12963 ATTRIBUTE_GCC_DIAG.
12964 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
12966 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
12967 * substring-locations.c: Include intl.h.
12968 (format_warning_va): Turned into small wrapper around
12969 format_warning_n_va, renamed to ...
12970 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
12971 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
12973 (format_warning_at_substring_n): New function.
12974 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
12975 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
12976 format_warning_at_substring with just a shorter name instead of
12977 const function pointer.
12978 (fmtwarn_n): New function.
12979 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
12980 appropriate, get rid of all the fmtstr temporaries, move conditionals
12981 with G_() wrapped string literals directly into fmtwarn arguments,
12982 cast dir.len to (int), formatting fixes.
12984 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
12986 * doc/invoke.texi: Remove "Cilk Plus" references.
12988 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12989 Richard Biener <rguenther@suse.de>
12992 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
12993 for error or warning attributes if CALL_FROM_THUNK_P is set.
12996 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12999 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
13000 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
13003 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
13004 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
13006 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
13007 instead of -1U in last predictors element's probability member.
13009 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
13012 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
13013 arguments if they are comparable.
13015 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
13017 PR tree-optimization/84634
13018 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
13019 masks and masked_loop_p with a single loop_masks, making sure it's
13020 null for bb vectorization.
13022 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
13024 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
13025 (vect_analyze_data_ref_access): Use loop->safe_len rather than
13026 loop->force_vectorize to check whether there is no alias.
13028 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13031 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
13033 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
13035 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
13036 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
13037 instead of a loop around prev_real_insn.
13038 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
13041 PR inline-asm/84625
13042 * config/i386/i386.c (ix86_print_operand): Use conditional
13043 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
13046 2018-03-02 Richard Biener <rguenther@suse.de>
13048 PR tree-optimization/84427
13049 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
13050 (bitmap_set_subtract_values): Rewrite to handle multiple
13053 (prune_clobbered_mems): Likewise.
13054 (phi_translate): Take edge instead of pred/phiblock.
13055 (phi_translate_1): Likewise.
13056 (phi_translate_set): Likewise. Insert all translated
13057 exprs for a value into the set, keeping possibly multiple
13058 expressions per value.
13059 (compute_antic_aux): Adjust for phi_translate changes.
13060 When intersecting union the expressions and prune those
13061 not in the final value set, keeping possibly multiple
13062 expressions per value. Do not use value-insertion
13063 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
13064 all expressions. Add verification that the value-sets
13065 only shrink during iteration.
13066 (compute_partial_antic_aux): Adjust for the phi_translate changes.
13067 (do_pre_regular_insertion): Likewise.
13068 (do_pre_partial_partial_insertion): Likewise.
13070 2018-03-02 Richard Biener <rguenther@suse.de>
13073 * config/darwin.c (saved_debug_info_level): New static global.
13074 (darwin_asm_lto_start): Disable debug info generation for LTO out.
13075 (darwin_asm_lto_end): Restore debug info generation settings.
13077 2018-03-01 Martin Liska <mliska@suse.cz>
13080 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13081 volatile arguments.
13083 2018-03-01 Richard Biener <rguenther@suse.de>
13086 * dwarf2out.c (gen_variable_die): Properly handle late VLA
13087 type annotation with LTO when debug was disabled at compile-time.
13089 2018-03-01 Matthew Fortune <mfortune@gmail.com>
13091 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
13093 (mips_final_postscan_insn): Likewise.
13095 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
13097 PR rtl-optimization/84528
13098 * alias.c (init_alias_target): Add commentary.
13099 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
13100 a unique base value if the frame pointer is not eliminated
13101 to the stack pointer.
13103 2018-03-01 Tom de Vries <tom@codesourcery.com>
13105 PR rtl-optimization/83327
13106 * lra-int.h (hard_regs_spilled_into): Declare.
13107 * lra.c (hard_regs_spilled_into): Define.
13108 (init_reg_info): Init hard_regs_spilled_into.
13109 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
13110 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
13111 (process_bb_lives): Handle hard_regs_spilled_into.
13112 (lra_create_live_ranges_1): Before doing liveness propagation, clear
13113 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
13115 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
13117 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
13118 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
13119 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
13120 * config/rs6000/aix72.h: New file.
13122 2018-02-28 Jakub Jelinek <jakub@redhat.com>
13124 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
13125 instead of warning_at with conditional singular and plural messages
13129 * stor-layout.c (update_alignment_for_field): For
13130 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
13131 && !DECL_PACKED (field), do the alignment update, just use
13132 only desired_align instead of MAX (type_align, desired_align)
13134 (place_field): Don't do known_align < desired_align handling
13135 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
13136 is non-NULL, instead do it after rli->prev_field handling and
13137 only if not within a bitfield word. For DECL_PACKED (field)
13138 use type_align of BITS_PER_UNIT.
13140 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
13142 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
13143 superfluous parentheses and trailing spaces.
13145 2018-02-28 Richard Biener <rguenther@suse.de>
13147 PR tree-optimization/84584
13148 * graphite-scop-detection.c (scop_detection::add_scop): Discard
13149 SCoPs with fake exit edge.
13151 2018-02-28 Martin Liska <mliska@suse.cz>
13154 * timevar.c (timer::print): Fix format to properly print 100%
13157 2018-02-28 Richard Biener <rguenther@suse.de>
13159 PR middle-end/84607
13160 * genmatch.c (capture_info::walk_match): Do not mark
13161 captured expressions without operands as expr_p given
13162 they act more like predicates and should be subject to
13163 "lost tail" side-effect preserving.
13165 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
13167 PR rtl-optimization/81611
13168 * auto-inc-dec.c (attempt_change): Move dead note from
13169 mem_insn if it's the next use of regno
13170 (find_address): Take address use of reg holding
13171 non-incremented value. Add parm to limit search to the named
13173 (merge_in_block): Attempt to use a mem insn that is the next
13174 use of the original regno.
13176 2018-02-27 Martin Sebor <msebor@redhat.com>
13179 * doc/invoke.texi (-Wmissing-attributes): New option.
13180 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
13182 2018-02-27 Martin Sebor <msebor@redhat.com>
13184 PR translation/84207
13185 * diagnostic-core.h (warning_n, error_n, inform_n): Change
13186 n argument to unsigned HOST_WIDE_INT.
13187 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
13188 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
13189 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
13190 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
13192 2018-02-27 Richard Biener <rguenther@suse.de>
13194 PR tree-optimization/84512
13195 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
13196 Do not use the estimate returned from record_stmt_cost for
13197 the scalar iteration cost but sum properly using add_stmt_cost.
13199 2018-02-27 Richard Biener <rguenther@suse.de>
13201 PR tree-optimization/84466
13202 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
13203 Adjust last change to less strictly validate use operands.
13205 2018-02-27 Martin Liska <mliska@suse.cz>
13207 PR gcov-profile/84548
13208 * gcov.c (process_file): Allow partial overlap and consider it
13209 also as group functions.
13210 (output_lines): Properly calculate range of lines for a group.
13212 2018-02-27 Martin Liska <mliska@suse.cz>
13214 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
13215 'ggc' suffixes. Change first column width.
13216 (timer::print): Fix formatting of the column.
13218 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
13220 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
13221 preserve inline entry blocks for the sake of debug inline
13222 entry point markers alone.
13223 (remove_unused_locals): Suggest in comments a better place to
13224 force the preservation of inline entry blocks that are
13225 otherwise unused, but do not preserve them.
13227 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
13229 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
13231 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
13234 * config/i386/constraints.md (Bs): Replace
13235 ix86_indirect_branch_register with
13236 TARGET_INDIRECT_BRANCH_REGISTER.
13238 * config/i386/i386.md (indirect_jump): Likewise.
13239 (tablejump): Likewise.
13240 (*sibcall_memory): Likewise.
13241 (*sibcall_value_memory): Likewise.
13242 Peepholes of indirect call and jump via memory: Likewise.
13243 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
13244 (*sibcall_value_GOT_32): Likewise.
13245 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13246 (GOT_memory_operand): Likewise.
13247 (call_insn_operand): Likewise.
13248 (sibcall_insn_operand): Likewise.
13249 (GOT32_symbol_operand): Likewise.
13250 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
13252 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
13254 PR rtl-optimization/83496
13255 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
13256 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
13257 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
13258 redundant insn, if any.
13259 (relax_delay_slots): Likewise.
13260 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
13262 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
13264 PR tree-optimization/83965
13265 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
13266 that grouped statements are part of a reduction chain. Return
13267 true if the statement is not marked as a reduction itself but
13268 is part of a group.
13269 (vect_recog_dot_prod_pattern): Don't check whether the statement
13270 is part of a group here.
13271 (vect_recog_sad_pattern): Likewise.
13272 (vect_recog_widen_sum_pattern): Likewise.
13274 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
13277 * final.c (rest_of_clean_state): Also look for calls inside sequences.
13279 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
13282 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
13284 (ix86_output_indirect_function_return): New prototype.
13285 (ix86_split_simple_return_pop_internal): Likewise.
13286 * config/i386/i386.c (indirect_return_via_cx): New.
13287 (indirect_return_via_cx_bnd): Likewise.
13288 (indirect_thunk_name): Handle return va CX_REG.
13289 (output_indirect_thunk_function): Create alias for
13290 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
13291 (ix86_output_indirect_jmp): Remove the bool argument.
13292 (ix86_output_indirect_function_return): New function.
13293 (ix86_split_simple_return_pop_internal): Likewise.
13294 * config/i386/i386.md (*indirect_jump): Don't pass false
13295 to ix86_output_indirect_jmp.
13296 (*tablejump_1): Likewise.
13297 (simple_return_pop_internal): Change it to define_insn_and_split.
13298 Call ix86_split_simple_return_pop_internal to split it for
13299 -mfunction-return=.
13300 (simple_return_indirect_internal): Call
13301 ix86_output_indirect_function_return instead of
13302 ix86_output_indirect_jmp.
13304 2018-02-26 Jakub Jelinek <jakub@redhat.com>
13307 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
13308 memset and value initialization afterwards.
13310 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
13312 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
13314 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
13317 * common/config/aarch64/aarch64-common.c
13318 (aarch_option_optimization_table[]): Switch
13319 off fomit-frame-pointer
13321 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
13322 Chung-Ju Wu <jasonwucj@gmail.com>
13324 * config/nds32/nds32-multiple.md (load_multiple): Disallow
13326 (store_multiple): Ditto.
13328 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
13330 * config.gcc: Add --with-cpu support for nds32 target.
13331 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
13332 * config/nds32/nds32.opt: Add -mcpu= option.
13334 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
13336 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
13337 isel=yes): Warn for these deprecated options.
13339 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
13341 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
13342 ISA_2_5_MASKS_EMBEDDED.
13344 2018-02-23 Jakub Jelinek <jakub@redhat.com>
13346 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
13347 p->max as pointers rather than using iterative_hash_expr.
13349 2018-02-23 Carl Love <cel@us.ibm.com>
13351 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
13352 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
13354 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
13355 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
13356 P8V_BUILTIN_VEC_VUNSIGNED2.
13358 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
13361 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
13362 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
13364 * lra-constraints.c (curr_insn_transform): Set up
13365 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
13367 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
13368 * lra-eliminations.c (spill_pseudos): Ditto.
13369 (process_insn_for_elimination): Ditto.
13370 * lra-lives.c (reg_early_clobber_p): Use the new macros.
13371 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
13372 LRA_NON_CLOBBERED_ALT.
13374 2018-02-22 Martin Sebor <msebor@redhat.com>
13376 PR tree-optimization/84480
13377 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
13378 to maybe_diag_stxncpy_trunc. Call it.
13379 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
13380 from gimple_fold_builtin_strcpy. Print inlining stack.
13381 (handle_builtin_stxncpy): Print inlining stack.
13382 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
13384 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
13387 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
13388 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
13389 and -fcheck-pointer-bounds are used together.
13390 (indirect_thunk_prefix): New enum.
13391 (indirect_thunk_need_prefix): New function.
13392 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
13393 "_nt" instead of "_bnd" for NOTRACK prefix.
13394 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
13395 (output_indirect_thunk_function): Likewise.
13397 (ix86_code_end): Update output_indirect_thunk_function calls.
13398 (ix86_output_indirect_branch_via_reg): Replace
13399 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
13400 (ix86_output_indirect_branch_via_push): Likewise.
13401 (ix86_output_function_return): Likewise.
13402 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
13403 incompatible with -fcf-protection=branch and
13404 -fcheck-pointer-bounds.
13406 2018-02-22 Steve Ellcey <sellcey@cavium.com>
13409 * config/aarch64/aarch64.c (aarch64_print_address_internal):
13410 Change gcc_assert call to output_operand_lossage.
13412 2018-02-22 Steve Ellcey <sellcey@cavium.com>
13414 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
13416 2018-02-22 DJ Delorie <dj@redhat.com>
13417 Sebastian Perta <sebastian.perta@renesas.com>
13418 Oleg Endo <olegendo@gcc.gnu.org>
13420 * config/rx/rx.c (rx_rtx_costs): New function.
13421 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
13423 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
13425 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
13427 2018-02-22 Martin Liska <mliska@suse.cz>
13430 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
13431 Add "native" as a possible value.
13433 2018-02-22 Martin Liska <mliska@suse.cz>
13436 * config/i386/i386.c (ix86_option_override_internal):
13437 Add "native" as a possible value for -march and -mtune.
13439 2018-02-22 Jakub Jelinek <jakub@redhat.com>
13442 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
13443 to all type variants.
13445 PR tree-optimization/84503
13446 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
13447 width as info->bitpos + info->bitsize - start.
13448 (merged_store_group::merge_overlapping): Simplify width computation.
13449 (check_no_overlap): New function.
13450 (imm_store_chain_info::try_coalesce_bswap): Compute expected
13451 start + width and last_order of the group, fail if check_no_overlap
13453 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
13454 to group if check_no_overlap fails.
13456 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
13458 * config/rs6000/altivec.md: Delete contraint arguments to
13459 define_expand, define_split, and define_peephole2, and in
13460 define_insn_and_split if always unused.
13461 * config/rs6000/darwin.md: Ditto.
13462 * config/rs6000/dfp.md: Ditto.
13463 * config/rs6000/rs6000.md: Ditto.
13464 * config/rs6000/sync.md: Ditto.
13465 * config/rs6000/vector.md: Ditto.
13466 * config/rs6000/vsx.md: Ditto.
13468 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
13470 * config/rs6000/altivec.md: Write output control strings as braced
13471 blocks instead of double-quoted strings.
13472 * config/rs6000/darwin.md: Ditto.
13473 * config/rs6000/rs6000.md: Ditto.
13474 * config/rs6000/vector.md: Ditto.
13475 * config/rs6000/vsx.md: Ditto.
13477 2018-02-21 Jason Merrill <jason@redhat.com>
13479 PR c++/84314 - ICE with templates and fastcall attribute.
13480 * attribs.c (build_type_attribute_qual_variant): Remove assert.
13482 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
13484 * ipa-cp.c (determine_versionability): Fix comment typos.
13486 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
13489 * ipa-cp.c (determine_versionability): Do not version functions caling
13492 2018-02-21 Martin Liska <mliska@suse.cz>
13495 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
13496 Add "native" as a possible value.
13497 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
13498 the macro when native cpu detection is available.
13500 2018-02-21 Martin Liska <mliska@suse.cz>
13503 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
13504 Add "native" as a possible value.
13505 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
13506 when native cpu detection is available.
13508 2018-02-21 Jakub Jelinek <jakub@redhat.com>
13509 Martin Sebor <msebor@redhat.com>
13511 PR tree-optimization/84478
13512 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
13514 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
13515 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
13516 support which is conservatively correct, for 2 only stay conservative
13517 for maxlen. Formatting and comment capitalization fixes. Add STRICT
13518 argument to the 2 argument get_range_strlen, adjust 6 arg
13519 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
13521 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
13522 (gimple_fold_builtin_strlen): Pass true as last argument to
13525 2018-02-20 Martin Sebor <msebor@redhat.com>
13527 PR middle-end/84095
13528 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
13529 (builtin_memref::set_base_and_offset): Same. Handle inner references.
13530 (builtin_memref::builtin_memref): Factor out parts into
13531 set_base_and_offset and call it.
13533 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
13535 PR middle-end/84406
13536 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
13537 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
13538 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
13539 search at the associated MODE_INT.
13541 2018-02-20 Jeff Law <law@redhat.com>
13543 PR middle-end/82123
13544 PR tree-optimization/81592
13545 PR middle-end/79257
13546 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
13547 for range data rather than using global data.
13548 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
13549 range data rather than using global data.
13550 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
13551 pass it to children as needed.
13552 (struct directive::fmtresult): Similarly.
13553 (struct directive::set_width): Similarly.
13554 (struct directive::set_precision): Similarly.
13555 (format_integer, format_directive, parse_directive): Similarly.
13556 (format_none): Accept unnamed vr_values parameter.
13557 (format_percent, format_floating, format_character): Similarly.
13558 (format_string, format_plain): Similarly.
13559 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
13560 the EVRP range analyzer for range data rather than using global data.
13561 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
13562 gimple-ssa-evrp-analyze.h
13563 (class sprintf_dom_walker): Add after_dom_children member function.
13564 Add evrp_range_analyzer member.
13565 (sprintf_dom_walker::before_dom_children): Call into the EVRP
13566 range analyzer as needed.
13567 (sprintf_dom_walker::after_dom_children): New member function.
13568 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
13570 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
13571 (evrp_range_analyzer::pop_to_marker): Likewise.
13573 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
13575 PR tree-optimization/84419
13576 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
13577 with the required type if its current type is compatible but
13580 2018-02-20 Jakub Jelinek <jakub@redhat.com>
13582 PR middle-end/82004
13583 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
13584 after vectorization.
13586 2018-02-20 Martin Liska <mliska@suse.cz>
13589 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
13590 possible values if we don't have a hint.
13592 2018-02-20 Martin Liska <mliska@suse.cz>
13596 * final.c (shorten_branches): Build align_tab array with one
13598 * opts.c (finish_options): Add alignment option limit check.
13599 (MAX_CODE_ALIGN): Likewise.
13600 (MAX_CODE_ALIGN_VALUE): Likewise.
13601 * doc/invoke.texi: Document maximum allowed option value for
13602 all -falign-* options.
13604 2018-02-19 Jakub Jelinek <jakub@redhat.com>
13607 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
13608 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
13609 * var-tracking.c (emit_note_insn_var_location): Remove all references
13610 to NOTE_INSN_CALL_ARG_LOCATION.
13611 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
13612 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
13613 Use copy_rtx_if_shared.
13614 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
13615 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
13616 (dwarf2out_var_location): Remove handling of
13617 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
13619 * final.c (final_scan_insn): Remove all references to
13620 NOTE_INSN_CALL_ARG_LOCATION.
13621 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
13622 before dumping final insns.
13623 * except.c (emit_note_eh_region_end): Remove all references to
13624 NOTE_INSN_CALL_ARG_LOCATION.
13625 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
13626 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
13627 * config/arc/arc.c (hwloop_optimize): Likewise.
13628 * config/arm/arm.c (create_fix_barrier): Likewise.
13629 * config/s390/s390.c (s390_chunkify_start): Likewise.
13630 * config/sh/sh.c (find_barrier): Likewise.
13631 * config/i386/i386.c (rest_of_insert_endbranch,
13632 ix86_seh_fixup_eh_fallthru): Likewise.
13633 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
13634 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
13635 * config/frv/frv.c (frv_function_prologue): Likewise.
13636 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
13638 (note_outside_basic_block_p): Remove all references to
13639 NOTE_INSN_CALL_ARG_LOCATION.
13640 * gengtype.c (adjust_field_rtx_def): Likewise.
13641 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
13643 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
13644 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
13647 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
13650 PR tree-optimization/84452
13651 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
13652 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
13655 2018-02-19 Martin Liska <mliska@suse.cz>
13658 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
13660 2018-02-19 Martin Liska <mliska@suse.cz>
13661 Richard Sandiford <richard.sandiford@linaro.org>
13663 PR tree-optimization/82491
13664 * gimple-fold.c (get_base_constructor): Make earlier bail out
13667 2018-02-19 Carl Love <cel@us.ibm.com>
13669 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
13670 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
13672 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
13673 P8V_BUILTIN_VEC_NEG.
13675 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
13677 * config/rl78/rl78.md (movdf): New define expand.
13679 2018-02-19 Martin Liska <mliska@suse.cz>
13682 * doc/invoke.texi: Fix typo.
13683 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
13685 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
13687 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
13688 handle rs6000_single_float and rs6000_double_float specially for
13691 2018-02-16 Jeff Law <law@redhat.com>
13693 * config/rx/rx.c (add_pop_cfi_notes): New function.;
13694 (pop_regs): Use it.
13696 2018-02-16 Jakub Jelinek <jakub@redhat.com>
13699 * ipa-inline.c (inline_small_functions): Fix a typo.
13701 2018-02-16 Nathan Sidwell <nathan@acm.org>
13703 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
13705 2018-02-16 Carl Love <cel@us.ibm.com>
13707 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
13708 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
13709 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
13710 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
13711 expansion to P8V_BUILTIN_VEC_FLOAT2.
13713 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
13715 PR rtl-optimization/70023
13716 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
13717 src_regno into account.
13719 2018-02-16 Carl Love <cel@us.ibm.com>
13721 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
13722 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
13723 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
13724 * config/rs6000/rs6000.c: Remove case statements for
13725 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
13726 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
13727 and P9V_BUILTIN_VEC_VINSERT4B.
13728 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
13729 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
13730 * config/rs6000/vsx.md:
13731 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
13734 2018-02-16 Carl Love <cel@us.ibm.com>
13736 * config/rs6000/altivec.h: Add builtin names vec_extract4b
13738 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
13740 * config/rs6000/rs6000-c.c: Add the definitions for
13741 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
13742 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
13743 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
13744 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
13745 definition for insert4b and define insn *insert3b_internal.
13746 * doc/extend.texi: Add documentation for vec_extract4b.
13748 2018-02-16 Nathan Sidwell <nathan@acm.org>
13750 * doc/extend.texi (Backwards Compatibility): Mention friend
13751 injection. Note for-scope is deprecated.
13752 * doc/invoke.texi (-ffriend-injection): Deprecate.
13754 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
13756 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
13757 that moved to I2, also allow destinations that are a paradoxical
13758 subreg (instead of a normal reg).
13760 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
13763 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
13766 2018-02-16 Richard Biener <rguenther@suse.de>
13768 PR tree-optimization/84037
13769 PR tree-optimization/84016
13771 * config/i386/i386.c (ix86_builtin_vectorization_cost):
13772 Adjust vec_construct for the fact we need additional higher latency
13773 128bit inserts for AVX256 and AVX512 vector builds.
13774 (ix86_add_stmt_cost): Scale vector construction cost for
13777 2018-02-16 Richard Biener <rguenther@suse.de>
13779 PR tree-optimization/84417
13780 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
13781 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
13782 (non_rewritable_lvalue_p): Likewise, use poly-ints.
13784 2018-02-16 Martin Liska <mliska@suse.cz>
13787 * internal-fn.def (ASAN_CHECK): Set proper flags.
13788 (ASAN_MARK): Likewise.
13790 2018-02-16 Julia Koval <julia.koval@intel.com>
13792 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
13793 from PTA_CANNONLAKE.
13795 2018-02-16 Jakub Jelinek <jakub@redhat.com>
13798 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
13799 Use ++iter rather than iter++ for std::list iterators.
13800 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
13801 defer deleting them until all nodes in the forest are processed. Do
13802 free even leaf nodes. Change to_process into auto_vec.
13805 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
13806 * vec.h (vec_default_construct): Use memset instead of placement new
13807 if BROKEN_VALUE_INITIALIZATION is defined.
13808 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
13809 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
13812 PR rtl-optimization/83723
13813 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
13814 * lra.c (lra_substitute_pseudo): Likewise. If true, use
13815 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
13817 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
13819 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
13821 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
13823 PR rtl-optimization/81443
13824 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
13825 from inner REGs to paradoxical SUBREGs.
13827 2018-02-16 Richard Biener <rguenther@suse.de>
13829 PR tree-optimization/84399
13830 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
13831 For operands we can analyze at their definition make sure we can
13832 analyze them at each use as well.
13834 2018-02-16 Richard Biener <rguenther@suse.de>
13836 PR tree-optimization/84190
13837 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
13838 volatile accesses if the decl isn't volatile.
13840 2018-02-15 Jason Merrill <jason@redhat.com>
13842 PR c++/84314 - ICE with templates and fastcall attribute.
13843 * attribs.c (build_type_attribute_qual_variant): Don't clobber
13844 TYPE_CANONICAL on an existing type.
13846 2018-02-15 Jakub Jelinek <jakub@redhat.com>
13848 PR tree-optimization/84383
13849 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
13850 dstoff nor call operand_equal_p if dstbase is NULL.
13852 PR tree-optimization/84334
13853 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
13854 also a CONSTANT_CLASS_P, punt.
13856 2018-02-14 Jim Wilson <jimw@sifive.com>
13858 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
13859 first SMALL_OPERAND check. New local min_second_step. Move assert
13860 to where locals are set. Add TARGET_RVC support.
13861 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
13863 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
13865 * doc/invoke.texi: Correct -Wformat-overflow code sample.
13867 2018-02-14 Martin Sebor <msebor@redhat.com>
13869 PR tree-optimization/83698
13870 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
13871 arrays constrain the offset range to their bounds.
13872 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
13873 (builtin_access::overlap): Avoid setting the size of overlap if it's
13875 (maybe_diag_overlap): Also consider arrays when deciding what values
13876 of offsets to include in diagnostics.
13878 2018-02-14 Martin Sebor <msebor@redhat.com>
13881 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
13882 that correspond to the kind of a declaration.
13884 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
13887 * config/pa/pa.md: Load address of PIC label using the linkage table
13888 if the label is nonlocal.
13890 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
13892 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
13893 warning message if user requests -maltivec=be.
13894 * doc/invoke.texi: Document deprecation of -maltivec=be.
13896 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
13899 * config/rs6000/rs6000-c.c: Update definitions for
13900 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
13901 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
13903 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13906 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
13907 add _get_ssp intrinsics. Remove argument from
13908 __builtin_ia32_rdssp[d|q].
13909 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
13910 * config/i386/i386-builtin.def: Remove argument from
13911 __builtin_ia32_rdssp[d|q].
13912 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
13913 ix86_expand_special_args_builtin for _rdssp[d|q].
13914 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
13915 Clear register before usage.
13916 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
13917 Add documentation for new _get_ssp and _inc_ssp intrinsics.
13919 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
13921 PR tree-optimization/84357
13922 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
13923 operand 1 of an ARRAY_REF too.
13925 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
13928 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
13929 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
13931 (set_of_reg): New struct.
13932 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
13933 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
13934 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
13936 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
13937 Split into bitclr, bitset, bitinvert patterns if appropriate.
13938 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
13939 use rx_fuse_in_memory_bitop.
13940 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
13941 to named insn, correct maximum insn length.
13943 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
13946 * machmode.def: Define a complex mode for PARTIAL_INT.
13947 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
13949 * doc/rtl.texi: Document CSPImode.
13950 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
13952 (msp430_hard_regno_nregs_with_padding): Likewise.
13954 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
13957 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
13959 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
13961 PR rtl-optimization/84169
13962 * combine.c (try_combine): New variable split_i2i3. Set it to true if
13963 we generated a parallel as new i3 and we split that to new i2 and i3
13964 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
13965 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
13966 those to i2, not i1. Partially rewrite this scan code.
13968 2018-02-13 Jakub Jelinek <jakub@redhat.com>
13971 * stor-layout.c (place_field): For variable length fields, adjust
13972 offset_align afterwards not just based on the field's alignment,
13973 but also on the size.
13975 PR middle-end/84309
13976 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
13977 of exps and logs in the use_exp2 case.
13979 2018-02-13 Jeff Law <law@redhat.com>
13981 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
13982 entry for "vector".
13984 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
13987 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
13991 * common.opt (gas-loc-support, gas-locview-support): New.
13992 (ginline-points, ginternal-reset-location-views): New.
13993 * doc/invoke.texi: Document them. Use @itemx where intended.
13994 (gvariable-location-views): Adjust.
13995 * target.def (reset_location_view): New.
13996 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
13997 (TARGET_RESET_LOCATION_VIEW): New.
13998 * doc/tm.texi: Rebuilt.
13999 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
14000 (dwarf2out_default_as_locview_support): New.
14001 (output_asm_line_debug_info): Use option variables.
14002 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
14003 (output_loc_list): Likewise.
14004 (add_high_low_attributes): Check option variables.
14005 Don't output entry view attribute in strict mode.
14006 (gen_inlined_subroutine_die): Check option variables.
14007 (dwarf2out_inline_entry): Likewise.
14008 (init_sections_and_labels): Likewise.
14009 (dwarf2out_early_finish): Likewise.
14010 (maybe_reset_location_view): New, from...
14011 (dwarf2out_var_location): ... here. Call it.
14012 * debug.h (dwarf2out_default_as_loc_support): Declare.
14013 (dwarf2out_default_as_locview_support): Declare.
14014 * hooks.c (hook_int_rtx_insn_0): New.
14015 * hooks.h (hook_int_rtx_insn_0): Declare.
14016 * toplev.c (process_options): Take -gas-loc-support and
14017 -gas-locview-support from dwarf2out. Enable
14018 -gvariable-location-views by default only with locview
14019 assembler support. Enable -ginternal-reset-location-views by
14020 default only if the target defines the corresponding hook.
14021 Enable -ginline-points by default if location views are
14022 enabled; force it disabled if statement frontiers are
14024 * tree-inline.c (expand_call_inline): Check option variables.
14025 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
14027 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
14029 PR tree-optimization/84321
14030 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
14031 handling. Also check whether the anti-range contains any values
14032 that satisfy the mask; switch to a VR_RANGE if not.
14034 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
14037 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
14039 2018-02-13 Martin Jambor <mjambor@suse.cz>
14042 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
14043 of call statements, also set location of a load to a temporary.
14045 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
14047 * config/rl78/rl78.c (add_vector_labels): New function.
14048 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
14049 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
14050 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
14051 which checks that no arguments are passed.
14052 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
14053 * doc/extend.texi: Documentation for the new attribute.
14055 2018-02-13 Andreas Schwab <schwab@suse.de>
14057 * config/riscv/linux.h (CPP_SPEC): Define.
14059 2018-02-13 Jakub Jelinek <jakub@redhat.com>
14062 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
14063 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
14064 OPTION_MASK_ISA_AES as first argument to def_builtin_const
14065 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
14066 instead of OPTION_MASK_ISA_PCLMUL as first argument to
14067 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
14068 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
14069 temporarily for AES and PCLMUL builtins.
14071 PR tree-optimization/84339
14072 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
14073 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
14076 PR middle-end/84309
14077 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
14078 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
14079 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
14081 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
14083 * omp-simd-clone.h: New file.
14084 * omp-simd-clone.c: Include omp-simd-clone.h.
14085 (expand_simd_clones): No longer static.
14086 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
14087 cgraph.h and omp-simd-clone.h.
14088 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
14089 (vect_recog_widen_shift_pattern): Formatting fix.
14090 (vect_pattern_recog_1): Don't check optab for calls.
14093 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
14094 operands[2] into a REG before using gen_lowpart on it.
14096 2018-02-12 Jeff Law <law@redhat.com>
14099 * config/sh/sh.c (find_barrier): Consider a sibling call
14102 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
14103 successfully back substituting a reg.
14105 2018-02-12 Richard Biener <rguenther@suse.de>
14107 PR tree-optimization/84037
14108 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
14109 parameter, move visited init to caller.
14110 (vect_slp_analyze_operations): Separate cost from validity
14111 check, initialize visited once for all instances.
14112 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
14114 * tree-vect-stmts.c (vect_model_simple_cost): Make early
14116 (vect_model_promotion_demotion_cost): Likewise.
14117 (vectorizable_bswap): Guard cost modeling with !slp_node
14118 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
14120 (vectorizable_call): Likewise.
14121 (vectorizable_conversion): Likewise.
14122 (vectorizable_assignment): Likewise.
14123 (vectorizable_shift): Likewise.
14124 (vectorizable_operation): Likewise.
14125 (vectorizable_store): Likewise.
14126 (vectorizable_load): Likewise.
14127 (vectorizable_condition): Likewise.
14128 (vectorizable_comparison): Likewise.
14130 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
14133 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
14134 (ASAN_MARK): Fix fnspec to account for return value, change pointer
14135 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
14137 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
14139 PR middle-end/83665
14140 * params.def (inline-min-speedup): Increase from 8 to 15.
14141 (max-inline-insns-auto): Decrease from 40 to 30.
14142 * ipa-split.c (consider_split): Add some buffer for function to
14143 be considered inlining candidate.
14144 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
14147 2018-02-12 Richard Biener <rguenther@suse.de>
14149 PR tree-optimization/84037
14150 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
14151 matched stmts if we cannot swap the non-matched ones.
14153 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
14155 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
14156 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
14157 _mm_maskz_scalef_round_ss): New intrinsics.
14158 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
14159 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
14160 __builtin_ia32_scalefss_round): Remove.
14161 (__builtin_ia32_scalefsd_mask_round,
14162 __builtin_ia32_scalefss_mask_round): New intrinsics.
14163 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
14164 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
14165 ((match_operand:VF_128 2 "<round_nimm_predicate>"
14166 "<round_constraint>")): Changed to ...
14167 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
14168 "<round_scalar_constraint>")): ... this.
14169 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
14170 %0, %1, %2<round_op3>}"): Changed to ...
14171 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
14172 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
14173 %2<round_scalar_mask_op3>}"): ... this.
14174 * config/i386/subst.md (round_scalar_nimm_predicate): New.
14176 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
14178 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
14179 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
14180 (_mm_maskz_sqrt_round_ss): New intrinsics.
14181 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
14182 (__builtin_ia32_sqrtsd_mask_round)
14183 (__builtin_ia32_sqrtss_mask_round): New builtins.
14184 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
14185 (__builtin_ia32_sqrtss_round): Remove.
14186 (__builtin_ia32_sqrtsd_mask_round)
14187 (__builtin_ia32_sqrtss_mask_round): New builtins.
14188 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
14189 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
14190 ((match_operand:VF_128 1 "vector_operand"
14191 "xBm,<round_constraint>")): Changed to ...
14192 ((match_operand:VF_128 1 "vector_operand"
14193 "xBm,<round_scalar_constraint>")): ... this.
14194 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
14195 %0, %2, %<iptr>1<round_op3>}): Changed to ...
14196 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
14197 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
14198 %<iptr>1<round_scalar_mask_op3>}): ... this.
14199 ((set_attr "prefix" "<round_prefix>")): Changed to ...
14200 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
14202 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
14205 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
14206 Cast vec_cmpeq result to correct type.
14207 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
14208 Cast vec_cmpgt result to correct type.
14210 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
14212 * final.c (final_scan_insn_1): Renamed from...
14213 (final_scan_insn): ... this. New wrapper, to recover
14214 seen from the outermost call in recursive ones.
14215 * config/sparc/sparc.c (output_return): Drop seen from call.
14216 (output_sibcall): Likewise.
14217 * config/visium/visium.c (output_branch): Likewise.
14219 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
14221 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
14224 2018-02-10 Alan Modra <amodra@gmail.com>
14227 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
14228 Specify LR as an input.
14230 2018-02-10 Jakub Jelinek <jakub@redhat.com>
14233 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
14234 remove_member_access_dummy_vars): New functions.
14235 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
14236 lower_omp_1, execute_lower_omp): Use them.
14238 PR rtl-optimization/84308
14239 * shrink-wrap.c (spread_components): Release todo vector.
14241 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
14243 PR rtl-optimization/57193
14244 * ira-color.c (struct allocno_color_data): Add member
14245 conflict_allocno_hard_prefs.
14246 (update_conflict_allocno_hard_prefs): New.
14247 (bucket_allocno_compare_func): Add a preference based on
14248 conflict_allocno_hard_prefs.
14249 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
14250 (color_allocnos): Remove a dead code. Initiate
14251 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
14253 2018-02-09 Jakub Jelinek <jakub@redhat.com>
14256 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
14257 constraint from =wa to wa. Avoid a subreg on the output operand,
14258 instead use a pseudo and subreg it in a move.
14259 (p9_xxbrd_<mode>): Changed to ...
14260 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
14261 (p9_xxbrd_v2df): New expander.
14262 (p9_xxbrw_<mode>): Changed to ...
14263 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
14264 (p9_xxbrw_v4sf): New expander.
14266 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
14268 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
14270 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
14273 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
14274 multiply in 32-bit mode.
14275 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
14276 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
14279 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
14281 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
14282 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
14283 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
14284 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
14286 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
14289 * dwarf2out.c (is_trivial_indirect_ref): New function.
14290 (dwarf2out_late_global_decl): Do not generate a location
14291 attribute for variables that have a non-trivial DECL_VALUE_EXPR
14292 and that are not defined in the current unit.
14294 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
14296 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
14297 instead of a libcall for UNORDERED.
14299 2018-02-09 Tamar Christina <tamar.christina@arm.com>
14302 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
14303 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
14305 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14308 * config/s390/s390.c (s390_set_current_function): Invoke
14309 s390_indirect_branch_settings also if fndecl didn't change.
14311 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
14313 * config/rs6000/rs6000.md (blockage): Set length to zero.
14315 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
14317 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
14319 2018-02-09 Jakub Jelinek <jakub@redhat.com>
14322 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
14323 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
14327 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
14328 PARALLEL incoming that failed vt_get_decl_and_offset check.
14330 PR middle-end/84237
14331 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
14332 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
14334 (get_variable_section): For decls in named .bss* sections pass true as
14335 second argument to bss_initializer_p.
14337 2018-02-09 Marek Polacek <polacek@redhat.com>
14338 Jakub Jelinek <jakub@redhat.com>
14341 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
14342 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
14343 Sync some changes from cxx_fold_indirect_ref.
14345 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
14347 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
14349 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
14350 (BLOCK_INLINE_ENTRY_LABEL): New.
14351 (dwarf2out_var_location): Disregard inline entry markers.
14352 (inline_entry_data): New struct.
14353 (inline_entry_data_hasher): New hashtable type.
14354 (inline_entry_data_hasher::hash): New.
14355 (inline_entry_data_hasher::equal): New.
14356 (inline_entry_data_table): New variable.
14357 (add_high_low_attributes): Add DW_AT_entry_pc and
14358 DW_AT_GNU_entry_view attributes if a pending entry is found
14359 in inline_entry_data_table. Add old entry_pc attribute only
14360 if debug nonbinding markers are disabled.
14361 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
14362 markers are enabled.
14363 (block_within_block_p, dwarf2out_inline_entry): New.
14364 (dwarf2out_finish): Check that no entries remained in
14365 inline_entry_data_table.
14366 * final.c (reemit_insn_block_notes): Handle inline entry notes.
14367 (final_scan_insn, notice_source_line): Likewise.
14368 (rest_of_clean_state): Skip inline entry markers.
14369 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
14371 * gimple.c (gimple_build_debug_inline_entry): New.
14372 * gimple.h (enum gimple_debug_subcode): Add
14373 GIMPLE_DEBUG_INLINE_ENTRY.
14374 (gimple_build_debug_inline_entry): Declare.
14375 (gimple_debug_inline_entry_p): New.
14376 (gimple_debug_nonbind_marker_p): Adjust.
14377 * insn-notes.def (INLINE_ENTRY): New.
14378 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
14379 inline entry marker notes.
14380 (print_insn): Likewise.
14381 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
14382 (INSN_DEBUG_MARKER_KIND): Likewise.
14383 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
14384 * tree-inline.c (expand_call_inline): Build and insert
14385 debug_inline_entry stmt.
14386 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
14387 inline entry blocks early, if nonbind markers are enabled.
14388 (dump_scope_block): Dump fragment info.
14389 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
14390 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
14391 (gimple_build_debug_inline_entry): New.
14392 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
14393 Enable/disable inline entry points too.
14394 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
14395 (DEBUG_INSN): Describe inline entry markers.
14397 * common.opt (gvariable-location-views): New.
14398 (gvariable-location-views=incompat5): New.
14399 * config.in: Rebuilt.
14400 * configure: Rebuilt.
14401 * configure.ac: Test assembler for view support.
14402 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
14403 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
14404 * dwarf2out.c (var_loc_view): New typedef.
14405 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
14406 (dwarf2out_locviews_in_attribute): New.
14407 (dwarf2out_locviews_in_loclist): New.
14408 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
14409 (enum dw_line_info_opcode): Add LI_adv_address.
14410 (struct dw_line_info_table): Add view.
14411 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
14412 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
14413 (zero_view_p): New variable.
14414 (ZERO_VIEW_P): New macro.
14415 (output_asm_line_debug_info): New.
14416 (struct var_loc_node): Add view.
14417 (add_AT_view_list, AT_loc_list): New.
14418 (add_var_loc_to_decl): Add view param. Test it against last.
14419 (new_loc_list): Add view params. Record them.
14420 (AT_loc_list_ptr): Handle loc and view lists.
14421 (view_list_to_loc_list_val_node): New.
14422 (print_dw_val): Handle dw_val_class_view_list.
14423 (size_of_die): Likewise.
14424 (value_format): Likewise.
14425 (loc_list_has_views): New.
14426 (gen_llsym): Set vl_symbol too.
14427 (maybe_gen_llsym, skip_loc_list_entry): New.
14428 (dwarf2out_maybe_output_loclist_view_pair): New.
14429 (output_loc_list): Output view list or entries too.
14430 (output_view_list_offset): New.
14431 (output_die): Handle dw_val_class_view_list.
14432 (output_dwarf_version): New.
14433 (output_compilation_unit_header): Use it.
14434 (output_skeleton_debug_sections): Likewise.
14435 (output_rnglists, output_line_info): Likewise.
14436 (output_pubnames, output_aranges): Update version comments.
14437 (output_one_line_info_table): Output view numbers in asm comments.
14438 (dw_loc_list): Determine current endview, pass it to new_loc_list.
14439 Call maybe_gen_llsym.
14440 (loc_list_from_tree_1): Adjust.
14441 (add_AT_location_description): Create view list attribute if
14442 needed, check it's absent otherwise.
14443 (convert_cfa_to_fb_loc_list): Adjust.
14444 (maybe_emit_file): Call output_asm_line_debug_info for test.
14445 (dwarf2out_var_location): Reset views as needed. Precompute
14446 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
14447 attribute. Set view.
14448 (new_line_info_table): Reset next view.
14449 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
14450 (dwarf2out_source_line): Likewise. Output view resets and labels to
14451 the assembler, or select appropriate line info opcodes.
14452 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
14453 (optimize_string_length): Catch it. Adjust.
14454 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
14455 dw_val_class_view_list, and remove it if no longer needed.
14456 (hash_loc_list): Hash view numbers.
14457 (loc_list_hasher::equal): Compare them.
14458 (optimize_location_lists): Check whether a view list symbol is
14459 needed, and whether the locview attribute is present, and
14460 whether they match. Remove the locview attribute if no longer
14462 (index_location_lists): Call skip_loc_list_entry for test.
14463 (dwarf2out_finish): Call output_asm_line_debug_info for test.
14464 Use output_dwarf_version.
14465 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
14466 (struct dw_val_node): Add val_view_list.
14467 * final.c (SEEN_NEXT_VIEW): New.
14468 (set_next_view_needed): New.
14469 (clear_next_view_needed): New.
14470 (maybe_output_next_view): New.
14471 (final_start_function): Rename to...
14472 (final_start_function_1): ... this. Take pointer to FIRST,
14473 add SEEN parameter. Emit param bindings in the initial view.
14474 (final_start_function): Reintroduce SEEN-less interface.
14475 (final): Rename to...
14476 (final_1): ... this. Take SEEN parameter. Output final pending
14477 next view at the end.
14478 (final): Reintroduce seen-less interface.
14479 (final_scan_insn): Output pending next view before switching
14480 sections or ending a block. Mark the next view as needed when
14481 outputting variable locations. Notify debug backend of section
14482 changes, and of location view changes.
14483 (rest_of_handle_final): Adjust.
14484 * toplev.c (process_options): Autodetect value for debug variable
14485 location views option. Warn on incompat5 without -gdwarf-5.
14486 * doc/invoke.texi (gvariable-location-views): New.
14487 (gvariable-location-views=incompat5): New.
14488 (gno-variable-location-views): New.
14490 2018-02-08 David Malcolm <dmalcolm@redhat.com>
14492 PR tree-optimization/84136
14493 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
14494 that the result of find_edge is non-NULL.
14496 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
14499 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
14500 storing integer register in SImode. Fix cost of 256 and 512
14501 byte aligned SSE register store.
14503 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
14505 * config/i386/i386.c (ix86_multiplication_cost): Fix
14506 multiplication cost for TARGET_AVX512DQ.
14508 2018-02-08 Marek Polacek <polacek@redhat.com>
14510 PR tree-optimization/84238
14511 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
14514 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14516 PR tree-optimization/84265
14517 * tree-vect-stmts.c (vectorizable_store): Don't treat
14518 VMAT_CONTIGUOUS accesses as grouped.
14519 (vectorizable_load): Likewise.
14521 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14523 PR tree-optimization/81635
14524 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
14525 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
14526 (test_round_for_mask): New functions.
14527 (wide_int_cc_tests): Call test_round_for_mask.
14528 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
14529 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
14530 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
14531 range returned by get_range_info.
14533 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
14536 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
14537 * symtab.c: Include builtins.h
14538 (symtab_node::output_to_lto_symbol_table_p): Move here
14539 from lto-streamer-out.c:output_symbol_p.
14540 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
14541 (output_symbol_p): Move all logic to symtab.c
14542 (produce_symtab): Update.
14544 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14546 * config/s390/s390-opts.h (enum indirect_branch): Define.
14547 * config/s390/s390-protos.h (s390_return_addr_from_memory)
14548 (s390_indirect_branch_via_thunk)
14549 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
14550 (enum s390_indirect_branch_type): Define.
14551 * config/s390/s390.c (struct s390_frame_layout, struct
14552 machine_function): Remove.
14553 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
14554 (indirect_branch_table_label_no, indirect_branch_table_name):
14556 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
14557 (enum s390_indirect_branch_option): Define.
14558 (s390_return_addr_from_memory): New function.
14559 (s390_handle_string_attribute): New function.
14560 (s390_attribute_table): Add new attribute handler.
14561 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
14562 (s390_indirect_branch_via_thunk): New function.
14563 (s390_indirect_branch_via_inline_thunk): New function.
14564 (s390_function_ok_for_sibcall): When jumping via thunk disallow
14565 sibling call optimization for non z10 compiles.
14566 (s390_emit_call): Force indirect branch target to be a single
14567 register. Add r1 clobber for non-z10 compiles.
14568 (s390_emit_epilogue): Emit return jump via return_use expander.
14569 (s390_reorg): Handle JUMP_INSNs as execute targets.
14570 (s390_option_override_internal): Perform validity checks for the
14571 new command line options.
14572 (s390_indirect_branch_attrvalue): New function.
14573 (s390_indirect_branch_settings): New function.
14574 (s390_set_current_function): Invoke s390_indirect_branch_settings.
14575 (s390_output_indirect_thunk_function): New function.
14576 (s390_code_end): Implement target hook.
14577 (s390_case_values_threshold): Implement target hook.
14578 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
14580 * config/s390/s390.h (struct s390_frame_layout)
14581 (struct machine_function): Move here from s390.c.
14582 (TARGET_INDIRECT_BRANCH_NOBP_RET)
14583 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
14584 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
14585 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
14586 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
14587 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
14588 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
14589 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
14590 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
14591 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
14592 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
14593 (mnemonic attribute): Add values which aren't recognized
14595 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
14596 pattern for branch conversion. Fix mnemonic attribute.
14597 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
14598 indirect branch via thunk if requested.
14599 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
14600 ("*indirect_jump"): Disable for branch conversion using out of
14602 ("indirect_jump_via_thunk<mode>_z10")
14603 ("indirect_jump_via_thunk<mode>")
14604 ("indirect_jump_via_inlinethunk<mode>_z10")
14605 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
14606 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
14607 ("casesi_jump_via_inlinethunk<mode>_z10")
14608 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
14609 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
14610 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
14611 ("*indirect2_jump"): Disable for branch conversion.
14612 ("casesi_jump"): Turn into expander and expand patterns for branch
14614 ("return_use"): New expander.
14615 ("*return"): Emit return via thunk and rename it to ...
14616 ("*return<mode>"): ... this one.
14617 * config/s390/s390.opt: Add new options and and enum for the
14620 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14622 * lra-constraints.c (match_reload): Unconditionally use
14623 gen_lowpart_SUBREG, rather than selecting between that
14624 and equivalent gen_rtx_SUBREG code.
14626 2018-02-08 Richard Biener <rguenther@suse.de>
14628 PR tree-optimization/84233
14629 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
14630 changed flag instead of boguously re-using phi_inserted.
14632 2018-02-08 Martin Jambor <mjambor@suse.cz>
14634 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
14635 static local variables.
14637 2018-02-08 Richard Biener <rguenther@suse.de>
14639 PR tree-optimization/84278
14640 * tree-vect-stmts.c (vectorizable_store): When looking for
14641 smaller vector types to perform grouped strided loads/stores
14642 make sure the mode is supported by the target.
14643 (vectorizable_load): Likewise.
14645 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
14647 * config/aarch64/aarch64.c (aarch64_components_for_bb):
14648 Increase LDP/STP opportunities by adding adjacent callee-saves.
14650 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
14652 PR rtl-optimization/84068
14653 PR rtl-optimization/83459
14654 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
14656 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
14658 PR tree-optimization/84224
14659 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
14660 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
14661 non-zero arguments.
14663 2018-02-07 Iain Sandoe <iain@codesourcery.com>
14666 * config/rs6000/altivec.md (*restore_world): Remove LR use.
14667 * config/rs6000/predicates.md (restore_world_operation): Adjust op
14668 count, remove one USE.
14670 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
14672 * doc/install.texi (Configuration): Document the
14673 --with-long-double-format={ibm,ieee} PowerPC configuration
14677 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
14678 Convert from define_expand to be define_insn_and_split. Rework
14679 float/double/_Float128 conversions to QI/HI/SImode to work with
14680 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
14681 conversions to QI/HImode types did a store and then a load to
14682 truncate the value. For conversions to VSX registers, don't split
14683 the insn, instead emit the code directly. Use the code iterator
14684 any_fix to combine signed and unsigned conversions.
14685 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
14686 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
14687 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
14688 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
14689 (fix_<mode>di2_hw): Likewise.
14690 (fixuns_<mode>di2_hw): Likewise.
14691 (fix_<mode>si2_hw): Likewise.
14692 (fixuns_<mode>si2_hw): Likewise.
14693 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
14694 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
14695 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
14696 fix<uns>_trunc<SFDF:mode>si2_p8.
14697 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
14698 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
14699 (fix<uns>_<mode>_mem): Likewise.
14700 (fctiw<u>z_<mode>_mem): Likewise.
14701 (fix<uns>_<mode>_mem): Likewise.
14702 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
14703 the register allocator from doing a direct move to the GPRs to do
14704 a store, and instead use the ISA 3.0 store byte/half-word from
14705 vector register instruction. For IEEE 128-bit floating point,
14706 also optimize stores of 32-bit ints.
14707 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
14709 2018-02-07 Alan Hayward <alan.hayward@arm.com>
14711 * genextract.c (push_pathstr_operand): New function to support
14713 (walk_rtx): Call push_pathstr_operand.
14714 (print_path): Support [a-zA-Z].
14716 2018-02-07 Richard Biener <rguenther@suse.de>
14718 PR tree-optimization/84037
14719 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
14720 (cse_and_gimplify_to_preheader): Declare.
14721 (vect_get_place_in_interleaving_chain): Likewise.
14722 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14724 (_loop_vec_info::~_loop_vec_info): Delete it.
14725 (cse_and_gimplify_to_preheader): New function.
14726 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
14727 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
14728 (vectorizable_load): Likewise. For grouped stores always base
14729 the IV on the first element.
14730 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
14731 condition before gimplifying.
14733 2018-02-07 Jakub Jelinek <jakub@redhat.com>
14735 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
14736 *DIV_EXPR and *MOD_EXPR.
14738 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
14741 * config/i386/i386.c (ix86_option_override_internal): Mask out
14742 the CF_SET bit when checking -fcf-protection.
14744 2018-02-07 Tom de Vries <tom@codesourcery.com>
14747 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
14750 2018-02-07 Richard Biener <rguenther@suse.de>
14752 PR tree-optimization/84204
14753 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
14756 PR tree-optimization/84205
14757 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
14758 special-case isl_ast_op_zdiv_r.
14760 PR tree-optimization/84223
14761 * graphite-scop-detection.c (gather_bbs::before_dom_children):
14762 Only add conditions from within the region.
14763 (gather_bbs::after_dom_children): Adjust.
14765 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
14768 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
14769 * config/avr/avr.md: Only post-reload split REG-REG moves if
14770 either register is GENERAL_REG_P.
14772 2018-02-07 Jakub Jelinek <jakub@redhat.com>
14774 PR tree-optimization/84235
14775 * tree-ssa-scopedtables.c
14776 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
14777 if the subtraction is performed in floating point type where NaNs are
14778 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
14779 build 1. Formatting fix.
14781 2018-02-06 Jakub Jelinek <jakub@redhat.com>
14784 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
14785 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
14786 and skip it regardless of bb boundaries. Use CALL_P macro,
14787 don't test INSN_P (insn) together with CALL_P or JUMP_P check
14788 unnecessarily, formatting fix.
14790 2018-02-06 Michael Collison <michael.collison@arm.com>
14792 * config/arm/thumb2.md:
14793 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
14794 (*thumb_mov_notscc): Ditto.
14796 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
14799 * config/rs6000/rs6000.md (su code attribute): Use "u" for
14800 unsigned_fix, not "s".
14802 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14804 * configure.ac (gcc_fn_eh_frame_ro): New function.
14805 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
14806 correct .eh_frame permissions.
14807 * configure: Regenerate.
14809 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
14811 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
14812 irrelevant options.
14814 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14816 * config/rs6000/rs6000.c (rs6000_option_override_internal):
14817 Display warning message for -mno-speculate-indirect-jumps.
14819 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
14821 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
14823 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
14825 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
14827 PR tree-optimization/84225
14828 * tree-eh.c (find_trapping_overflow): Only call
14829 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
14831 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14834 * config/i386/i386.c: Reimplement the check of possible options
14835 -mibt/-mshstk conbination. Change error messages.
14836 * doc/invoke.texi: Fix a typo: remove extra '='.
14838 2018-02-06 Marek Polacek <polacek@redhat.com>
14840 PR tree-optimization/84228
14841 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
14843 2018-02-06 Tamar Christina <tamar.christina@arm.com>
14846 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
14847 emitted arch directives.
14848 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
14849 __ARM_FEATURE_COPROC before changing architectures.
14851 2018-02-06 Richard Biener <rguenther@suse.de>
14853 * config/i386/i386.c (print_reg): Fix typo.
14854 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
14856 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
14858 * configure: Regenerate.
14860 2018-02-05 Martin Sebor <msebor@redhat.com>
14862 PR tree-optimization/83369
14863 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
14866 2018-02-05 Martin Liska <mliska@suse.cz>
14868 * doc/invoke.texi: Cherry-pick upstream r323995.
14870 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
14872 * ira.c (ira_init_register_move_cost): Adjust comment.
14874 2018-02-05 Martin Liska <mliska@suse.cz>
14876 PR gcov-profile/84137
14877 * doc/gcov.texi: Fix typo in documentation.
14879 2018-02-05 Martin Liska <mliska@suse.cz>
14881 PR gcov-profile/83879
14882 * doc/gcov.texi: Document necessity of --dynamic-list-data when
14883 using dlopen functionality.
14885 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
14887 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
14888 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
14889 _mm_maskz_range_ss, _mm_mask_range_round_ss,
14890 _mm_maskz_range_round_ss): New intrinsics.
14891 (__builtin_ia32_rangesd128_round)
14892 (__builtin_ia32_rangess128_round): Remove.
14893 (__builtin_ia32_rangesd128_mask_round,
14894 __builtin_ia32_rangess128_mask_round): New builtins.
14895 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
14896 __builtin_ia32_rangess128_round): Remove.
14897 (__builtin_ia32_rangesd128_mask_round,
14898 __builtin_ia32_rangess128_mask_round): New builtins.
14899 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
14900 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
14901 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
14902 "<round_saeonly_constraint>")): Changed to ...
14903 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
14904 "<round_saeonly_scalar_constraint>")): ... this.
14905 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
14906 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
14907 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
14908 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
14909 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
14911 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
14913 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
14915 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
14916 Remove all values except native, 8540 and 8548.
14918 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
14920 * config/i386/i386.c (ix86_output_function_return): Pass
14921 INVALID_REGNUM, instead of -1, as invalid register number to
14922 indirect_thunk_name and output_indirect_thunk.
14924 2018-02-02 Julia Koval <julia.koval@intel.com>
14926 * config.gcc: Add -march=icelake.
14927 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
14928 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
14929 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
14930 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
14931 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
14932 (processor_target_table): Add icelake.
14933 (ix86_option_override_internal): Handle new PTAs.
14934 (get_builtin_code_for_version): Handle icelake.
14935 (M_INTEL_COREI7_ICELAKE): New.
14936 (fold_builtin_cpu): Handle icelake.
14937 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
14938 * doc/invoke.texi: Add -march=icelake.
14940 2018-02-02 Julia Koval <julia.koval@intel.com>
14942 * config/i386/i386.c (ix86_option_override_internal): Change flags type
14943 to wide_int_bitmask.
14944 * wide-int-bitmask.h: New.
14946 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14949 * config/i386/i386.md: Replace Pmode with word_mode in
14950 builtin_setjmp_setup and builtin_longjmp to support x32.
14952 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
14956 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
14958 (rs6000_supported_cpu_names): New static variable.
14959 (linux_cpu_translation_table): Likewise.
14960 (elf_platform) <cpu>: Define new static variable and use it.
14961 Translate kernel AT_PLATFORM name to canonical name if needed.
14962 Error if platform name is unknown.
14964 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
14967 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
14969 2018-02-01 Jeff Law <law@redhat.com>
14972 * config/i386/i386.c (release_scratch_register_on_entry): Add new
14973 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
14974 the scratch if RELEASE_VIA_POP is false.
14975 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
14976 If we have to save a temporary register, decrement SIZE appropriately.
14977 Pass new arguments to release_scratch_register_on_entry.
14978 (ix86_adjust_stack_and_probe): Likewise.
14979 (ix86_emit_probe_stack_range): Pass new arguments to
14980 release_scratch_register_on_entry.
14982 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
14984 PR rtl-optimization/84157
14985 * combine.c (change_zero_ext): Use REG_P predicate in
14986 front of HARD_REGISTER_P predicate.
14988 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
14990 * config/avr/avr.c (avr_option_override): Move disabling of
14991 -fdelete-null-pointer-checks to...
14992 * common/config/avr/avr-common.c (avr_option_optimization_table):
14995 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14997 PR tree-optimization/81635
14998 * tree-data-ref.c (split_constant_offset_1): For types that
14999 wrap on overflow, try to use range info to prove that wrapping
15002 2018-02-01 Renlin Li <renlin.li@arm.com>
15005 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
15006 TAILCALL_ADDR_REGS.
15007 (aarch64_register_move_cost): Likewise.
15008 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
15009 TAILCALL_ADDR_REGS.
15010 (REG_CLASS_NAMES): Likewise.
15011 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
15012 TAILCALL_ADDR_REGS. Remove IP registers.
15013 * config/aarch64/aarch64.md (Ucs): Update register constraint.
15015 2018-02-01 Richard Biener <rguenther@suse.de>
15017 * domwalk.h (dom_walker::dom_walker): Add additional constructor
15018 for specifying RPO order and allow NULL for that.
15019 * domwalk.c (dom_walker::dom_walker): Likewise.
15020 (dom_walker::walk): Handle NULL RPO order.
15021 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
15023 (rewrite_update_dom_walker): Likewise.
15024 (mark_def_dom_walker): Likewise.
15026 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15028 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
15029 (aarch64_maybe_expand_sve_subreg_move): Declare.
15030 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
15031 * config/aarch64/predicates.md (aarch64_any_register_operand): New
15033 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
15034 that are semantically a reverse operation.
15035 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
15036 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
15037 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
15039 (aarch64_can_change_mode_class): For big-endian, forbid changes
15040 between two SVE modes if they have different element sizes.
15042 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15044 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
15045 the TImode handling for big-endian targets.
15047 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15049 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
15050 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
15052 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
15053 Remove BSWAP handing for big-endian targets and use the form of
15054 LD1RQ appropariate for the mode.
15056 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15058 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
15059 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
15060 duplicated element.
15062 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15065 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
15066 check for operands that need to go through aarch64_sve_reload_be.
15068 2018-02-01 Jakub Jelinek <jakub@redhat.com>
15070 PR tree-optimization/81661
15071 PR tree-optimization/84117
15072 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
15073 * tree-eh.c: Include gimplify.h.
15074 (find_trapping_overflow, replace_trapping_overflow,
15075 rewrite_to_non_trapping_overflow): New functions.
15076 * tree-vect-loop.c: Include tree-eh.h.
15077 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
15078 * tree-data-ref.c: Include tree-eh.h.
15079 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
15081 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
15083 PR rtl-optimization/84123
15084 * combine.c (change_zero_ext): Check if hard register satisfies
15085 can_change_dest_mode before calling gen_lowpart_SUBREG.
15087 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
15090 * ira.c (ira_init_register_move_cost): Remove assert.
15092 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15094 PR rtl-optimization/84071
15095 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
15096 * doc/tm.texi: Regenerate.
15098 2018-01-31 Richard Biener <rguenther@suse.de>
15100 PR tree-optimization/84132
15101 * tree-data-ref.c (analyze_miv_subscript): Properly
15102 check whether evolution_function_is_affine_multivariate_p
15103 before calling gcd_of_steps_may_divide_p.
15105 2018-01-31 Julia Koval <julia.koval@intel.com>
15108 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
15109 * config/i386/i386.md (rdpid_rex64) New.
15110 (rdpid): Make 32bit only.
15112 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
15115 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
15116 an IDENTIFIER_NODE for FUNCTION_TYPE's.
15118 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15121 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15123 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15125 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15127 PR rtl-optimization/84071
15128 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
15129 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
15131 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
15133 * config/arc/arc.c (arc_handle_aux_attribute): New function.
15134 (arc_attribute_table): Add 'aux' attribute.
15135 (arc_in_small_data_p): Consider aux like variables.
15136 (arc_is_aux_reg_p): New function.
15137 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
15138 (arc_get_aux_arg): New function.
15139 (prepare_move_operands): Handle aux-register access.
15140 (arc_handle_aux_attribute): New function.
15141 * doc/extend.texi (ARC Variable attributes): Add subsection.
15143 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
15145 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
15146 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
15147 (arc_attribute_table): Add 'uncached' attribute.
15148 (arc_print_operand): Print '.di' flag for uncached memory
15150 (arc_in_small_data_p): Do not consider for small data the uncached
15152 (arc_is_uncached_mem_p): New function.
15153 * config/arc/predicates.md (compact_store_memory_operand): Check
15154 for uncached memory accesses.
15155 (nonvol_nonimm_operand): Likewise.
15156 * doc/extend.texi (ARC Type Attribute): New subsection.
15158 2018-01-31 Jakub Jelinek <jakub@redhat.com>
15161 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
15162 falign-loops=): Add Optimization flag.
15164 2018-01-30 Jeff Law <law@redhat.com>
15167 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
15168 INT_REGISTERS_SAVED. Check it prior to calling
15169 get_scratch_register_on_entry.
15170 (ix86_adjust_stack_and_probe): Similarly.
15171 (ix86_emit_probe_stack_range): Similarly.
15172 (ix86_expand_prologue): Corresponding changes.
15174 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15177 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
15178 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
15180 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
15183 * lra-constraints.c (curr_insn_transform): Process AND in the
15186 2018-01-30 Jakub Jelinek <jakub@redhat.com>
15188 PR rtl-optimization/83986
15189 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
15190 dependence against last_pending_memory_flush in addition to
15191 pending_jump_insns.
15193 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
15195 PR tree-optimization/81611
15196 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
15199 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15202 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
15205 2018-01-30 Richard Biener <rguenther@suse.de>
15206 Jakub Jelinek <jakub@redhat.com>
15208 PR tree-optimization/84111
15209 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
15210 inner loops added during recursion, as they don't have up-to-date
15213 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
15216 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
15217 (can_inline_edge_by_limits_p): ... here.
15218 (can_early_inline_edge_p, check_callers,
15219 update_caller_keys, update_callee_keys, recursive_inlining,
15220 add_new_edges_to_heap, speculation_useful_p,
15221 inline_small_functions,
15222 inline_small_functions, flatten_function,
15223 inline_to_all_callers_1): Update.
15225 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
15227 * profile-count.c (profile_count::combine_with_ipa_count): Handle
15230 2018-01-30 Richard Biener <rguenther@suse.de>
15232 PR tree-optimization/83008
15233 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
15234 invariant and constant vector uses in stmts when they need
15235 more than one stmt.
15237 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15240 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
15241 * configure: Regenerate.
15243 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
15245 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
15247 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
15248 Use gen_rtx_REG rather than gen_lowpart.
15250 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
15252 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
15253 rather than 0 when creating partial subregs.
15255 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
15257 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
15260 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
15263 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
15264 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
15265 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
15266 flags. This restores the settings used before the 2017-07-24.
15267 Turning off pre increment/decrement/modify allows IVOPTS to
15268 optimize DF/SF loops where the index is an int.
15270 2018-01-29 Richard Biener <rguenther@suse.de>
15271 Kelvin Nilsen <kelvin@gcc.gnu.org>
15274 * tree-vect-stmts.c (vectorizable_call): Don't call
15275 targetm.vectorize_builtin_md_vectorized_function if callee is
15278 2018-01-22 Carl Love <cel@us.ibm.com>
15280 * doc/extend.tex: Fix typo in second arg in
15281 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
15283 2018-01-29 Richard Biener <rguenther@suse.de>
15285 PR tree-optimization/84086
15286 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
15287 (flush_ssaname_freelist): When SSA names were released reset
15288 the SCEV hash table.
15290 2018-01-29 Richard Biener <rguenther@suse.de>
15292 PR tree-optimization/84057
15293 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
15294 removed paths when removing edges.
15296 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
15298 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
15299 -mfunction-return=@var{choice}.
15301 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
15303 PR diagnostic/84034
15304 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
15305 Handle CR like TAB.
15306 (layout::print_source_line): Likewise.
15307 (test_get_line_width_without_trailing_whitespace): Add test cases.
15309 2018-01-27 Jakub Jelinek <jakub@redhat.com>
15311 PR middle-end/84040
15312 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
15315 2018-01-26 Jim Wilson <jimw@sifive.com>
15317 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
15319 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
15322 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15324 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
15325 and CMP + SUB-immediate -> SUBS.
15327 2018-01-26 Martin Sebor <msebor@redhat.com>
15329 PR tree-optimization/83896
15330 * tree-ssa-strlen.c (get_string_len): Rename...
15331 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
15332 Avoid assuming length is constant.
15333 (handle_char_store): Use HOST_WIDE_INT for string length.
15335 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
15338 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
15339 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
15341 2018-01-26 Richard Biener <rguenther@suse.de>
15343 PR rtl-optimization/84003
15344 * dse.c (record_store): Only record redundant stores when
15345 the earlier store aliases at least all accesses the later one does.
15347 2018-01-26 Jakub Jelinek <jakub@redhat.com>
15349 PR rtl-optimization/83985
15350 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
15351 REG_CFA_RESTORE insns.
15352 (delete_unmarked_insns): Don't ignore separate shrink wrapping
15353 REG_CFA_RESTORE insns here.
15356 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
15357 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
15359 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15361 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
15362 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
15363 (arc_init): Likewise.
15364 (arc_override_options): Likewise.
15365 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
15367 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
15369 * config/arc/arc.h (TARGET_DBNZ): Define.
15370 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
15371 properly set the tune attribute.
15372 (dbnz): Use TARGET_DBNZ guard.
15373 * config/arc/arc.opt (mtune): Add core3 option.
15375 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15377 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
15378 recognize new pic like addresses.
15379 (arc_delegitimize_address): Clean up.
15381 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15383 * config/arc/arc-arches.def: Option mrf16 valid for all
15385 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
15386 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
15387 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
15388 * config/arc/arc-tables.opt: Regenerate.
15389 * config/arc/arc.c (arc_conditional_register_usage): Handle
15390 reduced register file case.
15391 (arc_file_start): Set must have build attributes.
15392 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
15393 mrf16 option value.
15394 * config/arc/arc.opt (mrf16): Add new option.
15395 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
15396 * config/arc/genmultilib.awk: Handle new mrf16 option.
15397 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
15398 * config/arc/t-multilib: Regenerate.
15399 * doc/invoke.texi (ARC Options): Document mrf16 option.
15401 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15403 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
15404 * config/arc/arc.c (arc_handle_secure_attribute): New function.
15405 (arc_attribute_table): Add 'secure_call' attribute.
15406 (arc_print_operand): Print secure call operand.
15407 (arc_function_ok_for_sibcall): Don't optimize tail calls when
15409 (arc_is_secure_call_p): New function. * config/arc/arc.md
15410 (call_i): Add support for sjli instruction.
15411 (call_value_i): Likewise.
15412 * config/arc/constraints.md (Csc): New constraint.
15414 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15415 John Eric Martin <John.Martin@emmicro-us.com>
15417 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
15418 * config/arc/arc.c (_arc_jli_section): New struct.
15419 (arc_jli_section): New type.
15420 (rc_jli_sections): New static variable.
15421 (arc_handle_jli_attribute): New function.
15422 (arc_attribute_table): Add jli_always and jli_fixed attribute.
15423 (arc_file_end): New function.
15424 (TARGET_ASM_FILE_END): Define.
15425 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
15426 (arc_add_jli_section): New function.
15427 (jli_call_scan): Likewise.
15428 (arc_reorg): Call jli_call_scan.
15429 (arc_output_addsi): Remove 'S' from printing asm operand.
15430 (arc_is_jli_call_p): New function.
15431 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
15433 (movhi_insn): Likewise.
15434 (movsi_insn): Likewise.
15435 (movsi_set_cc_insn): Likewise.
15436 (loadqi_update): Likewise.
15437 (load_zeroextendqisi_update): Likewise.
15438 (load_signextendqisi_update): Likewise.
15439 (loadhi_update): Likewise.
15440 (load_zeroextendhisi_update): Likewise.
15441 (load_signextendhisi_update): Likewise.
15442 (loadsi_update): Likewise.
15443 (loadsf_update): Likewise.
15444 (movsicc_insn): Likewise.
15445 (bset_insn): Likewise.
15446 (bxor_insn): Likewise.
15447 (bclr_insn): Likewise.
15448 (bmsk_insn): Likewise.
15449 (bicsi3_insn): Likewise.
15450 (cmpsi_cc_c_insn): Likewise.
15451 (movsi_ne): Likewise.
15452 (movsi_cond_exec): Likewise.
15453 (clrsbsi2): Likewise.
15454 (norm_f): Likewise.
15464 (call_i): Remove 'S' asm letter, add jli instruction.
15465 (call_value_i): Likewise.
15466 * config/arc/arc.op (mjli-always): New option.
15467 * config/arc/constraints.md (Cji): New constraint.
15468 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
15470 (subsf3_fpx): Likewise.
15471 (mulsf3_fpx): Likewise.
15472 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
15474 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
15475 function attrbutes.
15476 * doc/invoke.texi (ARC): Document mjli-always option.
15478 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
15480 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
15481 avoid addition with 0 and use incw and decw where possible.
15483 2018-01-26 Richard Biener <rguenther@suse.de>
15485 PR tree-optimization/81082
15486 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
15487 association if it requires casting to unsigned.
15488 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
15489 from fold_plusminus_mult_expr to catch important cases late when
15490 range info is available.
15492 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15494 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
15495 * configure.ac (hidden_linkonce): New test.
15496 * configure: Regenerate.
15497 * config.in: Regenerate.
15499 2018-01-26 Julia Koval <julia.koval@intel.com>
15501 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
15502 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
15503 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
15504 _mm_mask_bitshuffle_epi64_mask): Fix type.
15505 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
15506 USI_FTYPE_V4DI_V4DI_USI): Remove.
15507 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
15508 __builtin_ia32_vpshufbitqmb256_mask,
15509 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
15510 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
15511 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
15513 2018-01-26 Alan Modra <amodra@gmail.com>
15516 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
15517 UNSPEC_VBPERMQ. Sort other unspecs.
15519 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
15521 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
15523 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
15525 PR middle-end/83055
15526 * predict.c (drop_profile): Do not push/pop cfun; update also
15528 (handle_missing_profiles): Fix logic looking for zero profiles.
15530 2018-01-25 Jakub Jelinek <jakub@redhat.com>
15532 PR middle-end/83977
15533 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
15534 on functions with #pragma omp declare simd or functions with simd
15536 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
15537 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
15538 Remove trailing \n from warning_at calls.
15540 2018-01-25 Tom de Vries <tom@codesourcery.com>
15543 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
15544 for neutered workers.
15546 2018-01-24 Joseph Myers <joseph@codesourcery.com>
15549 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
15550 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
15552 2018-01-24 Jeff Law <law@redhat.com>
15555 * i386.c (get_probe_interval): Move to earlier point.
15556 (ix86_compute_frame_layout): If -fstack-clash-protection and
15557 the frame is larger than the probe interval, then use pushes
15558 to save registers rather than reg->mem moves.
15559 (ix86_expand_prologue): Remove conditional for int_registers_saved
15562 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
15565 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
15566 min/max for never referenced object.
15568 2018-01-24 Jakub Jelinek <jakub@redhat.com>
15570 PR middle-end/83977
15571 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
15573 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
15574 attributes from DECL_ATTRIBUTES (decl) without affecting
15575 DECL_ATTRIBUTES (current_function_decl).
15576 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
15577 functions with non-NULL DECL_ABSTRACT_ORIGIN.
15579 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
15581 PR tree-optimization/83979
15582 * fold-const.c (fold_comparison): Use constant_boolean_node
15583 instead of boolean_{true,false}_node.
15585 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
15587 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
15590 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15592 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15593 Simplify the clause that sets the length attribute.
15594 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15595 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
15596 clause that sets the length attribute.
15597 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15599 2018-01-24 Tom de Vries <tom@codesourcery.com>
15602 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
15603 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
15604 Add strict parameter.
15605 (prevent_branch_around_nothing): Insert dummy insn between branch to
15606 label and label with no ptx insn inbetween.
15607 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
15609 2018-01-24 Tom de Vries <tom@codesourcery.com>
15612 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
15613 for neutered threads in warp.
15614 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
15616 2018-01-24 Richard Biener <rguenther@suse.de>
15618 PR tree-optimization/83176
15619 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
15622 2018-01-24 Richard Biener <rguenther@suse.de>
15624 PR tree-optimization/82819
15625 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
15626 code generating pluses that are no-ops in the target precision.
15628 2018-01-24 Richard Biener <rguenther@suse.de>
15630 PR middle-end/84000
15631 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
15633 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15635 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
15636 to merge probabilities.
15637 * predict.c (probably_never_executed): Also mark as cold functions
15638 with global 0 profile and guessed local profile.
15639 * profile-count.c (profile_probability::combine_with_count): New
15641 * profile-count.h (profile_probability::operator*,
15642 profile_probability::operator*=, profile_probability::operator/,
15643 profile_probability::operator/=): Reduce precision to adjusted
15644 and set value to guessed on contradictory divisions.
15645 (profile_probability::combine_with_freq): Remove.
15646 (profile_probability::combine_wiht_count): Declare.
15647 (profile_count::force_nonzero):: Set to adjusted.
15648 (profile_count::probability_in):: Set quality to adjusted.
15649 * tree-ssa-tail-merge.c (replace_block_by): Use
15650 combine_with_count.
15652 2018-01-23 Andrew Waterman <andrew@sifive.com>
15653 Jim Wilson <jimw@sifive.com>
15655 * config/riscv/riscv.c (riscv_stack_boundary): New.
15656 (riscv_option_override): Set riscv_stack_boundary. Handle
15657 riscv_preferred_stack_boundary_arg.
15658 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
15659 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
15660 (STACK_BOUNDARY): Set to riscv_stack_boundary.
15661 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
15662 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
15663 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
15665 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
15668 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
15669 of struct ix86_frame.
15670 (ix86_expand_epilogue): Likewise. Add a local variable for
15671 the reg_save_offset field in struct ix86_frame.
15673 2018-01-23 Bin Cheng <bin.cheng@arm.com>
15675 PR tree-optimization/82604
15676 * tree-loop-distribution.c (enum partition_kind): New enum item
15677 PKIND_PARTIAL_MEMSET.
15678 (partition_builtin_p): Support above new enum item.
15679 (generate_code_for_partition): Ditto.
15680 (compute_access_range): Differentiate cases that equality can be
15681 proven at all loops, the innermost loops or no loops.
15682 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
15683 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
15684 (finalize_partitions, distribute_loop): Don't fuse partition of
15685 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
15686 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
15687 parloop is enabled.
15689 2018-01-23 Martin Liska <mliska@suse.cz>
15691 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
15692 order to ignore the predictor.
15693 (PRED_POLYMORPHIC_CALL): Likewise.
15694 (PRED_RECURSIVE_CALL): Likewise.
15696 2018-01-23 Martin Liska <mliska@suse.cz>
15698 * tree-profile.c (tree_profiling): Print function header to
15699 aware reader which function we are working on.
15700 * value-prof.c (gimple_find_values_to_profile): Do not print
15701 not interesting value histograms.
15703 2018-01-23 Martin Liska <mliska@suse.cz>
15705 * profile-count.h (enum profile_quality): Add
15706 profile_uninitialized as the first value. Do not number values
15707 as they are zero based.
15708 (profile_count::verify): Update sanity check.
15709 (profile_probability::verify): Likewise.
15711 2018-01-23 Nathan Sidwell <nathan@acm.org>
15713 * doc/invoke.texi (ffor-scope): Deprecate.
15715 2018-01-23 David Malcolm <dmalcolm@redhat.com>
15717 PR tree-optimization/83510
15718 * domwalk.c (set_all_edges_as_executable): New function.
15719 (dom_walker::dom_walker): Convert bool param
15720 "skip_unreachable_blocks" to enum reachability. Move setup of
15721 edge flags to set_all_edges_as_executable and only do it when
15722 reachability is REACHABLE_BLOCKS.
15723 * domwalk.h (enum dom_walker::reachability): New enum.
15724 (dom_walker::dom_walker): Convert bool param
15725 "skip_unreachable_blocks" to enum reachability.
15726 (set_all_edges_as_executable): New decl.
15727 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
15728 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
15730 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
15731 but converting true to REACHABLE_BLOCKS.
15732 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
15734 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
15735 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
15736 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
15738 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
15739 if check_all_array_refs will be called.
15741 2018-01-23 David Malcolm <dmalcolm@redhat.com>
15743 * tree.c (selftest::test_location_wrappers): Add more test
15746 2018-01-23 David Malcolm <dmalcolm@redhat.com>
15748 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
15749 (selftest::test_bit_in_range): Likewise.
15751 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
15754 * doc/sourcebuild.texi (vect_float): Say that the selector
15755 only describes the situation when -funsafe-math-optimizations is on.
15756 (vect_float_strict): Document.
15758 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
15760 PR tree-optimization/83965
15761 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
15762 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
15763 instead of checking only for a reduction.
15764 (vect_recog_widen_sum_pattern): Likewise.
15766 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15768 * predict.c (probably_never_executed): Only use precise profile info.
15769 (compute_function_frequency): Skip after inlining hack since we now
15770 have quality checking.
15772 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15774 * profile-count.h (profile_probability::very_unlikely,
15775 profile_probability::unlikely, profile_probability::even): Set
15776 precision to guessed.
15778 2018-01-23 Richard Biener <rguenther@suse.de>
15780 PR tree-optimization/83963
15781 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
15782 Properly terminate dominator walk when crossing the exit edge not
15783 when visiting its source block.
15785 2018-01-23 Jakub Jelinek <jakub@redhat.com>
15788 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
15789 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
15791 2018-01-22 Jakub Jelinek <jakub@redhat.com>
15793 PR tree-optimization/83957
15794 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
15795 semicolon after for body surrounded by braces.
15797 PR tree-optimization/83081
15798 * profile-count.h (profile_probability::split): New method.
15799 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
15800 Use profile_probability::split.
15801 (do_compare_rtx_and_jump): Fix adjustment of probabilities
15802 when splitting a single conditional jump into 2.
15804 2018-01-22 David Malcolm <dmalcolm@redhat.com>
15806 PR tree-optimization/69452
15807 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
15810 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15812 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
15813 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
15814 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
15816 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15818 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
15820 * config/rl78/rl78.md (movdi): New define_expand.
15821 * config/rl78/rl78.c (rl78_split_movdi): New function.
15823 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
15826 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
15828 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
15829 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
15830 128-bit to produce an UNSPEC move to get the double word with the
15831 signbit and then a shift directly to do signbit.
15832 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
15833 implementation with a new version that just does either a direct
15834 move or a regular move. Move memory interface to separate insns.
15835 Move insns so they are next to the expander.
15836 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
15837 with signbit move. Split big and little endian case.
15838 (signbit<mode>2_dm_mem_le): Likewise.
15839 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
15840 (signbit<mode>2_dm2): Likewise.
15842 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15844 * config/rl78/rl78.md (anddi3): New define_expand.
15846 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15848 * config/rl78/rl78.md (umindi3): New define_expand.
15850 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15852 * config/rl78/rl78.md (smindi3): New define_expand.
15854 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15856 * config/rl78/rl78.md (smaxdi3): New define_expand.
15858 2018-01-22 Carl Love <cel@us.ibm.com>
15860 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
15861 LVX_V1TI): Add macro expansion.
15862 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
15863 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
15864 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
15865 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
15866 Change check to determine if the instruction is a byte reversing
15867 entry. Fix typo in comment.
15868 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
15869 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
15870 Add def_builtin calls for new builtins.
15871 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
15872 Add define_insn expansion.
15874 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15876 * config/rl78/rl78.md (umaxdi3): New define_expand.
15878 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15880 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
15881 for non-QImode registers.
15883 2018-01-22 Richard Biener <rguenther@suse.de>
15885 PR tree-optimization/83963
15886 * graphite-scop-detection.c (scop_detection::get_sese): Delay
15887 including the loop exit block.
15888 (scop_detection::merge_sese): Likewise.
15889 (scop_detection::add_scop): Do it here instead.
15891 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15893 * doc/sourcebuild.texi (arm_softfloat): Document.
15895 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
15898 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
15899 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
15900 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
15902 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15903 David Edelsohn <dje.gcc@gmail.com>
15906 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15907 Change "crset eq" to "crset 2".
15908 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15909 (*call_indirect_aix<mode>_nospec): Likewise.
15910 (*call_value_indirect_aix<mode>_nospec): Likewise.
15911 (*call_indirect_elfv2<mode>_nospec): Likewise.
15912 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
15913 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
15914 change assembly output from . to $.
15915 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15916 (indirect_jump<mode>_nospec): Change assembly output from . to $.
15917 (*tablejump<mode>_internal1_nospec): Likewise.
15919 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
15922 * config/sh/sh_optimize_sett_clrt.cc:
15923 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
15925 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
15927 PR tree-optimization/83940
15928 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
15929 offset_dt to vect_constant_def rather than vect_unknown_def_type.
15930 (vect_check_load_store_mask): Add a mask_dt_out parameter and
15931 use it to pass back the definition type.
15932 (vect_check_store_rhs): Likewise rhs_dt_out.
15933 (vect_build_gather_load_calls): Add a mask_dt argument and use
15934 it instead of a call to vect_is_simple_use.
15935 (vectorizable_store): Update calls to vect_check_load_store_mask
15936 and vect_check_store_rhs. Use the dt returned by the latter instead
15937 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
15938 instead of calls to vect_is_simple_use. Pass the scalar rather
15939 than the vector operand to vect_is_simple_use when handling
15940 second and subsequent copies of an rhs value.
15941 (vectorizable_load): Update calls to vect_check_load_store_mask
15942 and vect_build_gather_load_calls. Use the cached mask_dt and
15943 gs_info.offset_dt instead of calls to vect_is_simple_use.
15945 2018-01-20 Jakub Jelinek <jakub@redhat.com>
15947 PR middle-end/83945
15948 * tree-emutls.c: Include gimplify.h.
15949 (lower_emutls_2): New function.
15950 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
15951 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
15952 it before further processing.
15955 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
15956 UINTVAL (trueop1) instead of INTVAL (op1).
15958 2018-01-19 Jakub Jelinek <jakub@redhat.com>
15962 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
15963 INCOMING_FRAME_SP_OFFSET if not defined.
15964 (scan_trace): Add ENTRY argument. If true and
15965 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
15966 emit a note to adjust the CFA offset.
15967 (create_cfi_notes): Adjust scan_trace callers.
15968 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
15969 INCOMING_FRAME_SP_OFFSET in the CIE.
15970 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
15971 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
15973 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
15974 * doc/tm.texi: Regenerated.
15976 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15978 PR rtl-optimization/83147
15979 * lra-constraints.c (remove_inheritance_pseudos): Use
15980 lra_substitute_pseudo_within_insn.
15982 2018-01-19 Tom de Vries <tom@codesourcery.com>
15983 Cesar Philippidis <cesar@codesourcery.com>
15986 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
15988 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
15991 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
15992 spaces for function labels.
15994 2018-01-19 Martin Liska <mliska@suse.cz>
15996 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
15997 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
15998 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
15999 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
16000 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
16001 (PRED_CONST_RETURN): Change from 69 to 65.
16002 (PRED_NULL_RETURN): Change from 91 to 71.
16003 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
16004 (PRED_LOOP_GUARD): Change from 66 to 73.
16006 2018-01-19 Martin Liska <mliska@suse.cz>
16008 * predict.c (predict_insn_def): Add new assert.
16009 (struct branch_predictor): Change type to signed integer.
16010 (test_prediction_value_range): Amend test to cover
16011 PROB_UNINITIALIZED.
16012 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
16013 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
16014 (PRED_LOOP_ITERATIONS_MAX): Likewise.
16015 (PRED_LOOP_IV_COMPARE): Likewise.
16016 * predict.h (PROB_UNINITIALIZED): Define new constant.
16018 2018-01-19 Martin Liska <mliska@suse.cz>
16020 * predict.c (dump_prediction): Add new format for
16021 analyze_brprob.py script which is enabled with -details
16023 * profile-count.h (precise_p): New function.
16025 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
16027 PR tree-optimization/83922
16028 * tree-vect-loop.c (vect_verify_full_masking): Return false if
16029 there are no statements that need masking.
16030 (vect_active_double_reduction_p): New function.
16031 (vect_analyze_loop_operations): Use it when handling phis that
16032 are not in the loop header.
16034 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
16036 PR tree-optimization/83914
16037 * tree-vect-loop.c (vectorizable_induction): Don't convert
16038 init_expr or apply the peeling adjustment for inductions
16039 that are nested within the vectorized loop.
16041 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16043 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
16046 2018-01-18 Jakub Jelinek <jakub@redhat.com>
16050 * function.h (gimplify_parameters): Add gimple_seq * argument.
16051 * function.c: Include gimple.h and options.h.
16052 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
16053 for the added local temporaries if needed.
16054 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
16055 if there are any parameter cleanups, wrap whole body into a
16056 try/finally with the cleanups.
16058 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
16061 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
16062 Use GET_MODE_CLASS for scalar floating point.
16064 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
16068 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
16069 Fix call of call_cgraph_insertion_hooks.
16071 2018-01-18 Martin Sebor <msebor@redhat.com>
16073 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
16075 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
16078 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
16081 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
16084 * common.opt: (-ffile-prefix-map): New option.
16085 * opts.c (common_handle_option): Defer it.
16086 * opts-global.c (handle_common_deferred_options): Handle it.
16087 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
16088 * file-prefix-map.h: New file.
16089 (remap_debug_filename, add_debug_prefix_map): ...here.
16090 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
16091 * final.c (debug_prefix_map, add_debug_prefix_map
16092 remap_debug_filename): Move to...
16093 * file-prefix-map.c: New file.
16094 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
16095 generalize, get rid of alloca(), use strrchr() instead of strchr().
16096 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
16097 Implement in terms of add_prefix_map().
16098 (remap_macro_filename, remap_debug_filename): Implement in term of
16100 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
16101 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
16102 * dbxout.c: Include file-prefix-map.h.
16103 * varasm.c: Likewise.
16104 * vmsdbgout.c: Likewise.
16105 * xcoffout.c: Likewise.
16106 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
16107 * doc/cppopts.texi (-fmacro-prefix-map): Document.
16108 * doc/invoke.texi (-ffile-prefix-map): Document.
16109 (-fdebug-prefix-map): Update description.
16111 2018-01-18 Martin Liska <mliska@suse.cz>
16113 * config/i386/i386.c (indirect_thunk_name): Document that also
16115 (output_indirect_thunk): Document why both instructions
16116 (pause and lfence) are generated.
16118 2018-01-18 Richard Biener <rguenther@suse.de>
16120 PR tree-optimization/83887
16121 * graphite-scop-detection.c
16122 (scop_detection::get_nearest_dom_with_single_entry): Remove.
16123 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
16124 (scop_detection::merge_sese): Re-implement with a flood-fill
16125 algorithm that properly finds a SESE region if it exists.
16127 2018-01-18 Jakub Jelinek <jakub@redhat.com>
16130 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
16131 pointer_diff optimizations use view_convert instead of convert.
16133 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16135 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
16136 Generate different code for -mno-speculate-indirect-jumps.
16137 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
16138 (*call_indirect_aix<mode>): Disable for
16139 -mno-speculate-indirect-jumps.
16140 (*call_indirect_aix<mode>_nospec): New define_insn.
16141 (*call_value_indirect_aix<mode>): Disable for
16142 -mno-speculate-indirect-jumps.
16143 (*call_value_indirect_aix<mode>_nospec): New define_insn.
16144 (*sibcall_nonlocal_sysv<mode>): Generate different code for
16145 -mno-speculate-indirect-jumps.
16146 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
16148 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
16150 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
16151 long double type, set the flags for noting the default long double
16152 type, even if we don't pass or return a long double type.
16154 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
16157 * ipa-inline.c (flatten_function): Do not overwrite final inlining
16160 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
16162 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
16163 support for merge[hl].
16164 (fold_mergehl_helper): New helper function.
16165 (tree-vector-builder.h): New #include for tree_vector_builder usage.
16166 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
16167 (altivec_vmrglw_direct): Add xxmrglw insn.
16169 2018-01-17 Andrew Waterman <andrew@sifive.com>
16171 * config/riscv/riscv.c (riscv_conditional_register_usage): If
16172 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
16174 2018-01-17 David Malcolm <dmalcolm@redhat.com>
16177 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
16178 call the lto_location_cache before reading the
16179 DECL_SOURCE_LOCATION of the types.
16181 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
16182 Richard Sandiford <richard.sandiford@linaro.org>
16184 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
16185 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
16186 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
16187 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
16188 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
16190 * config/aarch64/constraints.md (aarch64_movti_operand):
16192 * config/aarch64/predicates.md (Uti): Add new constraint.
16194 2018-01-17 Carl Love <cel@us.ibm.com>
16196 * config/rs6000/vsx.md (define_expand xl_len_r,
16197 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
16198 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
16200 (define_expand, define_insn): Move the shift left from the
16201 define_insn to the define_expand for lxvl and stxvl instructions.
16202 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
16203 and XL_LEN_R definitions to PURE.
16205 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
16207 * config/i386/i386.c (indirect_thunk_name): Declare regno
16208 as unsigned int. Compare regno with INVALID_REGNUM.
16209 (output_indirect_thunk): Ditto.
16210 (output_indirect_thunk_function): Ditto.
16211 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
16212 in the call to output_indirect_thunk_function.
16214 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
16216 PR middle-end/83884
16217 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
16218 rather than the size of inner_type to determine the stack slot size
16219 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
16221 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
16224 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
16227 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
16229 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
16230 endian Linux systems to optionally enable multilibs for selecting
16231 the long double type if the user configured an explicit type.
16232 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
16233 have no long double multilibs if not defined.
16234 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
16235 warn if the user used -mabi={ieee,ibm}longdouble and we built
16236 multilibs for long double.
16237 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
16238 appropriate multilib option.
16239 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
16241 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
16242 for building long double multilibs.
16243 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
16245 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
16247 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
16250 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
16252 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
16255 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
16258 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
16261 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
16263 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
16264 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
16266 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
16268 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
16269 different rtl trees depending on TARGET_64BIT.
16270 (rs6000_gen_lvx): Likewise.
16272 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
16274 * config/visium/visium.md (nop): Tweak comment.
16275 (hazard_nop): Likewise.
16277 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16279 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
16280 -mspeculate-indirect-jumps.
16281 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
16282 for -mno-speculate-indirect-jumps.
16283 (*call_indirect_elfv2<mode>_nospec): New define_insn.
16284 (*call_value_indirect_elfv2<mode>): Disable for
16285 -mno-speculate-indirect-jumps.
16286 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
16287 (indirect_jump): Emit different RTL for
16288 -mno-speculate-indirect-jumps.
16289 (*indirect_jump<mode>): Disable for
16290 -mno-speculate-indirect-jumps.
16291 (*indirect_jump<mode>_nospec): New define_insn.
16292 (tablejump): Emit different RTL for
16293 -mno-speculate-indirect-jumps.
16294 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
16295 (tablejumpsi_nospec): New define_expand.
16296 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
16297 (tablejumpdi_nospec): New define_expand.
16298 (*tablejump<mode>_internal1): Disable for
16299 -mno-speculate-indirect-jumps.
16300 (*tablejump<mode>_internal1_nospec): New define_insn.
16301 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
16304 2018-01-16 Artyom Skrobov tyomitch@gmail.com
16306 * caller-save.c (insert_save): Drop unnecessary parameter. All
16309 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16310 Richard Biener <rguenth@suse.de>
16313 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
16314 return early, inline manually is_gimple_sizepos. Make sure if we
16315 call gimplify_expr we don't end up with a gimple constant.
16316 * tree.c (variably_modified_type_p): Don't return true for
16317 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
16318 * gimplify.h (is_gimple_sizepos): Remove.
16320 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16322 PR tree-optimization/83857
16323 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
16324 vectorizable_live_operation for pure SLP statements.
16325 (vectorizable_live_operation): Handle PHIs.
16327 2018-01-16 Richard Biener <rguenther@suse.de>
16329 PR tree-optimization/83867
16330 * tree-vect-stmts.c (vect_transform_stmt): Precompute
16331 nested_in_vect_loop_p since the scalar stmt may get invalidated.
16333 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16336 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
16337 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
16338 If off is not INTEGER_CST, issue a may not be aligned warning
16339 rather than isn't aligned. Use isn%'t rather than isn't.
16340 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
16342 <case MULT_EXPR>: Improve the case when bottom and one of the
16343 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
16344 operand, in that case check if the other operand is multiple of
16345 bottom divided by the INTEGER_CST operand.
16347 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16350 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
16351 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
16352 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
16353 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
16354 * config/pa/pa.c (pa_function_arg_advance): Likewise.
16355 (pa_function_arg, pa_arg_partial_bytes): Likewise.
16356 (pa_function_arg_size): New function.
16358 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16360 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
16361 in a separate statement.
16363 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16365 PR tree-optimization/83847
16366 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
16367 group gathers and scatters.
16369 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16371 PR rtl-optimization/86620
16372 * params.def (max-sched-ready-insns): Bump minimum value to 1.
16374 PR rtl-optimization/83213
16375 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
16376 to last if both are JUMP_INSNs.
16378 PR tree-optimization/83843
16379 * gimple-ssa-store-merging.c
16380 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
16381 store_immediate_info for bswap/nop orig_stores.
16383 2018-01-15 Andrew Waterman <andrew@sifive.com>
16385 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
16387 <UDIV>: Increase cost if !TARGET_DIV.
16389 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
16391 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
16392 (define_attr "cr_logical_3op"): New.
16393 (cceq_ior_compare): Adjust.
16394 (cceq_ior_compare_complement): Adjust.
16395 (*cceq_rev_compare): Adjust.
16396 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
16397 (is_cracked_insn): Adjust.
16398 (insn_must_be_first_in_group): Adjust.
16399 * config/rs6000/40x.md: Adjust.
16400 * config/rs6000/440.md: Adjust.
16401 * config/rs6000/476.md: Adjust.
16402 * config/rs6000/601.md: Adjust.
16403 * config/rs6000/603.md: Adjust.
16404 * config/rs6000/6xx.md: Adjust.
16405 * config/rs6000/7450.md: Adjust.
16406 * config/rs6000/7xx.md: Adjust.
16407 * config/rs6000/8540.md: Adjust.
16408 * config/rs6000/cell.md: Adjust.
16409 * config/rs6000/e300c2c3.md: Adjust.
16410 * config/rs6000/e500mc.md: Adjust.
16411 * config/rs6000/e500mc64.md: Adjust.
16412 * config/rs6000/e5500.md: Adjust.
16413 * config/rs6000/e6500.md: Adjust.
16414 * config/rs6000/mpc.md: Adjust.
16415 * config/rs6000/power4.md: Adjust.
16416 * config/rs6000/power5.md: Adjust.
16417 * config/rs6000/power6.md: Adjust.
16418 * config/rs6000/power7.md: Adjust.
16419 * config/rs6000/power8.md: Adjust.
16420 * config/rs6000/power9.md: Adjust.
16421 * config/rs6000/rs64.md: Adjust.
16422 * config/rs6000/titan.md: Adjust.
16424 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16426 * config/i386/predicates.md (indirect_branch_operand): Rewrite
16427 ix86_indirect_branch_register logic.
16429 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16431 * config/i386/constraints.md (Bs): Update
16432 ix86_indirect_branch_register check. Don't check
16433 ix86_indirect_branch_register with GOT_memory_operand.
16435 * config/i386/predicates.md (GOT_memory_operand): Don't check
16436 ix86_indirect_branch_register here.
16437 (GOT32_symbol_operand): Likewise.
16439 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16441 * config/i386/predicates.md (constant_call_address_operand):
16442 Rewrite ix86_indirect_branch_register logic.
16443 (sibcall_insn_operand): Likewise.
16445 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16447 * config/i386/constraints.md (Bs): Replace
16448 ix86_indirect_branch_thunk_register with
16449 ix86_indirect_branch_register.
16451 * config/i386/i386.md (indirect_jump): Likewise.
16452 (tablejump): Likewise.
16453 (*sibcall_memory): Likewise.
16454 (*sibcall_value_memory): Likewise.
16455 Peepholes of indirect call and jump via memory: Likewise.
16456 * config/i386/i386.opt: Likewise.
16457 * config/i386/predicates.md (indirect_branch_operand): Likewise.
16458 (GOT_memory_operand): Likewise.
16459 (call_insn_operand): Likewise.
16460 (sibcall_insn_operand): Likewise.
16461 (GOT32_symbol_operand): Likewise.
16463 2018-01-15 Jakub Jelinek <jakub@redhat.com>
16465 PR middle-end/83837
16466 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
16467 type rather than type addr's type points to.
16468 (expand_omp_atomic_mutex): Likewise.
16469 (expand_omp_atomic): Likewise.
16471 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16474 * config/i386/i386.c (output_indirect_thunk_function): Use
16475 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
16476 for __x86_return_thunk.
16478 2018-01-15 Richard Biener <rguenther@suse.de>
16480 PR middle-end/83850
16481 * expmed.c (extract_bit_field_1): Fix typo.
16483 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16486 * config/arm/iterators.md (VF): New mode iterator.
16487 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
16488 Remove integer-related logic from pattern.
16489 (neon_vabd<mode>_3): Likewise.
16491 2018-01-15 Jakub Jelinek <jakub@redhat.com>
16493 PR middle-end/82694
16494 * common.opt (fstrict-overflow): No longer an alias.
16495 (fwrapv-pointer): New option.
16496 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
16497 also for pointer types based on flag_wrapv_pointer.
16498 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
16499 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
16500 opts->x_flag_wrapv got set.
16501 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
16502 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
16503 POINTER_TYPE_OVERFLOW_UNDEFINED.
16504 * match.pd: Likewise in address comparison pattern.
16505 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
16507 2018-01-15 Richard Biener <rguenther@suse.de>
16510 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
16511 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
16512 Reset type names to their identifier if their TYPE_DECL doesn't
16513 have linkage (and thus is used for ODR and devirt).
16514 (save_debug_info_for_decl): Remove.
16515 (save_debug_info_for_type): Likewise.
16516 (add_tree_to_fld_list): Adjust.
16517 * tree-pretty-print.c (dump_generic_node): Make dumping of
16518 type names more robust.
16520 2018-01-15 Richard Biener <rguenther@suse.de>
16522 * BASE-VER: Bump to 8.0.1.
16524 2018-01-14 Martin Sebor <msebor@redhat.com>
16527 * builtins.c (check_access): Avoid warning when the no-warning bit
16530 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
16532 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
16533 * ira-color (allocno_hard_regs_compare): Likewise.
16535 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
16538 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
16539 Use .pushsection/.popsection.
16541 2018-01-14 Martin Sebor <msebor@redhat.com>
16544 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
16546 2018-01-14 Jakub Jelinek <jakub@redhat.com>
16548 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
16549 entry from extra_headers.
16550 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
16551 extra_headers, make the list bitwise identical to the i?86-*-* one.
16553 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16555 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
16556 -mcmodel=large with -mindirect-branch=thunk,
16557 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
16558 -mfunction-return=thunk-extern.
16559 * doc/invoke.texi: Document -mcmodel=large is incompatible with
16560 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
16561 -mfunction-return=thunk and -mfunction-return=thunk-extern.
16563 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16565 * config/i386/i386.c (print_reg): Print the name of the full
16566 integer register without '%'.
16567 (ix86_print_operand): Handle 'V'.
16568 * doc/extend.texi: Document 'V' modifier.
16570 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16572 * config/i386/constraints.md (Bs): Disallow memory operand for
16573 -mindirect-branch-register.
16575 * config/i386/predicates.md (indirect_branch_operand): Likewise.
16576 (GOT_memory_operand): Likewise.
16577 (call_insn_operand): Likewise.
16578 (sibcall_insn_operand): Likewise.
16579 (GOT32_symbol_operand): Likewise.
16580 * config/i386/i386.md (indirect_jump): Call convert_memory_address
16581 for -mindirect-branch-register.
16582 (tablejump): Likewise.
16583 (*sibcall_memory): Likewise.
16584 (*sibcall_value_memory): Likewise.
16585 Disallow peepholes of indirect call and jump via memory for
16586 -mindirect-branch-register.
16587 (*call_pop): Replace m with Bw.
16588 (*call_value_pop): Likewise.
16589 (*sibcall_pop_memory): Replace m with Bs.
16590 * config/i386/i386.opt (mindirect-branch-register): New option.
16591 * doc/invoke.texi: Document -mindirect-branch-register option.
16593 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16595 * config/i386/i386-protos.h (ix86_output_function_return): New.
16596 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
16597 set function_return_type.
16598 (indirect_thunk_name): Add ret_p to indicate thunk for function
16600 (output_indirect_thunk_function): Pass false to
16601 indirect_thunk_name.
16602 (ix86_output_indirect_branch_via_reg): Likewise.
16603 (ix86_output_indirect_branch_via_push): Likewise.
16604 (output_indirect_thunk_function): Create alias for function
16605 return thunk if regno < 0.
16606 (ix86_output_function_return): New function.
16607 (ix86_handle_fndecl_attribute): Handle function_return.
16608 (ix86_attribute_table): Add function_return.
16609 * config/i386/i386.h (machine_function): Add
16610 function_return_type.
16611 * config/i386/i386.md (simple_return_internal): Use
16612 ix86_output_function_return.
16613 (simple_return_internal_long): Likewise.
16614 * config/i386/i386.opt (mfunction-return=): New option.
16615 (indirect_branch): Mention -mfunction-return=.
16616 * doc/extend.texi: Document function_return function attribute.
16617 * doc/invoke.texi: Document -mfunction-return= option.
16619 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16621 * config/i386/i386-opts.h (indirect_branch): New.
16622 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
16623 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
16624 with local indirect jump when converting indirect call and jump.
16625 (ix86_set_indirect_branch_type): New.
16626 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
16627 (indirectlabelno): New.
16628 (indirect_thunk_needed): Likewise.
16629 (indirect_thunk_bnd_needed): Likewise.
16630 (indirect_thunks_used): Likewise.
16631 (indirect_thunks_bnd_used): Likewise.
16632 (INDIRECT_LABEL): Likewise.
16633 (indirect_thunk_name): Likewise.
16634 (output_indirect_thunk): Likewise.
16635 (output_indirect_thunk_function): Likewise.
16636 (ix86_output_indirect_branch_via_reg): Likewise.
16637 (ix86_output_indirect_branch_via_push): Likewise.
16638 (ix86_output_indirect_branch): Likewise.
16639 (ix86_output_indirect_jmp): Likewise.
16640 (ix86_code_end): Call output_indirect_thunk_function if needed.
16641 (ix86_output_call_insn): Call ix86_output_indirect_branch if
16643 (ix86_handle_fndecl_attribute): Handle indirect_branch.
16644 (ix86_attribute_table): Add indirect_branch.
16645 * config/i386/i386.h (machine_function): Add indirect_branch_type
16646 and has_local_indirect_jump.
16647 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
16649 (tablejump): Likewise.
16650 (*indirect_jump): Use ix86_output_indirect_jmp.
16651 (*tablejump_1): Likewise.
16652 (simple_return_indirect_internal): Likewise.
16653 * config/i386/i386.opt (mindirect-branch=): New option.
16654 (indirect_branch): New.
16657 (thunk-inline): Likewise.
16658 (thunk-extern): Likewise.
16659 * doc/extend.texi: Document indirect_branch function attribute.
16660 * doc/invoke.texi: Document -mindirect-branch= option.
16662 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
16665 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
16667 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
16669 * ipa-inline.c (want_inline_small_function_p): Return false if
16670 inlining has already failed with CIF_FINAL_ERROR.
16671 (update_caller_keys): Call want_inline_small_function_p before
16673 (update_callee_keys): Likewise.
16675 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
16677 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
16679 (rs6000_quadword_masked_address_p): Likewise.
16680 (quad_aligned_load_p): Likewise.
16681 (quad_aligned_store_p): Likewise.
16682 (const_load_sequence_p): Add comment to describe the outer-most loop.
16683 (mimic_memory_attributes_and_flags): New function.
16684 (rs6000_gen_stvx): Likewise.
16685 (replace_swapped_aligned_store): Likewise.
16686 (rs6000_gen_lvx): Likewise.
16687 (replace_swapped_aligned_load): Likewise.
16688 (replace_swapped_load_constant): Capitalize argument name in
16689 comment describing this function.
16690 (rs6000_analyze_swaps): Add a third pass to search for vector loads
16691 and stores that access quad-word aligned addresses and replace
16692 with stvx or lvx instructions when appropriate.
16693 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
16694 New function prototype.
16695 (rs6000_quadword_masked_address_p): Likewise.
16696 (rs6000_gen_lvx): Likewise.
16697 (rs6000_gen_stvx): Likewise.
16698 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
16699 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
16700 when memory address is aligned.
16701 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
16702 this split to select lvx instruction when memory address is aligned.
16703 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
16704 instruction when memory address is aligned.
16705 (*vsx_le_perm_load_v16qi): Likewise.
16706 (four unnamed splitters): Modify to select the stvx instruction
16707 when memory is aligned.
16709 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
16711 * predict.c (determine_unlikely_bbs): Handle correctly BBs
16712 which appears in the queue multiple times.
16714 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16715 Alan Hayward <alan.hayward@arm.com>
16716 David Sherwood <david.sherwood@arm.com>
16718 * tree-vectorizer.h (vec_lower_bound): New structure.
16719 (_loop_vec_info): Add check_nonzero and lower_bounds.
16720 (LOOP_VINFO_CHECK_NONZERO): New macro.
16721 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
16722 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
16723 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
16724 fields. Make seg_len the distance travelled, not including the
16726 (dr_direction_indicator): Declare.
16727 (dr_zero_step_indicator): Likewise.
16728 (dr_known_forward_stride_p): Likewise.
16729 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
16731 (runtime_alias_check_p): Allow runtime alias checks with
16733 (operator ==): Compare access_size and align.
16734 (prune_runtime_alias_test_list): Rework for new distinction between
16735 the access_size and seg_len.
16736 (create_intersect_range_checks_index): Likewise. Cope with polynomial
16738 (get_segment_min_max): New function.
16739 (create_intersect_range_checks): Use it.
16740 (dr_step_indicator): New function.
16741 (dr_direction_indicator): Likewise.
16742 (dr_zero_step_indicator): Likewise.
16743 (dr_known_forward_stride_p): Likewise.
16744 * tree-loop-distribution.c (data_ref_segment_size): Return
16745 DR_STEP * (niters - 1).
16746 (compute_alias_check_pairs): Update call to the dr_with_seg_len
16748 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
16749 (vect_preserves_scalar_order_p): New function, split out from...
16750 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
16751 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
16752 (vect_vfa_access_size): New function.
16753 (vect_vfa_align): Likewise.
16754 (vect_compile_time_alias): Take access_size_a and access_b arguments.
16755 (dump_lower_bound): New function.
16756 (vect_check_lower_bound): Likewise.
16757 (vect_small_gap_p): Likewise.
16758 (vectorizable_with_step_bound_p): Likewise.
16759 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
16760 depencies if the vectorization factor is 1. Convert the checks
16761 for nonzero steps into checks on the bounds of DR_STEP. Try using
16762 a bunds check for variable steps if the minimum required step is
16763 relatively small. Update calls to the dr_with_seg_len
16764 constructor and to vect_compile_time_alias.
16765 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
16767 (vect_loop_versioning): Call it.
16768 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
16770 (vect_estimate_min_profitable_iters): Account for any bounds checks.
16772 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16773 Alan Hayward <alan.hayward@arm.com>
16774 David Sherwood <david.sherwood@arm.com>
16776 * doc/sourcebuild.texi (vect_scatter_store): Document.
16777 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
16779 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
16781 * genopinit.c (main): Add supports_vec_scatter_store and
16782 supports_vec_scatter_store_cached to target_optabs.
16783 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
16784 IFN_MASK_SCATTER_STORE.
16785 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
16787 * internal-fn.h (internal_store_fn_p): Declare.
16788 (internal_fn_stored_value_index): Likewise.
16789 * internal-fn.c (scatter_store_direct): New macro.
16790 (expand_scatter_store_optab_fn): New function.
16791 (direct_scatter_store_optab_supported_p): New macro.
16792 (internal_store_fn_p): New function.
16793 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
16794 IFN_MASK_SCATTER_STORE.
16795 (internal_fn_mask_index): Likewise.
16796 (internal_fn_stored_value_index): New function.
16797 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
16798 for scatter stores.
16799 * optabs-query.h (supports_vec_scatter_store_p): Declare.
16800 * optabs-query.c (supports_vec_scatter_store_p): New function.
16801 * tree-vectorizer.h (vect_get_store_rhs): Declare.
16802 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
16803 true for scatter stores.
16804 (vect_gather_scatter_fn_p): Handle scatter stores too.
16805 (vect_check_gather_scatter): Consider using scatter stores if
16806 supports_vec_scatter_store_p.
16807 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
16808 scatter stores too.
16809 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
16810 internal_fn_stored_value_index.
16811 (check_load_store_masking): Handle scatter stores too.
16812 (vect_get_store_rhs): Make public.
16813 (vectorizable_call): Use internal_store_fn_p.
16814 (vectorizable_store): Handle scatter store internal functions.
16815 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
16816 when deciding whether the end of the group has been reached.
16817 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
16818 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
16819 (mask_scatter_store<mode>): New insns.
16821 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16822 Alan Hayward <alan.hayward@arm.com>
16823 David Sherwood <david.sherwood@arm.com>
16825 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
16826 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
16827 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
16829 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
16830 Use vect_truncate_gather_scatter_offset if we can't treat the
16831 operation as a normal gather load or scatter store.
16832 (get_group_load_store_type): Take the gather_scatter_info
16833 as argument. Try using a gather load or scatter store for
16834 single-element groups.
16835 (get_load_store_type): Update calls to get_group_load_store_type
16836 and vect_use_strided_gather_scatters_p.
16838 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16839 Alan Hayward <alan.hayward@arm.com>
16840 David Sherwood <david.sherwood@arm.com>
16842 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
16843 optional tree argument.
16844 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
16846 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
16847 but continue to use the current value as a fallback.
16848 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
16849 to compare the updates.
16850 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
16851 (get_load_store_type): Use it when handling a strided access.
16852 (vect_get_strided_load_store_ops): New function.
16853 (vect_get_data_ptr_increment): Likewise.
16854 (vectorizable_load): Handle strided gather loads. Always pass
16855 a step to vect_create_data_ref_ptr and bump_vector_ptr.
16857 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16858 Alan Hayward <alan.hayward@arm.com>
16859 David Sherwood <david.sherwood@arm.com>
16861 * doc/md.texi (gather_load@var{m}): Document.
16862 (mask_gather_load@var{m}): Likewise.
16863 * genopinit.c (main): Add supports_vec_gather_load and
16864 supports_vec_gather_load_cached to target_optabs.
16865 * optabs-tree.c (init_tree_optimization_optabs): Use
16866 ggc_cleared_alloc to allocate target_optabs.
16867 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
16868 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
16870 * internal-fn.h (internal_load_fn_p): Declare.
16871 (internal_gather_scatter_fn_p): Likewise.
16872 (internal_fn_mask_index): Likewise.
16873 (internal_gather_scatter_fn_supported_p): Likewise.
16874 * internal-fn.c (gather_load_direct): New macro.
16875 (expand_gather_load_optab_fn): New function.
16876 (direct_gather_load_optab_supported_p): New macro.
16877 (direct_internal_fn_optab): New function.
16878 (internal_load_fn_p): Likewise.
16879 (internal_gather_scatter_fn_p): Likewise.
16880 (internal_fn_mask_index): Likewise.
16881 (internal_gather_scatter_fn_supported_p): Likewise.
16882 * optabs-query.c (supports_at_least_one_mode_p): New function.
16883 (supports_vec_gather_load_p): Likewise.
16884 * optabs-query.h (supports_vec_gather_load_p): Declare.
16885 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
16886 and memory_type field.
16887 (NUM_PATTERNS): Bump to 15.
16888 * tree-vect-data-refs.c: Include internal-fn.h.
16889 (vect_gather_scatter_fn_p): New function.
16890 (vect_describe_gather_scatter_call): Likewise.
16891 (vect_check_gather_scatter): Try using internal functions for
16892 gather loads. Recognize existing calls to a gather load function.
16893 (vect_analyze_data_refs): Consider using gather loads if
16894 supports_vec_gather_load_p.
16895 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
16896 (vect_get_gather_scatter_offset_type): Likewise.
16897 (vect_convert_mask_for_vectype): Likewise.
16898 (vect_add_conversion_to_patterm): Likewise.
16899 (vect_try_gather_scatter_pattern): Likewise.
16900 (vect_recog_gather_scatter_pattern): New pattern recognizer.
16901 (vect_vect_recog_func_ptrs): Add it.
16902 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
16903 internal_fn_mask_index and internal_gather_scatter_fn_p.
16904 (check_load_store_masking): Take the gather_scatter_info as an
16905 argument and handle gather loads.
16906 (vect_get_gather_scatter_ops): New function.
16907 (vectorizable_call): Check internal_load_fn_p.
16908 (vectorizable_load): Likewise. Handle gather load internal
16910 (vectorizable_store): Update call to check_load_store_masking.
16911 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
16912 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
16913 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
16914 (aarch64_gather_scale_operand_d): New predicates.
16915 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
16916 (mask_gather_load<mode>): New insns.
16918 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16919 Alan Hayward <alan.hayward@arm.com>
16920 David Sherwood <david.sherwood@arm.com>
16922 * optabs.def (fold_left_plus_optab): New optab.
16923 * doc/md.texi (fold_left_plus_@var{m}): Document.
16924 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
16925 * internal-fn.c (fold_left_direct): Define.
16926 (expand_fold_left_optab_fn): Likewise.
16927 (direct_fold_left_optab_supported_p): Likewise.
16928 * fold-const-call.c (fold_const_fold_left): New function.
16929 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
16930 * tree-parloops.c (valid_reduction_p): New function.
16931 (gather_scalar_reductions): Use it.
16932 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
16933 (vect_finish_replace_stmt): Declare.
16934 * tree-vect-loop.c (fold_left_reduction_fn): New function.
16935 (needs_fold_left_reduction_p): New function, split out from...
16936 (vect_is_simple_reduction): ...here. Accept reductions that
16937 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
16938 (vect_force_simple_reduction): Also store the reduction type in
16939 the assignment's STMT_VINFO_REDUC_TYPE.
16940 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
16941 (merge_with_identity): New function.
16942 (vect_expand_fold_left): Likewise.
16943 (vectorize_fold_left_reduction): Likewise.
16944 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
16945 scalar phi in place for it. Check for target support and reject
16946 cases that would reassociate the operation. Defer the transform
16947 phase to vectorize_fold_left_reduction.
16948 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
16949 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
16950 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
16952 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16954 * tree-if-conv.c (predicate_mem_writes): Remove redundant
16955 call to ifc_temp_var.
16957 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16958 Alan Hayward <alan.hayward@arm.com>
16959 David Sherwood <david.sherwood@arm.com>
16961 * target.def (legitimize_address_displacement): Take the original
16962 offset as a poly_int.
16963 * targhooks.h (default_legitimize_address_displacement): Update
16965 * targhooks.c (default_legitimize_address_displacement): Likewise.
16966 * doc/tm.texi: Regenerate.
16967 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
16968 as an argument, moving assert of ad->disp == ad->disp_term to...
16969 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
16970 Try calling targetm.legitimize_address_displacement before expanding
16971 the address rather than afterwards, and adjust for the new interface.
16972 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
16973 Match the new hook interface. Handle SVE addresses.
16974 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
16975 new hook interface.
16977 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16979 * Makefile.in (OBJS): Add early-remat.o.
16980 * target.def (select_early_remat_modes): New hook.
16981 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
16982 * doc/tm.texi: Regenerate.
16983 * targhooks.h (default_select_early_remat_modes): Declare.
16984 * targhooks.c (default_select_early_remat_modes): New function.
16985 * timevar.def (TV_EARLY_REMAT): New timevar.
16986 * passes.def (pass_early_remat): New pass.
16987 * tree-pass.h (make_pass_early_remat): Declare.
16988 * early-remat.c: New file.
16989 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
16991 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
16993 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16994 Alan Hayward <alan.hayward@arm.com>
16995 David Sherwood <david.sherwood@arm.com>
16997 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
16998 vfm1 with a bound_epilog parameter.
16999 (vect_do_peeling): Update calls accordingly, and move the prologue
17000 call earlier in the function. Treat the base bound_epilog as 0 for
17001 fully-masked loops and retain vf - 1 for other loops. Add 1 to
17002 this base when peeling for gaps.
17003 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
17004 with fully-masked loops.
17005 (vect_estimate_min_profitable_iters): Handle the single peeled
17006 iteration in that case.
17008 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17009 Alan Hayward <alan.hayward@arm.com>
17010 David Sherwood <david.sherwood@arm.com>
17012 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
17013 single-element interleaving even if the size is not a power of 2.
17014 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
17015 accesses for single-element interleaving if the group size is
17018 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17019 Alan Hayward <alan.hayward@arm.com>
17020 David Sherwood <david.sherwood@arm.com>
17022 * doc/md.texi (fold_extract_last_@var{m}): Document.
17023 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
17024 * optabs.def (fold_extract_last_optab): New optab.
17025 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
17026 * internal-fn.c (fold_extract_direct): New macro.
17027 (expand_fold_extract_optab_fn): Likewise.
17028 (direct_fold_extract_optab_supported_p): Likewise.
17029 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
17030 * tree-vect-loop.c (vect_model_reduction_cost): Handle
17031 EXTRACT_LAST_REDUCTION.
17032 (get_initial_def_for_reduction): Do not create an initial vector
17033 for EXTRACT_LAST_REDUCTION reductions.
17034 (vectorizable_reduction): Leave the scalar phi in place for
17035 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
17036 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
17037 epilogue code for EXTRACT_LAST_REDUCTION and defer the
17038 transform phase to vectorizable_condition.
17039 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
17041 (vect_finish_stmt_generation): ...here.
17042 (vect_finish_replace_stmt): New function.
17043 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
17044 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
17046 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
17048 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17049 Alan Hayward <alan.hayward@arm.com>
17050 David Sherwood <david.sherwood@arm.com>
17052 * doc/md.texi (extract_last_@var{m}): Document.
17053 * optabs.def (extract_last_optab): New optab.
17054 * internal-fn.def (EXTRACT_LAST): New internal function.
17055 * internal-fn.c (cond_unary_direct): New macro.
17056 (expand_cond_unary_optab_fn): Likewise.
17057 (direct_cond_unary_optab_supported_p): Likewise.
17058 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
17059 loops using EXTRACT_LAST.
17060 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
17061 (extract_last_<mode>): ...this optab.
17062 (vec_extract<mode><Vel>): Update accordingly.
17064 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17065 Alan Hayward <alan.hayward@arm.com>
17066 David Sherwood <david.sherwood@arm.com>
17068 * target.def (empty_mask_is_expensive): New hook.
17069 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
17070 * doc/tm.texi: Regenerate.
17071 * targhooks.h (default_empty_mask_is_expensive): Declare.
17072 * targhooks.c (default_empty_mask_is_expensive): New function.
17073 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
17074 if the target says that empty masks are expensive.
17075 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
17077 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
17079 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17080 Alan Hayward <alan.hayward@arm.com>
17081 David Sherwood <david.sherwood@arm.com>
17083 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
17084 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
17085 (vect_use_loop_mask_for_alignment_p): New function.
17086 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
17087 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
17088 niters_skip argument. Make sure that the first niters_skip elements
17089 of the first iteration are inactive.
17090 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
17091 Update call to vect_set_loop_masks_directly.
17092 (get_misalign_in_elems): New function, split out from...
17093 (vect_gen_prolog_loop_niters): ...here.
17094 (vect_update_init_of_dr): Take a code argument that specifies whether
17095 the adjustment should be added or subtracted.
17096 (vect_update_init_of_drs): Likewise.
17097 (vect_prepare_for_masked_peels): New function.
17098 (vect_do_peeling): Skip prologue peeling if we're using a mask
17099 instead. Update call to vect_update_inits_of_drs.
17100 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17102 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
17103 alignment. Do not include the number of peeled iterations in
17104 the minimum threshold in that case.
17105 (vectorizable_induction): Adjust the start value down by
17106 LOOP_VINFO_MASK_SKIP_NITERS iterations.
17107 (vect_transform_loop): Call vect_prepare_for_masked_peels.
17108 Take the number of skipped iterations into account when calculating
17110 * tree-vect-stmts.c (vect_gen_while_not): New function.
17112 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17113 Alan Hayward <alan.hayward@arm.com>
17114 David Sherwood <david.sherwood@arm.com>
17116 * doc/sourcebuild.texi (vect_fully_masked): Document.
17117 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
17118 default value to 0.
17119 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
17121 (vect_analyze_loop_2): ...here. Don't check the vectorization
17122 factor against the number of loop iterations if the loop is
17125 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17126 Alan Hayward <alan.hayward@arm.com>
17127 David Sherwood <david.sherwood@arm.com>
17129 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
17130 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
17131 (dump_groups): Update accordingly.
17132 (iv_use::mem_type): New member variable.
17133 (address_p): New function.
17134 (record_use): Add a mem_type argument and initialize the new
17136 (record_group_use): Add a mem_type argument. Use address_p.
17137 Remove obsolete null checks of base_object. Update call to record_use.
17138 (find_interesting_uses_op): Update call to record_group_use.
17139 (find_interesting_uses_cond): Likewise.
17140 (find_interesting_uses_address): Likewise.
17141 (get_mem_type_for_internal_fn): New function.
17142 (find_address_like_use): Likewise.
17143 (find_interesting_uses_stmt): Try find_address_like_use before
17144 calling find_interesting_uses_op.
17145 (addr_offset_valid_p): Use the iv mem_type field as the type
17146 of the addressed memory.
17147 (add_autoinc_candidates): Likewise.
17148 (get_address_cost): Likewise.
17149 (split_small_address_groups_p): Use address_p.
17150 (split_address_groups): Likewise.
17151 (add_iv_candidate_for_use): Likewise.
17152 (autoinc_possible_for_pair): Likewise.
17153 (rewrite_groups): Likewise.
17154 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
17155 (determine_group_iv_cost): Update after split of USE_ADDRESS.
17156 (get_alias_ptr_type_for_ptr_address): New function.
17157 (rewrite_use_address): Rewrite address uses in calls that were
17158 identified by find_address_like_use.
17160 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17161 Alan Hayward <alan.hayward@arm.com>
17162 David Sherwood <david.sherwood@arm.com>
17164 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
17166 * gimple-expr.h (is_gimple_addressable: Likewise.
17167 * gimple-expr.c (is_gimple_address): Likewise.
17168 * internal-fn.c (expand_call_mem_ref): New function.
17169 (expand_mask_load_optab_fn): Use it.
17170 (expand_mask_store_optab_fn): Likewise.
17172 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17173 Alan Hayward <alan.hayward@arm.com>
17174 David Sherwood <david.sherwood@arm.com>
17176 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
17177 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
17178 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
17179 (cond_umax@var{mode}): Document.
17180 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
17181 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
17182 (cond_umin_optab, cond_umax_optab): New optabs.
17183 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
17184 (COND_IOR, COND_XOR): New internal functions.
17185 * internal-fn.h (get_conditional_internal_fn): Declare.
17186 * internal-fn.c (cond_binary_direct): New macro.
17187 (expand_cond_binary_optab_fn): Likewise.
17188 (direct_cond_binary_optab_supported_p): Likewise.
17189 (get_conditional_internal_fn): New function.
17190 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
17191 Cope with reduction statements that are vectorized as calls rather
17193 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
17194 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
17195 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
17196 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
17197 (UNSPEC_COND_EOR): New unspecs.
17198 (optab): Add mappings for them.
17199 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
17200 (sve_int_op, sve_fp_op): New int attributes.
17202 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17203 Alan Hayward <alan.hayward@arm.com>
17204 David Sherwood <david.sherwood@arm.com>
17206 * optabs.def (while_ult_optab): New optab.
17207 * doc/md.texi (while_ult@var{m}@var{n}): Document.
17208 * internal-fn.def (WHILE_ULT): New internal function.
17209 * internal-fn.h (direct_internal_fn_supported_p): New override
17210 that takes two types as argument.
17211 * internal-fn.c (while_direct): New macro.
17212 (expand_while_optab_fn): New function.
17213 (convert_optab_supported_p): Likewise.
17214 (direct_while_optab_supported_p): New macro.
17215 * wide-int.h (wi::udiv_ceil): New function.
17216 * tree-vectorizer.h (rgroup_masks): New structure.
17217 (vec_loop_masks): New typedef.
17218 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
17219 and fully_masked_p.
17220 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
17221 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
17222 (vect_max_vf): New function.
17223 (slpeel_make_loop_iterate_ntimes): Delete.
17224 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
17225 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
17226 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
17227 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
17228 internal-fn.h, stor-layout.h and optabs-query.h.
17229 (vect_set_loop_mask): New function.
17230 (add_preheader_seq): Likewise.
17231 (add_header_seq): Likewise.
17232 (interleave_supported_p): Likewise.
17233 (vect_maybe_permute_loop_masks): Likewise.
17234 (vect_set_loop_masks_directly): Likewise.
17235 (vect_set_loop_condition_masked): Likewise.
17236 (vect_set_loop_condition_unmasked): New function, split out from
17237 slpeel_make_loop_iterate_ntimes.
17238 (slpeel_make_loop_iterate_ntimes): Rename to..
17239 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
17240 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
17241 (vect_do_peeling): Update call accordingly.
17242 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
17244 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17245 mask_compare_type, can_fully_mask_p and fully_masked_p.
17246 (release_vec_loop_masks): New function.
17247 (_loop_vec_info): Use it to free the loop masks.
17248 (can_produce_all_loop_masks_p): New function.
17249 (vect_get_max_nscalars_per_iter): Likewise.
17250 (vect_verify_full_masking): Likewise.
17251 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
17252 retries, and free the mask rgroups before retrying. Check loop-wide
17253 reasons for disallowing fully-masked loops. Make the final decision
17254 about whether use a fully-masked loop or not.
17255 (vect_estimate_min_profitable_iters): Do not assume that peeling
17256 for the number of iterations will be needed for fully-masked loops.
17257 (vectorizable_reduction): Disable fully-masked loops.
17258 (vectorizable_live_operation): Likewise.
17259 (vect_halve_mask_nunits): New function.
17260 (vect_double_mask_nunits): Likewise.
17261 (vect_record_loop_mask): Likewise.
17262 (vect_get_loop_mask): Likewise.
17263 (vect_transform_loop): Handle the case in which the final loop
17264 iteration might handle a partial vector. Call vect_set_loop_condition
17265 instead of slpeel_make_loop_iterate_ntimes.
17266 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
17267 (check_load_store_masking): New function.
17268 (prepare_load_store_mask): Likewise.
17269 (vectorizable_store): Handle fully-masked loops.
17270 (vectorizable_load): Likewise.
17271 (supportable_widening_operation): Use vect_halve_mask_nunits for
17273 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
17274 (vect_gen_while): New function.
17275 * config/aarch64/aarch64.md (umax<mode>3): New expander.
17276 (aarch64_uqdec<mode>): New insn.
17278 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17279 Alan Hayward <alan.hayward@arm.com>
17280 David Sherwood <david.sherwood@arm.com>
17282 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
17283 (reduc_xor_scal_optab): New optabs.
17284 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
17285 (reduc_xor_scal_@var{m}): Document.
17286 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
17287 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
17288 internal functions.
17289 * fold-const-call.c (fold_const_call): Handle them.
17290 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
17291 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
17292 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
17293 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
17294 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
17295 (UNSPEC_XORV): New unspecs.
17296 (optab): Add entries for them.
17297 (BITWISEV): New int iterator.
17298 (bit_reduc_op): New int attributes.
17300 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17301 Alan Hayward <alan.hayward@arm.com>
17302 David Sherwood <david.sherwood@arm.com>
17304 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
17305 * internal-fn.def (VEC_SHL_INSERT): New internal function.
17306 * optabs.def (vec_shl_insert_optab): New optab.
17307 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
17308 (duplicate_and_interleave): Likewise.
17309 * tree-vect-loop.c: Include internal-fn.h.
17310 (neutral_op_for_slp_reduction): New function, split out from
17311 get_initial_defs_for_reduction.
17312 (get_initial_def_for_reduction): Handle option 2 for variable-length
17313 vectors by loading the neutral value into a vector and then shifting
17314 the initial value into element 0.
17315 (get_initial_defs_for_reduction): Replace the code argument with
17316 the neutral value calculated by neutral_op_for_slp_reduction.
17317 Use gimple_build_vector for constant-length vectors.
17318 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
17319 but the first group_size elements have a neutral value.
17320 Use duplicate_and_interleave otherwise.
17321 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
17322 Update call to get_initial_defs_for_reduction. Handle SLP
17323 reductions for variable-length vectors by creating one vector
17324 result for each scalar result, with the elements associated
17325 with other scalar results stubbed out with the neutral value.
17326 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
17327 Require IFN_VEC_SHL_INSERT for double reductions on
17328 variable-length vectors, or SLP reductions that have
17329 a neutral value. Require can_duplicate_and_interleave_p
17330 support for variable-length unchained SLP reductions if there
17331 is no neutral value, such as for MIN/MAX reductions. Also require
17332 the number of vector elements to be a multiple of the number of
17333 SLP statements when doing variable-length unchained SLP reductions.
17334 Update call to vect_create_epilog_for_reduction.
17335 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
17336 and remove initial values.
17337 (duplicate_and_interleave): Make public.
17338 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
17339 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
17341 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17342 Alan Hayward <alan.hayward@arm.com>
17343 David Sherwood <david.sherwood@arm.com>
17345 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
17346 (can_duplicate_and_interleave_p): New function.
17347 (vect_get_and_check_slp_defs): Take the vector of statements
17348 rather than just the current one. Remove excess parentheses.
17349 Restriction rejectinon of vect_constant_def and vect_external_def
17350 for variable-length vectors to boolean types, or types for which
17351 can_duplicate_and_interleave_p is false.
17352 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
17353 (duplicate_and_interleave): New function.
17354 (vect_get_constant_vectors): Use gimple_build_vector for
17355 constant-length vectors and suitable variable-length constant
17356 vectors. Use duplicate_and_interleave for other variable-length
17357 vectors. Don't defer the update when inserting new statements.
17359 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17360 Alan Hayward <alan.hayward@arm.com>
17361 David Sherwood <david.sherwood@arm.com>
17363 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
17364 min_profitable_iters doesn't go negative.
17366 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17367 Alan Hayward <alan.hayward@arm.com>
17368 David Sherwood <david.sherwood@arm.com>
17370 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
17371 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
17372 * optabs.def (vec_mask_load_lanes_optab): New optab.
17373 (vec_mask_store_lanes_optab): Likewise.
17374 * internal-fn.def (MASK_LOAD_LANES): New internal function.
17375 (MASK_STORE_LANES): Likewise.
17376 * internal-fn.c (mask_load_lanes_direct): New macro.
17377 (mask_store_lanes_direct): Likewise.
17378 (expand_mask_load_optab_fn): Handle masked operations.
17379 (expand_mask_load_lanes_optab_fn): New macro.
17380 (expand_mask_store_optab_fn): Handle masked operations.
17381 (expand_mask_store_lanes_optab_fn): New macro.
17382 (direct_mask_load_lanes_optab_supported_p): Likewise.
17383 (direct_mask_store_lanes_optab_supported_p): Likewise.
17384 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
17386 (vect_load_lanes_supported): Likewise.
17387 * tree-vect-data-refs.c (strip_conversion): New function.
17388 (can_group_stmts_p): Likewise.
17389 (vect_analyze_data_ref_accesses): Use it instead of checking
17390 for a pair of assignments.
17391 (vect_store_lanes_supported): Take a masked_p parameter.
17392 (vect_load_lanes_supported): Likewise.
17393 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
17394 vect_store_lanes_supported and vect_load_lanes_supported.
17395 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
17396 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
17397 parameter. Don't allow gaps for masked accesses.
17398 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
17399 and vect_load_lanes_supported.
17400 (get_load_store_type): Take a masked_p parameter and update
17401 call to get_group_load_store_type.
17402 (vectorizable_store): Update call to get_load_store_type.
17403 Handle IFN_MASK_STORE_LANES.
17404 (vectorizable_load): Update call to get_load_store_type.
17405 Handle IFN_MASK_LOAD_LANES.
17407 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17408 Alan Hayward <alan.hayward@arm.com>
17409 David Sherwood <david.sherwood@arm.com>
17411 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
17413 * config/aarch64/aarch64-protos.h
17414 (aarch64_sve_struct_memory_operand_p): Declare.
17415 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
17416 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
17417 (VPRED, vpred): Handle SVE structure modes.
17418 * config/aarch64/constraints.md (Utx): New constraint.
17419 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
17420 (aarch64_sve_struct_nonimmediate_operand): New predicates.
17421 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
17422 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
17423 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
17424 structure modes. Split into pieces after RA.
17425 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
17426 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
17428 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
17429 SVE structure modes.
17430 (aarch64_classify_address): Likewise.
17431 (sizetochar): Move earlier in file.
17432 (aarch64_print_operand): Handle SVE register lists.
17433 (aarch64_array_mode): New function.
17434 (aarch64_sve_struct_memory_operand_p): Likewise.
17435 (TARGET_ARRAY_MODE): Redefine.
17437 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17438 Alan Hayward <alan.hayward@arm.com>
17439 David Sherwood <david.sherwood@arm.com>
17441 * target.def (array_mode): New target hook.
17442 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
17443 * doc/tm.texi: Regenerate.
17444 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
17445 * hooks.c (hook_optmode_mode_uhwi_none): New function.
17446 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
17447 targetm.array_mode.
17448 * stor-layout.c (mode_for_array): Likewise. Support polynomial
17451 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17452 Alan Hayward <alan.hayward@arm.com>
17453 David Sherwood <david.sherwood@arm.com>
17455 * fold-const.c (fold_binary_loc): Check the argument types
17456 rather than the result type when testing for a vector operation.
17458 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17460 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
17461 * doc/tm.texi: Regenerate.
17463 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17464 Alan Hayward <alan.hayward@arm.com>
17465 David Sherwood <david.sherwood@arm.com>
17467 * doc/invoke.texi (-msve-vector-bits=): Document new option.
17468 (sve): Document new AArch64 extension.
17469 * doc/md.texi (w): Extend the description of the AArch64
17470 constraint to include SVE vectors.
17471 (Upl, Upa): Document new AArch64 predicate constraints.
17472 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
17474 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
17475 (msve-vector-bits=): New option.
17476 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
17477 SVE when these are disabled.
17478 (sve): New extension.
17479 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
17480 modes. Adjust their number of units based on aarch64_sve_vg.
17481 (MAX_BITSIZE_MODE_ANY_MODE): Define.
17482 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
17483 aarch64_addr_query_type.
17484 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
17485 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
17486 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
17487 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
17488 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
17489 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
17490 (aarch64_simd_imm_zero_p): Delete.
17491 (aarch64_check_zero_based_sve_index_immediate): Declare.
17492 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
17493 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
17494 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
17495 (aarch64_sve_float_mul_immediate_p): Likewise.
17496 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
17497 rather than an rtx.
17498 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
17499 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
17500 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
17501 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
17502 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
17503 (aarch64_regmode_natural_size): Likewise.
17504 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
17505 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
17507 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
17508 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
17509 for VG and the SVE predicate registers.
17510 (V_ALIASES): Add a "z"-prefixed alias.
17511 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
17512 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
17513 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
17514 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
17515 (REG_CLASS_NAMES): Add entries for them.
17516 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
17517 and the predicate registers.
17518 (aarch64_sve_vg): Declare.
17519 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
17520 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
17521 (REGMODE_NATURAL_SIZE): Define.
17522 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
17524 * config/aarch64/aarch64.c: Include cfgrtl.h.
17525 (simd_immediate_info): Add a constructor for series vectors,
17526 and an associated step field.
17527 (aarch64_sve_vg): New variable.
17528 (aarch64_dbx_register_number): Handle VG and the predicate registers.
17529 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
17530 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
17531 (VEC_ANY_DATA, VEC_STRUCT): New constants.
17532 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
17533 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
17534 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
17535 (aarch64_get_mask_mode): New functions.
17536 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
17537 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
17538 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
17539 predicate modes and predicate registers. Explicitly restrict
17540 GPRs to modes of 16 bytes or smaller. Only allow FP registers
17541 to store a vector mode if it is recognized by
17542 aarch64_classify_vector_mode.
17543 (aarch64_regmode_natural_size): New function.
17544 (aarch64_hard_regno_caller_save_mode): Return the original mode
17546 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
17547 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
17548 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
17549 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
17551 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
17552 does not overlap dest if the function is frame-related. Handle
17554 (aarch64_split_add_offset): New function.
17555 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
17556 them aarch64_add_offset.
17557 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
17558 and update call to aarch64_sub_sp.
17559 (aarch64_add_cfa_expression): New function.
17560 (aarch64_expand_prologue): Pass extra temporary registers to the
17561 functions above. Handle the case in which we need to emit new
17562 DW_CFA_expressions for registers that were originally saved
17563 relative to the stack pointer, but now have to be expressed
17564 relative to the frame pointer.
17565 (aarch64_output_mi_thunk): Pass extra temporary registers to the
17567 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
17568 IP0 and IP1 values for SVE frames.
17569 (aarch64_expand_vec_series): New function.
17570 (aarch64_expand_sve_widened_duplicate): Likewise.
17571 (aarch64_expand_sve_const_vector): Likewise.
17572 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
17573 Handle SVE constants. Use emit_move_insn to move a force_const_mem
17574 into the register, rather than emitting a SET directly.
17575 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
17576 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
17577 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
17578 (offset_9bit_signed_scaled_p): New functions.
17579 (aarch64_replicate_bitmask_imm): New function.
17580 (aarch64_bitmask_imm): Use it.
17581 (aarch64_cannot_force_const_mem): Reject expressions involving
17582 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
17583 (aarch64_classify_index): Handle SVE indices, by requiring
17584 a plain register index with a scale that matches the element size.
17585 (aarch64_classify_address): Handle SVE addresses. Assert that
17586 the mode of the address is VOIDmode or an integer mode.
17587 Update call to aarch64_classify_symbol.
17588 (aarch64_classify_symbolic_expression): Update call to
17589 aarch64_classify_symbol.
17590 (aarch64_const_vec_all_in_range_p): New function.
17591 (aarch64_print_vector_float_operand): Likewise.
17592 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
17593 "vN" for FP registers with SVE modes. Handle (const ...) vectors
17594 and the FP immediates 1.0 and 0.5.
17595 (aarch64_print_address_internal): Handle SVE addresses.
17596 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
17597 (aarch64_regno_regclass): Handle predicate registers.
17598 (aarch64_secondary_reload): Handle big-endian reloads of SVE
17600 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
17601 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
17602 (aarch64_convert_sve_vector_bits): New function.
17603 (aarch64_override_options): Use it to handle -msve-vector-bits=.
17604 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
17605 rather than an rtx.
17606 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
17607 Handle SVE vector and predicate modes. Accept VL-based constants
17608 that need only one temporary register, and VL offsets that require
17609 no temporary registers.
17610 (aarch64_conditional_register_usage): Mark the predicate registers
17611 as fixed if SVE isn't available.
17612 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
17613 Return true for SVE vector and predicate modes.
17614 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
17615 rather than an unsigned int. Handle SVE modes.
17616 (aarch64_preferred_simd_mode): Update call accordingly. Handle
17618 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
17620 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
17621 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
17622 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
17623 (aarch64_sve_float_mul_immediate_p): New functions.
17624 (aarch64_sve_valid_immediate): New function.
17625 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
17626 Explicitly reject structure modes. Check for INDEX constants.
17627 Handle PTRUE and PFALSE constants.
17628 (aarch64_check_zero_based_sve_index_immediate): New function.
17629 (aarch64_simd_imm_zero_p): Delete.
17630 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
17631 vector modes. Accept constants in the range of CNT[BHWD].
17632 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
17633 ask for an Advanced SIMD mode.
17634 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
17635 (aarch64_simd_vector_alignment): Handle SVE predicates.
17636 (aarch64_vectorize_preferred_vector_alignment): New function.
17637 (aarch64_simd_vector_alignment_reachable): Use it instead of
17639 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
17640 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
17642 (MAX_VECT_LEN): Delete.
17643 (expand_vec_perm_d): Add a vec_flags field.
17644 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
17645 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
17646 (aarch64_evpc_ext): Don't apply a big-endian lane correction
17648 (aarch64_evpc_rev): Rename to...
17649 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
17650 (aarch64_evpc_rev_global): New function.
17651 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
17652 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
17654 (aarch64_evpc_sve_tbl): New function.
17655 (aarch64_expand_vec_perm_const_1): Update after rename of
17656 aarch64_evpc_rev. Handle SVE permutes too, trying
17657 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
17658 than aarch64_evpc_tbl.
17659 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
17660 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
17661 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
17662 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
17663 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
17664 (aarch64_expand_sve_vcond): New functions.
17665 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
17666 of aarch64_vector_mode_p.
17667 (aarch64_dwarf_poly_indeterminate_value): New function.
17668 (aarch64_compute_pressure_classes): Likewise.
17669 (aarch64_can_change_mode_class): Likewise.
17670 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
17671 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
17672 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
17673 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
17674 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
17675 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
17676 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
17677 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
17679 (Dn, Dl, Dr): Accept const as well as const_vector.
17680 (Dz): Likewise. Compare against CONST0_RTX.
17681 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
17682 of "vector" where appropriate.
17683 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
17684 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
17685 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
17686 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
17687 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
17688 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
17689 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
17690 (v_int_equiv): Extend to SVE modes.
17691 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
17693 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
17694 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
17695 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
17696 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
17697 (SVE_COND_FP_CMP): New int iterators.
17698 (perm_hilo): Handle the new unpack unspecs.
17699 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
17701 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
17702 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
17703 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
17704 (aarch64_equality_operator, aarch64_constant_vector_operand)
17705 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
17706 (aarch64_sve_nonimmediate_operand): Likewise.
17707 (aarch64_sve_general_operand): Likewise.
17708 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
17709 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
17710 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
17711 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
17712 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
17713 (aarch64_sve_float_arith_immediate): Likewise.
17714 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
17715 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
17716 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
17717 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
17718 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
17719 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
17720 (aarch64_sve_float_arith_operand): Likewise.
17721 (aarch64_sve_float_arith_with_sub_operand): Likewise.
17722 (aarch64_sve_float_mul_operand): Likewise.
17723 (aarch64_sve_vec_perm_operand): Likewise.
17724 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
17725 (aarch64_mov_operand): Accept const_poly_int and const_vector.
17726 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
17727 as well as const_vector.
17728 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
17729 in file. Use CONST0_RTX and CONSTM1_RTX.
17730 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
17731 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
17732 Use aarch64_simd_imm_zero.
17733 * config/aarch64/aarch64-sve.md: New file.
17734 * config/aarch64/aarch64.md: Include it.
17735 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
17736 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
17737 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
17738 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
17739 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
17740 (sve): New attribute.
17741 (enabled): Disable instructions with the sve attribute unless
17743 (movqi, movhi): Pass CONST_POLY_INT operaneds through
17744 aarch64_expand_mov_immediate.
17745 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
17746 CNT[BHSD] immediates.
17747 (movti): Split CONST_POLY_INT moves into two halves.
17748 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
17749 Split additions that need a temporary here if the destination
17750 is the stack pointer.
17751 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
17752 (*add<mode>3_poly_1): New instruction.
17753 (set_clobber_cc): New expander.
17755 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17757 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
17758 parameter and use it instead of GET_MODE_SIZE (innermode). Use
17759 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
17760 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
17761 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
17762 Change innermode from fixed_mode_size to machine_mode.
17763 (simplify_subreg): Update call accordingly. Handle a constant-sized
17764 subreg of a variable-length CONST_VECTOR.
17766 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17767 Alan Hayward <alan.hayward@arm.com>
17768 David Sherwood <david.sherwood@arm.com>
17770 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
17771 (add_offset_to_base): New function, split out from...
17772 (create_mem_ref): ...here. When handling a scale other than 1,
17773 check first whether the address is valid without the offset.
17774 Add it into the base if so, leaving the index and scale as-is.
17776 2018-01-12 Jakub Jelinek <jakub@redhat.com>
17779 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
17780 fold_for_warn before checking if arg2 is INTEGER_CST.
17782 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
17784 * config/rs6000/predicates.md (load_multiple_operation): Delete.
17785 (store_multiple_operation): Delete.
17786 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
17787 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
17788 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
17789 guarded by TARGET_STRING.
17790 (rs6000_output_load_multiple): Delete.
17791 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
17792 OPTION_MASK_STRING / TARGET_STRING handling.
17793 (print_operand) <'N', 'O'>: Add comment that these are unused now.
17794 (const rs6000_opt_masks) <"string">: Change mask to 0.
17795 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
17796 (MASK_STRING): Delete.
17797 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
17799 (load_multiple): Delete.
17806 (store_multiple): Delete.
17813 (movmemsi_8reg): Delete.
17814 (corresponding unnamed define_insn): Delete.
17815 (movmemsi_6reg): Delete.
17816 (corresponding unnamed define_insn): Delete.
17817 (movmemsi_4reg): Delete.
17818 (corresponding unnamed define_insn): Delete.
17819 (movmemsi_2reg): Delete.
17820 (corresponding unnamed define_insn): Delete.
17821 (movmemsi_1reg): Delete.
17822 (corresponding unnamed define_insn): Delete.
17823 * config/rs6000/rs6000.opt (mno-string): New.
17824 (mstring): Replace by deprecation warning stub.
17825 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
17827 2018-01-12 Jakub Jelinek <jakub@redhat.com>
17829 * regrename.c (regrename_do_replace): If replacing the same
17830 reg multiple times, try to reuse last created gen_raw_REG.
17833 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
17834 main to workaround a bug in GDB.
17836 2018-01-12 Tom de Vries <tom@codesourcery.com>
17839 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
17841 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
17843 PR rtl-optimization/80481
17844 * ira-color.c (get_cap_member): New function.
17845 (allocnos_conflict_by_live_ranges_p): Use it.
17846 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
17847 (setup_slot_coalesced_allocno_live_ranges): Ditto.
17849 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
17852 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
17853 (*saddl_se_1): Ditto.
17854 (*ssubsi_1): Ditto.
17855 (*ssubl_se_1): Ditto.
17857 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17859 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
17860 rather than wi::to_widest for DR_INITs.
17861 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
17862 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
17863 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
17865 (vect_analyze_group_access_1): Note that here.
17867 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17869 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
17870 polynomial type sizes.
17872 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17874 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
17875 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
17876 (gimple_add_tmp_var): Likewise.
17878 2018-01-12 Martin Liska <mliska@suse.cz>
17880 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
17881 (gimple_alloc_sizes): Likewise.
17882 (dump_gimple_statistics): Use PRIu64 in printf format.
17883 * gimple.h: Change uint64_t to int.
17885 2018-01-12 Martin Liska <mliska@suse.cz>
17887 * tree-core.h: Use uint64_t instead of int.
17888 * tree.c (tree_node_counts): Likewise.
17889 (tree_node_sizes): Likewise.
17890 (dump_tree_statistics): Use PRIu64 in printf format.
17892 2018-01-12 Martin Liska <mliska@suse.cz>
17894 * Makefile.in: As qsort_chk is implemented in vec.c, add
17895 vec.o to linkage of gencfn-macros.
17896 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
17897 passing the info to record_node_allocation_statistics.
17898 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
17900 * ggc-common.c (struct ggc_usage): Add operator== and use
17901 it in operator< and compare function.
17902 * mem-stats.h (struct mem_usage): Likewise.
17903 * vec.c (struct vec_usage): Remove operator< and compare
17904 function. Can be simply inherited.
17906 2018-01-12 Martin Jambor <mjambor@suse.cz>
17909 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
17910 * tree-ssa-math-opts.c: Include domwalk.h.
17911 (convert_mult_to_fma_1): New function.
17912 (fma_transformation_info): New type.
17913 (fma_deferring_state): Likewise.
17914 (cancel_fma_deferring): New function.
17915 (result_of_phi): Likewise.
17916 (last_fma_candidate_feeds_initial_phi): Likewise.
17917 (convert_mult_to_fma): Added deferring logic, split actual
17918 transformation to convert_mult_to_fma_1.
17919 (math_opts_dom_walker): New type.
17920 (math_opts_dom_walker::after_dom_children): New method, body moved
17921 here from pass_optimize_widening_mul::execute, added deferring logic
17923 (pass_optimize_widening_mul::execute): Moved most of code to
17924 math_opts_dom_walker::after_dom_children.
17925 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
17926 * config/i386/i386.c (ix86_option_override_internal): Added
17927 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
17929 2018-01-12 Richard Biener <rguenther@suse.de>
17932 * dwarf2out.c (gen_variable_die): Do not reset old_die for
17933 inline instance vars.
17935 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
17938 * config/rx/rx.c (rx_is_restricted_memory_address):
17939 Handle SUBREG case.
17941 2018-01-12 Richard Biener <rguenther@suse.de>
17943 PR tree-optimization/80846
17944 * target.def (split_reduction): New target hook.
17945 * targhooks.c (default_split_reduction): New function.
17946 * targhooks.h (default_split_reduction): Declare.
17947 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
17948 target requests first reduce vectors by combining low and high
17950 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
17951 (get_vectype_for_scalar_type_and_size): Export.
17952 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
17953 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
17954 * doc/tm.texi: Regenerate.
17955 * config/i386/i386.c (ix86_split_reduction): Implement
17956 TARGET_VECTORIZE_SPLIT_REDUCTION.
17958 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17961 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
17962 in PIC mode except for TARGET_VXWORKS_RTP.
17963 * config/sparc/sparc.c: Include cfgrtl.h.
17964 (TARGET_INIT_PIC_REG): Define.
17965 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
17966 (sparc_pic_register_p): New predicate.
17967 (sparc_legitimate_address_p): Use it.
17968 (sparc_legitimize_pic_address): Likewise.
17969 (sparc_delegitimize_address): Likewise.
17970 (sparc_mode_dependent_address_p): Likewise.
17971 (gen_load_pcrel_sym): Remove 4th parameter.
17972 (load_got_register): Adjust call to above. Remove obsolete stuff.
17973 (sparc_expand_prologue): Do not call load_got_register here.
17974 (sparc_flat_expand_prologue): Likewise.
17975 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
17976 (sparc_use_pseudo_pic_reg): New function.
17977 (sparc_init_pic_reg): Likewise.
17978 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
17979 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
17981 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
17983 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
17984 Add item for branch_cost.
17986 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17988 PR rtl-optimization/83565
17989 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
17990 not extend the result to a larger mode for rotate operations.
17991 (num_sign_bit_copies1): Likewise.
17993 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17996 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
17998 Use values-Xc.o for -pedantic.
17999 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
18001 2018-01-12 Martin Liska <mliska@suse.cz>
18004 * ipa-devirt.c (final_warning_record::grow_type_warnings):
18006 (possible_polymorphic_call_targets): Use it.
18007 (ipa_devirt): Likewise.
18009 2018-01-12 Martin Liska <mliska@suse.cz>
18011 * profile-count.h (enum profile_quality): Use 0 as invalid
18012 enum value of profile_quality.
18014 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
18016 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
18017 -mext-string options.
18019 2018-01-12 Richard Biener <rguenther@suse.de>
18021 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
18022 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
18023 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
18025 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
18027 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
18029 * configure.ac (--with-long-double-format): Add support for the
18030 configuration option to change the default long double format on
18032 * config.gcc (powerpc*-linux*-*): Likewise.
18033 * configure: Regenerate.
18034 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
18035 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
18036 used without modification.
18038 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18040 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
18041 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
18042 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
18043 MISC_BUILTIN_SPEC_BARRIER.
18044 (rs6000_init_builtins): Likewise.
18045 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
18047 (speculation_barrier): New define_insn.
18048 * doc/extend.texi: Document __builtin_speculation_barrier.
18050 2018-01-11 Jakub Jelinek <jakub@redhat.com>
18053 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
18054 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
18055 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
18057 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
18058 integral modes instead of "ss" and "sd".
18059 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
18060 vectors with 32-bit and 64-bit elements.
18061 (vecdupssescalarmodesuffix): New mode attribute.
18062 (vec_dup<mode>): Use it.
18064 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
18067 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
18068 frame if argument is passed on stack.
18070 2018-01-11 Jakub Jelinek <jakub@redhat.com>
18073 * ree.c (combine_reaching_defs): Optimize also
18074 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
18075 reg2=any_extend(exp); reg1=reg2;, formatting fix.
18077 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
18079 PR middle-end/83189
18080 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
18082 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
18084 PR middle-end/83718
18085 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
18086 after they are computed.
18088 2018-01-11 Bin Cheng <bin.cheng@arm.com>
18090 PR tree-optimization/83695
18091 * gimple-loop-linterchange.cc
18092 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
18093 reset cached scev information after interchange.
18094 (pass_linterchange::execute): Remove call to scev_reset_htab.
18096 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18098 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
18099 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
18100 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
18101 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
18102 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
18103 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
18104 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
18105 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
18106 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
18107 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
18108 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
18109 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
18110 (V_lane_reg): Likewise.
18111 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
18113 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
18114 (vfmal_lane_low<mode>_intrinsic,
18115 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
18116 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
18117 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
18118 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
18119 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
18120 vfmsl_lane_high<mode>_intrinsic): New define_insns.
18122 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18124 * config/arm/arm-cpus.in (fp16fml): New feature.
18125 (ALL_SIMD): Add fp16fml.
18126 (armv8.2-a): Add fp16fml as an option.
18127 (armv8.3-a): Likewise.
18128 (armv8.4-a): Add fp16fml as part of fp16.
18129 * config/arm/arm.h (TARGET_FP16FML): Define.
18130 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
18132 * config/arm/arm-modes.def (V2HF): Define.
18133 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
18134 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
18135 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
18136 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
18137 vfmsl_low, vfmsl_high): New set of builtins.
18138 * config/arm/iterators.md (PLUSMINUS): New code iterator.
18139 (vfml_op): New code attribute.
18140 (VFMLHALVES): New int iterator.
18141 (VFML, VFMLSEL): New mode attributes.
18142 (V_reg): Define mapping for V2HF.
18143 (V_hi, V_lo): New mode attributes.
18144 (VF_constraint): Likewise.
18145 (vfml_half, vfml_half_selector): New int attributes.
18146 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
18148 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
18149 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
18151 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
18152 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
18153 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
18154 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
18156 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
18157 Document new effective target and option set.
18159 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18161 * config/arm/arm-cpus.in (armv8_4): New feature.
18162 (ARMv8_4a): New fgroup.
18163 (armv8.4-a): New arch.
18164 * config/arm/arm-tables.opt: Regenerate.
18165 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
18166 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
18167 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
18168 Add matching rules for -march=armv8.4-a and extensions.
18169 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
18171 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
18174 * config/rx/rx.md (BW): New mode attribute.
18175 (sync_lock_test_and_setsi): Add mode suffix to insn output.
18177 2018-01-11 Richard Biener <rguenther@suse.de>
18179 PR tree-optimization/83435
18180 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
18181 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
18182 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
18184 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18185 Alan Hayward <alan.hayward@arm.com>
18186 David Sherwood <david.sherwood@arm.com>
18188 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
18190 (aarch64_classify_address): Initialize it. Track polynomial offsets.
18191 (aarch64_print_address_internal): Use it to check for a zero offset.
18193 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18194 Alan Hayward <alan.hayward@arm.com>
18195 David Sherwood <david.sherwood@arm.com>
18197 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
18198 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
18199 Return a poly_int64 rather than a HOST_WIDE_INT.
18200 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
18201 rather than a HOST_WIDE_INT.
18202 * config/aarch64/aarch64.h (aarch64_frame): Protect with
18203 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
18204 hard_fp_offset, frame_size, initial_adjust, callee_offset and
18205 final_offset from HOST_WIDE_INT to poly_int64.
18206 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
18207 to_constant when getting the number of units in an Advanced SIMD
18209 (aarch64_builtin_vectorized_function): Check for a constant number
18211 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
18213 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
18214 attribute instead of GET_MODE_NUNITS.
18215 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
18216 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
18217 GET_MODE_SIZE for fixed-size registers.
18218 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
18219 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
18220 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
18221 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
18222 (aarch64_print_operand, aarch64_print_address_internal)
18223 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
18224 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
18225 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
18226 Handle polynomial GET_MODE_SIZE.
18227 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
18228 wider than SImode without modification.
18229 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
18230 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
18231 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
18232 passing and returning SVE modes.
18233 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
18234 rather than GEN_INT.
18235 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
18236 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
18237 (aarch64_allocate_and_probe_stack_space): Likewise.
18238 (aarch64_layout_frame): Cope with polynomial offsets.
18239 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
18240 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
18241 polynomial offsets.
18242 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
18243 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
18244 poly_int64 rather than a HOST_WIDE_INT.
18245 (aarch64_get_separate_components, aarch64_process_components)
18246 (aarch64_expand_prologue, aarch64_expand_epilogue)
18247 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
18248 (aarch64_anchor_offset): New function, split out from...
18249 (aarch64_legitimize_address): ...here.
18250 (aarch64_builtin_vectorization_cost): Handle polynomial
18251 TYPE_VECTOR_SUBPARTS.
18252 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
18254 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
18255 number of elements from the PARALLEL rather than the mode.
18256 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
18257 rather than GET_MODE_BITSIZE.
18258 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
18259 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
18260 (aarch64_expand_vec_perm_const_1): Handle polynomial
18261 d->perm.length () and d->perm elements.
18262 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
18263 Apply to_constant to d->perm elements.
18264 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
18265 polynomial CONST_VECTOR_NUNITS.
18266 (aarch64_move_pointer): Take amount as a poly_int64 rather
18268 (aarch64_progress_pointer): Avoid temporary variable.
18269 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
18270 the mode attribute instead of GET_MODE.
18272 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18273 Alan Hayward <alan.hayward@arm.com>
18274 David Sherwood <david.sherwood@arm.com>
18276 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
18277 x exists before using it.
18278 (aarch64_add_constant_internal): Rename to...
18279 (aarch64_add_offset_1): ...this. Replace regnum with separate
18280 src and dest rtxes. Handle the case in which they're different,
18281 including when the offset is zero. Replace scratchreg with an rtx.
18282 Use 2 additions if there is no spare register into which we can
18283 move a 16-bit constant.
18284 (aarch64_add_constant): Delete.
18285 (aarch64_add_offset): Replace reg with separate src and dest
18286 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
18287 Use aarch64_add_offset_1.
18288 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
18289 an rtx rather than an int. Take the delta as a poly_int64
18290 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
18291 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
18292 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
18293 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
18294 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
18295 and aarch64_add_sp.
18296 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
18297 aarch64_add_constant.
18299 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18301 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
18302 Use scalar_float_mode.
18304 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18306 * config/aarch64/aarch64-simd.md
18307 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
18308 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
18309 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
18310 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
18311 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
18312 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
18313 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
18314 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
18315 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
18316 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
18318 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18321 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
18322 targ_options->x_arm_arch_string is non NULL.
18324 2018-01-11 Tamar Christina <tamar.christina@arm.com>
18326 * config/aarch64/aarch64.h
18327 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
18329 2018-01-11 Sudakshina Das <sudi.das@arm.com>
18332 * expmed.c (emit_store_flag_force): Swap if const op0
18333 and change VOIDmode to mode of op0.
18335 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18337 PR rtl-optimization/83761
18338 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
18339 than bytes to mode_for_size.
18341 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18343 PR middle-end/83189
18344 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
18345 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
18348 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18350 PR middle-end/83575
18351 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
18352 when in layout mode.
18353 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
18354 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
18357 2018-01-10 Michael Collison <michael.collison@arm.com>
18359 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
18360 * config/aarch64/aarch64-option-extension.def: Add
18361 AARCH64_OPT_EXTENSION of 'fp16fml'.
18362 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18363 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
18364 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
18365 * config/aarch64/constraints.md (Ui7): New constraint.
18366 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
18367 (VFMLA_SEL_W): Ditto.
18370 (VFMLA16_LOW): New int iterator.
18371 (VFMLA16_HIGH): Ditto.
18372 (UNSPEC_FMLAL): New unspec.
18373 (UNSPEC_FMLSL): Ditto.
18374 (UNSPEC_FMLAL2): Ditto.
18375 (UNSPEC_FMLSL2): Ditto.
18376 (f16mac): New code attribute.
18377 * config/aarch64/aarch64-simd-builtins.def
18378 (aarch64_fmlal_lowv2sf): Ditto.
18379 (aarch64_fmlsl_lowv2sf): Ditto.
18380 (aarch64_fmlalq_lowv4sf): Ditto.
18381 (aarch64_fmlslq_lowv4sf): Ditto.
18382 (aarch64_fmlal_highv2sf): Ditto.
18383 (aarch64_fmlsl_highv2sf): Ditto.
18384 (aarch64_fmlalq_highv4sf): Ditto.
18385 (aarch64_fmlslq_highv4sf): Ditto.
18386 (aarch64_fmlal_lane_lowv2sf): Ditto.
18387 (aarch64_fmlsl_lane_lowv2sf): Ditto.
18388 (aarch64_fmlal_laneq_lowv2sf): Ditto.
18389 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
18390 (aarch64_fmlalq_lane_lowv4sf): Ditto.
18391 (aarch64_fmlsl_lane_lowv4sf): Ditto.
18392 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
18393 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
18394 (aarch64_fmlal_lane_highv2sf): Ditto.
18395 (aarch64_fmlsl_lane_highv2sf): Ditto.
18396 (aarch64_fmlal_laneq_highv2sf): Ditto.
18397 (aarch64_fmlsl_laneq_highv2sf): Ditto.
18398 (aarch64_fmlalq_lane_highv4sf): Ditto.
18399 (aarch64_fmlsl_lane_highv4sf): Ditto.
18400 (aarch64_fmlalq_laneq_highv4sf): Ditto.
18401 (aarch64_fmlsl_laneq_highv4sf): Ditto.
18402 * config/aarch64/aarch64-simd.md:
18403 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
18404 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
18405 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
18406 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
18407 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
18408 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
18409 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
18410 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
18411 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
18412 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
18413 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
18414 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
18415 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
18416 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
18417 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
18418 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
18419 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
18420 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
18421 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
18422 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
18423 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
18424 (vfmlsl_low_u32): Ditto.
18425 (vfmlalq_low_u32): Ditto.
18426 (vfmlslq_low_u32): Ditto.
18427 (vfmlal_high_u32): Ditto.
18428 (vfmlsl_high_u32): Ditto.
18429 (vfmlalq_high_u32): Ditto.
18430 (vfmlslq_high_u32): Ditto.
18431 (vfmlal_lane_low_u32): Ditto.
18432 (vfmlsl_lane_low_u32): Ditto.
18433 (vfmlal_laneq_low_u32): Ditto.
18434 (vfmlsl_laneq_low_u32): Ditto.
18435 (vfmlalq_lane_low_u32): Ditto.
18436 (vfmlslq_lane_low_u32): Ditto.
18437 (vfmlalq_laneq_low_u32): Ditto.
18438 (vfmlslq_laneq_low_u32): Ditto.
18439 (vfmlal_lane_high_u32): Ditto.
18440 (vfmlsl_lane_high_u32): Ditto.
18441 (vfmlal_laneq_high_u32): Ditto.
18442 (vfmlsl_laneq_high_u32): Ditto.
18443 (vfmlalq_lane_high_u32): Ditto.
18444 (vfmlslq_lane_high_u32): Ditto.
18445 (vfmlalq_laneq_high_u32): Ditto.
18446 (vfmlslq_laneq_high_u32): Ditto.
18447 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
18448 (AARCH64_FL_FOR_ARCH8_4): New.
18449 (AARCH64_ISA_F16FML): New ISA flag.
18450 (TARGET_F16FML): New feature flag for fp16fml.
18451 (doc/invoke.texi): Document new fp16fml option.
18453 2018-01-10 Michael Collison <michael.collison@arm.com>
18455 * config/aarch64/aarch64-builtins.c:
18456 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
18457 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18458 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
18459 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
18460 (AARCH64_ISA_SHA3): New ISA flag.
18461 (TARGET_SHA3): New feature flag for sha3.
18462 * config/aarch64/iterators.md (sha512_op): New int attribute.
18463 (CRYPTO_SHA512): New int iterator.
18464 (UNSPEC_SHA512H): New unspec.
18465 (UNSPEC_SHA512H2): Ditto.
18466 (UNSPEC_SHA512SU0): Ditto.
18467 (UNSPEC_SHA512SU1): Ditto.
18468 * config/aarch64/aarch64-simd-builtins.def
18469 (aarch64_crypto_sha512hqv2di): New builtin.
18470 (aarch64_crypto_sha512h2qv2di): Ditto.
18471 (aarch64_crypto_sha512su0qv2di): Ditto.
18472 (aarch64_crypto_sha512su1qv2di): Ditto.
18473 (aarch64_eor3qv8hi): Ditto.
18474 (aarch64_rax1qv2di): Ditto.
18475 (aarch64_xarqv2di): Ditto.
18476 (aarch64_bcaxqv8hi): Ditto.
18477 * config/aarch64/aarch64-simd.md:
18478 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
18479 (aarch64_crypto_sha512su0qv2di): Ditto.
18480 (aarch64_crypto_sha512su1qv2di): Ditto.
18481 (aarch64_eor3qv8hi): Ditto.
18482 (aarch64_rax1qv2di): Ditto.
18483 (aarch64_xarqv2di): Ditto.
18484 (aarch64_bcaxqv8hi): Ditto.
18485 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
18486 (vsha512h2q_u64): Ditto.
18487 (vsha512su0q_u64): Ditto.
18488 (vsha512su1q_u64): Ditto.
18489 (veor3q_u16): Ditto.
18490 (vrax1q_u64): Ditto.
18491 (vxarq_u64): Ditto.
18492 (vbcaxq_u16): Ditto.
18493 * config/arm/types.md (crypto_sha512): New type attribute.
18494 (crypto_sha3): Ditto.
18495 (doc/invoke.texi): Document new sha3 option.
18497 2018-01-10 Michael Collison <michael.collison@arm.com>
18499 * config/aarch64/aarch64-builtins.c:
18500 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
18501 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18502 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
18503 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
18504 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
18505 (AARCH64_ISA_SM4): New ISA flag.
18506 (TARGET_SM4): New feature flag for sm4.
18507 * config/aarch64/aarch64-simd-builtins.def
18508 (aarch64_sm3ss1qv4si): Ditto.
18509 (aarch64_sm3tt1aq4si): Ditto.
18510 (aarch64_sm3tt1bq4si): Ditto.
18511 (aarch64_sm3tt2aq4si): Ditto.
18512 (aarch64_sm3tt2bq4si): Ditto.
18513 (aarch64_sm3partw1qv4si): Ditto.
18514 (aarch64_sm3partw2qv4si): Ditto.
18515 (aarch64_sm4eqv4si): Ditto.
18516 (aarch64_sm4ekeyqv4si): Ditto.
18517 * config/aarch64/aarch64-simd.md:
18518 (aarch64_sm3ss1qv4si): Ditto.
18519 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
18520 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
18521 (aarch64_sm4eqv4si): Ditto.
18522 (aarch64_sm4ekeyqv4si): Ditto.
18523 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
18524 (sm3part_op): Ditto.
18525 (CRYPTO_SM3TT): Ditto.
18526 (CRYPTO_SM3PART): Ditto.
18527 (UNSPEC_SM3SS1): New unspec.
18528 (UNSPEC_SM3TT1A): Ditto.
18529 (UNSPEC_SM3TT1B): Ditto.
18530 (UNSPEC_SM3TT2A): Ditto.
18531 (UNSPEC_SM3TT2B): Ditto.
18532 (UNSPEC_SM3PARTW1): Ditto.
18533 (UNSPEC_SM3PARTW2): Ditto.
18534 (UNSPEC_SM4E): Ditto.
18535 (UNSPEC_SM4EKEY): Ditto.
18536 * config/aarch64/constraints.md (Ui2): New constraint.
18537 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
18538 * config/arm/types.md (crypto_sm3): New type attribute.
18539 (crypto_sm4): Ditto.
18540 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
18541 (vsm3tt1aq_u32): Ditto.
18542 (vsm3tt1bq_u32): Ditto.
18543 (vsm3tt2aq_u32): Ditto.
18544 (vsm3tt2bq_u32): Ditto.
18545 (vsm3partw1q_u32): Ditto.
18546 (vsm3partw2q_u32): Ditto.
18547 (vsm4eq_u32): Ditto.
18548 (vsm4ekeyq_u32): Ditto.
18549 (doc/invoke.texi): Document new sm4 option.
18551 2018-01-10 Michael Collison <michael.collison@arm.com>
18553 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
18554 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
18555 (AARCH64_FL_FOR_ARCH8_4): New.
18556 (AARCH64_FL_V8_4): New flag.
18557 (doc/invoke.texi): Document new armv8.4-a option.
18559 2018-01-10 Michael Collison <michael.collison@arm.com>
18561 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18562 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
18563 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
18564 * config/aarch64/aarch64-option-extension.def: Add
18565 AARCH64_OPT_EXTENSION of 'sha2'.
18566 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
18567 (crypto): Disable sha2 and aes if crypto disabled.
18568 (crypto): Enable aes and sha2 if enabled.
18569 (simd): Disable sha2 and aes if simd disabled.
18570 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
18572 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
18573 (TARGET_SHA2): New feature flag for sha2.
18574 (TARGET_AES): New feature flag for aes.
18575 * config/aarch64/aarch64-simd.md:
18576 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
18577 conditional on TARGET_AES.
18578 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
18579 (aarch64_crypto_sha1hsi): Make pattern conditional
18581 (aarch64_crypto_sha1hv4si): Ditto.
18582 (aarch64_be_crypto_sha1hv4si): Ditto.
18583 (aarch64_crypto_sha1su1v4si): Ditto.
18584 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
18585 (aarch64_crypto_sha1su0v4si): Ditto.
18586 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
18587 (aarch64_crypto_sha256su0v4si): Ditto.
18588 (aarch64_crypto_sha256su1v4si): Ditto.
18589 (doc/invoke.texi): Document new aes and sha2 options.
18591 2018-01-10 Martin Sebor <msebor@redhat.com>
18593 PR tree-optimization/83781
18594 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
18597 2018-01-11 Martin Sebor <msebor@gmail.com>
18598 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18600 PR tree-optimization/83501
18601 PR tree-optimization/81703
18603 * tree-ssa-strlen.c (get_string_cst): Rename...
18604 (get_string_len): ...to this. Handle global constants.
18605 (handle_char_store): Adjust.
18607 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
18608 Jim Wilson <jimw@sifive.com>
18610 * config/riscv/riscv-protos.h (riscv_output_return): New.
18611 * config/riscv/riscv.c (struct machine_function): New naked_p field.
18612 (riscv_attribute_table, riscv_output_return),
18613 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
18614 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
18615 (riscv_compute_frame_info): Only compute frame->mask if not a naked
18617 (riscv_expand_prologue): Add early return for naked function.
18618 (riscv_expand_epilogue): Likewise.
18619 (riscv_function_ok_for_sibcall): Return false for naked function.
18620 (riscv_set_current_function): New.
18621 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
18622 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
18623 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
18624 * doc/extend.texi (RISC-V Function Attributes): New.
18626 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
18628 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
18629 check for 128-bit long double before checking TCmode.
18630 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
18631 128-bit long doubles before checking TFmode or TCmode.
18632 (FLOAT128_IBM_P): Likewise.
18634 2018-01-10 Martin Sebor <msebor@redhat.com>
18636 PR tree-optimization/83671
18637 * builtins.c (c_strlen): Unconditionally return zero for the empty
18639 Use -Warray-bounds for warnings.
18640 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
18641 for non-constant array indices with COMPONENT_REF, arrays of
18642 arrays, and pointers to arrays.
18643 (gimple_fold_builtin_strlen): Determine and set length range for
18644 non-constant character arrays.
18646 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
18648 PR middle-end/81897
18649 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
18652 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
18654 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
18656 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
18659 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
18660 VECTOR_MEM_ALTIVEC_OR_VSX_P.
18661 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
18662 indexed_or_indirect_operand predicate.
18663 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
18664 (*vsx_le_perm_load_v8hi): Likewise.
18665 (*vsx_le_perm_load_v16qi): Likewise.
18666 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
18667 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
18668 (*vsx_le_perm_store_v8hi): Likewise.
18669 (*vsx_le_perm_store_v16qi): Likewise.
18670 (eight unnamed splitters): Likewise.
18672 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
18674 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
18675 * config/rs6000/emmintrin.h: Likewise.
18676 * config/rs6000/mmintrin.h: Likewise.
18677 * config/rs6000/xmmintrin.h: Likewise.
18679 2018-01-10 David Malcolm <dmalcolm@redhat.com>
18682 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
18684 * tree.c (tree_nop_conversion): Return true for location wrapper
18686 (maybe_wrap_with_location): New function.
18687 (selftest::check_strip_nops): New function.
18688 (selftest::test_location_wrappers): New function.
18689 (selftest::tree_c_tests): Call it.
18690 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
18691 (maybe_wrap_with_location): New decl.
18692 (EXPR_LOCATION_WRAPPER_P): New macro.
18693 (location_wrapper_p): New inline function.
18694 (tree_strip_any_location_wrapper): New inline function.
18696 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
18699 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
18700 stack_realign_offset for the largest alignment of stack slot
18702 (ix86_find_max_used_stack_alignment): New function.
18703 (ix86_finalize_stack_frame_flags): Use it. Set
18704 max_used_stack_alignment if we don't realign stack.
18705 * config/i386/i386.h (machine_function): Add
18706 max_used_stack_alignment.
18708 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
18710 * config/arm/arm.opt (-mbranch-cost): New option.
18711 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
18714 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
18717 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
18718 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
18720 2018-01-10 Richard Biener <rguenther@suse.de>
18723 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
18724 early out so it also covers the case where we have a non-NULL
18727 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
18729 PR tree-optimization/83753
18730 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
18731 for non-strided grouped accesses if the number of elements is 1.
18733 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18736 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
18737 * i386.h (TARGET_USE_GATHER): Define.
18738 * x86-tune.def (X86_TUNE_USE_GATHER): New.
18740 2018-01-10 Martin Liska <mliska@suse.cz>
18743 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
18744 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
18746 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
18747 CLEANUP_NO_PARTITIONING is not set.
18749 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
18751 * doc/rtl.texi: Remove documentation of (const ...) wrappers
18752 for vectors, as a partial revert of r254296.
18753 * rtl.h (const_vec_p): Delete.
18754 (const_vec_duplicate_p): Don't test for vector CONSTs.
18755 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
18756 * expmed.c (make_tree): Likewise.
18759 * common.md (E, F): Use CONSTANT_P instead of checking for
18761 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
18762 checking for CONST_VECTOR.
18764 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18766 PR middle-end/83575
18767 * predict.c (force_edge_cold): Handle in more sane way edges
18768 with no prediction.
18770 2018-01-09 Carl Love <cel@us.ibm.com>
18772 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
18774 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
18775 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
18776 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
18777 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
18778 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
18779 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
18780 * config/rs6000/rs6000-protos.h: Add extern defition for
18781 rs6000_generate_float2_double_code.
18782 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
18784 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
18785 (float2_v2df): Add define_expand.
18787 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
18790 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
18791 op_mode in the force_to_mode call.
18793 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18795 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
18796 instead of checking each element individually.
18797 (aarch64_evpc_uzp): Likewise.
18798 (aarch64_evpc_zip): Likewise.
18799 (aarch64_evpc_ext): Likewise.
18800 (aarch64_evpc_rev): Likewise.
18801 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
18802 instead of checking each element individually. Return true without
18804 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
18805 whether all selected elements come from the same input, instead of
18806 checking each element individually. Remove calls to gen_rtx_REG,
18807 start_sequence and end_sequence and instead assert that no rtl is
18810 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18812 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
18813 order of HIGH and CONST checks.
18815 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18817 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
18818 if the destination isn't an SSA_NAME.
18820 2018-01-09 Richard Biener <rguenther@suse.de>
18822 PR tree-optimization/83668
18823 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
18825 (canonicalize_loop_form): ... here, renamed from ...
18826 (canonicalize_loop_closed_ssa_form): ... this and amended to
18827 swap successor edges for loop exit blocks to make us use
18828 the RPO order we need for initial schedule generation.
18830 2018-01-09 Joseph Myers <joseph@codesourcery.com>
18832 PR tree-optimization/64811
18833 * match.pd: When optimizing comparisons with Inf, avoid
18834 introducing or losing exceptions from comparisons with NaN.
18836 2018-01-09 Martin Liska <mliska@suse.cz>
18839 * asan.c (shadow_mem_size): Add gcc_assert.
18841 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
18843 Don't save registers in main().
18846 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
18847 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
18848 * config/avr/avr.c (avr_set_current_function): Don't error if
18849 naked, OS_task or OS_main are specified at the same time.
18850 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
18852 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
18854 * common/config/avr/avr-common.c (avr_option_optimization_table):
18855 Switch on -mmain-is-OS_task for optimizing compilations.
18857 2018-01-09 Richard Biener <rguenther@suse.de>
18859 PR tree-optimization/83572
18860 * graphite.c: Include cfganal.h.
18861 (graphite_transform_loops): Connect infinite loops to exit
18862 and remove fake edges at the end.
18864 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18866 * ipa-inline.c (edge_badness): Revert accidental checkin.
18868 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18871 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
18872 symbols; not inline clones.
18874 2018-01-09 Jakub Jelinek <jakub@redhat.com>
18877 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
18878 hard registers. Formatting fixes.
18880 PR preprocessor/83722
18881 * gcc.c (try_generate_repro): Pass
18882 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
18883 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
18886 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
18887 Kito Cheng <kito.cheng@gmail.com>
18889 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
18890 (riscv_leaf_function_p): Delete.
18891 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
18893 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18895 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
18897 (do_ifelse): New function.
18898 (do_isel): New function.
18899 (do_sub3): New function.
18900 (do_add3): New function.
18901 (do_load_mask_compare): New function.
18902 (do_overlap_load_compare): New function.
18903 (expand_compare_loop): New function.
18904 (expand_block_compare): Call expand_compare_loop() when appropriate.
18905 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
18906 option description.
18907 (-mblock-compare-inline-loop-limit): New option.
18909 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18912 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
18913 Reverse order of second and third operands in first alternative.
18914 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
18915 of first and second elements in UNSPEC_VPERMR vector.
18916 (altivec_expand_vec_perm_le): Likewise.
18918 2018-01-08 Jeff Law <law@redhat.com>
18920 PR rtl-optimizatin/81308
18921 * tree-switch-conversion.c (cfg_altered): New file scoped static.
18922 (process_switch): If group_case_labels makes a change, then set
18924 (pass_convert_switch::execute): If a switch is converted, then
18925 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
18927 PR rtl-optimization/81308
18928 * recog.c (split_all_insns): Conditionally cleanup the CFG after
18931 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
18933 PR target/83663 - Revert r255946
18934 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
18935 generation for cases where splatting a value is not useful.
18936 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
18937 across a vec_duplicate and a paradoxical subreg forming a vector
18938 mode to a vec_concat.
18940 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18942 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
18943 -march=armv8.3-a variants.
18944 * config/arm/t-multilib: Likewise.
18945 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
18947 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18949 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
18951 (cceq_ior_compare_complement): Give it a name so I can use it, and
18952 change boolean_or_operator predicate to boolean_operator so it can
18953 be used to generate a crand.
18954 (eqne): New code iterator.
18955 (bd/bd_neg): New code_attrs.
18956 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
18957 a single define_insn.
18958 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
18959 decrement (bdnzt/bdnzf/bdzt/bdzf).
18960 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
18961 with the new names of the branch decrement patterns, and added the
18962 names of the branch decrement conditional patterns.
18964 2018-01-08 Richard Biener <rguenther@suse.de>
18966 PR tree-optimization/83563
18967 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
18970 2018-01-08 Richard Biener <rguenther@suse.de>
18972 PR middle-end/83713
18973 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
18975 2018-01-08 Richard Biener <rguenther@suse.de>
18977 PR tree-optimization/83685
18978 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
18979 references to abnormals.
18981 2018-01-08 Richard Biener <rguenther@suse.de>
18984 * dwarf2out.c (output_indirect_strings): Handle empty
18985 skeleton_debug_str_hash.
18986 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
18988 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
18990 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
18991 (emit_store_direct): Likewise.
18992 (arc_trampoline_adjust_address): Likewise.
18993 (arc_asm_trampoline_template): New function.
18994 (arc_initialize_trampoline): Use asm_trampoline_template.
18995 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
18996 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
18997 * config/arc/arc.md (flush_icache): Delete pattern.
18999 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
19001 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
19002 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
19005 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
19008 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
19009 by not USED_FOR_TARGET.
19010 (make_pass_resolve_sw_modes): Likewise.
19012 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
19014 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
19017 2018-01-08 Richard Biener <rguenther@suse.de>
19019 PR middle-end/83580
19020 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
19022 2018-01-08 Richard Biener <rguenther@suse.de>
19024 PR middle-end/83517
19025 * match.pd ((t * 2) / 2) -> t): Add missing :c.
19027 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
19029 PR middle-end/81897
19030 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
19031 basic blocks with a small number of successors.
19032 (convert_control_dep_chain_into_preds): Improve handling of
19034 (dump_predicates): Split apart into...
19035 (dump_pred_chain): ...here...
19036 (dump_pred_info): ...and here.
19037 (can_one_predicate_be_invalidated_p): Add debugging printfs.
19038 (can_chain_union_be_invalidated_p): Improve check for invalidation
19040 (uninit_uses_cannot_happen): Avoid unnecessary if
19041 convert_control_dep_chain_into_preds yielded nothing.
19043 2018-01-06 Martin Sebor <msebor@redhat.com>
19045 PR tree-optimization/83640
19046 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
19047 subtracting negative offset from size.
19048 (builtin_access::overlap): Adjust offset bounds of the access to fall
19049 within the size of the object if possible.
19051 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
19053 PR rtl-optimization/83699
19054 * expmed.c (extract_bit_field_1): Restrict the vector usage of
19055 extract_bit_field_as_subreg to cases in which the extracted
19056 value is also a vector.
19058 * lra-constraints.c (process_alt_operands): Test for the equivalence
19059 substitutions when detecting a possible reload cycle.
19061 2018-01-06 Jakub Jelinek <jakub@redhat.com>
19064 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
19065 by default if flag_selective_schedling{,2}. Formatting fixes.
19067 PR rtl-optimization/83682
19068 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
19069 if it has non-VECTOR_MODE element mode.
19070 (vec_duplicate_p): Likewise.
19072 PR middle-end/83694
19073 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
19074 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
19076 2018-01-05 Jakub Jelinek <jakub@redhat.com>
19079 * config/i386/i386-builtin.def
19080 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
19081 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
19082 Require also OPTION_MASK_ISA_AVX512F in addition to
19083 OPTION_MASK_ISA_GFNI.
19084 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
19085 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
19086 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
19087 to OPTION_MASK_ISA_GFNI.
19088 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
19089 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
19090 OPTION_MASK_ISA_AVX512BW.
19091 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
19092 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
19093 addition to OPTION_MASK_ISA_GFNI.
19094 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
19095 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
19096 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
19097 to OPTION_MASK_ISA_GFNI.
19098 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
19099 a requirement for all ISAs rather than any of them with a few
19101 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
19103 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
19104 bitmasks to be enabled with 3 exceptions, instead of requiring any
19105 enabled ISA with lots of exceptions.
19106 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
19107 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
19108 Change avx512bw in isa attribute to avx512f.
19109 * config/i386/sgxintrin.h: Add license boilerplate.
19110 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
19111 to __AVX512F__ and __AVX512VL to __AVX512VL__.
19112 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
19113 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
19115 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
19116 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
19117 temporarily sse2 rather than sse if not enabled already.
19120 * config/i386/sse.md (VI248_VLBW): Rename to ...
19121 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
19122 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
19123 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
19124 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
19125 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
19126 mode iterator instead of VI248_VLBW.
19128 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
19130 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
19131 (record_modified): Skip clobbers; add debug output.
19132 (param_change_prob): Use sreal frequencies.
19134 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
19136 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
19137 punt for user-aligned variables.
19139 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
19141 * tree-chrec.c (chrec_contains_symbols): Return true for
19144 2018-01-05 Sudakshina Das <sudi.das@arm.com>
19147 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
19148 of (x|y) == x for BICS pattern.
19150 2018-01-05 Jakub Jelinek <jakub@redhat.com>
19152 PR tree-optimization/83605
19153 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
19154 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
19157 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
19159 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
19160 * config/epiphany/rtems.h: New file.
19162 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19163 Uros Bizjak <ubizjak@gmail.com>
19166 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
19167 QIreg_operand instead of register_operand predicate.
19168 * config/i386/i386.c (ix86_rop_should_change_byte_p,
19169 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
19170 comments instead of -fmitigate[-_]rop.
19172 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19175 * cgraphunit.c (symbol_table::compile): Switch to text_section
19176 before calling assembly_start debug hook.
19177 * run-rtl-passes.c (run_rtl_passes): Likewise.
19180 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19182 * tree-vrp.c (extract_range_from_binary_expr_1): Check
19183 range_int_cst_p rather than !symbolic_range_p before calling
19184 extract_range_from_multiplicative_op_1.
19186 2018-01-04 Jeff Law <law@redhat.com>
19188 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
19189 redundant test in assertion.
19191 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19193 * doc/rtl.texi: Document machine_mode wrapper classes.
19195 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19197 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
19198 using tree_to_uhwi.
19200 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19202 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
19203 the VEC_PERM_EXPR fold to fail.
19205 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19208 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
19209 to switched_sections.
19211 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19214 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
19215 test for d.testing.
19217 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
19220 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
19221 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
19223 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19226 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
19227 is BLKmode and bitpos not zero or mode change is needed.
19229 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19232 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
19235 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
19238 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
19239 instead of MULT rtx. Update all corresponding splitters.
19240 (*saddl_se): Ditto.
19241 (*ssub<modesuffix>): Ditto.
19242 (*ssubl_se): Ditto.
19243 (*cmp_sadd_di): Update split patterns.
19244 (*cmp_sadd_si): Ditto.
19245 (*cmp_sadd_sidi): Ditto.
19246 (*cmp_ssub_di): Ditto.
19247 (*cmp_ssub_si): Ditto.
19248 (*cmp_ssub_sidi): Ditto.
19249 * config/alpha/predicates.md (const23_operand): New predicate.
19250 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
19251 Look for ASHIFT, not MULT inner operand.
19252 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
19254 2018-01-04 Martin Liska <mliska@suse.cz>
19256 PR gcov-profile/83669
19257 * gcov.c (output_intermediate_file): Add version to intermediate
19259 * doc/gcov.texi: Document new field 'version' in intermediate
19260 file format. Fix location of '-k' option of gcov command.
19262 2018-01-04 Martin Liska <mliska@suse.cz>
19265 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
19267 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19269 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
19271 2018-01-03 Martin Sebor <msebor@redhat.com>
19273 PR tree-optimization/83655
19274 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
19275 checking calls with invalid arguments.
19277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19279 * tree-vect-stmts.c (vect_get_store_rhs): New function.
19280 (vectorizable_mask_load_store): Delete.
19281 (vectorizable_call): Return false for masked loads and stores.
19282 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
19283 instead of gimple_assign_rhs1.
19284 (vectorizable_load): Handle IFN_MASK_LOAD.
19285 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
19287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19289 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
19291 (vectorizable_mask_load_store): ...here.
19292 (vectorizable_load): ...and here.
19294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19296 * tree-vect-stmts.c (vect_build_all_ones_mask)
19297 (vect_build_zero_merge_argument): New functions, split out from...
19298 (vectorizable_load): ...here.
19300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19302 * tree-vect-stmts.c (vect_check_store_rhs): New function,
19304 (vectorizable_mask_load_store): ...here.
19305 (vectorizable_store): ...and here.
19307 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19309 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
19311 (vectorizable_mask_load_store): ...here.
19313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19315 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
19316 (vect_model_store_cost): Take a vec_load_store_type instead of a
19318 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
19319 (vect_model_store_cost): Take a vec_load_store_type instead of a
19321 (vectorizable_mask_load_store): Update accordingly.
19322 (vectorizable_store): Likewise.
19323 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
19325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19327 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
19328 IFN_MASK_LOAD calls here rather than...
19329 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
19331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19332 Alan Hayward <alan.hayward@arm.com>
19333 David Sherwood <david.sherwood@arm.com>
19335 * expmed.c (extract_bit_field_1): For vector extracts,
19336 fall back to extract_bit_field_as_subreg if vec_extract
19339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19340 Alan Hayward <alan.hayward@arm.com>
19341 David Sherwood <david.sherwood@arm.com>
19343 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
19344 they are variable or constant sized.
19345 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
19346 slots for constant-sized data.
19348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19349 Alan Hayward <alan.hayward@arm.com>
19350 David Sherwood <david.sherwood@arm.com>
19352 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
19353 handling COND_EXPRs with boolean comparisons, try to find a better
19354 basis for the mask type than the boolean itself.
19356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19358 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
19359 is calculated and how it can be overridden.
19360 * genmodes.c (max_bitsize_mode_any_mode): New variable.
19361 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
19363 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
19366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19367 Alan Hayward <alan.hayward@arm.com>
19368 David Sherwood <david.sherwood@arm.com>
19370 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
19371 Remove the mode argument.
19372 (aarch64_simd_valid_immediate): Remove the mode and inverse
19374 * config/aarch64/iterators.md (bitsize): New iterator.
19375 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
19376 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
19377 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
19378 aarch64_simd_valid_immediate.
19379 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
19380 (aarch64_reg_or_bic_imm): Likewise.
19381 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
19382 with an insn_type enum and msl with a modifier_type enum.
19383 Replace element_width with a scalar_mode. Change the shift
19384 to unsigned int. Add constructors for scalar_float_mode and
19385 scalar_int_mode elements.
19386 (aarch64_vect_float_const_representable_p): Delete.
19387 (aarch64_can_const_movi_rtx_p)
19388 (aarch64_simd_scalar_immediate_valid_for_move)
19389 (aarch64_simd_make_constant): Update call to
19390 aarch64_simd_valid_immediate.
19391 (aarch64_advsimd_valid_immediate_hs): New function.
19392 (aarch64_advsimd_valid_immediate): Likewise.
19393 (aarch64_simd_valid_immediate): Remove mode and inverse
19394 arguments. Rewrite to use the above. Use const_vec_duplicate_p
19395 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
19396 and aarch64_float_const_representable_p on the result.
19397 (aarch64_output_simd_mov_immediate): Remove mode argument.
19398 Update call to aarch64_simd_valid_immediate and use of
19399 simd_immediate_info.
19400 (aarch64_output_scalar_simd_mov_immediate): Update call
19403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19404 Alan Hayward <alan.hayward@arm.com>
19405 David Sherwood <david.sherwood@arm.com>
19407 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
19408 (mode_nunits): Likewise CONST_MODE_NUNITS.
19409 * machmode.def (ADJUST_NUNITS): Document.
19410 * genmodes.c (mode_data::need_nunits_adj): New field.
19411 (blank_mode): Update accordingly.
19412 (adj_nunits): New variable.
19413 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
19415 (emit_mode_size_inline): Set need_bytesize_adj for all modes
19416 listed in adj_nunits.
19417 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
19418 listed in adj_nunits. Don't emit case statements for such modes.
19419 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
19420 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
19421 nothing if adj_nunits is nonnull.
19422 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
19423 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
19424 (emit_mode_fbit): Update use of print_maybe_const_decl.
19425 (emit_move_size): Likewise. Treat the array as non-const
19427 (emit_mode_adjustments): Handle adj_nunits.
19429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19431 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
19432 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
19433 (VECTOR_MODES): Use it.
19434 (make_vector_modes): Take the prefix as an argument.
19436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19437 Alan Hayward <alan.hayward@arm.com>
19438 David Sherwood <david.sherwood@arm.com>
19440 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
19441 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
19442 for MODE_VECTOR_BOOL.
19443 * machmode.def (VECTOR_BOOL_MODE): Document.
19444 * genmodes.c (VECTOR_BOOL_MODE): New macro.
19445 (make_vector_bool_mode): New function.
19446 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
19448 * lto-streamer-in.c (lto_input_mode_table): Likewise.
19449 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
19451 * stor-layout.c (int_mode_for_mode): Likewise.
19452 * tree.c (build_vector_type_for_mode): Likewise.
19453 * varasm.c (output_constant_pool_2): Likewise.
19454 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
19455 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
19456 for MODE_VECTOR_BOOL.
19457 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
19458 of mode class checks.
19459 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
19460 instead of a list of mode class checks.
19461 (expand_vector_scalar_condition): Likewise.
19462 (type_for_widest_vector_mode): Handle BImode as an inner mode.
19464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19465 Alan Hayward <alan.hayward@arm.com>
19466 David Sherwood <david.sherwood@arm.com>
19468 * machmode.h (mode_size): Change from unsigned short to
19470 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
19471 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
19472 or if measurement_type is not polynomial.
19473 (fixed_size_mode::includes_p): Check for constant-sized modes.
19474 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
19475 return a poly_uint16 rather than an unsigned short.
19476 (emit_mode_size): Change the type of mode_size from unsigned short
19477 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
19478 (emit_mode_adjustments): Cope with polynomial vector sizes.
19479 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19481 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19483 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
19484 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
19485 * caller-save.c (setup_save_areas): Likewise.
19486 (replace_reg_with_saved_mem): Likewise.
19487 * calls.c (emit_library_call_value_1): Likewise.
19488 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
19489 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
19490 (gen_lowpart_for_combine): Likewise.
19491 * convert.c (convert_to_integer_1): Likewise.
19492 * cse.c (equiv_constant, cse_insn): Likewise.
19493 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
19494 (cselib_subst_to_values): Likewise.
19495 * dce.c (word_dce_process_block): Likewise.
19496 * df-problems.c (df_word_lr_mark_ref): Likewise.
19497 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
19498 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
19499 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
19500 (rtl_for_decl_location): Likewise.
19501 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
19502 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
19503 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
19504 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
19505 (expand_expr_real_1): Likewise.
19506 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
19507 (pad_below): Likewise.
19508 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
19509 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
19510 * ira.c (get_subreg_tracking_sizes): Likewise.
19511 * ira-build.c (ira_create_allocno_objects): Likewise.
19512 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
19513 (ira_sort_regnos_for_alter_reg): Likewise.
19514 * ira-costs.c (record_operand_costs): Likewise.
19515 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
19516 (resolve_simple_move): Likewise.
19517 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
19518 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
19519 (lra_constraints): Likewise.
19520 (CONST_POOL_OK_P): Reject variable-sized modes.
19521 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
19522 (add_pseudo_to_slot, lra_spill): Likewise.
19523 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19524 * optabs-query.c (get_best_extraction_insn): Likewise.
19525 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
19526 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
19527 (expand_mult_highpart, valid_multiword_target_p): Likewise.
19528 * recog.c (offsettable_address_addr_space_p): Likewise.
19529 * regcprop.c (maybe_mode_change): Likewise.
19530 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
19531 * regrename.c (build_def_use): Likewise.
19532 * regstat.c (dump_reg_info): Likewise.
19533 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
19534 (find_reloads, find_reloads_subreg_address): Likewise.
19535 * reload1.c (eliminate_regs_1): Likewise.
19536 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
19537 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
19538 (simplify_binary_operation_1, simplify_subreg): Likewise.
19539 * targhooks.c (default_function_arg_padding): Likewise.
19540 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
19541 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
19542 (verify_gimple_assign_ternary): Likewise.
19543 * tree-inline.c (estimate_move_cost): Likewise.
19544 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19545 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
19546 (get_address_cost_ainc): Likewise.
19547 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
19548 (vect_supportable_dr_alignment): Likewise.
19549 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
19550 (vectorizable_reduction): Likewise.
19551 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
19552 (vectorizable_operation, vectorizable_load): Likewise.
19553 * tree.c (build_same_sized_truth_vector_type): Likewise.
19554 * valtrack.c (cleanup_auto_inc_dec): Likewise.
19555 * var-tracking.c (emit_note_insn_var_location): Likewise.
19556 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
19557 (ADDR_VEC_ALIGN): Likewise.
19559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19560 Alan Hayward <alan.hayward@arm.com>
19561 David Sherwood <david.sherwood@arm.com>
19563 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
19565 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
19566 or if measurement_type is polynomial.
19567 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
19568 * combine.c (make_extraction): Likewise.
19569 * dse.c (find_shift_sequence): Likewise.
19570 * dwarf2out.c (mem_loc_descriptor): Likewise.
19571 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
19572 (extract_bit_field, extract_low_bits): Likewise.
19573 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
19574 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
19575 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
19576 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
19577 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
19578 * reload.c (find_reloads): Likewise.
19579 * reload1.c (alter_reg): Likewise.
19580 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
19581 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
19582 * tree-if-conv.c (predicate_mem_writes): Likewise.
19583 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19584 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
19585 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
19586 * valtrack.c (dead_debug_insert_temp): Likewise.
19587 * varasm.c (mergeable_constant_section): Likewise.
19588 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
19590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19591 Alan Hayward <alan.hayward@arm.com>
19592 David Sherwood <david.sherwood@arm.com>
19594 * expr.c (expand_assignment): Cope with polynomial mode sizes
19595 when assigning to a CONCAT.
19597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19598 Alan Hayward <alan.hayward@arm.com>
19599 David Sherwood <david.sherwood@arm.com>
19601 * machmode.h (mode_precision): Change from unsigned short to
19603 (mode_to_precision): Return a poly_uint16 rather than an unsigned
19605 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
19606 or if measurement_type is not polynomial.
19607 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
19608 in which the mode is already known to be a scalar_int_mode.
19609 * genmodes.c (emit_mode_precision): Change the type of mode_precision
19610 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
19612 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19613 for GET_MODE_PRECISION.
19614 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19615 for GET_MODE_PRECISION.
19616 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
19618 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
19619 (expand_field_assignment, make_extraction): Likewise.
19620 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
19621 (get_last_value): Likewise.
19622 * convert.c (convert_to_integer_1): Likewise.
19623 * cse.c (cse_insn): Likewise.
19624 * expr.c (expand_expr_real_1): Likewise.
19625 * lra-constraints.c (simplify_operand_subreg): Likewise.
19626 * optabs-query.c (can_atomic_load_p): Likewise.
19627 * optabs.c (expand_atomic_load): Likewise.
19628 (expand_atomic_store): Likewise.
19629 * ree.c (combine_reaching_defs): Likewise.
19630 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
19631 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
19632 * tree.h (type_has_mode_precision_p): Likewise.
19633 * ubsan.c (instrument_si_overflow): Likewise.
19635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19636 Alan Hayward <alan.hayward@arm.com>
19637 David Sherwood <david.sherwood@arm.com>
19639 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
19640 polynomial numbers of units.
19641 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
19642 (valid_vector_subparts_p): New function.
19643 (build_vector_type): Remove temporary shim and take the number
19644 of units as a poly_uint64 rather than an int.
19645 (build_opaque_vector_type): Take the number of units as a
19646 poly_uint64 rather than an int.
19647 * tree.c (build_vector_from_ctor): Handle polynomial
19648 TYPE_VECTOR_SUBPARTS.
19649 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
19650 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
19651 (build_vector_from_val): If the number of units is variable,
19652 use build_vec_duplicate_cst for constant operands and
19653 VEC_DUPLICATE_EXPR otherwise.
19654 (make_vector_type): Remove temporary is_constant ().
19655 (build_vector_type, build_opaque_vector_type): Take the number of
19656 units as a poly_uint64 rather than an int.
19657 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
19659 * cfgexpand.c (expand_debug_expr): Likewise.
19660 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
19661 (store_constructor, expand_expr_real_1): Likewise.
19662 (const_scalar_mask_from_tree): Likewise.
19663 * fold-const-call.c (fold_const_reduction): Likewise.
19664 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
19665 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
19666 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
19667 (fold_relational_const): Likewise.
19668 (native_interpret_vector): Likewise. Change the size from an
19669 int to an unsigned int.
19670 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
19671 TYPE_VECTOR_SUBPARTS.
19672 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
19673 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
19674 duplicating a non-constant operand into a variable-length vector.
19675 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
19676 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
19677 * ipa-icf.c (sem_variable::equals): Likewise.
19678 * match.pd: Likewise.
19679 * omp-simd-clone.c (simd_clone_subparts): Likewise.
19680 * print-tree.c (print_node): Likewise.
19681 * stor-layout.c (layout_type): Likewise.
19682 * targhooks.c (default_builtin_vectorization_cost): Likewise.
19683 * tree-cfg.c (verify_gimple_comparison): Likewise.
19684 (verify_gimple_assign_binary): Likewise.
19685 (verify_gimple_assign_ternary): Likewise.
19686 (verify_gimple_assign_single): Likewise.
19687 * tree-pretty-print.c (dump_generic_node): Likewise.
19688 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19689 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
19690 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
19691 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
19692 (vect_shift_permute_load_chain): Likewise.
19693 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
19694 (expand_vector_condition, optimize_vector_constructor): Likewise.
19695 (lower_vec_perm, get_compute_type): Likewise.
19696 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
19697 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
19698 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
19699 (vect_recog_mask_conversion_pattern): Likewise.
19700 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
19701 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
19702 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19703 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
19704 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
19705 (vectorizable_shift, vectorizable_operation, vectorizable_store)
19706 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
19707 (supportable_widening_operation): Likewise.
19708 (supportable_narrowing_operation): Likewise.
19709 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
19711 * varasm.c (output_constant): Likewise.
19713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19714 Alan Hayward <alan.hayward@arm.com>
19715 David Sherwood <david.sherwood@arm.com>
19717 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
19718 so that both the length == 3 and length != 3 cases set up their
19719 own permute vectors. Add comments explaining why we know the
19720 number of elements is constant.
19721 (vect_permute_load_chain): Likewise.
19723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19724 Alan Hayward <alan.hayward@arm.com>
19725 David Sherwood <david.sherwood@arm.com>
19727 * machmode.h (mode_nunits): Change from unsigned char to
19729 (ONLY_FIXED_SIZE_MODES): New macro.
19730 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
19731 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
19732 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
19734 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
19735 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
19736 or if measurement_type is not polynomial.
19737 * genmodes.c (ZERO_COEFFS): New macro.
19738 (emit_mode_nunits_inline): Make mode_nunits_inline return a
19740 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
19741 Use ZERO_COEFFS when emitting initializers.
19742 * data-streamer.h (bp_pack_poly_value): New function.
19743 (bp_unpack_poly_value): Likewise.
19744 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19745 for GET_MODE_NUNITS.
19746 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19747 for GET_MODE_NUNITS.
19748 * tree.c (make_vector_type): Remove temporary shim and make
19749 the real function take the number of units as a poly_uint64
19750 rather than an int.
19751 (build_vector_type_for_mode): Handle polynomial nunits.
19752 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
19753 * emit-rtl.c (const_vec_series_p_1): Likewise.
19754 (gen_rtx_CONST_VECTOR): Likewise.
19755 * fold-const.c (test_vec_duplicate_folding): Likewise.
19756 * genrecog.c (validate_pattern): Likewise.
19757 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
19758 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
19759 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
19760 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
19761 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
19762 * rtlanal.c (subreg_get_info): Likewise.
19763 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19764 (vect_grouped_load_supported): Likewise.
19765 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19766 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
19767 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19768 (simplify_const_unary_operation, simplify_binary_operation_1)
19769 (simplify_const_binary_operation, simplify_ternary_operation)
19770 (test_vector_ops_duplicate, test_vector_ops): Likewise.
19771 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
19772 instead of CONST_VECTOR_NUNITS.
19773 * varasm.c (output_constant_pool_2): Likewise.
19774 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
19775 explicit-encoded elements in the XVEC for variable-length vectors.
19777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19779 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
19781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19782 Alan Hayward <alan.hayward@arm.com>
19783 David Sherwood <david.sherwood@arm.com>
19785 * coretypes.h (fixed_size_mode): Declare.
19786 (fixed_size_mode_pod): New typedef.
19787 * builtins.h (target_builtins::x_apply_args_mode)
19788 (target_builtins::x_apply_result_mode): Change type to
19789 fixed_size_mode_pod.
19790 * builtins.c (apply_args_size, apply_result_size, result_vector)
19791 (expand_builtin_apply_args_1, expand_builtin_apply)
19792 (expand_builtin_return): Update accordingly.
19794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19796 * cse.c (hash_rtx_cb): Hash only the encoded elements.
19797 * cselib.c (cselib_hash_rtx): Likewise.
19798 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
19799 CONST_VECTOR encoding.
19801 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19802 Jeff Law <law@redhat.com>
19805 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
19806 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
19807 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
19808 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
19811 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
19812 explicitly probe *sp in a noreturn function if there were any callee
19813 register saves or frame pointer is needed.
19815 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19818 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
19819 BLKmode for ternary, binary or unary expressions.
19822 * var-tracking.c (delete_vta_debug_insn): New inline function.
19823 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
19824 insns from get_insns () to NULL instead of each bb separately.
19825 Use delete_vta_debug_insn. No longer static.
19826 (vt_debug_insns_local, variable_tracking_main_1): Adjust
19827 delete_vta_debug_insns callers.
19828 * rtl.h (delete_vta_debug_insns): Declare.
19829 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
19830 instead of variable_tracking_main.
19832 2018-01-03 Martin Sebor <msebor@redhat.com>
19834 PR tree-optimization/83603
19835 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
19836 arguments past the endof the argument list in functions declared
19837 without a prototype.
19838 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
19839 Avoid checking when arguments are null.
19841 2018-01-03 Martin Sebor <msebor@redhat.com>
19844 * doc/extend.texi (attribute const): Fix a typo.
19845 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
19846 issuing -Wsuggest-attribute for void functions.
19848 2018-01-03 Martin Sebor <msebor@redhat.com>
19850 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
19851 offset_int::from instead of wide_int::to_shwi.
19852 (maybe_diag_overlap): Remove assertion.
19853 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
19854 * gimple-ssa-sprintf.c (format_directive): Same.
19855 (parse_directive): Same.
19856 (sprintf_dom_walker::compute_format_length): Same.
19857 (try_substitute_return_value): Same.
19859 2018-01-03 Jeff Law <law@redhat.com>
19861 PR middle-end/83654
19862 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
19863 non-constant residual for zero at runtime and avoid probing in
19864 that case. Reorganize code for trailing problem to mirror handling
19867 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19869 PR tree-optimization/83501
19870 * tree-ssa-strlen.c (get_string_cst): New.
19871 (handle_char_store): Call get_string_cst.
19873 2018-01-03 Martin Liska <mliska@suse.cz>
19875 PR tree-optimization/83593
19876 * tree-ssa-strlen.c: Include tree-cfg.h.
19877 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
19878 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
19879 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
19881 (strlen_dom_walker::before_dom_children): Call
19882 gimple_purge_dead_eh_edges. Dump tranformation with details
19884 (strlen_dom_walker::before_dom_children): Update call by adding
19885 new argument cleanup_eh.
19886 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
19888 2018-01-03 Martin Liska <mliska@suse.cz>
19891 * cif-code.def (VARIADIC_THUNK): New enum value.
19892 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
19895 2018-01-03 Jan Beulich <jbeulich@suse.com>
19897 * sse.md (mov<mode>_internal): Tighten condition for when to use
19898 vmovdqu<ssescalarsize> for TI and OI modes.
19900 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19902 Update copyright years.
19904 2018-01-03 Martin Liska <mliska@suse.cz>
19907 * ipa-visibility.c (function_and_variable_visibility): Skip
19908 functions with noipa attribure.
19910 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19912 * gcc.c (process_command): Update copyright notice dates.
19913 * gcov-dump.c (print_version): Ditto.
19914 * gcov.c (print_version): Ditto.
19915 * gcov-tool.c (print_version): Ditto.
19916 * gengtype.c (create_file): Ditto.
19917 * doc/cpp.texi: Bump @copying's copyright year.
19918 * doc/cppinternals.texi: Ditto.
19919 * doc/gcc.texi: Ditto.
19920 * doc/gccint.texi: Ditto.
19921 * doc/gcov.texi: Ditto.
19922 * doc/install.texi: Ditto.
19923 * doc/invoke.texi: Ditto.
19925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19927 * vector-builder.h (vector_builder::m_full_nelts): Change from
19928 unsigned int to poly_uint64.
19929 (vector_builder::full_nelts): Update prototype accordingly.
19930 (vector_builder::new_vector): Likewise.
19931 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
19932 (vector_builder::operator ==): Likewise.
19933 (vector_builder::finalize): Likewise.
19934 * int-vector-builder.h (int_vector_builder::int_vector_builder):
19935 Take the number of elements as a poly_uint64 rather than an
19937 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
19938 from unsigned int to poly_uint64.
19939 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
19940 (vec_perm_indices::new_vector): Likewise.
19941 (vec_perm_indices::length): Likewise.
19942 (vec_perm_indices::nelts_per_input): Likewise.
19943 (vec_perm_indices::input_nelts): Likewise.
19944 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
19945 number of elements per input as a poly_uint64 rather than an
19946 unsigned int. Use the original encoding for variable-length
19947 vectors, rather than clamping each individual element.
19948 For the second and subsequent elements in each pattern,
19949 clamp the step and base before clamping their sum.
19950 (vec_perm_indices::series_p): Handle polynomial element counts.
19951 (vec_perm_indices::all_in_range_p): Likewise.
19952 (vec_perm_indices_to_tree): Likewise.
19953 (vec_perm_indices_to_rtx): Likewise.
19954 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19955 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
19956 (tree_vector_builder::new_binary_operation): Handle polynomial
19957 element counts. Return false if we need to know the number
19958 of elements at compile time.
19959 * fold-const.c (fold_vec_perm): Punt if the number of elements
19960 isn't known at compile time.
19962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19964 * vec-perm-indices.h (vec_perm_builder): Change element type
19965 from HOST_WIDE_INT to poly_int64.
19966 (vec_perm_indices::element_type): Update accordingly.
19967 (vec_perm_indices::clamp): Handle polynomial element_types.
19968 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19969 (vec_perm_indices::all_in_range_p): Likewise.
19970 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
19972 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
19973 polynomial vec_perm_indices element types.
19974 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
19975 * fold-const.c (fold_vec_perm): Likewise.
19976 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
19977 * tree-vect-generic.c (lower_vec_perm): Likewise.
19978 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19979 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
19980 element type to HOST_WIDE_INT.
19982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19983 Alan Hayward <alan.hayward@arm.com>
19984 David Sherwood <david.sherwood@arm.com>
19986 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
19987 rather than an int. Use plus_constant.
19988 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
19989 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
19991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19992 Alan Hayward <alan.hayward@arm.com>
19993 David Sherwood <david.sherwood@arm.com>
19995 * calls.c (emit_call_1, expand_call): Change struct_value_size from
19996 a HOST_WIDE_INT to a poly_int64.
19998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19999 Alan Hayward <alan.hayward@arm.com>
20000 David Sherwood <david.sherwood@arm.com>
20002 * calls.c (load_register_parameters): Cope with polynomial
20003 mode sizes. Require a constant size for BLKmode parameters
20004 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
20005 forces a parameter to be padded at the lsb end in order to
20006 fill a complete number of words, require the parameter size
20007 to be ordered wrt UNITS_PER_WORD.
20009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20010 Alan Hayward <alan.hayward@arm.com>
20011 David Sherwood <david.sherwood@arm.com>
20013 * reload1.c (spill_stack_slot_width): Change element type
20014 from unsigned int to poly_uint64_pod.
20015 (alter_reg): Treat mode sizes as polynomial.
20017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20018 Alan Hayward <alan.hayward@arm.com>
20019 David Sherwood <david.sherwood@arm.com>
20021 * reload.c (complex_word_subreg_p): New function.
20022 (reload_inner_reg_of_subreg, push_reload): Use it.
20024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20025 Alan Hayward <alan.hayward@arm.com>
20026 David Sherwood <david.sherwood@arm.com>
20028 * lra-constraints.c (process_alt_operands): Reject matched
20029 operands whose sizes aren't ordered.
20030 (match_reload): Refer to this check here.
20032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20033 Alan Hayward <alan.hayward@arm.com>
20034 David Sherwood <david.sherwood@arm.com>
20036 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
20037 that the mode size is in the set {1, 2, 4, 8, 16}.
20039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20040 Alan Hayward <alan.hayward@arm.com>
20041 David Sherwood <david.sherwood@arm.com>
20043 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
20044 Use plus_constant instead of gen_rtx_PLUS.
20046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20047 Alan Hayward <alan.hayward@arm.com>
20048 David Sherwood <david.sherwood@arm.com>
20050 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
20051 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
20052 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
20053 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
20054 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
20055 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
20056 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
20057 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
20058 * config/i386/i386.c (ix86_push_rounding): ...this new function.
20059 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
20061 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
20062 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
20063 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
20064 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
20065 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
20066 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
20067 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
20068 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
20069 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
20070 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
20072 * expr.c (emit_move_resolve_push): Treat the input and result
20073 of PUSH_ROUNDING as a poly_int64.
20074 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
20075 (emit_push_insn): Likewise.
20076 * lra-eliminations.c (mark_not_eliminable): Likewise.
20077 * recog.c (push_operand): Likewise.
20078 * reload1.c (elimination_effects): Likewise.
20079 * rtlanal.c (nonzero_bits1): Likewise.
20080 * calls.c (store_one_arg): Likewise. Require the padding to be
20081 known at compile time.
20083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20084 Alan Hayward <alan.hayward@arm.com>
20085 David Sherwood <david.sherwood@arm.com>
20087 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
20088 Use plus_constant instead of gen_rtx_PLUS.
20090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20091 Alan Hayward <alan.hayward@arm.com>
20092 David Sherwood <david.sherwood@arm.com>
20094 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
20095 rather than an int.
20097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20098 Alan Hayward <alan.hayward@arm.com>
20099 David Sherwood <david.sherwood@arm.com>
20101 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
20102 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
20103 via stack temporaries. Treat the mode size as polynomial too.
20105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20106 Alan Hayward <alan.hayward@arm.com>
20107 David Sherwood <david.sherwood@arm.com>
20109 * expr.c (expand_expr_real_2): When handling conversions involving
20110 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
20111 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
20112 as a poly_uint64 too.
20114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20115 Alan Hayward <alan.hayward@arm.com>
20116 David Sherwood <david.sherwood@arm.com>
20118 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
20120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20121 Alan Hayward <alan.hayward@arm.com>
20122 David Sherwood <david.sherwood@arm.com>
20124 * combine.c (can_change_dest_mode): Handle polynomial
20125 REGMODE_NATURAL_SIZE.
20126 * expmed.c (store_bit_field_1): Likewise.
20127 * expr.c (store_constructor): Likewise.
20128 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
20129 and polynomial REGMODE_NATURAL_SIZE.
20130 (gen_lowpart_common): Likewise.
20131 * reginfo.c (record_subregs_of_mode): Likewise.
20132 * rtlanal.c (read_modify_subreg_p): Likewise.
20134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20135 Alan Hayward <alan.hayward@arm.com>
20136 David Sherwood <david.sherwood@arm.com>
20138 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
20139 numbers of elements.
20141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20142 Alan Hayward <alan.hayward@arm.com>
20143 David Sherwood <david.sherwood@arm.com>
20145 * match.pd: Cope with polynomial numbers of vector elements.
20147 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20148 Alan Hayward <alan.hayward@arm.com>
20149 David Sherwood <david.sherwood@arm.com>
20151 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
20152 in a POINTER_PLUS_EXPR.
20154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20155 Alan Hayward <alan.hayward@arm.com>
20156 David Sherwood <david.sherwood@arm.com>
20158 * omp-simd-clone.c (simd_clone_subparts): New function.
20159 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
20160 (ipa_simd_modify_function_body): Likewise.
20162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20163 Alan Hayward <alan.hayward@arm.com>
20164 David Sherwood <david.sherwood@arm.com>
20166 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
20167 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
20168 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
20169 (expand_vector_condition, vector_element): Likewise.
20170 (subparts_gt): New function.
20171 (get_compute_type): Use subparts_gt.
20172 (count_type_subparts): Delete.
20173 (expand_vector_operations_1): Use subparts_gt instead of
20174 count_type_subparts.
20176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20177 Alan Hayward <alan.hayward@arm.com>
20178 David Sherwood <david.sherwood@arm.com>
20180 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
20181 (vect_compile_time_alias): ...this new function. Do the calculation
20182 on poly_ints rather than trees.
20183 (vect_prune_runtime_alias_test_list): Update call accordingly.
20185 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20186 Alan Hayward <alan.hayward@arm.com>
20187 David Sherwood <david.sherwood@arm.com>
20189 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
20191 (vect_schedule_slp_instance): Likewise.
20193 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20194 Alan Hayward <alan.hayward@arm.com>
20195 David Sherwood <david.sherwood@arm.com>
20197 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
20198 constant and extern definitions for variable-length vectors.
20199 (vect_get_constant_vectors): Note that the number of units
20200 is known to be constant.
20202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20203 Alan Hayward <alan.hayward@arm.com>
20204 David Sherwood <david.sherwood@arm.com>
20206 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
20207 of units as polynomial. Choose between WIDE and NARROW based
20210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20211 Alan Hayward <alan.hayward@arm.com>
20212 David Sherwood <david.sherwood@arm.com>
20214 * tree-vect-stmts.c (simd_clone_subparts): New function.
20215 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
20217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20218 Alan Hayward <alan.hayward@arm.com>
20219 David Sherwood <david.sherwood@arm.com>
20221 * tree-vect-stmts.c (vectorizable_call): Treat the number of
20222 vectors as polynomial. Use build_index_vector for
20223 IFN_GOMP_SIMD_LANE.
20225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20226 Alan Hayward <alan.hayward@arm.com>
20227 David Sherwood <david.sherwood@arm.com>
20229 * tree-vect-stmts.c (get_load_store_type): Treat the number of
20230 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
20231 for variable-length vectors.
20232 (vectorizable_mask_load_store): Treat the number of units as
20233 polynomial, asserting that it is constant if the condition has
20234 already been enforced.
20235 (vectorizable_store, vectorizable_load): Likewise.
20237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20238 Alan Hayward <alan.hayward@arm.com>
20239 David Sherwood <david.sherwood@arm.com>
20241 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
20242 of units as polynomial. Punt if we can't tell at compile time
20243 which vector contains the final result.
20245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20246 Alan Hayward <alan.hayward@arm.com>
20247 David Sherwood <david.sherwood@arm.com>
20249 * tree-vect-loop.c (vectorizable_induction): Treat the number
20250 of units as polynomial. Punt on SLP inductions. Use an integer
20251 VEC_SERIES_EXPR for variable-length integer reductions. Use a
20252 cast of such a series for variable-length floating-point
20255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20256 Alan Hayward <alan.hayward@arm.com>
20257 David Sherwood <david.sherwood@arm.com>
20259 * tree.h (build_index_vector): Declare.
20260 * tree.c (build_index_vector): New function.
20261 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
20262 of units as polynomial, forcibly converting it to a constant if
20263 vectorizable_reduction has already enforced the condition.
20264 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
20265 to create a {1,2,3,...} vector.
20266 (vectorizable_reduction): Treat the number of units as polynomial.
20267 Choose vectype_in based on the largest scalar element size rather
20268 than the smallest number of units. Enforce the restrictions
20271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20272 Alan Hayward <alan.hayward@arm.com>
20273 David Sherwood <david.sherwood@arm.com>
20275 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
20276 number of units as polynomial.
20278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20279 Alan Hayward <alan.hayward@arm.com>
20280 David Sherwood <david.sherwood@arm.com>
20282 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
20283 * target.def (autovectorize_vector_sizes): Return the vector sizes
20284 by pointer, using vector_sizes rather than a bitmask.
20285 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
20286 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
20287 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
20289 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
20290 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
20291 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
20292 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
20293 * omp-general.c (omp_max_vf): Likewise.
20294 * omp-low.c (omp_clause_aligned_alignment): Likewise.
20295 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
20296 * tree-vect-loop.c (vect_analyze_loop): Likewise.
20297 * tree-vect-slp.c (vect_slp_bb): Likewise.
20298 * doc/tm.texi: Regenerate.
20299 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
20301 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
20302 the vector size as a poly_uint64 rather than an unsigned int.
20303 (current_vector_size): Change from an unsigned int to a poly_uint64.
20304 (get_vectype_for_scalar_type): Update accordingly.
20305 * tree.h (build_truth_vector_type): Take the size and number of
20306 units as a poly_uint64 rather than an unsigned int.
20307 (build_vector_type): Add a temporary overload that takes
20308 the number of units as a poly_uint64 rather than an unsigned int.
20309 * tree.c (make_vector_type): Likewise.
20310 (build_truth_vector_type): Take the number of units as a poly_uint64
20311 rather than an unsigned int.
20313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20314 Alan Hayward <alan.hayward@arm.com>
20315 David Sherwood <david.sherwood@arm.com>
20317 * target.def (get_mask_mode): Take the number of units and length
20318 as poly_uint64s rather than unsigned ints.
20319 * targhooks.h (default_get_mask_mode): Update accordingly.
20320 * targhooks.c (default_get_mask_mode): Likewise.
20321 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
20322 * doc/tm.texi: Regenerate.
20324 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20325 Alan Hayward <alan.hayward@arm.com>
20326 David Sherwood <david.sherwood@arm.com>
20328 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
20329 * omp-general.c (omp_max_vf): Likewise.
20330 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
20331 (expand_omp_simd): Handle polynomial safelen.
20332 * omp-low.c (omplow_simd_context): Add a default constructor.
20333 (omplow_simd_context::max_vf): Change from int to poly_uint64.
20334 (lower_rec_simd_input_clauses): Update accordingly.
20335 (lower_rec_input_clauses): Likewise.
20337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20338 Alan Hayward <alan.hayward@arm.com>
20339 David Sherwood <david.sherwood@arm.com>
20341 * tree-vectorizer.h (vect_nunits_for_cost): New function.
20342 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
20343 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
20344 (vect_analyze_slp_cost): Likewise.
20345 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
20346 (vect_model_load_cost): Likewise.
20348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20349 Alan Hayward <alan.hayward@arm.com>
20350 David Sherwood <david.sherwood@arm.com>
20352 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
20353 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
20354 from an unsigned int * to a poly_uint64_pod *.
20355 (calculate_unrolling_factor): New function.
20356 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
20358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20359 Alan Hayward <alan.hayward@arm.com>
20360 David Sherwood <david.sherwood@arm.com>
20362 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
20363 from an unsigned int to a poly_uint64.
20364 (_loop_vec_info::slp_unrolling_factor): Likewise.
20365 (_loop_vec_info::vectorization_factor): Change from an int
20367 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
20368 (vect_get_num_vectors): New function.
20369 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
20370 (vect_get_num_copies): Use vect_get_num_vectors.
20371 (vect_analyze_data_ref_dependences): Change max_vf from an int *
20372 to an unsigned int *.
20373 (vect_analyze_data_refs): Change min_vf from an int * to a
20375 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
20376 than an unsigned HOST_WIDE_INT.
20377 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
20378 (vect_analyze_data_ref_dependence): Change max_vf from an int *
20379 to an unsigned int *.
20380 (vect_analyze_data_ref_dependences): Likewise.
20381 (vect_compute_data_ref_alignment): Handle polynomial vf.
20382 (vect_enhance_data_refs_alignment): Likewise.
20383 (vect_prune_runtime_alias_test_list): Likewise.
20384 (vect_shift_permute_load_chain): Likewise.
20385 (vect_supportable_dr_alignment): Likewise.
20386 (dependence_distance_ge_vf): Take the vectorization factor as a
20387 poly_uint64 rather than an unsigned HOST_WIDE_INT.
20388 (vect_analyze_data_refs): Change min_vf from an int * to a
20390 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
20391 vfm1 as a poly_uint64 rather than an int. Make the same change
20392 for the returned bound_scalar.
20393 (vect_gen_vector_loop_niters): Handle polynomial vf.
20394 (vect_do_peeling): Likewise. Update call to
20395 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
20396 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
20398 * tree-vect-loop.c (vect_determine_vectorization_factor)
20399 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
20400 (vect_get_known_peeling_cost): Likewise.
20401 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
20402 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
20403 (vect_transform_loop): Likewise. Use the lowest possible VF when
20404 updating the upper bounds of the loop.
20405 (vect_min_worthwhile_factor): Make static. Return an unsigned int
20406 rather than an int.
20407 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
20408 polynomial unroll factors.
20409 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
20410 (vect_make_slp_decision): Likewise.
20411 (vect_supported_load_permutation_p): Likewise, and polynomial
20413 (vect_analyze_slp_cost): Handle polynomial vf.
20414 (vect_slp_analyze_node_operations): Likewise.
20415 (vect_slp_analyze_bb_1): Likewise.
20416 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
20417 than an unsigned HOST_WIDE_INT.
20418 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
20419 (vectorizable_load): Handle polynomial vf.
20420 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
20422 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
20424 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20425 Alan Hayward <alan.hayward@arm.com>
20426 David Sherwood <david.sherwood@arm.com>
20428 * match.pd: Handle bit operations involving three constants
20429 and try to fold one pair.
20431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20433 * tree-vect-loop-manip.c: Include gimple-fold.h.
20434 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
20435 niters_maybe_zero parameters. Handle other cases besides a step of 1.
20436 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
20437 Add a path that uses a step of VF instead of 1, but disable it
20439 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
20440 and niters_no_overflow parameters. Update calls to
20441 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
20442 Create a new SSA name if the latter choses to use a ste other
20443 than zero, and return it via niters_vector_mult_vf_var.
20444 * tree-vect-loop.c (vect_transform_loop): Update calls to
20445 vect_do_peeling, vect_gen_vector_loop_niters and
20446 slpeel_make_loop_iterate_ntimes.
20447 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
20448 (vect_gen_vector_loop_niters): Update declarations after above changes.
20450 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
20452 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
20453 128-bit round to integer instructions.
20454 (ceil<mode>2): Likewise.
20455 (btrunc<mode>2): Likewise.
20456 (round<mode>2): Likewise.
20458 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20460 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
20461 unaligned VSX load/store on P8/P9.
20462 (expand_block_clear): Allow the use of unaligned VSX
20463 load/store on P8/P9.
20465 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20467 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
20469 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
20470 swap associated with both a load and a store.
20472 2018-01-02 Andrew Waterman <andrew@sifive.com>
20474 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
20475 * config/riscv/riscv.md (clear_cache): Use it.
20477 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
20479 * web.c: Remove out-of-date comment.
20481 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20483 * expr.c (fixup_args_size_notes): Check that any existing
20484 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
20485 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
20486 (emit_single_push_insn): ...here.
20488 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20490 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
20491 (const_vector_encoded_nelts): New function.
20492 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
20493 (const_vector_int_elt, const_vector_elt): Declare.
20494 * emit-rtl.c (const_vector_int_elt_1): New function.
20495 (const_vector_elt): Likewise.
20496 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
20497 of CONST_VECTOR_ELT.
20499 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20501 * expr.c: Include rtx-vector-builder.h.
20502 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
20503 directly on the tree encoding.
20504 (const_vector_from_tree): Likewise.
20505 * optabs.c: Include rtx-vector-builder.h.
20506 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
20507 sequence of "u" values.
20508 * vec-perm-indices.c: Include rtx-vector-builder.h.
20509 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
20510 directly on the vec_perm_indices encoding.
20512 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20514 * doc/rtl.texi (const_vector): Describe new encoding scheme.
20515 * Makefile.in (OBJS): Add rtx-vector-builder.o.
20516 * rtx-vector-builder.h: New file.
20517 * rtx-vector-builder.c: Likewise.
20518 * rtl.h (rtx_def::u2): Add a const_vector field.
20519 (CONST_VECTOR_NPATTERNS): New macro.
20520 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
20521 (CONST_VECTOR_DUPLICATE_P): Likewise.
20522 (CONST_VECTOR_STEPPED_P): Likewise.
20523 (CONST_VECTOR_ENCODED_ELT): Likewise.
20524 (const_vec_duplicate_p): Check for a duplicated vector encoding.
20525 (unwrap_const_vec_duplicate): Likewise.
20526 (const_vec_series_p): Check for a non-duplicated vector encoding.
20527 Say that the function only returns true for integer vectors.
20528 * emit-rtl.c: Include rtx-vector-builder.h.
20529 (gen_const_vec_duplicate_1): Delete.
20530 (gen_const_vector): Call gen_const_vec_duplicate instead of
20531 gen_const_vec_duplicate_1.
20532 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
20533 (gen_const_vec_duplicate): Use rtx_vector_builder.
20534 (gen_const_vec_series): Likewise.
20535 (gen_rtx_CONST_VECTOR): Likewise.
20536 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
20537 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
20538 Build a new vector rather than modifying a CONST_VECTOR in-place.
20539 (handle_special_swappables): Update call accordingly.
20540 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
20541 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
20542 Build a new vector rather than modifying a CONST_VECTOR in-place.
20543 (handle_special_swappables): Update call accordingly.
20545 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20547 * simplify-rtx.c (simplify_const_binary_operation): Use
20548 CONST_VECTOR_ELT instead of XVECEXP.
20550 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20552 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
20553 the selector elements to be different from the data elements
20554 if the selector is a VECTOR_CST.
20555 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
20556 ssizetype for the selector.
20558 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20560 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
20561 before testing each element individually.
20562 * tree-vect-generic.c (lower_vec_perm): Likewise.
20564 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20566 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
20567 * selftest-run-tests.c (selftest::run_tests): Call it.
20568 * vector-builder.h (vector_builder::operator ==): New function.
20569 (vector_builder::operator !=): Likewise.
20570 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
20571 (vec_perm_indices::all_from_input_p): New function.
20572 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
20573 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
20574 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
20575 instead of reading the VECTOR_CST directly. Detect whether both
20576 vector inputs are the same before constructing the vec_perm_indices,
20577 and update the number of inputs argument accordingly. Use the
20578 utility functions added above. Only construct sel2 if we need to.
20580 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20582 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
20583 the broadcast of the low byte.
20584 (expand_mult_highpart): Use an explicit encoding for the permutes.
20585 * optabs-query.c (can_mult_highpart_p): Likewise.
20586 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
20587 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20588 (vectorizable_bswap): Likewise.
20589 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
20590 explicit encoding for the power-of-2 permutes.
20591 (vect_permute_store_chain): Likewise.
20592 (vect_grouped_load_supported): Likewise.
20593 (vect_permute_load_chain): Likewise.
20595 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20597 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
20598 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
20599 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
20600 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
20601 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
20602 (vect_gen_perm_mask_any): Likewise.
20604 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20606 * int-vector-builder.h: New file.
20607 * vec-perm-indices.h: Include int-vector-builder.h.
20608 (vec_perm_indices): Redefine as an int_vector_builder.
20609 (auto_vec_perm_indices): Delete.
20610 (vec_perm_builder): Redefine as a stand-alone class.
20611 (vec_perm_indices::vec_perm_indices): New function.
20612 (vec_perm_indices::clamp): Likewise.
20613 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
20614 (vec_perm_indices::new_vector): New function.
20615 (vec_perm_indices::new_expanded_vector): Update for new
20616 vec_perm_indices class.
20617 (vec_perm_indices::rotate_inputs): New function.
20618 (vec_perm_indices::all_in_range_p): Operate directly on the
20619 encoded form, without computing elided elements.
20620 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
20621 encoding. Update for new vec_perm_indices class.
20622 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
20623 the given vec_perm_builder.
20624 (expand_vec_perm_var): Update vec_perm_builder constructor.
20625 (expand_mult_highpart): Use vec_perm_builder instead of
20626 auto_vec_perm_indices.
20627 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
20628 vec_perm_indices instead of auto_vec_perm_indices. Use a single
20629 or double series encoding as appropriate.
20630 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
20631 vec_perm_indices instead of auto_vec_perm_indices.
20632 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20633 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20634 (vect_permute_store_chain): Likewise.
20635 (vect_grouped_load_supported): Likewise.
20636 (vect_permute_load_chain): Likewise.
20637 (vect_shift_permute_load_chain): Likewise.
20638 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
20639 (vect_transform_slp_perm_load): Likewise.
20640 (vect_schedule_slp_instance): Likewise.
20641 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20642 (vectorizable_mask_load_store): Likewise.
20643 (vectorizable_bswap): Likewise.
20644 (vectorizable_store): Likewise.
20645 (vectorizable_load): Likewise.
20646 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
20647 vec_perm_indices instead of auto_vec_perm_indices. Use
20648 tree_to_vec_perm_builder to read the vector from a tree.
20649 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
20650 vec_perm_builder instead of a vec_perm_indices.
20651 (have_whole_vector_shift): Use vec_perm_builder and
20652 vec_perm_indices instead of auto_vec_perm_indices. Leave the
20653 truncation to calc_vec_perm_mask_for_shift.
20654 (vect_create_epilog_for_reduction): Likewise.
20655 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
20656 from auto_vec_perm_indices to vec_perm_indices.
20657 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
20658 instead of changing individual elements.
20659 (aarch64_vectorize_vec_perm_const): Use new_vector to install
20660 the vector in d.perm.
20661 * config/arm/arm.c (expand_vec_perm_d::perm): Change
20662 from auto_vec_perm_indices to vec_perm_indices.
20663 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
20664 instead of changing individual elements.
20665 (arm_vectorize_vec_perm_const): Use new_vector to install
20666 the vector in d.perm.
20667 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
20668 Update vec_perm_builder constructor.
20669 (rs6000_expand_interleave): Likewise.
20670 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
20671 (rs6000_expand_interleave): Likewise.
20673 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20675 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
20676 to qimode could truncate the indices.
20677 * optabs.c (expand_vec_perm_var): Likewise.
20679 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20681 * Makefile.in (OBJS): Add vec-perm-indices.o.
20682 * vec-perm-indices.h: New file.
20683 * vec-perm-indices.c: Likewise.
20684 * target.h (vec_perm_indices): Replace with a forward class
20686 (auto_vec_perm_indices): Move to vec-perm-indices.h.
20687 * optabs.h: Include vec-perm-indices.h.
20688 (expand_vec_perm): Delete.
20689 (selector_fits_mode_p, expand_vec_perm_var): Declare.
20690 (expand_vec_perm_const): Declare.
20691 * target.def (vec_perm_const_ok): Replace with...
20692 (vec_perm_const): ...this new hook.
20693 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
20694 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
20695 * doc/tm.texi: Regenerate.
20696 * optabs.def (vec_perm_const): Delete.
20697 * doc/md.texi (vec_perm_const): Likewise.
20698 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
20699 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
20700 expand_vec_perm for constant permutation vectors. Assert that
20701 the mode of variable permutation vectors is the integer equivalent
20702 of the mode that is being permuted.
20703 * optabs-query.h (selector_fits_mode_p): Declare.
20704 * optabs-query.c: Include vec-perm-indices.h.
20705 (selector_fits_mode_p): New function.
20706 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
20707 is defined, instead of checking whether the vec_perm_const_optab
20708 exists. Use targetm.vectorize.vec_perm_const instead of
20709 targetm.vectorize.vec_perm_const_ok. Check whether the indices
20710 fit in the vector mode before using a variable permute.
20711 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
20712 vec_perm_indices instead of an rtx.
20713 (expand_vec_perm): Replace with...
20714 (expand_vec_perm_const): ...this new function. Take the selector
20715 as a vec_perm_indices rather than an rtx. Also take the mode of
20716 the selector. Update call to shift_amt_for_vec_perm_mask.
20717 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
20718 Use vec_perm_indices::new_expanded_vector to expand the original
20719 selector into bytes. Check whether the indices fit in the vector
20720 mode before using a variable permute.
20721 (expand_vec_perm_var): Make global.
20722 (expand_mult_highpart): Use expand_vec_perm_const.
20723 * fold-const.c: Includes vec-perm-indices.h.
20724 * tree-ssa-forwprop.c: Likewise.
20725 * tree-vect-data-refs.c: Likewise.
20726 * tree-vect-generic.c: Likewise.
20727 * tree-vect-loop.c: Likewise.
20728 * tree-vect-slp.c: Likewise.
20729 * tree-vect-stmts.c: Likewise.
20730 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
20732 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
20733 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
20734 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
20735 (aarch64_vectorize_vec_perm_const): ...this new function.
20736 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
20737 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20738 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
20739 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
20740 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
20741 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20742 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
20744 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
20745 check for NEON modes.
20746 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
20747 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
20748 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
20749 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
20751 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
20752 the old VEC_PERM_CONST conditions.
20753 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
20754 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
20755 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
20756 (ia64_vectorize_vec_perm_const_ok): Merge into...
20757 (ia64_vectorize_vec_perm_const): ...this new function.
20758 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
20759 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
20760 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
20761 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
20762 * config/mips/mips.c (mips_expand_vec_perm_const)
20763 (mips_vectorize_vec_perm_const_ok): Merge into...
20764 (mips_vectorize_vec_perm_const): ...this new function.
20765 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
20766 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
20767 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
20768 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
20769 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
20770 (rs6000_expand_vec_perm_const): Delete.
20771 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
20773 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20774 (altivec_expand_vec_perm_const_le): Take each operand individually.
20775 Operate on constant selectors rather than rtxes.
20776 (altivec_expand_vec_perm_const): Likewise. Update call to
20777 altivec_expand_vec_perm_const_le.
20778 (rs6000_expand_vec_perm_const): Delete.
20779 (rs6000_vectorize_vec_perm_const_ok): Delete.
20780 (rs6000_vectorize_vec_perm_const): New function.
20781 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
20782 an element count and rtx array.
20783 (rs6000_expand_extract_even): Update call accordingly.
20784 (rs6000_expand_interleave): Likewise.
20785 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
20786 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
20787 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
20788 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
20789 (rs6000_expand_vec_perm_const): Delete.
20790 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
20791 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20792 (altivec_expand_vec_perm_const_le): Take each operand individually.
20793 Operate on constant selectors rather than rtxes.
20794 (altivec_expand_vec_perm_const): Likewise. Update call to
20795 altivec_expand_vec_perm_const_le.
20796 (rs6000_expand_vec_perm_const): Delete.
20797 (rs6000_vectorize_vec_perm_const_ok): Delete.
20798 (rs6000_vectorize_vec_perm_const): New function. Remove stray
20799 reference to the SPE evmerge intructions.
20800 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
20801 an element count and rtx array.
20802 (rs6000_expand_extract_even): Update call accordingly.
20803 (rs6000_expand_interleave): Likewise.
20804 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
20805 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
20807 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20809 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20811 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
20812 vector mode and that that mode matches the mode of the data
20814 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
20815 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
20816 directly using expand_vec_perm_1 when forcing selectors into
20818 (expand_vec_perm_var): New function, split out from expand_vec_perm.
20820 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20822 * optabs-query.h (can_vec_perm_p): Delete.
20823 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
20824 * optabs-query.c (can_vec_perm_p): Split into...
20825 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
20826 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
20827 particular selector is valid.
20828 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20829 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20830 (vect_grouped_load_supported): Likewise.
20831 (vect_shift_permute_load_chain): Likewise.
20832 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
20833 (vect_transform_slp_perm_load): Likewise.
20834 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20835 (vectorizable_bswap): Likewise.
20836 (vect_gen_perm_mask_checked): Likewise.
20837 * fold-const.c (fold_ternary_loc): Likewise. Don't take
20838 implementations of variable permutation vectors into account
20839 when deciding which selector to use.
20840 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
20841 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
20842 with a false third argument.
20843 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
20844 to test whether the constant selector is valid and can_vec_perm_var_p
20845 to test whether a variable selector is valid.
20847 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20849 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
20850 * optabs-query.c (can_vec_perm_p): Likewise.
20851 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
20852 instead of vec_perm_indices.
20853 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
20854 (vect_gen_perm_mask_checked): Likewise,
20855 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
20856 (vect_gen_perm_mask_checked): Likewise,
20858 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20860 * optabs-query.h (qimode_for_vec_perm): Declare.
20861 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
20862 (qimode_for_vec_perm): ...this new function.
20863 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
20865 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20867 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
20868 does not have a conditional at the top.
20870 2018-01-02 Richard Biener <rguenther@suse.de>
20872 * ipa-inline.c (big_speedup_p): Fix expression.
20874 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
20877 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
20880 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
20884 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
20885 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
20886 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
20887 cond_taken_branch_cost 3->4.
20889 2018-01-01 Jakub Jelinek <jakub@redhat.com>
20891 PR tree-optimization/83581
20892 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
20893 TODO_cleanup_cfg if any changes have been made.
20895 PR middle-end/83608
20896 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
20897 convert_modes if target mode has the right side, but different mode
20900 PR middle-end/83609
20901 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
20902 last argument when extracting from CONCAT. If either from_real or
20903 from_imag is NULL, use expansion through memory. If result is not
20904 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
20905 the parts directly to inner mode, if even that fails, use expansion
20908 PR middle-end/83623
20909 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
20910 check for bswap in mode rather than HImode and use that in expand_unop
20913 Copyright (C) 2018 Free Software Foundation, Inc.
20915 Copying and distribution of this file, with or without modification,
20916 are permitted in any medium without royalty provided the copyright
20917 notice and this notice are preserved.