1 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
4 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
5 STMT_VINFO_NUM_SLP_USES when it's true.
6 (vect_free_slp_instance): Add a final_p parameter and pass it to
8 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
9 (vect_analyze_slp_instance): Likewise.
10 (vect_slp_analyze_operations): Likewise.
11 (vect_slp_analyze_bb_1): Likewise.
12 * tree-vectorizer.c (vec_info): Likewise.
13 * tree-vect-loop.c (vect_transform_loop): Likewise.
15 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
17 * tree-vect-loop.c (vectorizable_reduction): Assert that the
18 function is not called for second and subsequent members of
21 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
23 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
24 cases for nested loops from here to ...
25 (vect_create_epilog_for_reduction): ...here. Only call
26 vect_is_simple_use for inner-loop reductions.
28 2018-07-31 Martin Liska <mliska@suse.cz>
33 * profile.c (struct location_triplet): New.
34 (struct location_triplet_hash): Likewise.
35 (output_location): Do not output a BB that
36 is already recorded for a line.
37 (branch_prob): Use streamed_locations.
39 2018-07-31 Martin Liska <mliska@suse.cz>
42 * coverage.c (coverage_begin_function): Do not mark target
43 clones as artificial functions.
45 2018-07-31 Martin Liska <mliska@suse.cz>
51 * profile.c (branch_prob): Do not record GOTO expressions
52 for GIMPLE statements which locations are already streamed.
54 2018-07-31 Olivier Hainque <hainque@adacore.com>
56 * gcc.c (handle_spec_function): Accept a soft_matched_part
57 argument, as do_spec_1. Pass it down to ...
58 (eval_spec_function): Accept a soft_matched_part argument,
59 and pass it down to ...
60 (do_spec_2): Accept a soft_matched_part argument, and pass
62 (do_spec_1): Pass soft_matched_part to handle_spec_function.
63 (handle_braces): Update call to handle_spec_function.
64 (driver::set_up_specs): Update calls to do_spec_2.
65 (compare_debug_dump_opt_spec_function): Likewise.
66 (compare_debug_self_opt_spec_function): Likewise.
68 2018-07-31 Olivier Hainque <hainque@adacore.com>
70 * common.opt (nolibc): New option.
71 * doc/invoke.texi (Link Options): Document it.
72 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
73 * config/alpha/linux.h: Likewise.
74 * config/arc/elf.h: Likewise.
75 * config/arm/uclinux-elf.h: Likewise.
76 * config/arm/unknown-elf.h: Likewise.
77 * config/avr/avrlibc.h: Likewise.
78 * config/bfin/bfin.h: Likewise.
79 * config/bfin/linux.h: Likewise.
80 * config/bfin/uclinux.h: Likewise.
81 * config/darwin.h: Likewise.
82 * config/darwin10.h: Likewise.
83 * config/darwin12.h: Likewise.
84 * config/gnu-user.h: Likewise.
85 * config/lm32/uclinux-elf.h: Likewise.
86 * config/pa/pa-hpux11.h: Likewise.
87 * config/pa/pa64-hpux.h: Likewise.
88 * config/sparc/sparc.h: Likewise.
90 2018-07-31 Olivier Hainque <hainque@adacore.com>
92 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
95 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
98 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
101 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
103 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
104 terminated string literal.
106 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
108 PR rtl-optimization/85160
109 * combine.c (is_just_move): New function.
110 (try_combine): Allow combining two instructions into two if neither of
111 the original instructions was a move.
113 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
116 * doc/extend.texi (Global Register Variables): Discourage use of type
118 (Local Register Variables): Likewise.
120 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
122 PR tree-optimization/86506
123 * hwint.h (ceil_log2): Resync with hwint.c implementation.
125 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
128 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
129 hard_regno, make sure no insn between `from` and `to` clobbers it.
131 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
132 Tom de Vries <tdevries@suse.de>
134 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
135 (PTX_DEFAULT_RUNTIME_DIM): ... this.
136 (nvptx_goacc_validate_dims): Set default worker and gang dims to
137 PTX_DEFAULT_RUNTIME_DIM.
138 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
140 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
142 * config/pa/pa.c (pa_output_addr_vec): Align address table.
143 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
144 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
146 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
148 * config/rs6000/constraints.md (wG constraint): Delete, no longer
150 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
151 predicate to reflect toc fusion has been deleted.
152 (toc_fusion_mem_raw): Delete, no longer used.
153 (toc_fusion_mem_wrapped): Likewise.
154 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
156 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
157 Delete, no longer used.
158 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
159 meant to be used for toc fusion.
160 (rs6000_debug_print_mode): Delete toc fusion debugging.
161 (rs6000_debug_reg_global): Likewise.
162 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
163 fusion and secondary reload support that were never used.
164 (rs6000_option_override_internal): Delete TOC fusion, that was only
165 partially defined, and it did not work unless you also used the
167 (rs6000_legitimate_address_p): Delete TOC fusion support.
168 (rs6000_opt_masks): Likewise.
169 (fusion_wrap_memory_address): Delete function, no longer used.
170 (fusion_split_address); Delete TOC fusion support.
171 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
172 longer used with toc fusion being deleted.
173 (TARGET_TOC_FUSION_FP): Likewise.
174 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
176 (toc fusion spliter): Delete TOC fusion support.
177 (toc_fusionload_<mode>): Likewise.
178 (toc_fusionload_di): Likewise.
179 (fusion_gpr_load_<mode>): Delete generator function, this insn no
180 longer needs to be named. Rename predicate to delete TOC fusion.
181 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
182 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
183 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
184 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
185 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
187 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
189 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
190 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
191 __int128 in built-in function prototypes.
192 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
193 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
195 2018-07-27 Martin Sebor <msebor@redhat.com>
197 PR tree-optimization/86696
198 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
199 types, including enums.
200 (handle_char_store): Be prepared for the above function to fail.
202 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
204 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
205 when optimization level is lower than 2 or optimize for size.
207 2018-07-26 Martin Sebor <msebor@redhat.com>
209 PR tree-optimization/86043
210 PR tree-optimization/86042
211 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
212 (get_string_cst_length): Rename...
213 (get_min_string_length): ...to this. Add argument.
214 (handle_char_store): Extend to handle multi-character stores by
216 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
217 * tree.h (initializer_zerop): Add argument.
219 2018-07-26 Jakub Jelinek <jakub@redhat.com>
222 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
223 declare target to variables if they have always,{to,from,tofrom} map
226 2018-07-26 Martin Liska <mliska@suse.cz>
229 * lto-wrapper.c: Add linker_output as prefix
230 for ltrans_output_file.
232 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
234 PR rtl-optimization/85805
235 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
236 value for hard registers if that was written in the same mode.
238 2018-07-26 Martin Liska <mliska@suse.cz>
240 PR gcov-profile/86536
241 * gcov.c (format_gcov): Use printf format %.*f directly
242 and do not handle special values.
244 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
246 * common/config/arc/arc-common.c (arc_option_optimization_table):
247 Update default optimizations for size.
249 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
251 * config/arc/arc.md (movsf_insn): Add short instruction selection.
252 * config/arc/constraints.md (CfZ): New constraint.
253 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
254 (subsf3_fpu): Likewise.
255 (cmpsf_fpu): Likewise.
256 (cmpsf_fpu_uneq): Likewise.
258 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
260 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
262 (arc_is_uncached_mem_p): uncached applies to both the variable and
265 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
267 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
270 2018-07-25 David Malcolm <dmalcolm@redhat.com>
272 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
273 field "m_scopes" from vec to auto_vec.
275 2018-07-25 Martin Liska <mliska@suse.cz>
277 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
280 2018-07-25 Richard Biener <rguenther@suse.de>
283 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
284 special wrt context_die late.
285 (gen_subprogram_die): Re-use DIEs in local scope.
287 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
289 PR tree-optimization/86644
290 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
292 2018-07-25 Martin Liska <mliska@suse.cz>
295 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
296 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
298 2018-07-25 Martin Liska <mliska@suse.cz>
301 * params.def: Explain ASan abbreviation and provide
302 a documentation link.
304 2018-07-24 Martin Sebor <msebor@redhat.com>
306 PR tree-optimization/86622
307 PR tree-optimization/86532
308 * builtins.h (string_length): Declare.
309 * builtins.c (c_strlen): Correct handling of non-constant offsets.
310 (check_access): Be prepared for non-constant length ranges.
311 (string_length): Make extern.
312 * expr.c (string_constant): Only handle the minor non-constant
313 array index. Use string_constant to compute the length of
314 a generic string constant.
316 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
318 PR tree-optimization/86618
319 * tree-vect-stmts.c (vectorizable_call): Don't take the address
320 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
322 2018-07-24 David Malcolm <dmalcolm@redhat.com>
324 PR tree-optimization/86636
325 * json.cc (json::object::set): Fix comment. Add assertions.
326 (json::array::append): Move here from json.h. Add comment and an
328 (json::string::string): Likewise.
329 * json.h (json::array::append): Move to json.cc.
330 (json::string::string): Likewise.
331 * optinfo-emit-json.cc
332 (optrecord_json_writer::impl_location_to_json): Assert that we
333 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
334 wrapper around it. Expand the location once, rather than three
336 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
337 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
339 (optrecord_json_writer::optinfo_to_json): Likewise, in four
340 places. Fix some overlong lines.
342 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
344 * config/aarch64/aarch64-simd.md
345 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
346 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
347 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
348 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
349 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
350 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
351 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
352 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
353 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
355 2018-07-24 Jakub Jelinek <jakub@redhat.com>
358 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
359 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
360 and abs_d == d, do the power of two handling if profitable.
362 2018-07-24 Richard Biener <rguenther@suse.de>
364 * match.pd: Add BIT_FIELD_REF canonicalizations.
366 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
369 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
371 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
373 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
374 terminated STRING_CST object.
376 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
378 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
380 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
382 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
383 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
384 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
385 the elements into a register.
386 (rs6000_split_v4si_init_di_reg): Delete.
387 (rs6000_split_v4si_init): Delete.
388 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
389 (vsx_init_v4si): Rewrite as a define_expand.
391 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
393 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
394 zero_extend argument from memory): New.
396 2018-07-22 Martin Sebor <msebor@redhat.com>
399 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
400 diagnosing calls with unknown arguments unless -Walloca-larger-than
401 is restricted to less than PTRDIFF_MAX bytes.
403 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
405 * doc/gcov.texi (Invoking Gcov): Editorial changes.
407 2018-07-20 David Malcolm <dmalcolm@redhat.com>
409 * pretty-print.c (text_info::set_location): Remove redundant
410 "line_table" parameter from call to rich_location::set_range.
412 2018-07-20 Martin Sebor <msebor@redhat.com>
415 * builtins.c (expand_builtin_alloca): Adjust.
416 * calls.c (alloc_max_size): Simplify.
417 * cgraphunit.c (cgraph_node::expand): Adjust.
418 * common.opt (larger_than_size, warn_frame_larger_than): Remove
420 (frame_larger_than_size): Same.
421 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
422 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
424 * doc/invoke.texi (GCC Command Options): Document option arguments.
425 Explain byte-size arguments and suffixes.
426 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
427 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
428 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
429 * doc/options.texi (UInteger): Expand.
430 (Host_Wide_Int, ByteSize): Document new properties.
431 * final.c (final_start_function_1): Include sizes in an error message.
432 * function.c (frame_offset_overflow): Same.
433 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
434 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
435 Diagnose unbounded alloca calls only for limits of less than
437 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
438 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
440 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
441 only for limits of less than PTRDIFF_MAX.
442 * langhooks-def.h (lhd_handle_option): Change function argument
444 * langhooks.c (lhd_handle_option): Same.
445 * langhooks.h (handle_option): Same.
446 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
448 (var_type, var_type_struct): Same.
449 (var_set): Handle ByteSize flag.
450 * optc-gen.awk: Add comments to output to ease debugging. Make
451 use of HOST_WIDE_INT where appropriate.
452 * opts-gen-save.awk: Use %lx to format unsigned long.
453 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
454 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
455 arguments. Parse bytes-size suffixes.
456 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
457 (enum_value_to_arg): Same.
458 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
459 (handle_option): Adjust.
460 (generate_option): Change function argument to HOST_WIDE_INT.
461 (cmdline_handle_error): Adjust.
462 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
463 (set_option): Change function argument to HOST_WIDE_INT.
464 (option_enabled): Handle cl_host_wide_int.
465 (get_option_state): Handle CLVC_SIZE.
466 (control_warning_option): Same.
467 * opts.c (common_handle_option): Change function argument to
468 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
469 OPT_Wvla_larger_than_.
470 * opts.h (enum cl_var_type): Add an enumerator.
471 * stor-layout.c (layout_decl): Print a more meaningful warning.
472 * toplev.c (output_stack_usage): Adjust.
474 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
476 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
477 call to inline_expand_builtin_string_cmp.
478 (expand_builtin_strcmp): Likewise.
479 (expand_builtin_strncmp): Likewise.
480 (inline_string_cmp): Delete the last parameter, change char_type_node
481 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
483 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
484 the inlining expansion on target where the type of the call has same or
485 narrower precision than unsigned char.
487 2018-07-20 David Malcolm <dmalcolm@redhat.com>
489 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
490 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
491 * common.opt (fsave-optimization-record): New option.
492 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
493 * doc/invoke.texi (-fsave-optimization-record): New option.
494 * dumpfile.c: Include "optinfo-emit-json.h".
495 (struct kv_pair): Move to coretypes.h.
496 (optgroup_options): Make non-static.
497 (dump_context::end_scope): Call
498 optimization_records_maybe_pop_dump_scope.
499 * dumpfile.h (optgroup_options): New decl.
502 * optinfo-emit-json.cc: New file.
503 * optinfo-emit-json.h: New file.
504 * optinfo.cc: Include "optinfo-emit-json.h".
505 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
506 (optinfo_enabled_p): Check optimization_records_enabled_p.
507 (optinfo_wants_inlining_info_p): Likewise.
508 * optinfo.h: Update comment.
509 * profile-count.c (profile_quality_as_string): New function.
510 * profile-count.h (profile_quality_as_string): New decl.
511 (profile_count::quality): New accessor.
512 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
513 and optinfo_emit_json_cc_tests.
514 * selftest.h (selftest::json_cc_tests): New decl.
515 (selftest::optinfo_emit_json_cc_tests): New decl.
516 * toplev.c: Include "optinfo-emit-json.h".
517 (compile_file): Call optimization_records_finish.
518 (do_compile): Call optimization_records_start.
519 * tree-ssa-live.c: Include optinfo.h.
520 (remove_unused_scope_block_p): Retain inlining information if
521 optinfo_wants_inlining_info_p returns true.
523 2018-07-20 Richard Biener <rguenther@suse.de>
526 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
527 to cover -flto-partition=none.
529 2018-07-20 Martin Liska <mliska@suse.cz>
531 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
532 (get_decl_source_range): Remove unused function.
534 2018-07-20 Richard Biener <rguenther@suse.de>
536 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
537 (struct vn_phi_s): Likewise.
538 (struct vn_reference_s): Likewise.
539 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
540 for searching the slot of an entry known to be in the hash itself.
541 (vn_phi_hasher::equal): Likewise.
542 (vn_reference_hasher::equal): Likewise.
543 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
545 (optimistic_info, current_info): Remove, keeping only valid_info.
546 (vn_reference_lookup_1): Remove fallback lookup.
547 (vn_reference_lookup_2): Likewise.
548 (vn_nary_op_lookup_1): Likewise.
549 (vn_phi_lookup): Likewise.
550 (vn_nary_build_or_lookup_1): Make sure to not chain the built
552 (vn_reference_insert): Adjust, chain the inserted hash element
553 at last_inserted_ref.
554 (vn_reference_insert_pieces): Likewise.
555 (visit_reference_op_call): Likewise.
556 (vn_nary_op_insert_into): Chain the inserted hash element at
558 (vn_nary_op_insert_pieces): Adjust.
559 (vn_nary_op_insert): Likewise.
560 (vn_nary_op_insert_stmt): Likewise.
561 (vn_phi_insert): Adjust, chain the inserted hash element at
563 (process_scc): Remove clearing and copying the optimistic
564 table. Instead remove elements inserted during an optimistic
565 iteration from the single table we maintain.
566 (init_scc_vn): Adjust.
567 (free_scc_vn): Likewise.
568 (sccvn_dom_walker::record_cond): Likewise.
569 (sccvn_dom_walker::after_dom_children): Likewise.
571 2018-07-19 Martin Sebor <msebor@redhat.com>
573 PR tree-optimization/84047
574 PR tree-optimization/83776
575 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
576 (check_array_bounds): Call it.
578 2018-07-19 Martin Sebor <msebor@redhat.com>
580 * align.h (align_flags): Use member initialization.
582 2018-07-19 David Malcolm <dmalcolm@redhat.com>
584 * Makefile.in (OBJS): Add optinfo.o.
585 * coretypes.h (class symtab_node): New forward decl.
586 (struct cgraph_node): New forward decl.
587 (class varpool_node): New forward decl.
588 * dump-context.h: New file.
589 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
591 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
592 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
593 (set_alt_dump_file): Likewise.
594 (dump_context::~dump_context): New dtor.
595 (dump_gimple_stmt): Move implementation to...
596 (dump_context::dump_gimple_stmt): ...this new member function.
597 Add the stmt to any pending optinfo, creating one if need be.
598 (dump_gimple_stmt_loc): Move implementation to...
599 (dump_context::dump_gimple_stmt_loc): ...this new member function.
600 Start a new optinfo and add the stmt to it.
601 (dump_gimple_expr): Move implementation to...
602 (dump_context::dump_gimple_expr): ...this new member function.
603 Add the stmt to any pending optinfo, creating one if need be.
604 (dump_gimple_expr_loc): Move implementation to...
605 (dump_context::dump_gimple_expr_loc): ...this new member function.
606 Start a new optinfo and add the stmt to it.
607 (dump_generic_expr): Move implementation to...
608 (dump_context::dump_generic_expr): ...this new member function.
609 Add the tree to any pending optinfo, creating one if need be.
610 (dump_generic_expr_loc): Move implementation to...
611 (dump_context::dump_generic_expr_loc): ...this new member
612 function. Add the tree to any pending optinfo, creating one if
614 (dump_printf): Move implementation to...
615 (dump_context::dump_printf_va): ...this new member function. Add
616 the text to any pending optinfo, creating one if need be.
617 (dump_printf_loc): Move implementation to...
618 (dump_context::dump_printf_loc_va): ...this new member function.
619 Start a new optinfo and add the stmt to it.
620 (dump_dec): Move implementation to...
621 (dump_context::dump_dec): ...this new member function. Add the
622 value to any pending optinfo, creating one if need be.
623 (dump_context::dump_symtab_node): New member function.
624 (dump_context::get_scope_depth): New member function.
625 (dump_context::begin_scope): New member function.
626 (dump_context::end_scope): New member function.
627 (dump_context::ensure_pending_optinfo): New member function.
628 (dump_context::begin_next_optinfo): New member function.
629 (dump_context::end_any_optinfo): New member function.
630 (dump_context::s_current): New global.
631 (dump_context::s_default): New global.
632 (dump_scope_depth): Delete global.
633 (dumpfile_ensure_any_optinfo_are_flushed): New function.
634 (dump_symtab_node): New function.
635 (get_dump_scope_depth): Reimplement in terms of dump_context.
636 (dump_begin_scope): Likewise.
637 (dump_end_scope): Likewise.
638 (selftest::temp_dump_context::temp_dump_context): New ctor.
639 (selftest::temp_dump_context::~temp_dump_context): New dtor.
640 (selftest::verify_item): New function.
641 (ASSERT_IS_TEXT): New macro.
642 (ASSERT_IS_TREE): New macro.
643 (ASSERT_IS_GIMPLE): New macro.
644 (selftest::test_capture_of_dump_calls): New test.
645 (selftest::dumpfile_c_tests): Call it.
646 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
647 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
648 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
650 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
651 (dump_node, dump_bb): Move these unrelated decls.
652 (class dump_manager): Add leading comment.
653 * optinfo.cc: New file.
654 * optinfo.h: New file.
656 2018-07-19 Michael Collison <michael.collison@arm.com>
657 Richard Henderson <rth@redhat.com>
659 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
660 (subti): Handle op1 zero.
661 (subvti4, usub4ti4): New.
662 (*sub<GPI>3_compare1_imm): New.
663 (sub<GPI>3_carryinCV): New.
664 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
665 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
667 2018-07-19 Michael Collison <michael.collison@arm.com>
668 Richard Henderson <rth@redhat.com>
670 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
671 (addti3): Create simpler code if low part is already known to be 0.
672 (addvti4, uaddvti4): New.
673 (*add<GPI>3_compareC_cconly_imm): New.
674 (*add<GPI>3_compareC_cconly): New.
675 (*add<GPI>3_compareC_imm): New.
676 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
677 handle constants within this pattern..
678 (*add<GPI>3_compareV_cconly_imm): New.
679 (*add<GPI>3_compareV_cconly): New.
680 (*add<GPI>3_compareV_imm): New.
681 (add<GPI>3_compareV): New.
682 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
683 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
684 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
685 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
686 with 'comparison' operator.
687 (*add<GPI>3_compareV_cconly_imm): Ditto.
688 (*add<GPI>3_compareV_cconly): Ditto.
689 (*add<GPI>3_compareV_imm): Ditto.
690 (add<GPI>3_compareV): Ditto.
691 (add<mode>3_carryinC): Ditto.
692 (*add<mode>3_carryinC_zero): Ditto.
693 (*add<mode>3_carryinC): Ditto.
694 (add<mode>3_carryinV): Ditto.
695 (*add<mode>3_carryinV_zero): Ditto.
696 (*add<mode>3_carryinV): Ditto.
698 2018-07-19 Michael Collison <michael.collison@arm.com>
699 Richard Henderson <rth@redhat.com>
701 * config/aarch64/aarch64-modes.def (CC_V): New.
702 * config/aarch64/aarch64-protos.h
703 (aarch64_addti_scratch_regs): Declare
704 (aarch64_subvti_scratch_regs): Declare.
705 (aarch64_expand_subvti): Declare.
706 (aarch64_gen_unlikely_cbranch): Declare
707 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
708 for signed overflow using CC_Vmode.
709 (aarch64_get_condition_code_1): Handle CC_Vmode.
710 (aarch64_gen_unlikely_cbranch): New function.
711 (aarch64_addti_scratch_regs): New function.
712 (aarch64_subvti_scratch_regs): New function.
713 (aarch64_expand_subvti): New function.
715 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
717 * config/aarch64/aarch64-option-extensions.def: New entry for profile
719 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
720 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
723 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
726 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
727 address check not strict.
729 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
731 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
733 (store_pair_lanes<mode>): Likewise.
734 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
735 enum value 'ADDR_QUERY_LDP_STP_N'.
736 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
737 (aarch64_print_address_internal): Add declaration.
738 (aarch64_print_ldpstp_address): Remove.
739 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
740 (aarch64_print_operand): Change printing of 'y'.
741 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
742 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
743 'true' rather than '1'.
744 * config/aarch64/constraints.md (Uml): Likewise.
745 (Uml): Rename to Umn.
748 2018-07-19 Richard Biener <rguenther@suse.de>
750 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
752 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
753 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
754 (vn_reference_hasher): Likewise.
755 (struct vn_tables_s): Remove obstack and alloc-pool members.
756 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
757 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
758 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
759 (vn_reference_insert_pieces): Likewise.
760 (alloc_vn_nary_op_noinit): Adjust.
761 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
763 (shared_lookup_phiargs): Remove.
764 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
765 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
766 (visit_reference_op_call): Likewise.
767 (copy_nary, copy_phi, copy_reference): Remove.
768 (process_scc): Rewind the obstack when iterating. Do not
769 copy the elements to valid_info but just move them from one
770 hashtable to the other.
771 (allocate_vn_table): Adjust.
772 (free_vn_table): Likewise.
773 (init_scc_vn): Likewise.
774 (free_scc_vn): Likewise.
776 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
779 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
780 indirect_return as function type attribute.
781 (ix86_attribute_table): Change indirect_return to function
783 * doc/extend.texi: Update indirect_return attribute.
785 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
787 * wide-int.h (widest2_int): New.
788 * gimple-fold.c (arith_overflowed_p): Use it.
789 * tree.h (widest2_int_cst): New.
790 * tree-vrp.c (wide_int_binop_overflow): Rename from
792 Rewrite to work on trees.
793 (extract_range_from_multiplicative_op_1): Abstract code to...
794 (wide_int_range_min_max): ...here.
795 (wide_int_range_cross_product): ...and here.
796 (extract_range_from_binary_expr_1): Abstract overflow code to...
797 (wide_int_range_mult_wrapping): ...here.
798 * tree-vrp.h (wide_int_range_cross_product): New.
799 (wide_int_range_mult_wrapping): New.
801 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
802 Julia Koval <julia.koval@intel.com>
804 * config/i386/x86-tune-costs.h (skylake_memcpy,
805 skylake_memset): Replace rep_prefix with unrolling for size 512.
807 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
810 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle comparision with EQ_EXPR
813 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
815 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
816 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
817 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
818 previously known as "PowerPC AltiVec Built-in Functions". Move
819 some material to new subsubsections "PowerPC AltiVec Built-in
820 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
822 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
823 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
824 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
825 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
827 2018-07-18 Richard Biener <rguenther@suse.de>
829 PR tree-optimization/86557
830 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
833 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
835 * config/s390/s390.c (s390_function_profiler): Generate CFI.
837 2018-07-17 Jeff Law <law@redhat.com>
839 * config/arm/arm.c (get_label_padding): Update for recent
840 changes to label_to_alignment.
842 PR tree-optimization/86010
843 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
845 * config/mips/mips.c (vr4130_align_insns): Update for recent
846 changes to label_to_alignment.
848 * config/frv/frv.c (frv_label_align): Update for recent changes
849 to label_to_alignment.
851 * config/nios2/nios2.c (nios2_label_align): Update for recent
852 changes which dropped ALIGN_LABELS_LOG.
854 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
856 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
857 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
859 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
861 * config/arc/arc.c (arc_label_align): Use align_labels instead of
862 deprecated align_labels_log.
864 2018-07-17 Richard Biener <rguenther@suse.de>
867 * dwarf2out.c (init_sections_and_labels): Always generate
868 a debug_line_str_section for early LTO debug.
869 (dwarf2out_finish): Reset debug_line_str_hash output early.
870 Bump counter for extra dwarf5 .debug_loc labels to not conflict
872 (dwarf2out_early_finish): Output debug_line_str.
874 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
876 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
877 index register on z196 or later.
879 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
881 * config/s390/s390.c (s390_default_align): Set default function
883 (s390_override_options_after_change): Call s390_default align.
884 (s390_option_override_internal): Call s390_default_align.
885 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
887 2018-07-17 Jakub Jelinek <jakub@redhat.com>
890 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
891 to _looptemp_ clauses, other than the first two.
893 2018-07-17 Martin Liska <mliska@suse.cz>
895 * opts.c: Do not enable OPT_falign_* for -Os.
897 2018-07-17 Martin Liska <mliska@suse.cz>
899 * align.h (MAX_CODE_ALIGN): New.
900 (MAX_CODE_ALIGN_VALUE): New.
901 * common/config/i386/i386-common.c (ix86_handle_option):
902 (MAX_CODE_ALIGN): Moved to align.h.
903 * final.c (MAX_CODE_ALIGN): Likewise.
904 * opts.c (parse_and_check_align_values):
905 (MAX_CODE_ALIGN): Likewise.
906 (MAX_CODE_ALIGN_VALUE): Likewise.
908 2018-07-17 Martin Liska <mliska@suse.cz>
910 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
911 in order to fulfil coding style.
912 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
913 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
914 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
915 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
916 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
917 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
918 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
919 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
920 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
922 2018-07-17 Martin Liska <mliska@suse.cz>
925 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
926 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
927 align_flags of label_to_alignment.
928 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
930 * config/m68k/m68k.c: Do not use removed align_labels_value and
932 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
933 (LOOP_ALIGN): Likewise.
934 (LABEL_ALIGN): Likewise.
935 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
936 Remove not used macro.
937 (rs6000_loop_align): Change return type to align_flags.
938 (rs6000_loop_align_max_skip): Remove.
939 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
940 Change return type to align_flags.
941 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
942 Remove not used macro.
943 (rs6000_loop_align): Change return type to align_flags.
944 (rs6000_loop_align_max_skip): Remove.
945 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
946 * config/rx/rx-protos.h (rx_align_for_label): Make it
948 * config/rx/rx.c (rx_align_for_label): Change return type
950 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
952 into align_flags class.
953 (LABEL_ALIGN): Likewise.
954 (LOOP_ALIGN): Likewise.
955 * config/s390/s390.c (s390_label_align): Use align_flags
957 (s390_asm_output_function_label): Likewise.
958 * config/sh/sh.c (sh_override_options_after_change):
959 Use align_flags class directly without macros.
960 (find_barrier): Likewise.
961 (barrier_align): Likewise.
962 (sh_loop_align): Likewise.
963 * config/spu/spu.c (spu_option_override):
964 Use align_flags_tuple::get_value instead of removed macros.
965 (spu_sched_init): Likewise.
966 * config/spu/spu.h (GTY): Likewise.
967 * config/visium/visium.c (visium_option_override):
968 Set "8" as default secondary alignment.
969 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
970 in order to guarantee secondary alignment of 8.
971 * coretypes.h: Include align.h header file.
972 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
973 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
974 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
975 * doc/tm.texi.in: Likewise.
976 * final.c (struct label_alignment): Remove not used structure.
977 (LABEL_ALIGN): Change type to align_flags.
978 (LOOP_ALIGN): Likewise.
979 (JUMP_ALIGN): Likewise.
980 (default_loop_align_max_skip): Remove.
981 (default_label_align_max_skip): Likewise.
982 (default_jump_align_max_skip): Likewise.
983 (default_label_align_after_barrier_max_skip):
984 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
985 (LABEL_TO_MAX_SKIP): Remove.
986 (label_to_alignment): Return align_flags type instead of integer.
987 (label_to_max_skip): Remove.
988 (align_fuzz): Use align_flags type.
989 (compute_alignments): Use align_flags type and use align_flags::max
990 to combine multiple alignments.
991 (grow_label_align): Grow vec instead of C array.
992 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
993 (shorten_branches): Use align_flags type and use align_flags::max
994 to combine multiple alignments.
995 (final_scan_insn_1): Remove usage of secondary alignment that comes
996 from label alignment, but instead use proper secondary alignment
997 which is computed in grow_label_align.
998 * flags.h (struct align_flags_tuple): Move to align.h.
999 (struct align_flags): Likewise.
1000 (state_align_loops): Rename to align_loops.
1001 (state_align_jumps): Rename to align_jumps.
1002 (state_align_labels): Rename to align_labels.
1003 (state_align_functions): Rename to align_functions.
1004 (align_loops_log): Remove.
1005 (align_jumps_log): Remove.
1006 (align_labels_log): Remove.
1007 (align_functions_log): Remove.
1008 (align_loops_max_skip): Remove.
1009 (align_jumps_max_skip): Remove.
1010 (align_labels_max_skip): Remove.
1011 (align_functions_max_skip): Remove.
1012 (align_loops_value): Remove.
1013 (align_jumps_value): Remove.
1014 (align_labels_value): Remove.
1015 (align_functions_value): Remove.
1016 * output.h (label_to_alignment): Change return type to align_flags.
1017 (label_to_max_skip): Remove.
1018 * target.def: Remove loop_align_max_skip, label_align_max_skip,
1019 jump_align_max_skip macros.
1020 * targhooks.h (default_loop_align_max_skip): Remove.
1021 (default_label_align_max_skip): Likewise.
1022 (default_jump_align_max_skip): Likewise.
1023 (default_label_align_after_barrier_max_skip): Remove.
1024 * toplev.c (read_log_maxskip): Use ::normalize function.
1025 (parse_N_M): Remove not used argument and also call ::normalize.
1026 (parse_alignment_opts): Do not pass unused arguments.
1027 * varasm.c (assemble_start_function): Use directly align_functions
1028 instead of removed macros.
1029 * system.h: Do not poison removed macros.
1031 2018-07-17 Jakub Jelinek <jakub@redhat.com>
1034 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
1035 and cond temporaries don't have reference type if iterator has
1036 pointer type. For init use &for_pre_body instead of pre_p if
1037 for_pre_body is non-empty.
1039 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1041 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
1042 double-double modes to SFmode directly directly.
1043 (trunc<mode>sf2_fprs): Delete.
1045 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1047 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
1048 for conversions between IFmode and the decimal floating point modes.
1049 (init_float128_ieee): Use the correct names for conversions between
1050 KFmode and the decimal floating point modes.
1052 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1054 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
1055 for the conversions between TDmode and IFmode.
1056 (init_float128_ieee): Use more correct names for the conversions
1057 between TDmode and KFmode.
1059 2018-07-16 Jakub Jelinek <jakub@redhat.com>
1061 PR tree-optimization/86526
1062 * builtins.c (expand_builtin_memcmp): Formatting fixes.
1063 (inline_expand_builtin_string_cmp): Likewise.
1064 (inline_string_cmp): Likewise. Use c_readstr instead of
1065 builtin_memcpy_read_str. Add unit_mode temporary.
1067 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
1070 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
1071 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
1073 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
1075 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
1076 Alphabetize prototypes of built-in functions, separating out
1077 built-in functions that are listed in this section but should be
1078 described elsewhere.
1080 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
1083 * expmed.c (emit_store_flag): Do not emit setcc followed by a
1084 conditional move when trapping comparison was split to a
1085 non-trapping one (and vice versa).
1087 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1089 * config/s390/s390.c (s390_function_profiler): Generate nops
1090 instead of profiler call sequences.
1091 * config/s390/s390.opt: Add the new option.
1093 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1095 * config/s390/s390.c (s390_function_profiler): Generate
1096 __mcount_loc section.
1097 * config/s390/s390.opt: Add the new option.
1099 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1101 * common.opt: Add the new warning.
1102 * config/s390/s390.c (s390_function_profiler): Emit "brasl
1103 %r0,__fentry__" when -mfentry is specified.
1104 (s390_option_override_internal): Disallow -mfentry for 31-bit
1106 * config/s390/s390.opt: Add the new option.
1108 2018-07-16 Richard Biener <rguenther@suse.de>
1111 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
1112 for function-local FUNCTION_DECL and RESULT_DECL immediately.
1114 2018-07-16 Martin Liska <mliska@suse.cz>
1117 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
1120 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
1122 * config/arc/arcHS.md: Update ARCHS scheduling rules.
1124 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
1126 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
1128 * config/arc/arc-cpus.def (hs4x): New cpu.
1130 * config/arc/arc-tables.opt: Regenerate.
1131 * config/arc/arc.c (arc_sched_issue_rate): New function.
1132 (TARGET_SCHED_ISSUE_RATE): Define.
1133 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
1134 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
1136 (attr tune): Add ARCHS4x tune values.
1137 (attr tune_dspmpy): Define.
1138 (*tst): Correct instruction type.
1139 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
1140 * config/arc/arcHS4x.md: New file.
1141 * config/arc/fpu.md: Update instruction type attributes.
1142 * config/arc/t-multilib: Regenerate.
1144 2018-07-16 Tom de Vries <tdevries@suse.de>
1147 * var-tracking.c (vt_initialize): Fix pre_dec handling.
1149 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1151 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
1154 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
1156 PR tree-optimization/86514
1157 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
1158 conversion to a boolean type from a type with greater precision.
1160 2018-07-16 Tom de Vries <tdevries@suse.de>
1162 * var-tracking.c (vt_initialize): Print adjusted insn slim if
1163 dump_flags request TDF_SLIM.
1165 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
1167 * fold-const.c (int_const_binop_1): Abstract...
1168 (wide_int_binop): ...wide int code here.
1169 (poly_int_binop): ...poly int code here.
1170 Abstract the rest of int_const_binop_1 into int_const_binop.
1171 * fold-const.h (wide_int_binop): New.
1172 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
1173 Remove useless PLUS/MINUS_EXPR case.
1174 (zero_nonzero_bits_from_vr): Move wide int code...
1175 (zero_nonzero_bits_from_bounds): ...here.
1176 (extract_range_from_binary_expr_1): Move mask optimization code...
1177 (range_easy_mask_min_max): ...here.
1178 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
1179 (range_easy_mask_min_max): New.
1181 2018-07-15 Jeff Law <law@redhat.com>
1184 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
1187 2018-07-14 Jim Wilson <jimw@sifive.com>
1189 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
1191 2018-07-14 Paul Koning <ni1d@arrl.net>
1193 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
1195 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
1197 * lto-streamer-out.c (copy_function_or_variable): Dump info about
1200 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
1201 Steve Munroe <munroesj52@gmail.com>
1203 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
1204 (_mm_andnot_si128): Likewise.
1205 (_mm_or_si128): Likewise.
1206 (_mm_xor_si128): Likewise.
1208 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
1211 * builtins.c (expand_builtin_memcmp): Inline the calls first
1212 when result_eq is false.
1213 (expand_builtin_strcmp): Inline the calls first.
1214 (expand_builtin_strncmp): Likewise.
1215 (inline_string_cmp): New routine. Expand a string compare
1216 call by using a sequence of char comparison.
1217 (inline_expand_builtin_string_cmp): New routine. Inline expansion
1218 a call to str(n)cmp/memcmp.
1219 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
1221 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
1223 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
1225 * config/arm/driver-arm.c: Include arm-native.h.
1226 (host_detect_local_cpu): Use auto-generated data tables.
1227 (vendors, arm_cpu_table): Delete. Move part information to ...
1228 * config/arm/arm-cpus.in: ... here.
1229 * config/arm/parsecpu.awk (gen_native): New function.
1230 (vendor, part): New CPU fields.
1231 (END): Add support for building the native CPU detection tables.
1232 * config/arm/t-arm (arm-native.h): Add build rule.
1233 (driver-arm.o): Add dependency on arm-native.h.
1235 2018-07-13 Richard Biener <rguenther@suse.de>
1238 * match.pd (addr1 - addr2): Allow either of the operand to
1241 2018-07-13 Tom de Vries <tdevries@suse.de>
1243 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
1246 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
1248 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
1249 arrays instead of numbered variables.
1251 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
1253 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
1254 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
1256 2018-07-13 Richard Biener <rguenther@suse.de>
1259 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
1260 instead of get_context_die.
1262 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
1263 Richard Biener <rguenther@suse.de>
1266 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
1267 that the loop latch destination where phi is defined.
1269 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
1271 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
1272 (riscv_expand_epilogue): Add assertion to check interrupt mode.
1273 (riscv_set_current_function): Extract getting interrupt type to new
1275 (riscv_get_interrupt_type): New function.
1276 (riscv_merge_decl_attributes): New function, checking interrupt type is
1278 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
1280 2018-07-12 Paul Koning <ni1d@arrl.net>
1282 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
1285 2018-07-12 Paul Koning <ni1d@arrl.net>
1287 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
1288 zero reference, add doloop_end instead.
1289 * doc/md.texi (decrement_and_branch_until_zero): Remove.
1290 (Looping patterns): Remove decrement_and_branch_until_zero. Add
1291 detail for doloop_end.
1293 2018-07-12 Martin Sebor <msebor@redhat.com>
1296 * attribs.c (decl_attributes): Reject conflicting attributes before
1297 calling attribute handlers.
1299 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
1301 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
1303 (gcc::dump_manager::get_dump_file_name): likewise.
1304 (dump_begin): Likewise.
1305 * dumpfile.h (dump_begin): Update prototype.
1306 (gcc::dump_manager::get_dump_file_name,
1307 gcc::dump_manager::get_dump_file_name): Update prototype.
1309 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1311 * internal-fn.h (vectorizable_internal_fn_p): New function.
1312 * tree-vect-slp.c (compatible_calls_p): Likewise.
1313 (vect_build_slp_tree_1): Remove nops argument. Handle calls
1314 to internal functions.
1315 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
1317 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1319 * fold-const.h (inverse_conditions_p): Declare.
1320 * fold-const.c (inverse_conditions_p): New function.
1321 * match.pd: Use inverse_conditions_p. Add folds of view_converts
1322 that test the inverse condition of a conditional internal function.
1323 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
1324 * internal-fn.c (internal_fn_mask_index): Handle conditional
1326 (vectorized_internal_fn_supported_p): New function.
1327 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
1328 (any_pred_load_store): Replace with...
1329 (need_to_predicate): ...this new variable.
1330 (redundant_ssa_names): New variable.
1331 (ifcvt_can_use_mask_load_store): Move initial checks to...
1332 (ifcvt_can_predicate): ...this new function. Handle tree codes
1333 for which a conditional internal function exists.
1334 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
1335 instead of ifcvt_can_use_mask_load_store. Update after variable
1337 (predicate_load_or_store): New function, split out from
1338 predicate_mem_writes.
1339 (check_redundant_cond_expr): New function.
1340 (value_available_p): Likewise.
1341 (predicate_rhs_code): Likewise.
1342 (predicate_mem_writes): Rename to...
1343 (predicate_statements): ...this. Use predicate_load_or_store
1344 and predicate_rhs_code.
1345 (combine_blocks, tree_if_conversion): Update after above name changes.
1346 (ifcvt_local_dce): Handle redundant_ssa_names.
1347 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
1348 general conditional functions.
1349 * tree-vect-stmts.c (vectorizable_call): Likewise.
1351 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1352 Alan Hayward <alan.hayward@arm.com>
1353 David Sherwood <david.sherwood@arm.com>
1355 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
1356 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
1357 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
1358 plus and minus and convert them into IFN_COND_FMA-based sequences.
1359 (convert_mult_to_fma): Handle conditional plus and minus.
1361 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1363 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
1364 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
1365 (cond_fnms_optab): New optabs.
1366 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
1368 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
1369 * internal-fn.h (get_conditional_internal_fn): Declare.
1370 (get_unconditional_internal_fn): Likewise.
1371 * internal-fn.c (cond_ternary_direct): New macro.
1372 (expand_cond_ternary_optab_fn): Likewise.
1373 (direct_cond_ternary_optab_supported_p): Likewise.
1374 (FOR_EACH_COND_FN_PAIR): Likewise.
1375 (get_conditional_internal_fn): New function.
1376 (get_unconditional_internal_fn): Likewise.
1377 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
1378 (gimple_match_op::gimple_match_op): Add a new overload for 5
1380 (gimple_match_op::set_op): Likewise.
1381 (gimple_resimplify5): Declare.
1382 * genmatch.c (decision_tree::gen): Generate simplifications for
1384 * gimple-match-head.c (gimple_simplify): Define an overload for
1385 5 operands. Handle calls with 5 arguments in the top-level overload.
1386 (convert_conditional_op): Handle conversions from unconditional
1387 internal functions to conditional ones.
1388 (gimple_resimplify5): New function.
1389 (build_call_internal): Pass a fifth operand.
1390 (maybe_push_res_to_seq): Likewise.
1391 (try_conditional_simplification): Try converting conditional
1392 internal functions to unconditional internal functions.
1393 Handle 3-operand unconditional forms.
1394 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
1395 Define ternary equivalents of the current rules for binary conditional
1397 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
1399 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
1400 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
1401 (optab): Handle them.
1402 (SVE_COND_FP_TERNARY): New int iterator.
1403 (sve_fmla_op, sve_fmad_op): New int attributes.
1404 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
1405 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
1406 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
1408 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1410 * target.def (preferred_else_value): New target hook.
1411 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
1412 * doc/tm.texi: Regenerate.
1413 * targhooks.h (default_preferred_else_value): Declare.
1414 * targhooks.c (default_preferred_else_value): New function.
1415 * internal-fn.h (conditional_internal_fn_code): Declare.
1416 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
1417 (get_conditional_internal_fn): Use it.
1418 (conditional_internal_fn_code): New function.
1419 * gimple-match.h (gimple_match_cond): New struct.
1420 (gimple_match_op): Add a cond member function.
1421 (gimple_match_op::gimple_match_op): Update all forms to take a
1423 * genmatch.c (expr::gen_transform): Use the same condition as res_op
1424 for the suboperation, but don't specify a particular else_value.
1425 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
1426 (visit_nary_op, visit_reference_op_load): Pass
1427 gimple_match_cond::UNCOND to the gimple_match_op constructor.
1428 * gimple-match-head.c: Include tree-eh.h
1429 (convert_conditional_op): New function.
1430 (maybe_resimplify_conditional_op): Likewise.
1431 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
1432 (gimple_resimplify2): Likewise.
1433 (gimple_resimplify3): Likewise.
1434 (gimple_resimplify4): Likewise.
1435 (maybe_push_res_to_seq): Return null for conditional operations.
1436 (try_conditional_simplification): New function.
1437 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
1439 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
1441 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
1443 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
1445 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
1447 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1449 (hash_tree): Do not hash DECL_FCONTEXT
1450 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
1451 Do not stream DECL_FCONTEXT.
1452 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
1453 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
1455 2018-07-12 Richard Biener <rguenther@suse.de>
1458 * dwarf2out.c (gen_block_die): Only output blocks when they have
1459 at least one !DECL_IGNORED_P variable.
1461 2018-07-12 Richard Biener <rguenther@suse.de>
1464 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
1465 Remove -mieee-fp handling.
1467 2018-07-12 Richard Biener <rguenther@suse.de>
1469 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
1470 left-over from last patch.
1472 2018-07-12 Jakub Jelinek <jakub@redhat.com>
1474 PR tree-optimization/86492
1475 * gimple-ssa-store-merging.c
1476 (imm_store_chain_info::coalesce_immediate_stores): Call
1477 check_no_overlap even for the merge_overlapping case. Formatting fix.
1479 2018-07-12 Richard Biener <rguenther@suse.de>
1482 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
1483 move possibly trapping operations into the conditional.
1485 2018-07-12 Richard Biener <rguenther@suse.de>
1487 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
1488 (vn_lookup_simplify_result): Remove recursion limit applied
1490 (vn_nary_build_or_lookup_1): Adjust.
1491 (try_to_simplify): Likewise.
1492 * gimple-match-head.c (gimple_resimplify1): Instead apply one
1494 (gimple_resimplify2): Likewise.
1495 (gimple_resimplify3): Likewise.
1496 (gimple_resimplify4): Likewise.
1498 2018-07-11 Jakub Jelinek <jakub@redhat.com>
1500 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
1501 Use __mmask64 type instead of __mmask8 for __M argument.
1502 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
1503 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
1505 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
1506 __mmask16 for __M argument.
1507 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
1508 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
1509 to __mmask16 instead of __mmask8.
1510 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
1511 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
1512 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
1513 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
1514 instead of __mmask16 for __U argument.
1515 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
1516 __mmask16 instead of __mmask8 for __U argument.
1517 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
1519 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
1521 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
1523 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
1525 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
1527 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
1529 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
1531 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
1533 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
1535 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
1537 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
1538 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
1539 return type as well as __M argument type and all casts from __mmask8
1541 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
1542 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
1543 return type as well as __M argument type and all casts from __mmask8
1545 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
1546 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
1547 return type as well as __M argument type and all casts from __mmask8
1549 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
1550 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
1551 return type as well as __M argument type and all casts from __mmask8
1553 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
1554 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
1557 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
1559 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
1560 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
1563 2018-07-11 Paul Koning <ni1d@arrl.net>
1565 * doc/md.texi (define_subst): Document how multiple occurrences of
1566 the same argument in the replacement pattern are handled.
1568 2018-07-11 Paul Koning <ni1d@arrl.net>
1570 * doc/extend.texi (Common Variable Attributes): Move "mode" into
1572 (Common Type Attributes): Add "mode" attribute.
1574 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
1576 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
1577 stream DECL_ORIGINAL_TYPE.
1578 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
1579 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
1580 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
1581 Do not walk original type.
1582 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
1584 (write_ts_decl_non_common_tree_pointers): Do not stream
1586 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
1587 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
1589 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
1591 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
1592 thread twice from the same starting edge.
1594 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
1596 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
1597 * gimple.c (gimple_call_nonnull_result_p): ...here...
1598 (gimple_call_nonnull_arg): ...and here.
1599 * gimple.h (gimple_call_nonnull_result_p): New.
1600 (gimple_call_nonnull_arg): New.
1602 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
1604 * config/arm/arm-cpus.in: Move information from fpu field of each
1605 cpu definition to the isa field.
1606 * config/arm/parsecpu.awk (fpu): Delete match rule.
1607 (gen_comm_data): Don't add bits from the CPU's FPU entry.
1609 2018-07-11 Richard Biener <rguenther@suse.de>
1612 * dwarf2out.c (init_sections_and_labels): Use
1613 output_asm_line_debug_info consistently.
1614 (dwarf2out_early_finish): Likewise.
1615 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
1618 2018-07-11 Richard Biener <rguenther@suse.de>
1620 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
1621 Rework father_bb setting in a way to avoid propagating constants
1622 multiple times on a loop body.
1624 2018-07-10 Mark Wielaard <mark@klomp.org>
1627 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
1629 2018-07-10 Richard Biener <rguenther@suse.de>
1631 * hash-map.h (hash_map::iterator::operator*): Return
1632 references to key and value.
1634 2018-07-10 Jakub Jelinek <jakub@redhat.com>
1637 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
1638 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
1639 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
1640 (gimplify_omp_for): For composite loops, move outer
1641 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
1642 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
1643 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
1644 TREE_LIST for both the original class iterator and the "last" helper
1645 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
1646 loop, remember has_decl_expr from outer composite loops for the
1647 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
1649 2018-07-09 Martin Sebor <msebor@redhat.com>
1653 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
1654 accessing implicitly initialized array elements.
1655 * expr.c (string_constant): Handle string initializers of
1656 character arrays within aggregates.
1657 * gimple-fold.c (fold_array_ctor_reference): Add argument.
1658 Store element offset. As a special case, handle zero size.
1659 (fold_nonarray_ctor_reference): Same.
1660 (fold_ctor_reference): Add argument. Store subobject offset.
1661 * gimple-fold.h (fold_ctor_reference): Add argument.
1663 2018-07-09 Paul Koning <ni1d@arrl.net>
1665 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
1666 (pdp11_insn_cost): New function.
1667 (pdp11_md_asm_adjust): New function.
1668 (TARGET_INVALID_WITHIN_DOLOOP): Define.
1669 (pdp11_rtx_costs): Update to match machine better.
1670 (output_addr_const_pdp11): Correct format mismatch warnings.
1671 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
1672 * config/pdp11/pdp11.md: General change to add base_cost and/or
1673 length attributes for use by new pdp11_insn_cost function.
1674 (MIN_BRANCH): Correct definition.
1676 (doloop_end): Use standard pattern name for looping pattern.
1677 (doloop_end_nocc): New.
1678 (movsf): Add another constraint alternative.
1679 (zero_extendqihi2): Add constraint alternatives for not in place
1681 (zero_extendhisi2): Remove.
1682 (shift patterns): Add CC handling variants.
1686 (define_peephole2): New peephole to recognize mov that sets CC for
1689 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1691 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
1692 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
1695 2018-07-09 Jakub Jelinek <jakub@redhat.com>
1698 * real.c (real_nextafter): Return true if result is denormal.
1700 2018-07-09 Martin Liska <mliska@suse.cz>
1702 * common.opt: Add back wrongly removed attribute.
1704 2018-07-09 Richard Biener <rguenther@suse.de>
1707 * dwarf2out.c (gen_block_die): For an early generated DIE
1708 always output high/low PC attributes.
1710 2018-07-09 Tom de Vries <tdevries@suse.de>
1712 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
1713 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
1714 onto VAR_DECL with abstract origin.
1716 2018-07-07 Jim Wilson <jimw@sifive.com>
1718 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
1720 2018-07-07 Tom de Vries <tdevries@suse.de>
1722 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
1724 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
1726 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
1728 (combine_bound): Use wide-int overflow calculation instead of
1730 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
1732 * fold-const.c (int_const_binop_2): Same.
1733 (extract_muldiv_1): Same.
1734 (fold_div_compare): Same.
1735 (fold_abs_const): Same.
1737 * poly-int.h (add): Same.
1741 * predict.c (predict_iv_comparison): Same.
1742 * profile-count.c (slow_safe_scale_64bit): Same.
1743 * simplify-rtx.c (simplify_const_binary_operation): Same.
1744 * tree-chrec.c (tree_fold_binomial): Same.
1745 * tree-data-ref.c (split_constant_offset_1): Same.
1746 * tree-if-conv.c (idx_within_array_bound): Same.
1747 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
1748 * tree-ssa-phiopt.c (minmax_replacement): Same.
1749 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
1750 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
1751 * vr-values.c (vr_values::adjust_range_with_scev): Same.
1752 * wide-int.cc (wi::add_large): Same.
1753 (wi::mul_internal): Same.
1754 (wi::sub_large): Same.
1755 (wi::divmod_internal): Same.
1756 * wide-int.h: Change overflow type to overflow_type for neg, add,
1757 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
1758 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
1759 mul_internal, divmod_internal.
1760 (overflow_type): New enum.
1761 (accumulate_overflow): New.
1763 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
1765 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
1766 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
1768 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
1770 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
1771 argument is checked for zero before entering loop, avoid checking again.
1773 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
1775 * gimplify.h (generic_expr_could_trap_p): Set as global function.
1776 * gimplify.h (generic_expr_could_trap_p): Likwise.
1777 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
1779 2018-07-06 Jakub Jelinek <jakub@redhat.com>
1781 PR tree-optimization/86401
1782 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
1783 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
1784 (fold_bit_and_mask): ... here. New helper function for match.pd.
1785 * fold-const.h (fold_bit_and_mask): Declare.
1786 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
1788 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
1791 * target.def (translate_mode_attribute): New hook.
1792 * targhooks.h (default_translate_mode_attribute): Declare.
1793 * targhooks.c (default_translate_mode_attribute): New function.
1794 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
1795 * doc/tm.texi: Regenerate.
1796 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
1797 (rs6000_translate_mode_attribute): New function.
1799 2018-07-06 Paul Koning <ni1d@arrl.net>
1801 * doc/md.texi (define_split): Document DONE and FAIL.
1802 (define_peephole2): Ditto.
1804 2018-07-05 Jeff Law <law@redhat.com>
1806 PR tree-optimization/86010
1807 * tree-ssa-dse.c (compute_trims): More aggressively trim at
1808 both the head and tail of mem* and str* calls.
1810 2018-07-05 Jim Wilson <jimw@sifive.com>
1812 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
1814 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
1816 * config/aarch64/aarch64-simd.md: correct flags text for
1817 MIN_EXPR replacement.
1819 2018-07-05 James Clarke <jrtc27@jrtc27.com>
1821 * configure: Regenerated.
1823 2018-07-05 Carl Love <cel@us.ibm.com>
1825 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
1826 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
1827 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
1828 VSX_BUILTIN_DOUBLEL_V4SF.
1830 2018-07-05 Martin Sebor <msebor@redhat.com>
1833 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
1834 than its domain to compute its the upper bound of a char array.
1836 2018-07-05 Nathan Sidwell <nathan@acm.org>
1838 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
1839 * doc/cpp.texi: Update comment.
1840 * doc/tm.texi: Rebuilt.
1841 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
1842 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
1843 * doc/extend.texi (Backwards Compatibility): Clarify it is system
1844 headers affected by extern "C".
1845 * system.h: Poison NO_IMPLICIT_EXTERN_C.
1846 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
1847 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
1848 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
1849 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
1850 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
1851 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
1852 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
1853 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
1854 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
1855 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
1856 config/sparc/sp64-elf.h, config/spu/spu.h,
1857 config/stormy16/stormy16.h, config/v850/v850.h,
1858 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
1859 define NO_IMPLICIT_EXTERN_C.
1860 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
1862 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1865 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1866 instead of GET_MODE_SIZE when comparing Units.
1868 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1871 * rtlanal.c (set_noop_p): Constrain on mode change,
1872 include hard-reg-set.h
1874 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1876 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
1878 2018-07-05 Jakub Jelinek <jakub@redhat.com>
1881 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
1884 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
1886 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
1888 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
1891 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
1893 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
1895 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
1898 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
1900 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
1902 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
1903 neg_*_op* variables.
1905 2018-07-04 Martin Liska <mliska@suse.cz>
1907 * tree-switch-conversion.c: Define
1908 max_ratio_for_speed and max_ratio_for_size constants.
1910 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
1911 Martin Liska <mliska@suse.cz>
1916 * common.opt: Rename align options with 'str_' prefix.
1917 * common/config/i386/i386-common.c (set_malign_value): New
1919 (ix86_handle_option): Use it to set -falign-* options/
1920 * config/aarch64/aarch64-protos.h (struct tune_params): Change
1921 type from int to string.
1922 * config/aarch64/aarch64.c: Update default values from int
1924 * config/alpha/alpha.c (alpha_override_options_after_change):
1926 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
1927 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1928 max skip conditionally.
1929 * config/i386/freebsd.h (SUBALIGN_LOG): New.
1930 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1931 max skip conditionally.
1932 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1933 max skip conditionally.
1934 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
1935 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1936 max skip conditionally.
1937 * config/i386/i386.c (struct ptt): Change type from int to
1939 (ix86_default_align): Set default values.
1940 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
1941 max skip conditionally.
1942 * config/i386/iamcu.h (SUBALIGN_LOG): New.
1943 (ASM_OUTPUT_MAX_SKIP_ALIGN):
1944 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
1945 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1946 max skip conditionally.
1947 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
1948 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
1949 * config/i386/x86-64.h (SUBALIGN_LOG): New.
1950 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1951 max skip conditionally.
1952 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1953 * config/ia64/ia64.c (ia64_option_override): Set default values
1954 for alignment options.
1955 * config/m68k/m68k.c: Handle new str_align_* options.
1956 * config/mips/mips.c (mips_set_compression_mode): Change
1958 (mips_option_override): Set default values for options.
1959 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
1961 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1963 * config/rx/rx.c (rx_option_override): Likewise.
1964 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
1965 (LABEL_ALIGN): Use align_labels_log.
1966 (LOOP_ALIGN): Use align_loops_align.
1967 * config/s390/s390.c (s390_asm_output_function_label): Use new
1969 * config/sh/sh.c (sh_override_options_after_change):
1970 Change type of constants.
1971 * config/spu/spu.c (spu_sched_init): Likewise.
1972 * config/sparc/sparc.c (sparc_option_override): Set default
1974 * config/visium/visium.c (visium_option_override): Likewise.
1975 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
1976 emit p2align format with last argument if it's not needed.
1977 * doc/invoke.texi: Document extended format of -falign-*.
1978 * final.c: Use align_labels alignment.
1979 * flags.h (struct target_flag_state): Change type to use
1981 (struct align_flags_tuple): New.
1982 (struct align_flags): Likewise.
1983 (align_loops_log): Redefine macro to use new types.
1984 (align_loops_max_skip): Redefine macro to use new types.
1985 (align_jumps_log): Redefine macro to use new types.
1986 (align_jumps_max_skip): Redefine macro to use new types.
1987 (align_labels_log): Redefine macro to use new types.
1988 (align_labels_max_skip): Redefine macro to use new types.
1989 (align_functions_log): Redefine macro to use new types.
1990 (align_loops): Redefine macro to use new types.
1991 (align_jumps): Redefine macro to use new types.
1992 (align_labels): Redefine macro to use new types.
1993 (align_functions): Redefine macro to use new types.
1994 (align_functions_max_skip): Redefine macro to use new types.
1995 (align_loops_value): New macro.
1996 (align_jumps_value): New macro.
1997 (align_labels_value): New macro.
1998 (align_functions_value): New macro.
1999 * function.c (invoke_set_current_function_hook): Propagate
2000 alignment values from flags to global variables default in
2002 * ipa-icf.c (sem_function::equals_wpa): Use
2003 cl_optimization_option_eq instead of memcmp.
2004 * lto-streamer.h (cl_optimization_stream_out): Support streaming
2006 (cl_optimization_stream_in): Likewise.
2007 * optc-save-gen.awk: Support strings in cl_optimization.
2008 * opth-gen.awk: Likewise.
2009 * opts.c (finish_options): Remove error checking of invalid
2011 (MAX_CODE_ALIGN): Remove.
2012 (MAX_CODE_ALIGN_VALUE): Likewise.
2013 (parse_and_check_align_values): New function.
2014 (check_alignment_argument): Likewise.
2015 (common_handle_option): Use check_alignment_argument.
2016 * opts.h (parse_and_check_align_values): Declare.
2017 * toplev.c (init_alignments): Remove.
2018 (read_log_maxskip): New.
2019 (parse_N_M): Likewise.
2020 (parse_alignment_opts): Likewise.
2021 (backend_init_target): Remove usage of init_alignments.
2022 * toplev.h (parse_alignment_opts): Declare.
2023 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
2025 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
2026 * tree.c (cl_option_hasher::equal): New.
2027 * varasm.c: Use new global macros.
2029 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
2031 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
2032 Use a simpler align directive also if MAXSKIP = ALIGN-1.
2033 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2034 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2035 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2036 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
2037 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
2038 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
2039 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
2040 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2041 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2042 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2043 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2045 2018-07-04 Martin Liska <mliska@suse.cz>
2046 Jonathan Wakely <jwakely@redhat.com>
2048 * coverage.c: Use correct type.
2049 * doc/invoke.texi: Language correction.
2051 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
2054 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
2055 ENDBRANCH for non-tail call which may return via indirect branch.
2056 * doc/extend.texi: Document indirect_return attribute.
2058 2018-07-03 Martin Sebor <msebor@redhat.com>
2060 PR tree-optimization/86274
2061 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
2063 (format_floating): Correct handling of infinities and NaNs.
2065 2018-07-03 Martin Sebor <msebor@redhat.com>
2067 * print-tree.c (print_real_cst): New function.
2068 (print_node_brief): Call it.
2069 (print_node): Ditto.
2071 2018-07-03 Jeff Law <law@redhat.com>
2073 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
2074 into a single pattern.
2076 * config/h8300/h8300.md (ors code_iterator): New.
2077 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
2078 a single pattern and single splitter.
2079 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
2080 (iorqi3_1, xorqi3_1): Likewise.
2081 (iorqi3, xorqi3 expanders): Similarly.
2083 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
2084 (movmd_internal) into a single pattern using the P mode iterator.
2085 (movmd splitters): Similarly.
2086 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
2087 (movsd splitters): Similarly.
2089 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
2090 ADDB, ADDW and ADDL into a single ADD attribute which selects the
2091 right table based on the size of the operand.
2092 * config/h8300/h8300.md (length_table): Corresponding changes. All
2093 references to "addb", "addw" and "addl" changed to "add".
2094 (btst patterns): Merge two variants into a single pattern.
2095 (tstqi, tsthi): Likewise.
2096 (addhi3_incdec, addsi3_incdec): Likewise.
2097 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
2098 (mulhi3, mulsi3): Likewise.
2099 (udivhi3, udivsi3): Likewise.
2100 (divhi3, divsi3): Likewise.
2101 (andorqi3, andorhi3, andorsi3): Likewise.
2103 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
2106 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
2107 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
2109 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2111 PR tree-optimization/85694
2112 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
2113 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
2115 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
2116 (<u>avg<mode>3_ceil): New patterns.
2118 2018-07-03 David Malcolm <dmalcolm@redhat.com>
2120 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
2121 scan-tree-dump directive.
2122 * gcc.dg/vect/slp-perm-2.c: Likewise.
2123 * gcc.dg/vect/slp-perm-3.c: Likewise.
2124 * gcc.dg/vect/slp-perm-5.c: Likewise.
2125 * gcc.dg/vect/slp-perm-6.c: Likewise.
2126 * gcc.dg/vect/slp-perm-7.c: Likewise.
2127 * gcc.dg/vect/slp-perm-8.c: Likewise.
2129 2018-07-03 Marek Polacek <polacek@redhat.com>
2132 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
2134 2018-07-03 Richard Biener <rguenther@suse.de>
2137 * tree-ssa-structalias.c (find_func_clobbers): Properly
2138 handle indirect calls.
2140 2018-07-03 Jeff Law <law@redhat.com>
2142 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
2143 (shifts): New code iterator.
2144 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
2145 expander. Fix HImode handling on H8/SX.
2146 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
2147 (subqi3, subhi3, subsi3 expanders): Likewise.
2148 (andqi3, andhi3, andsi3 expanders): Likewise.
2149 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
2150 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
2151 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
2152 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
2153 (zero_extendqihi2, zero_extendqisi2): Likewise.
2154 (extendqihi2, extendqisi2): Likewise.
2155 (rotlqi3, rotlhi3, rotlsi3): Likewise.
2156 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
2157 (rotlqi3_1, rotlhi3_1): Likewise.
2158 (logicalhi3_sn, logicalsi3_sn): Likewise.
2159 (logicalhi3, logicalsi3): Likewise.
2161 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2163 * tree-vect-patterns.c (vect_recog_rotate_pattern)
2164 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
2165 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
2166 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
2167 type to append_pattern_def_seq instead of creating a stmt_vec_info
2169 (build_mask_conversion): Likewise. Remove vinfo argument.
2170 (vect_add_conversion_to_patterm): Likewise, renaming to...
2171 (vect_add_conversion_to_pattern): ...this.
2172 (vect_recog_mask_conversion_pattern): Update call to
2173 build_mask_conversion. Pass the vector type to
2174 append_pattern_def_seq here too.
2175 (vect_recog_gather_scatter_pattern): Update call to
2176 vect_add_conversion_to_pattern.
2178 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2180 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
2181 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2182 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
2183 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
2184 STMT_VINFO_PATTERN_DEF_SEQ to null here.
2185 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
2186 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
2187 append_pattern_def_seq instead of new_pattern_def_seq.
2188 (vect_recog_divmod_pattern): Do both of the above.
2189 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
2192 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2194 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
2195 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
2196 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
2197 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
2198 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
2199 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
2200 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
2201 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2202 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2203 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
2204 parameter with a single stmt_vec_info.
2205 (vect_recog_func_ptr): Likewise.
2206 (vect_recog_gather_scatter_pattern): Likewise, folding in...
2207 (vect_try_gather_scatter_pattern): ...this.
2208 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
2209 the stmt_vec_info of the statement to be matched. Don't clear
2210 STMT_VINFO_RELATED_STMT.
2211 (vect_pattern_recog): Update call accordingly.
2213 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2215 PR tree-optimization/85694
2216 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
2217 (uavgM3_ceil): Document new optabs.
2218 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
2219 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
2221 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
2222 (savg_ceil_optab): New optabs.
2223 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
2224 (vect_vect_recog_func_ptrs): Add it.
2225 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
2226 constant directly from the associated lhs.
2228 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2230 * tree-vect-patterns.c (vect_split_statement): New function.
2231 (vect_convert_input): Use it to try to split an existing cast.
2233 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2235 * poly-int.h (print_hex): New function.
2236 * dumpfile.h (dump_dec, dump_hex): Declare.
2237 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
2238 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
2239 min_input_precision, operation_precision and operation_sign.
2240 * tree-vect-patterns.c (vect_get_range_info): New function.
2241 (vect_same_loop_or_bb_p, vect_single_imm_use)
2242 (vect_operation_fits_smaller_type): Delete.
2243 (vect_look_through_possible_promotion): Add an optional
2244 single_use_p parameter.
2245 (vect_recog_over_widening_pattern): Rewrite to use new
2246 stmt_vec_info infomration. Handle one operation at a time.
2247 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
2248 (vect_truncatable_operation_p, vect_set_operation_type)
2249 (vect_set_min_input_precision): New functions.
2250 (vect_determine_min_output_precision_1): Likewise.
2251 (vect_determine_min_output_precision): Likewise.
2252 (vect_determine_precisions_from_range): Likewise.
2253 (vect_determine_precisions_from_users): Likewise.
2254 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
2255 (vect_vect_recog_func_ptrs): Put over_widening first.
2257 (vect_pattern_recog): Call vect_determine_precisions.
2259 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2261 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
2262 statements that have been replaced by further pattern statements.
2263 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
2265 2018-07-03 Richard Biener <rguenther@suse.de>
2267 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
2268 always set *dt. Dump vectype in vectype overload.
2269 * dumpfile.h (dump_gimple_expr): New function.
2270 (dump_gimple_expr_loc): Likewise.
2271 * dumpfile.c (dump_gimple_expr): New function.
2272 (dump_gimple_expr_loc): Likewise.
2274 2018-07-02 Jeff Law <law@redhat.com>
2276 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
2277 the H8/300, H8/300H and H8/S variants into a single pattern.
2278 (movhi_h8300, movqi_h8300hs): Similarly.
2279 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
2280 (QHI mode iterator): New.
2282 * config/h8300/h8300.md: Remove trailing whitespace.
2284 2018-07-02 Jim Wilson <jimw@sifive.com>
2286 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
2287 instead of emit_insn for interrupt returns.
2288 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
2289 (riscv_sret, riscv_uret): Likewise.
2291 2018-07-02 David Malcolm <dmalcolm@redhat.com>
2293 * pretty-print.c (selftest::test_pp_format): Move save and restore
2294 of quotes to class auto_fix_quotes, and add an instance.
2295 * selftest.c: Include "intl.h".
2296 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
2297 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
2298 * selftest.h (selftest::auto_fix_quotes): New class.
2300 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
2302 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
2303 (aarch64_sve_prepare_conditional_op): Remove.
2304 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
2305 Allow aarch64_simd_reg_or_zero as select operand; remove
2306 the aarch64_sve_prepare_conditional_op call.
2307 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
2308 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
2309 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
2310 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
2311 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
2312 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
2313 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
2314 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
2315 and a splitters to match all of the *_any patterns.
2316 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
2318 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
2319 (SVE_COND_FP_BINARY_REV): Remove.
2320 (sve_int_op_rev, sve_fp_op_rev): New.
2321 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
2322 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
2323 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
2324 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
2325 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
2326 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
2327 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
2328 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
2329 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
2331 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
2332 Remove match_dup 1 from the inner unspec.
2333 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
2335 * config/aarch64/aarch64.md (movprfx): New attr.
2336 (length): Default movprfx to 8.
2337 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
2338 (*madd<SVE_I>, *msub<SVE_I): Likewise.
2339 (*<su>mul<SVE_I>3_highpart): Likewise.
2340 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
2341 (*v<ASHIFT><SVE_I>3): Likewise.
2342 (*<su><MAXMIN><SVE_I>3): Likewise.
2343 (*<su><MAXMIN><SVE_F>3): Likewise.
2344 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
2345 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
2346 (*div<SVE_F>4): Likewise.
2348 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
2350 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
2353 2018-07-02 Richard Biener <rguenther@suse.de>
2355 PR tree-optimization/86363
2356 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
2357 memset argument refers to a non-variable address.
2359 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
2361 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
2362 {PLUS,MINUS}_EXPR code to...
2363 (adjust_symbolic_bound): ...here,
2364 (combine_bound): ...here,
2365 (set_value_range_with_overflow): ...and here.
2367 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
2369 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
2371 (extract_range_from_abs_expr): ...here.
2373 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
2375 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
2376 -fno-omit-frame-pointer when not optimizing.
2378 2018-07-02 Martin Liska <mliska@suse.cz>
2381 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
2382 (propagate_nothrow): Likewise.
2384 2018-07-02 Martin Liska <mliska@suse.cz>
2387 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
2389 2018-07-02 David Malcolm <dmalcolm@redhat.com>
2391 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
2392 function in r262149, changing "loc" param from source_location to
2393 const dump_location_t &.
2394 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
2395 declaration, as above.
2397 2018-07-01 Paul Koning <ni1d@arrl.net>
2399 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
2400 -munit-asm, -mgnu-asm, -mdec-asm.
2401 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
2402 (pdp11_output_labelref): New.
2403 (pdp11_output_def): New.
2404 (pdp11_output_addr_vec_elt): New.
2405 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
2406 %# and %@ format codes.
2407 (pdp11_option_override): New.
2408 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
2409 (pdp11_output_ident): New.
2410 (pdp11_asm_named_section): New.
2411 (pdp11_asm_init_sections): New.
2412 (pdp11_file_start): New.
2413 (pdp11_file_end): New.
2414 (output_ascii): Use .ascii/.asciz for -mdec-asm.
2415 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
2416 %o, like %c but octal.
2417 (pdp11_option_override): New.
2418 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
2420 (DATA_SECTION_ASM_OP): Ditto.
2421 (READONLY_DATA_SECTION_ASM_OP): New.
2422 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
2423 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
2424 (ASM_OUTPUT_LABELREF): Ditto.
2425 (ASM_OUTPUT_DEF): Ditto.
2426 (ASM_OUTPUT_EXTERNAL): New.
2427 (ASM_OUTPUT_SOURCE_FILENAME): New.
2428 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
2429 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
2430 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
2431 %# and %@ format codes.
2432 * config/pdp11/pdp11.opt (mgnu-asm): New.
2433 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
2434 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
2435 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
2437 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
2439 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
2440 dereferencing path[] beyond its length.
2442 (debug_all_paths): New.
2443 (rewire_first_differing_edge): New.
2444 (adjust_paths_after_duplication): New.
2445 (duplicate_thread_path): Call adjust_paths_after_duplication.
2447 (thread_through_all_blocks): Add new argument to
2448 duplicate_thread_path.
2450 2018-06-30 Jim Wilson <jimw@sifive.com>
2452 * config/riscv/predicates.md (p2m1_shift_operand): New.
2453 (high_mask_shift_operand): New.
2454 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
2455 pattern using p2m1_shift_operand.
2456 (lshsi3_zero_extend_3+2): New combiner pattern using
2457 high_mask_shift_operand.
2459 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2461 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
2463 (vect_recog_rotate_pattern): ...here.
2464 (vect_convert_input): Try to insert casts of invariants in the
2466 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
2467 preheader to be empty.
2469 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2471 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
2472 vector type. If given, install it in the new statement's
2474 (vect_element_precision): New function.
2475 (vect_unpromoted_value): New struct.
2476 (vect_unpromoted_value::vect_unpromoted_value): New function.
2477 (vect_unpromoted_value::set_op): Likewise.
2478 (vect_look_through_possible_promotion): Likewise.
2479 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
2480 (vect_widened_op_tree, vect_convert_input): Likewise.
2481 (vect_convert_inputs, vect_convert_output): Likewise.
2482 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
2483 to handle the optional cast of the multiplication result and
2484 vect_widened_op_tree to detect the widened multiplication itself.
2485 Do not require the input and output of promotion casts to have
2486 the same sign, but base the signedness of the operation on the
2487 input rather than the result. If the pattern includes two
2488 promotions, check that those promotions have the same sign.
2489 Do not restrict the MULT_EXPR handling to a double-width result;
2490 handle quadruple-width results and wider. Use vect_convert_inputs
2491 to convert the inputs to the common type.
2492 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
2493 to handle the optional cast of the ABS result. Also allow a sign
2494 change or a sign extension between the ABS and MINUS.
2495 Use vect_widened_op_tree to detect the widened subtraction and use
2496 vect_convert_inputs to convert the inputs to the common type.
2497 (vect_handle_widen_op_by_const): Delete.
2498 (vect_recog_widen_op_pattern): New function.
2499 (vect_recog_widen_mult_pattern): Use it.
2500 (vect_recog_widen_shift_pattern): Likewise.
2501 (vect_recog_widen_sum_pattern): Use
2502 vect_look_through_possible_promotion to handle the promoted
2505 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2507 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
2508 the containing gimple_seq *.
2509 * gimple-iterator.h (gsi_for_stmt): Declare it.
2510 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
2511 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
2512 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
2513 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
2514 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
2516 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
2518 (vect_mark_pattern_stmts): ...here. Handle cases in which the
2519 statement being replaced is part of an existing pattern
2520 definition sequence, inserting the new pattern statements before
2522 (vect_pattern_recog_1): Don't return a bool. If the statement
2523 is already part of a pattern, instead apply pattern matching
2524 to the pattern definition statements. Don't clear the
2525 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
2526 (vect_pattern_recog): Don't break after the first match;
2527 continue processing the pattern definition statements instead.
2528 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
2530 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2532 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2533 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2534 (vect_recog_widen_sum_pattern): Use it.
2536 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2538 * tree-vect-loop.c (vectorizable_reduction): Assert that the
2539 phi is not a pattern statement and has not been replaced by
2540 a pattern statement.
2541 * tree-vect-patterns.c (type_conversion_p): Don't check
2542 STMT_VINFO_IN_PATTERN_P.
2543 (vect_recog_vector_vector_shift_pattern): Likewise.
2544 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
2545 the pattern statement rather than the original statement; check
2546 directly for a WIDEN_MULT_EXPR here.
2547 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
2548 vect_is_simple_use to return the pattern statement rather
2549 than the original statement; use is_pattern_stmt_p to check
2550 for such a pattern statement.
2551 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
2552 to return the pattern statement rather than the original statement;
2553 don't do the same transformation here.
2554 (vect_is_simple_use): If the defining statement has been replaced
2555 by a pattern statement, return the pattern statement instead.
2556 Remove the corresponding (local) transformation from the vectype
2559 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2561 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
2562 end and default to null.
2563 * tree-vect-loop.c (vect_create_epilog_for_reduction)
2564 (vectorizable_reduction): Update calls accordingly, dropping the
2565 gimple ** argument if the passed-back statement isn't needed.
2566 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
2567 (vect_recog_rotate_pattern): Likewise.
2568 (vect_recog_mask_conversion_pattern): Likewise.
2569 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2570 (vect_mask_constant_operand_p): Likewise.
2571 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
2572 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
2573 (get_group_load_store_type, get_load_store_type): Likewise.
2574 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
2575 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
2576 (vectorizable_conversion, vectorizable_assignment): Likewise.
2577 (vectorizable_shift, vectorizable_operation): Likewise.
2578 (vectorizable_store, vect_is_simple_cond): Likewise.
2579 (vectorizable_condition, vectorizable_comparison): Likewise.
2580 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
2581 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
2582 and move it to the end. Cope with null def_stmt_outs.
2584 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2586 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
2588 2018-06-29 Jeff Law <law@redhat.com>
2590 * config/v850/v850.c (v850_legitimate_address_p): Handle large
2591 displacements for TARGET_V850E2V3 and newer.
2592 (TARGET_LRA_P): Remove. Defaults to LRA now.
2593 * config/v850/v850.md (sign23byte_load): Remove.
2594 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
2595 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
2597 2018-06-29 Martin Liska <mliska@suse.cz>
2600 * coverage.c (coverage_init): Mangle full path name.
2601 * doc/invoke.texi: Document the change.
2602 * gcov-io.c (mangle_path): New.
2603 * gcov-io.h (mangle_path): Likewise.
2604 * gcov.c (mangle_name): Use mangle_path for path mangling.
2606 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2608 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
2609 if starting source register is not even.
2611 2018-06-29 Martin Liska <mliska@suse.cz>
2613 PR tree-optimization/86263
2614 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
2615 Make edge redirection.
2617 2018-06-29 David Malcolm <dmalcolm@redhat.com>
2619 * dumpfile.c (dump_loc): Add indentation based on scope depth.
2620 (dump_scope_depth): New variable.
2621 (get_dump_scope_depth): New function.
2622 (dump_begin_scope): New function.
2623 (dump_end_scope): New function.
2624 * dumpfile.h (get_dump_scope_depth): New declaration.
2625 (dump_begin_scope): New declaration.
2626 (dump_end_scope): New declaration.
2627 (class auto_dump_scope): New class.
2628 (AUTO_DUMP_SCOPE): New macro.
2629 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
2632 2018-06-29 Richard Biener <rguenther@suse.de>
2634 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
2635 compute_all_dependences succeeds.
2636 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
2637 exceed --param loop-max-datarefs-for-datadeps.
2639 2018-06-29 Jakub Jelinek <jakub@redhat.com>
2641 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
2643 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
2646 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
2647 alternative 0 in preferred_for_speed attribute.
2649 2018-06-28 Paul Koning <ni1d@arrl.net>
2651 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
2652 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
2653 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
2654 * config/pdp11/pdp11.md: Correct "length" attribute calculation
2655 for shift insn patterns.
2657 2018-06-28 David Malcolm <dmalcolm@redhat.com>
2659 * cgraph.c (cgraph_node::get_body): Replace assignments to
2660 "dump_file" with calls to set_dump_file.
2661 * dumpfile.c (alt_dump_file): Make static, and group with...
2662 (alt_flags): ...this definition.
2663 (dumps_are_enabled): New variable.
2664 (refresh_dumps_are_enabled): New function.
2665 (set_dump_file): New function.
2666 (set_alt_dump_file): New function.
2667 (gcc::dump_manager::dump_start): Replace assignments to
2668 "dump_file" and "alt_dump_file" with calls to set_dump_file and
2670 (gcc::dump_manager::dump_finish): Likewise.
2671 * dumpfile.h (alt_dump_file): Delete decl.
2672 (dumps_are_enabled): New variable decl.
2673 (set_dump_file): New function decl.
2674 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
2676 * tree-nested.c (lower_nested_functions): Replace assignments to
2677 "dump_file" with calls to set_dump_file.
2679 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
2681 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
2682 goto_locus of each outgoing edge of each basic block.
2684 2018-06-28 Richard Biener <rguenther@suse.de>
2686 * dwarf2out.c (decl_scope_table): Remove.
2687 (push_decl_scope): Likewise.
2688 (pop_decl_scope): Likewise.
2689 (gen_type_die_for_member): Do not call push/pop_decl_scope.
2690 (gen_struct_or_union_type_die): Likewise.
2691 (gen_tagged_type_die): Likewise.
2692 (dwarf2out_init): Do not initialize decl_scope_table.
2693 (dwarf2out_c_finalize): Do not free it.
2695 2018-06-28 Richard Biener <rguenther@suse.de>
2697 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
2698 deciding whether to not re-use a DIE.
2700 2018-06-28 Richard Biener <rguenther@suse.de>
2702 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
2703 DW_AT_abstract_origin attribute.
2705 2018-06-28 Martin Liska <mliska@suse.cz>
2707 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2708 Use newly introduced constants.
2709 * tree-switch-conversion.h (struct jump_table_cluster):
2710 Define max_ratio_for_size and max_ratio_for_speed.
2712 2018-06-28 Martin Liska <mliska@suse.cz>
2714 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2715 Add new checking assert to catch invalid state.
2716 (jump_table_cluster::can_be_handled): Handle single case
2718 (jump_table_cluster::is_beneficial): Bail out for such case.
2719 (bit_test_cluster::find_bit_tests):
2720 Add new checking assert to catch invalid state.
2721 (bit_test_cluster::can_be_handled): Handle single case
2723 (bit_test_cluster::is_beneficial): Bail out for such case.
2724 (switch_decision_tree::analyze_switch_statement):
2727 2018-06-28 Martin Liska <mliska@suse.cz>
2729 * common.opt: Introduce -completion option.
2730 * gcc.c (driver_handle_option): Handle it.
2731 (driver::main): Print completions if completion
2733 * opt-suggestions.c (option_proposer::get_completions):
2735 (option_proposer::suggest_completion): Likewise.
2736 (option_proposer::find_param_completions): Likewise.
2737 (verify_autocompletions): Likewise.
2738 (test_completion_valid_options): Likewise.
2739 (test_completion_valid_params): Likewise.
2740 (in_completion_p): Likewise.
2741 (empty_completion_p): Likewise.
2742 (test_completion_partial_match): Likewise.
2743 (test_completion_garbage): Likewise.
2744 (opt_proposer_c_tests): Likewise.
2745 * opt-suggestions.h: Declare new functions.
2746 * opts.c (common_handle_option): Handle OPT__completion_.
2747 * selftest-run-tests.c (selftest::run_tests): Add
2748 opt_proposer_c_tests.
2749 * selftest.c (assert_str_startswith): New.
2750 * selftest.h (assert_str_startswith): Likewise.
2751 (opt_proposer_c_tests): New.
2752 (ASSERT_STR_STARTSWITH): Likewise.
2754 2018-06-28 Martin Liska <mliska@suse.cz>
2756 * Makefile.in: Add opt-suggestions.o.
2757 * gcc-main.c: Include opt-suggestions.h.
2758 * gcc.c (driver::driver): Likewise.
2759 (driver::~driver): Remove m_option_suggestions.
2760 (driver::build_option_suggestions): Moved to option_proposer.
2761 (driver::suggest_option): Likewise.
2762 (driver::handle_unrecognized_options): Use option_proposer.
2763 * gcc.h (class driver): Add new memver m_option_proposer.
2764 * opt-suggestions.c: New file.
2765 * opt-suggestions.h: New file.
2767 2018-06-28 Martin Liska <mliska@suse.cz>
2769 * vec.h (class auto_string_vec): New (moved from auto_argvec).
2770 (auto_string_vec::~auto_string_vec): Likewise.
2772 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
2774 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
2775 prevent_decl_creation_for_types fields up and add reset_location field.
2776 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
2777 statement if id->reset_location is true.
2778 (copy_edges_for_bb): Do not set goto_locus on the new edges if
2779 id->reset_location is true.
2780 (copy_phis_for_bb): Force input_location on the arguments if
2781 id->reset_location is true.
2782 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
2783 is set on the function to be inlined.
2785 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
2787 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
2789 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
2791 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
2792 registers for Pmode.
2793 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
2794 hard registers for the clobbered pseudo.
2796 2018-06-27 Paul Koning <ni1d@arrl.net>
2798 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
2799 mutually exclusive options.
2800 * config/pdp11/constraints.md (h): New constraint.
2801 (O): Update definition to match shift code generation.
2802 (D): New constraint.
2803 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
2805 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
2807 (output_jump): Change arguments.
2808 (pdp11_fixed_cc_regs): New function.
2809 (pdp11_cc_mode): Ditto.
2810 (pdp11_expand_shift): Ditto.
2811 (pdp11_assemble_shift): Ditto.
2812 (pdp11_small_shift): Ditto.
2813 (pdp11_branch_cost): Remove.
2814 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
2816 (pdp11_register_move_cost): Update for CC registers.
2817 (pdp11_rtx_costs): Add case for LSHIFTRT.
2818 (pdp11_output_jump): Add CCNZ mode conditional branches.
2819 (notice_update_cc_on_set): Remove.
2820 (pdp11_cc_mode): New function.
2821 (simple_memory_operand): Correct pre/post decrement case.
2822 (no_side_effect_operand): New function.
2823 (pdp11_regno_reg_class): Add CC_REGS class.
2824 (pdp11_fixed_cc_regs): New function.
2825 (pdp11_small_shift): New function.
2826 (pdp11_expand_shift): New function to expand shift insns.
2827 (pdp11_assemble_shift): New function to output shifts.
2828 (pdp11_branch_cost): Remove.
2829 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
2830 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
2831 (WCHAR_TYPE): Ditto.
2832 (PTRDIFF_TYPE): Ditto.
2833 (ADJUST_INSN_LENGTH): New macro.
2834 (FIXED_REGISTERS): Add CC registers.
2835 (CALL_USED_REGISTERS): Ditto.
2837 (REG_CLASS_NAMES): Ditto.
2838 (REG_CLASS_CONTENTS): Ditto.
2839 (SELECT_CC_MODE): Use new function.
2840 (TARGET_FLAGS_REGNUM): New macro.
2841 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
2842 (cc0_reg_rtx): Remove.
2843 (CC_STATUS_MDEP): Remove.
2844 (CC_STATUS_MDEFP_INIT): Remove.
2845 (CC_IN_FPU): Remove.
2846 (NOTICE_UPDATE_CC): Remove.
2847 (REGISTER_NAMES): Add CC registers.
2848 (BRANCH_COST): Change to constant 1.
2849 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
2851 * config/pdp11/pdp11.opt (mbcopy): Remove.
2852 (mbcopy-builtin): Remove.
2853 (mbranch-cheap): Remove.
2854 (mbranch-expensive): Remove.
2855 * config/pdp11/predicates.md (expand_shift_operand): Update to
2856 match shift code generation.
2857 (ccnz_operator): New predicate.
2858 * doc/invoke.texi (PDP-11 Options): Remove deleted options
2859 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
2860 Remove non-existent option -mabshi, -mno-abshi. Document mutually
2862 * doc/md.texi (PDP-11): Document new D and h constraints. Update
2863 description of O constraint.
2865 2018-06-27 Jeff Law <law@redhat.com>
2866 Austin Law <austinklaw@gmail.com>
2868 * config/v850/v850.md (addsi3_set_flags): New pattern.
2869 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
2870 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
2871 (zero_extendhisi2_v850_set_flags): Likewise.
2872 (zero_extendqisi2_v850_set_flags): Likewise.
2873 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
2874 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
2875 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
2877 * config/v850/v850-protos.h (notice_update_cc): Remove.
2878 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
2879 (v850_print_operand): Handle 'D' and "d".
2880 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
2881 Add handling of arithmetic/logical operations compared against zero.
2882 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
2883 Do not look at v850_compare_op, instead get mode from last argument.
2884 (v850_gen_compare): Remove
2885 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
2886 after reload for prologue insns.
2887 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
2889 (construct_save_jarl): Likewise.
2890 (TARGET_FLAGS_REGNUM): Define.
2891 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
2892 (NOTICE_UPDATE_CC): Remove.
2893 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
2894 than cc0. Conditionalize on reload_completed.
2895 (cmpsi_insn, setfcc_insn): Likewise.
2896 (tst1 splitter): Turn into define_and_split which sets the flags
2898 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
2899 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
2900 (cstoresf4, cstoredf4): Clobber the flags.
2901 (cmpsi, cmpsf, cmpdf): Remove expanders.
2902 (setf_insn): Remove pattern.
2903 (addsi3): Turn into define_and_split which clobbers the flags after
2904 reload and a suitable pattern (addsi3_clobber_flags) for use after
2906 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
2907 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
2908 (ashrsi3, ashrsi3_v850e2): Likewise.
2909 (bins): Clobber the flags.
2910 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
2911 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
2912 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
2913 (call_value_internal_short, call_value_internal_long): Likewise.
2914 (callt_save_interrupt, callt_return_interrupt): Likewise.
2915 (save_interrupt, return_interrupt): Likewise.
2916 (callt_save_all_interrupt, save_all_interrupt): Likewise.
2917 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
2918 (restore_all_interrupt, _restore_all_interrupt): Likewise.
2919 (All FP comparisons): Only allow after reload has completed.
2921 (divh, divhu): Tweak output template.
2922 (branch_z_normal, branch_z_invert): Remove
2923 (branch_nz_normal, branch_nz_invert): Likewise.
2924 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
2926 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
2927 * config/v850/v850.c (notice_update_cc): Remove.
2928 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
2929 (CC_NO_CARRY): Likewise.
2930 (NOTICE_UPDATE_CC): Define to nothing.
2931 * config/v850/v850.md: Remove block comment on cc0 handling
2932 Remove "cc" attribute from all patterns. Remove cc_status handling
2933 from all patterns. Minor formatting fixes.
2935 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2937 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
2938 (cortex-a76.cortex-a55): Likewise.
2939 * config/aarch64/aarch64-tune.md: Regenerate.
2940 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
2941 cortex-a76.cortex-a55.
2943 2018-06-27 Jeff Law <law@redhat.com>
2945 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
2946 (MULTILIB_DIRNAMES): Similarly.
2948 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
2950 * gimple.h (gimple_return_retbnd): Delete.
2951 (gimple_return_set_retbnd): Likewise.
2952 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
2953 gimple_return_set_retbnd.
2954 * gimple-pretty-print.c (dump_gimple_return): Remove call to
2955 gimple_return_retbnd and adjust.
2956 * tree-inline.h (struct copy_body_data): Remove retbnd field.
2957 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
2958 Explicitly return NULL in a couple more cases. Move assertion
2959 on debug statements and remove unreachable code.
2960 (reset_debug_binding): Do not test id->retbnd.
2961 (expand_call_inline): Do not set it.
2963 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2965 * configure.ac: Add --disable-gcov option.
2966 * configure: Regenerate.
2967 * Makefile.in: Honour @enable_gcov@.
2968 * doc/install.texi: Document --disable-gcov.
2970 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2972 * config/arm/arm-cpus.in (cortex-a76): New entry.
2973 (cortex-a76.cortex-a55): Likewise.
2974 * config/arm/arm-tables.opt: Regenerate.
2975 * config/arm/arm-tune.md: Likewise.
2976 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
2977 * doc/invoke.texi (ARM Options): Document cortex-a76 and
2978 cortex-a76.cortex-a55.
2980 2018-06-27 Tamar Christina <tamar.christina@arm.com>
2983 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
2985 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
2987 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
2989 (EPILOGUE_USES): Likewise.
2991 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
2993 * tree-inline.c (remap_location): New function extracted from...
2994 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
2995 (copy_phis_for_bb): ...here. Call remap_location.
2996 (copy_cfg_body): Adjust call to copy_edges_for_bb.
2998 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
3000 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
3001 unaligned vsx for 16B memset.
3003 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
3006 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
3007 ieee128_float_type_node to long_double_type_node unless
3008 TARGET_LONG_DOUBLE_128 is set.
3010 2018-06-26 David Malcolm <dmalcolm@redhat.com>
3012 * cfgloop.c (get_loop_location): Convert return type from
3013 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
3014 by implicit construction from rtx_insn *, and using
3015 dump_user_location_t::from_function_decl for the fallback case.
3016 * cfgloop.h (get_loop_location): Convert return type from
3017 location_t to dump_user_location_t.
3018 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
3019 dump_printf_loc to pass in a dump_location_t rather than a
3020 location_t, via the gimple stmt.
3021 * coverage.c (get_coverage_counts): Update calls to
3022 dump_printf_loc to pass in dump_location_t rather than a
3024 * doc/optinfo.texi (Dump types): Convert example of
3025 dump_printf_loc from taking "locus" to taking "insn". Update
3026 description of the "_loc" calls to cover dump_location_t.
3027 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
3029 (dump_user_location_t::dump_user_location_t): New constructors,
3030 from gimple *stmt and rtx_insn *.
3031 (dump_user_location_t::from_function_decl): New function.
3032 (dump_loc): Make static.
3033 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
3034 const dump_location_t &.
3035 (dump_generic_expr_loc): Delete.
3036 (dump_printf_loc): Convert param "loc" from location_t to
3037 const dump_location_t &.
3038 (selftest::test_impl_location): New function.
3039 (selftest::dumpfile_c_tests): New function.
3040 * dumpfile.h: Include "profile-count.h".
3041 (class dump_user_location_t): New class.
3042 (struct dump_impl_location_t): New struct.
3043 (class dump_location_t): New class.
3044 (dump_printf_loc): Convert 2nd param from source_location to
3045 const dump_location_t &.
3046 (dump_generic_expr_loc): Delete.
3047 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
3048 const dump_location_t &.
3049 * gimple-fold.c (fold_gimple_assign): Update call to
3050 dump_printf_loc to pass in a dump_location_t rather than a
3051 location_t, via the gimple stmt.
3052 (gimple_fold_call): Likewise.
3053 * gimple-loop-interchange.cc
3054 (loop_cand::analyze_iloop_reduction_var): Update for change to
3055 check_reduction_path.
3056 (tree_loop_interchange::interchange): Update for change to
3058 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
3059 change in return-type of find_loop_location.
3060 (graphite_regenerate_ast_isl): Likewise.
3061 * graphite-optimize-isl.c (optimize_isl): Likewise.
3062 * graphite.c (graphite_transform_loops): Likewise.
3063 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
3064 pass in a dump_location_t rather than a location_t, via the
3066 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3067 * ipa.c (walk_polymorphic_call_targets): Likewise.
3068 * loop-unroll.c (report_unroll): Convert "locus" param from
3069 location_t to dump_location_t.
3070 (decide_unrolling): Update for change to get_loop_location's
3072 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
3073 location_t to dump_user_location_t.
3074 (grid_find_single_omp_among_assignments_1): Updates calls to
3075 dump_printf_loc to pass in a dump_location_t rather than a
3076 location_t, via the gimple stmt.
3077 (grid_parallel_clauses_gridifiable): Convert "tloc" from
3078 location_t to dump_location_t. Updates calls to dump_printf_loc
3079 to pass in a dump_location_t rather than a location_t, via the
3081 (grid_inner_loop_gridifiable_p): Likewise.
3082 (grid_dist_follows_simple_pattern): Likewise.
3083 (grid_gfor_follows_tiling_pattern): Likewise.
3084 (grid_target_follows_gridifiable_pattern): Likewise.
3085 (grid_attempt_target_gridification): Convert initialization
3086 of local "grid" from memset to zero-initialization; FIXME: does
3087 this require C++11? Update call to dump_printf_loc to pass in a
3088 optinfo_location rather than a location_t, via the gimple stmt.
3089 * profile.c (read_profile_edge_counts): Updates call to
3090 dump_printf_loc to pass in a dump_location_t rather than a
3092 (compute_branch_probabilities): Likewise.
3093 * selftest-run-tests.c (selftest::run_tests): Call
3095 * selftest.h (dumpfile_c_tests): New decl.
3096 * tree-loop-distribution.c (pass_loop_distribution::execute):
3097 Update for change in return type of find_loop_location.
3098 * tree-parloops.c (parallelize_loops): Likewise.
3099 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
3100 "locus" from location_t to dump_user_location_t.
3101 (canonicalize_loop_induction_variables): Likewise.
3102 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
3103 for change in return type of find_loop_location.
3104 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
3105 to dump_printf_loc to pass in a dump_location_t rather than a
3106 location_t, via the stmt.
3107 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
3109 * tree-vect-loop-manip.c (find_loop_location): Convert return
3110 type from source_location to dump_user_location_t.
3111 (vect_do_peeling): Update for above change.
3112 (vect_loop_versioning): Update for change in type of
3114 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
3115 from location_t to dump_user_location_t.
3116 (vect_estimate_min_profitable_iters): Update for change in type
3118 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
3119 location_t to dump_location_t.
3120 (vect_slp_bb): Update for change in type of vect_location.
3121 * tree-vectorizer.c (vect_location): Convert from source_location
3122 to dump_user_location_t.
3123 (try_vectorize_loop_1): Update for change in vect_location's type.
3124 (vectorize_loops): Likewise.
3125 (increase_alignment): Likewise.
3126 * tree-vectorizer.h (vect_location): Convert from source_location
3127 to dump_user_location_t.
3128 (find_loop_location): Convert return type from source_location to
3129 dump_user_location_t.
3130 (check_reduction_path): Convert 1st param from location_t to
3131 dump_user_location_t.
3132 * value-prof.c (check_counter): Update call to dump_printf_loc to
3133 pass in a dump_user_location_t rather than a location_t; update
3134 call to error_at for change in type of "locus".
3135 (check_ic_target): Update call to dump_printf_loc to
3136 pass in a dump_user_location_t rather than a location_t, via the
3139 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
3141 * config/s390/s390.h (enum processor_flags): Do not use
3142 default tune parameter when -march was specified.
3144 2018-06-26 Jakub Jelinek <jakub@redhat.com>
3147 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
3148 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
3151 2018-06-26 Richard Biener <rguenther@suse.de>
3153 PR tree-optimization/86287
3155 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
3156 (vect_analyze_loop): Initialize n_stmts.
3158 2018-06-26 Richard Biener <rguenther@suse.de>
3161 * fold-const.c (fold_convertible_p): Pointer extension
3164 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
3167 * dwarf2out.c (loc_list_has_views): Adjust comments.
3168 (dw_loc_list): Split single cross-partition range with
3171 2018-06-25 Jeff Law <law@redhat.com>
3173 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
3174 on -mbig-switch by default.
3176 * config/v850/predicates.md (const_float_1_operand): Fix match_code
3178 (const_float_0_operand): Remove unused predicate.
3179 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
3180 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
3181 (recipsf2): New expander. Original pattern now called
3183 (recipdf2, recipdf2_insn): Similarly.
3184 (rsqrtsf2, rsqrtsf2_insn): Similarly
3185 (rsqrtdf2, rsqrtdf2_insn): Similarly
3187 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
3189 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
3190 Simplify logic for FreeBSD (twice).
3192 2018-06-25 Martin Sebor <msebor@redhat.com>
3194 PR tree-optimization/86204
3195 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
3196 a strnlen result if it's less than the length of the string.
3198 2018-06-25 Martin Sebor <msebor@redhat.com>
3200 PR tree-optimization/85700
3201 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
3202 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
3203 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
3205 2018-06-25 Martin Sebor <msebor@redhat.com>
3207 * doc/extend.texi (Zero-length arrays): Update and clarify.
3209 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
3211 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
3212 added IEEE/IBM long double multilib support on PowerPC little
3213 endian Linux systems.
3214 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
3215 (MULTILIB_DEFAULTS): Likewise.
3216 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3218 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
3219 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
3220 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
3222 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
3225 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
3226 (REORDER_45): Likewise.
3228 2018-06-25 Jeff Law <law@redhat.com>
3230 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
3231 dividend to 32 bits. Adjust length.
3232 (udivmodhi4): Cleanup output template. Fix length.
3234 2018-06-25 Carl Love <cel@us.ibm.com>
3236 * config/rs6000/vsx.md: Change word selector to prefered location.
3238 2018-06-25 Richard Biener <rguenther@suse.de>
3240 PR tree-optimization/86304
3241 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
3242 epilogue-if-converted loops as well.
3244 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
3246 * lto-section-out.c (lto_begin_section): Do not print section
3247 name for noaddr and unnumbered dumps.
3249 2018-06-25 Richard Biener <rguenther@suse.de>
3251 * tree-vectorizer.h (struct vec_info_shared): New structure
3252 with parts split out from struct vec_info and loop_nest from
3253 struct _loop_vec_info.
3254 (struct vec_info): Adjust accordingly.
3255 (struct _loop_vec_info): Likewise.
3256 (LOOP_VINFO_LOOP_NEST): Adjust.
3257 (LOOP_VINFO_DATAREFS): Likewise.
3258 (LOOP_VINFO_DDRS): Likewise.
3259 (struct _bb_vec_info): Likewise.
3260 (BB_VINFO_DATAREFS): Likewise.
3261 (BB_VINFO_DDRS): Likewise.
3262 (struct _stmt_vec_info): Add dr_aux member.
3263 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
3264 (DR_MISALIGNMENT_UNINITIALIZED): New.
3265 (set_dr_misalignment): Adjust.
3266 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
3267 (vect_analyze_loop): Adjust prototype.
3268 (vect_analyze_loop_form): Likewise.
3269 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
3270 Compute dependences lazily.
3271 (vect_record_base_alignments): Use shared datarefs/ddrs.
3272 (vect_verify_datarefs_alignment): Likewise.
3273 (vect_analyze_data_refs_alignment): Likewise.
3274 (vect_analyze_data_ref_accesses): Likewise.
3275 (vect_analyze_data_refs): Likewise.
3276 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
3277 constructor parameter for shared part.
3278 (vect_analyze_loop_form): Pass in shared part and adjust.
3279 (vect_analyze_loop_2): Pass in storage for the number of
3280 stmts. Move loop nest finding to the caller. Compute
3282 (vect_analyze_loop): Pass in shared part.
3283 (vect_transform_loop): Verify shared datarefs are unchanged.
3284 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
3285 constructor parameter for shared part.
3286 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
3287 (vect_slp_bb): Verify shared datarefs are unchanged before
3289 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
3291 (new_stmt_vec_info): Initialize DR_AUX misalignment to
3292 DR_MISALIGNMENT_UNINITIALIZED.
3293 * tree-vectorizer.c (vec_info::vec_info): Add constructor
3294 parameter for shared part.
3295 (vec_info::~vec_info): Adjust.
3296 (vec_info_shared::vec_info_shared): New.
3297 (vec_info_shared::~vec_info_shared): Likewise.
3298 (vec_info_shared::save_datarefs): Likewise.
3299 (vec_info_shared::check_datarefs): Likewise.
3300 (try_vectorize_loop_1): Construct shared part live for analyses
3301 of a single loop for multiple vector sizes.
3302 * tree-parloops.c (gather_scalar_reductions): Adjust.
3304 2018-06-25 Richard Biener <rguenther@suse.de>
3306 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
3307 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
3308 (vect_analyze_data_refs): Remove similar code from here and
3309 simplify accordingly.
3311 2018-06-25 Richard Biener <rguenther@suse.de>
3313 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
3314 for reverse storage order accesses rather than asserting
3315 they cannot happen here.
3317 2018-06-25 Tom de Vries <tdevries@suse.de>
3320 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
3321 Use data16 instead of .byte for insn prefix.
3323 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
3326 * parser.c (make_char_string_pack): Pass this literal chars
3327 through cpp_interpret_string.
3328 (cp_parser_userdef_numeric_literal): Check the result of
3329 make_char_string_pack.
3331 2018-06-24 Maya Rashish <coypu@sdf.org>
3333 * ginclude/stddef.h: Simplify conditions around avoiding
3334 re-definition of __size_t.
3336 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3338 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
3339 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
3341 2018-06-22 Maya Rashish <coypu@sdf.org>
3343 * doc/invoke.texi (mno-fancy-math-387): Update for changes
3344 made to OpenBSD and NetBSD through the years.
3346 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3348 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3349 behavior of vec_pack (vector double, vector double) to match
3350 behavior of vec_float2 (vector double, vector double).
3352 2018-06-22 Olivier Hainque <hainque@adacore.com>
3354 * gimplify.c (gimplify_function_tree): Prevent creation
3355 of a trampoline for the address of the current function
3356 passed to entry/exit instrumentation hooks.
3358 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
3361 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
3364 2018-06-22 Martin Liska <mliska@suse.cz>
3366 PR tree-optimization/86263
3367 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3368 Bail out if is_enabled is false.
3369 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
3371 (jump_table_cluster::is_enabled): New function.
3373 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3375 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
3376 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
3377 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
3378 (lto_input_ts_binfo_tree_pointers): Likewise.
3379 * tree-streamer-out.c (streamer_write_tree_bitfields,
3380 write_ts_binfo_tree_pointers): Likewise.
3381 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
3383 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3385 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
3387 2018-06-22 Martin Liska <mliska@suse.cz>
3389 * symbol-summary.h (get): Make it pure and inline move
3390 functionality from ::get function.
3391 (get): Remove and inline into ::get and ::get_create.
3392 (get_create): Move code from ::get function.
3394 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3397 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
3398 -x assembler-with-cpp.
3400 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3402 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
3403 _FILE_OFFSET_BITS=64 for C++.
3405 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
3407 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
3408 conversion insn that shows up when pr85657-3.c is compiled using
3409 IEEE 128-bit long double.
3410 (neg<mode>2_internal): Use the correct mode to check whether the
3411 mode is IBM extended.
3412 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
3413 multiply and divide external functions from being created more
3416 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3418 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
3420 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
3421 the edge can be forwarded.
3422 (cfg_layout_merge_blocks): Likewise.
3424 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3426 * except.c (finish_eh_generation): Commit edge insertions only after
3427 the EH edges have been redirected from post-landing to landing pads.
3429 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3431 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
3432 create_tmp_var_for to create the FRAME decl.
3433 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
3435 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3437 * tree-inline.c (copy_edges_for_bb): Minor tweak.
3438 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
3439 debug statement when resetting its value.
3440 (expand_call_inline): Copy the locus of the call onto the assignment
3441 of the return value, if any. Use local variable in more cases.
3443 2018-06-21 Martin Liska <mliska@suse.cz>
3445 * ipa-pure-const.c (propagate_nothrow): Use
3446 funct_state_summaries->get.
3447 (dump_malloc_lattice): Likewise.
3448 (propagate_malloc): Likewise.
3450 2018-06-21 Richard Biener <rguenther@suse.de>
3452 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
3453 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
3454 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
3456 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
3457 BLOCK_ABSTRACT_ORIGIN unconditionally.
3459 2018-06-21 David Malcolm <dmalcolm@redhat.com>
3461 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
3463 * ipa-reference.c (ipa_reference_c_finalize): Delete
3464 ipa_ref_opt_sum_summaries and set it to NULL.
3466 2018-06-21 Tom de Vries <tdevries@suse.de>
3468 PR tree-optimization/85859
3469 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
3470 test with comment from bb_no_side_effects_p.
3472 2018-06-21 Richard Biener <rguenther@suse.de>
3474 PR tree-optimization/86232
3475 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
3476 max for constant niter.
3478 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
3480 * config/aarch64/aarch64-simd.md
3481 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
3483 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
3485 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
3486 Make opernads of the unspec commutative.
3488 2018-06-21 Richard Biener <rguenther@suse.de>
3490 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
3491 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
3492 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
3493 (vect_analyze_data_ref_dependence): Re-order checks to deal with
3495 (vect_record_base_alignments): Do not record base alignment
3496 for gathers or scatters.
3497 (vect_compute_data_ref_alignment): Drop return value that is always
3498 true. Bail out early for gathers or scatters.
3499 (vect_enhance_data_refs_alignment): Bail out early for gathers
3501 (vect_find_same_alignment_drs): Likewise.
3502 (vect_analyze_data_refs_alignment): Remove dead code.
3503 (vect_slp_analyze_and_verify_node_alignment): Likewise.
3504 (vect_analyze_data_refs): For possible gathers or scatters do
3505 not create an alternate DR, just check their possible validity
3506 and mark them. Adjust DECL_NONALIASED handling to not rely
3508 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
3509 update inits of gathers or scatters.
3510 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
3511 Also copy gather/scatter flag to pattern vinfo.
3513 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
3515 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3516 behavior of vec_packsu (vector unsigned long long, vector unsigned
3517 long long) to match behavior of vec_packs with same signature.
3519 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
3520 Thomas Schwinge <thomas@codesourcery.com>
3521 Cesar Philippidis <cesar@codesourcery.com>
3523 * gimplify.c (gimplify_scan_omp_clauses): Add support for
3524 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
3525 (gimplify_adjust_omp_clauses): Likewise.
3526 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
3527 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
3528 (gimplify_omp_target_update): Update handling of acc update and
3530 * omp-low.c (install_var_field): Remove unused parameter
3531 base_pointers_restrict.
3532 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
3533 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
3535 (omp_target_base_pointers_restrict_p): Delete.
3536 (scan_omp_target): Update call to scan_sharing_clauses.
3537 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
3539 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
3540 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
3541 (convert_local_omp_clauses): Likewise.
3542 * tree-pretty-print.c (dump_omp_clause): Likewise.
3543 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
3545 (omp_clause_code_name): Likewise.
3547 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3550 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
3553 PR tree-optimization/86231
3554 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
3555 anti-range don't overwrite *vr0min before using it to compute *vr0max.
3557 2018-06-20 Tom de Vries <tdevries@suse.de>
3559 PR tree-optimization/86097
3560 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
3561 iv type if signedness of iv type is not the same as that of *nit.
3563 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3565 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
3566 EDGE_EH edges, verify they are all EDGE_EH.
3568 2018-06-20 Maya Rashish <coypu@sdf.org>
3570 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
3572 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3574 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
3575 * config/aarch64/aarch64.c (xgene1_tunings): Add
3576 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
3577 (aarch64_mode_valid_for_sched_fusion_p):
3578 Allow 16-byte modes.
3579 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
3580 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
3582 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
3584 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
3585 * config/aarch64/iterators.md (VQ2): New mode iterator.
3587 2018-06-20 Martin Liska <mliska@suse.cz>
3589 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
3590 Change default ratio from 10 to 8.
3592 2018-06-20 Martin Liska <mliska@suse.cz>
3594 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3596 (bit_test_cluster::find_bit_tests): Likewise.
3597 (switch_decision_tree::analyze_switch_statement): Find clusters.
3598 * tree-switch-conversion.h (struct jump_table_cluster): Document
3601 2018-06-20 Martin Liska <mliska@suse.cz>
3603 * tree-switch-conversion.c (switch_conversion::collect):
3604 Record m_uniq property.
3605 (switch_conversion::expand): Bail out for special conditions.
3606 (group_cluster::~group_cluster): New.
3607 (group_cluster::group_cluster): Likewise.
3608 (group_cluster::dump): Likewise.
3609 (jump_table_cluster::emit): New.
3610 (switch_decision_tree::fix_phi_operands_for_edges): New.
3611 (struct case_node): Remove struct.
3612 (jump_table_cluster::can_be_handled): New.
3613 (case_values_threshold): Moved to header.
3614 (reset_out_edges_aux): Likewise.
3615 (jump_table_cluster::is_beneficial): New.
3616 (bit_test_cluster::can_be_handled): Likewise.
3617 (add_case_node): Remove.
3618 (bit_test_cluster::is_beneficial): New.
3619 (case_bit_test::cmp): New.
3620 (bit_test_cluster::emit): New.
3621 (expand_switch_as_decision_tree_p): Remove.
3622 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
3623 (fix_phi_operands_for_edge): Likewise.
3624 (switch_decision_tree::analyze_switch_statement): New.
3625 (compute_cases_per_edge): Move ...
3626 (switch_decision_tree::compute_cases_per_edge): ... here.
3627 (try_switch_expansion): Likewise.
3628 (switch_decision_tree::try_switch_expansion): Likewise.
3629 (record_phi_operand_mapping): Likewise.
3630 (switch_decision_tree::record_phi_operand_mapping): Likewise.
3631 (emit_case_decision_tree): Likewise.
3632 (switch_decision_tree::emit): Likewise.
3633 (balance_case_nodes): Likewise.
3634 (switch_decision_tree::balance_case_nodes): Likewise.
3635 (dump_case_nodes): Likewise.
3636 (switch_decision_tree::dump_case_nodes): Likewise.
3637 (emit_jump): Likewise.
3638 (switch_decision_tree::emit_jump): Likewise.
3639 (emit_cmp_and_jump_insns): Likewise.
3640 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
3641 (emit_case_nodes): Likewise.
3642 (switch_decision_tree::emit_case_nodes): Likewise.
3643 (conditional_probability): Remove.
3644 * tree-switch-conversion.h (enum cluster_type): New.
3646 (struct cluster): Likewise.
3647 (cluster::cluster): Likewise.
3648 (struct simple_cluster): Likewise.
3649 (simple_cluster::simple_cluster): Likewise.
3650 (struct group_cluster): Likewise.
3651 (struct jump_table_cluster): Likewise.
3652 (struct bit_test_cluster): Likewise.
3653 (struct min_cluster_item): Likewise.
3654 (struct case_tree_node): Likewise.
3655 (case_tree_node::case_tree_node): Likewise.
3656 (jump_table_cluster::case_values_threshold): Likewise.
3657 (struct case_bit_test): Likewise.
3658 (struct switch_decision_tree): Likewise.
3659 (struct switch_conversion): Likewise.
3660 (switch_decision_tree::reset_out_edges_aux): Likewise.
3662 2018-06-20 Martin Liska <mliska@suse.cz>
3664 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
3665 (hoist_edge_and_branch_if_true): Likewise.
3666 (expand_switch_using_bit_tests_p): Likewise.
3667 (struct case_bit_test): Likewise.
3668 (case_bit_test_cmp): Likewise.
3669 (emit_case_bit_tests): Likewise.
3670 (switch_conversion::switch_conversion): New class.
3671 (struct switch_conv_info): Remove old struct.
3672 (collect_switch_conv_info): More to ...
3673 (switch_conversion::collect): ... this.
3674 (check_range): Likewise.
3675 (switch_conversion::check_range): Likewise.
3676 (check_all_empty_except_final): Likewise.
3677 (switch_conversion::check_all_empty_except_final): Likewise.
3678 (check_final_bb): Likewise.
3679 (switch_conversion::check_final_bb): Likewise.
3680 (create_temp_arrays): Likewise.
3681 (switch_conversion::create_temp_arrays): Likewise.
3682 (free_temp_arrays): Likewise.
3683 (gather_default_values): Likewise.
3684 (switch_conversion::gather_default_values): Likewise.
3685 (build_constructors): Likewise.
3686 (switch_conversion::build_constructors): Likewise.
3687 (constructor_contains_same_values_p): Likewise.
3688 (switch_conversion::contains_same_values_p): Likewise.
3689 (array_value_type): Likewise.
3690 (switch_conversion::array_value_type): Likewise.
3691 (build_one_array): Likewise.
3692 (switch_conversion::build_one_array): Likewise.
3693 (build_arrays): Likewise.
3694 (switch_conversion::build_arrays): Likewise.
3695 (gen_def_assigns): Likewise.
3696 (switch_conversion::gen_def_assigns): Likewise.
3697 (prune_bbs): Likewise.
3698 (switch_conversion::prune_bbs): Likewise.
3699 (fix_phi_nodes): Likewise.
3700 (switch_conversion::fix_phi_nodes): Likewise.
3701 (gen_inbound_check): Likewise.
3702 (switch_conversion::gen_inbound_check): Likewise.
3703 (process_switch): Use the newly created class.
3704 (switch_conversion::expand): New.
3705 (switch_conversion::~switch_conversion): New.
3706 * tree-switch-conversion.h: New file.
3708 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3710 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
3711 tree-vect-patterns.c.
3712 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
3713 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
3714 (vect_recog_sad_pattern): Likewise.
3715 (vect_recog_widen_sum_pattern): Likewise.
3716 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
3717 (vect_recog_widen_shift_pattern): Remove the type_in argument.
3718 (vect_recog_rotate_pattern): Likewise.
3719 (vect_recog_mult_pattern): Likewise.
3720 (vect_recog_vector_vector_shift_pattern): Likewise.
3721 (vect_recog_divmod_pattern): Likewise.
3722 (vect_recog_mixed_size_cond_pattern): Likewise.
3723 (vect_recog_bool_pattern): Likewise.
3724 (vect_recog_mask_conversion_pattern): Likewise.
3725 (vect_try_gather_scatter_pattern): Likewise.
3726 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
3727 (vect_recog_over_widening_pattern): Likewise.
3728 (vect_recog_gather_scatter_pattern): Likewise.
3729 (vect_recog_func_ptr): Move from tree-vectorizer.h
3730 (vect_vect_recog_func_ptrs): Move further down the file.
3731 (vect_recog_func): Likewise. Remove the third argument.
3732 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
3733 (vect_pattern_recog_1): Expect the pattern function to do any
3734 necessary target tests. Also expect it to provide a vector type.
3735 Remove the type_in handling.
3737 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3739 * tree-vect-patterns.c (vect_pattern_detected): New function.
3740 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
3741 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
3742 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
3743 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
3744 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
3745 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
3746 (vect_recog_mask_conversion_pattern)
3747 (vect_try_gather_scatter_pattern): Likewise.
3749 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3751 * tree-vect-patterns.c (vect_get_internal_def): New function.
3752 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
3753 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
3754 (search_type_for_mask_1): Use it.
3756 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3758 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
3759 redundant WIDEN_SUM_EXPR handling.
3760 (vect_recog_sad_pattern): Likewise.
3762 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3764 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
3765 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
3766 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
3767 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
3768 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
3770 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3772 * tree-vect-stmts.c (vectorizable_call): Make sure that we
3773 use the stmt_vec_info of the original bb statement for the
3774 new zero assignment, even if the call is part of a pattern.
3776 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3778 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
3779 that the sequence is attached to the original statement rather
3780 than the pattern statement.
3781 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
3782 PATTERN_DEF_SEQ from the original statement rather than
3783 the main pattern statement.
3784 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
3785 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
3786 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
3788 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3790 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
3791 definition statements before the early exit for statements that aren't
3793 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
3795 (vect_transform_loop): ...here. Process pattern definition
3796 statements without first checking whether the main pattern
3797 statement is live or relevant.
3799 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
3801 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
3802 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
3804 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
3806 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
3807 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
3808 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
3809 (expand_block_compare): Change select_block_compare_mode call.
3810 (expand_strncmp_align_check): Use new functions, fix comment.
3811 (emit_final_str_compare_gpr): New function.
3812 (expand_strn_compare): Refactor and clean up code.
3813 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
3815 2018-06-19 Tony Reix <tony.reix@atos.com>
3816 Damien Bergamini <damien.bergamini@atos.com>
3817 David Edelsohn <dje.gcc@gmail.com>
3819 * collect2.c (static_obj): New variable.
3820 (static_libs): New variable.
3821 (is_in_list): Uncomment declaration.
3822 (main): Track AIX libraries linked statically.
3823 (is_in_list): Uncomment definition.
3824 (scan_prog_file): Don't add AIX shared libraries initializer
3825 to constructor list if linking statically.
3827 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
3829 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
3831 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
3833 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
3835 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
3838 2018-06-19 Martin Liska <mliska@suse.cz>
3840 * config/i386/i386.c (ix86_can_inline_p): Do not use
3841 ipa_fn_summaries::get_create.
3842 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
3844 (devirtualization_time_bonus): Likewise.
3845 (ipcp_propagate_stage): Likewise.
3846 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3847 (edge_set_predicate): Likewise.
3848 (evaluate_conditions_for_known_args): Likewise.
3849 (evaluate_properties_for_edge): Likewise.
3850 (ipa_call_summary::reset): Tranform to ...
3851 (ipa_call_summary::~ipa_call_summary): ... this.
3852 (ipa_fn_summary::reset): Transform to ...
3853 (ipa_fn_summary::~ipa_fn_summary): ... this.
3854 (ipa_fn_summary_t::remove): Rename to ...
3855 (ipa_fn_summary_t::remove_callees): ... this.
3856 (ipa_fn_summary_t::duplicate): Use placement new
3857 instead of memory copy.
3858 (ipa_call_summary_t::duplicate): Likewise.
3859 (ipa_call_summary_t::remove): Remove.
3860 (dump_ipa_call_summary): Change get_create to get.
3861 (ipa_dump_fn_summary): Dump only when summary exists.
3862 (analyze_function_body): Use symbol_summary::get instead
3864 (compute_fn_summary): Likewise.
3865 (estimate_edge_devirt_benefit): Likewise.
3866 (estimate_edge_size_and_time): Likewise.
3867 (inline_update_callee_summaries): Likewise.
3868 (remap_edge_change_prob): Likewise.
3869 (remap_edge_summaries): Likewise.
3870 (ipa_merge_fn_summary_after_inlining): Likewise.
3871 (write_ipa_call_summary): Likewise.
3872 (ipa_fn_summary_write): Likewise.
3873 (ipa_free_fn_summary): Likewise.
3874 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
3875 (struct ipa_call_summary): Likewise.
3876 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
3878 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
3879 (estimate_size_after_inlining): Likewise.
3880 (estimate_growth): Likewise.
3881 (growth_likely_positive): Likewise.
3882 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3883 (inline_call): Likewise.
3884 * ipa-inline.c (caller_growth_limits): Likewise.
3885 (can_inline_edge_p): Likewise.
3886 (can_inline_edge_by_limits_p): Likewise.
3887 (compute_uninlined_call_time): Likewise.
3888 (compute_inlined_call_time): Likewise.
3889 (want_inline_small_function_p): Likewise.
3890 (edge_badness): Likewise.
3891 (update_caller_keys): Likewise.
3892 (update_callee_keys): Likewise.
3893 (inline_small_functions): Likewise.
3894 (inline_to_all_callers_1): Likewise.
3895 (dump_overall_stats): Likewise.
3896 (early_inline_small_functions): Likewise.
3897 (early_inliner): Likewise.
3898 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3899 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3900 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3901 * ipa-split.c (execute_split_functions): Likewise.
3902 * symbol-summary.h: Likewise.
3903 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3905 2018-06-19 Richard Biener <rguenther@suse.de>
3907 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
3908 (vectorize_loops): ... here. Fix dbgcnt handling.
3909 (try_vectorize_loop): Wrap try_vectorize_loop_1.
3911 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
3914 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
3915 ieee128 argument takes up only one (vector) register, not two (floating
3918 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
3920 * gimplify.c (gimplify_init_constructor): Really never clear for an
3921 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
3923 2018-06-19 Richard Biener <rguenther@suse.de>
3925 PR tree-optimization/86179
3926 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
3927 after failed recognition.
3929 2018-06-18 Martin Sebor <msebor@redhat.com>
3932 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
3933 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
3934 Handle integer subtraction.
3935 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
3936 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
3938 2018-06-18 David Malcolm <dmalcolm@redhat.com>
3940 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
3941 param from rtx to rtx_insn *.
3942 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
3944 (frv_ifcvt_modify_insn): Likwise.
3945 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
3946 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
3947 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
3948 as_a <rtx_insn *> cast to local "unprotected_region" once
3949 it's been established that it's not NULL or pc_rtx.
3950 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
3951 param "sethi" from rtx to rtx_insn *.
3952 (nds32_group_float_insns): Likewise for param "insn".
3953 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
3955 (vax_output_int_subtract): Likewise.
3956 * config/vax/vax.c (vax_output_int_add): Likewise for param
3958 (vax_output_int_subtract): Likewise.
3959 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
3960 (emit_pattern_after): Likewise for param "after".
3961 (emit_insn_after): Likewise.
3962 (emit_jump_insn_after): Likewise.
3963 (emit_call_insn_after): Likewise.
3964 (emit_debug_insn_after): Likewise.
3965 (emit_pattern_before): Likewise for param "before".
3966 (emit_insn_before): Likewise.
3967 (emit_jump_insn_before): Likewise.
3968 * final.c (get_insn_template): Likewise for param "insn", removing
3970 * output.h (get_insn_template): Likewise for 2nd param.
3971 * rtl.h (emit_insn_before): Likewise.
3972 (emit_jump_insn_before): Likewise.
3973 (emit_debug_insn_before_noloc): Likewise.
3974 (emit_insn_after): Likewise.
3975 (emit_jump_insn_after): Likewise.
3976 (emit_call_insn_after): Likewise.
3977 (emit_debug_insn_after): Likewise.
3978 (set_insn_deleted): Likewise for param.
3980 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
3983 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
3984 floating point modes, so that IFmode is numerically greater than
3985 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
3986 to declare the ordering. This prevents IFmode from being
3987 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
3988 machine. Include rs6000-modes.h to share the fractional values
3989 between genmodes* and the rest of the compiler.
3993 * config/rs6000/rs6000-modes.h: New file.
3994 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
3995 meaning of rs6000_long_double_size so that 126..128 selects an
3996 appropriate 128-bit floating point type.
3997 (rs6000_option_override_internal): Likewise.
3998 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
3999 (TARGET_LONG_DOUBLE_128): Change the meaning of
4000 rs6000_long_double_size so that 126..128 selects an appropriate
4001 128-bit floating point type.
4002 (LONG_DOUBLE_TYPE_SIZE): Update comment.
4003 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
4004 source and destination to match the standard usage.
4005 (truncifkf2): Likewise.
4006 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
4007 ISA 2.07 to use an explicit clobber, instead of passing in a
4009 (copysign<mode>3_soft): Likewise.
4011 2018-06-18 David Malcolm <dmalcolm@redhat.com>
4013 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
4014 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
4015 (vect_slp_analyze_instance_dependence): Likewise.
4016 (vect_enhance_data_refs_alignment): Likewise.
4017 (vect_analyze_data_refs_alignment): Likewise.
4018 (vect_slp_analyze_and_verify_instance_alignment
4019 (vect_analyze_data_ref_accesses): Likewise.
4020 (vect_prune_runtime_alias_test_list): Likewise.
4021 (vect_analyze_data_refs): Likewise.
4022 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
4023 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
4024 (vect_analyze_scalar_cycles_1): Likewise.
4025 (vect_get_loop_niters): Likewise.
4026 (vect_analyze_loop_form_1): Likewise.
4027 (vect_update_vf_for_slp): Likewise.
4028 (vect_analyze_loop_operations): Likewise.
4029 (vect_analyze_loop): Likewise.
4030 (vectorizable_induction): Likewise.
4031 (vect_transform_loop): Likewise.
4032 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
4033 * tree-vect-slp.c (vect_analyze_slp): Likewise.
4034 (vect_make_slp_decision): Likewise.
4035 (vect_detect_hybrid_slp): Likewise.
4036 (vect_slp_analyze_operations): Likewise.
4037 (vect_slp_bb): Likewise.
4038 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
4039 (vectorizable_bswap): Likewise.
4040 (vectorizable_call): Likewise.
4041 (vectorizable_simd_clone_call): Likewise.
4042 (vectorizable_conversion): Likewise.
4043 (vectorizable_assignment): Likewise.
4044 (vectorizable_shift): Likewise.
4045 (vectorizable_operation): Likewise.
4046 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
4048 2018-06-18 Martin Sebor <msebor@redhat.com>
4050 PR tree-optimization/81384
4051 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
4052 * builtins.c (expand_builtin_strnlen): New function.
4053 (expand_builtin): Call it.
4054 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
4055 * builtins.def (BUILT_IN_STRNLEN): New.
4056 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
4057 Warn for bounds in excess of maximum object size.
4058 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
4059 single-value ranges. Handle strnlen.
4060 (handle_builtin_strlen): Handle strnlen.
4061 (strlen_check_and_optimize_stmt): Same.
4062 * doc/extend.texi (Other Builtins): Document strnlen.
4064 2018-06-18 Maya Rashish <coypu@sdf.org>
4066 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
4067 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
4068 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
4070 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
4072 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
4074 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4076 * tree.c (escaped_string::escape): Replace cast to char * by
4077 const_cast<char *> (unescaped).
4079 2018-06-18 Nick Clifton <nickc@redhat.com>
4082 * tree.c (escaped_string): New class. Converts an unescaped
4083 string into its escaped equivalent.
4084 (warn_deprecated_use): Use the new class to convert the
4085 deprecation message, if present.
4086 (test_escaped_strings): New self test.
4087 (test_c_tests): Add test_escaped_strings.
4088 * doc/extend.texi (deprecated): Add a note that the
4089 deprecation message is affected by the -fmessage-length
4090 option, and that control characters will be escaped.
4091 (#pragma GCC error): Document this pragma.
4092 (#pragma GCC warning): Likewise.
4093 * doc/invoke.texi (-fmessage-length): Document this option's
4094 effect on the #warning and #error preprocessor directives and
4095 the deprecated attribute.
4097 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
4099 * tree.c (decl_value_expr_lookup): Revert latest change.
4100 (decl_value_expr_insert): Likewise.
4102 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
4104 * gimplify.c (nonlocal_vlas): Delete.
4105 (nonlocal_vla_vars): Likewise.
4106 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
4108 (gimplify_body): Do not create and destroy nonlocal_vlas.
4109 * tree-nested.c: Include diagnostic.h.
4110 (use_pointer_in_frame): Tweak.
4111 (lookup_field_for_decl): Add assertion and declare the transformation.
4112 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
4113 internal error when the reference is in a wrong context. Do not
4114 create a debug decl by default.
4115 (note_nonlocal_block_vlas): Delete.
4116 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
4117 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
4118 create a debug decl by default.
4119 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
4120 call is in a wrong context.
4121 (fixup_vla_decls): New function.
4122 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
4123 debug variables were created.
4124 * tree.c (decl_value_expr_lookup): Add checking assertion.
4125 (decl_value_expr_insert): Likewise.
4127 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
4130 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
4131 * tree-scalar-evolution.c (interpret_expr): Likewise.
4132 (expression_expensive_p): Likewise.
4133 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
4134 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
4135 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
4136 (ssa_defined_by_minus_one_stmt_p): New.
4138 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
4141 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
4142 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
4143 * dojump.c (do_jump): Likewise.
4144 * expr.c (expand_expr_real_2): Check operand type's sign.
4145 * fold-const.c (const_unop): Handle ABSU_EXPR.
4146 (fold_abs_const): Likewise.
4147 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
4148 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
4149 (strip_sign_op_1): Likesise.
4150 * match.pd: Add new pattern to generate ABSU_EXPR.
4151 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
4152 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4153 * tree-eh.c (operation_could_trap_helper_p): Likewise.
4154 * tree-inline.c (estimate_operator_cost): Likewise.
4155 * tree-pretty-print.c (dump_generic_node): Likewise.
4156 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
4157 * tree.def (ABSU_EXPR): New.
4159 2018-06-16 Jakub Jelinek <jakub@redhat.com>
4162 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
4163 documented as preserved for backward compatibility only.
4164 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
4166 PR rtl-optimization/86108
4167 * bb-reorder.c (create_forwarder_block): Renamed to ...
4168 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
4169 jump from new landing pad to the second part.
4170 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
4173 2018-06-15 Jakub Jelinek <jakub@redhat.com>
4176 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
4177 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
4178 Only call store_expr for halves if the mode is the same.
4181 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
4182 Fix up comment formatting.
4184 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
4186 * typed-splay-tree.h (typed_splay_tree::remove): New function.
4187 (typed_splay_tree::closure,
4188 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
4189 (typed_splay_tree::typed_splay_tree,
4190 typed_splay_tree::operator =): Declared private.
4191 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
4192 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
4193 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
4194 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
4195 typed_splay_tree::splay_tree_splay,
4196 typed_splay_tree::splay_tree_foreach_helper,
4197 typed_splay_tree::splay_tree_insert,
4198 typed_splay_tree::splay_tree_remove,
4199 typed_splay_tree::splay_tree_lookup,
4200 typed_splay_tree::splay_tree_predecessor,
4201 typed_splay_tree::splay_tree_successor,
4202 typed_splay_tree::splay_tree_min,
4203 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
4204 (typed_splay_tree::root, typed_splay_tree::comp,
4205 typed_splay_tree::delete_key,
4206 typed_splay_tree::delete_value): New data members.
4207 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
4208 typed_splay_tree::remove.
4210 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
4212 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
4213 -mginv and -mno-ginv to the assembler.
4214 * config/mips/mips.opt (-mcrc): New option.
4216 * doc/invoke.text (-mcrc): Document.
4219 2018-06-15 Nick Clifton <nickc@redhat.com>
4222 * tree.c (escaped_string): New class. Converts an unescaped
4223 string into its escaped equivalent.
4224 (warn_deprecated_use): Use the new class to convert the
4225 deprecation message, if present.
4226 (test_escaped_strings): New self test.
4227 (test_c_tests): Add test_escaped_strings.
4228 * doc/extend.texi (deprecated): Add a note that the
4229 deprecation message is affected by the -fmessage-length
4230 option, and that control characters will be escaped.
4231 (#pragma GCC error): Document this pragma.
4232 (#pragma GCC warning): Likewise.
4233 * doc/invoke.texi (-fmessage-length): Document this option's
4234 effect on the #warning and #error preprocessor directives and
4235 the deprecated attribute.
4237 2018-06-15 Richard Biener <rguenther@suse.de>
4239 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
4240 here, also noting vector size used.
4241 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
4242 size used in MSG_OPTIMIZED_LOCATIONS dump.
4243 (pass_slp_vectorize::execute): Adjust.
4245 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
4248 * config/arc/arc.c (arc_return_address_register): Fix
4251 2018-06-15 Richard Biener <rguenther@suse.de>
4254 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
4255 leave useless conversion stripping to force_gimple_operand_gsi.
4256 (gimplify_build2): Likewise.
4257 (gimplify_build1): Likewise.
4259 2018-06-15 Richard Biener <rguenther@suse.de>
4262 * tree-cfg.c (move_stmt_op): unshare invariant addresses
4263 before adjusting their block.
4265 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
4267 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
4268 multilibs for *-*-rtems*.
4269 * config/riscv/t-rtems: New file.
4271 2018-06-14 Jakub Jelinek <jakub@redhat.com>
4274 * match.pd ((A +- CST1) +- CST2): Punt if last resort
4275 unsigned_type_for returns NULL.
4278 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
4279 subregs of multi-word pseudos unless the float mode has word size.
4281 2018-06-14 Richard Biener <rguenther@suse.de>
4284 * tree-vect-generic.c (build_word_mode_vector_type): Remove
4285 duplicate and harmful type_hash_canon.
4286 * tree.c (type_hash_canon): Assert we didn't find ourselves.
4288 2018-06-14 Richard Biener <rguenther@suse.de>
4291 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
4294 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
4296 * config/rtems.h (STDINT_LONG32): Define.
4298 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
4299 Prachi Godbole <prachi.godbole@imgtec.com>
4301 * config/mips/mips-cpus.def: Define P6600.
4302 * config/mips/mips-tables.opt: Regenerate.
4303 * config/mips/mips.c (mips_ucbranch_type): New enum.
4304 (mips_rtx_cost_data): Add support for P6600.
4305 (mips_issue_rate): Likewise.
4306 (mips_multipass_dfa_lookahead): Likewise.
4307 (mips_avoid_hazard): Likewise.
4308 (mips_reorg_process_insns): Likewise.
4309 (mips_classify_branch_p6600): New function.
4310 * config/mips/mips.h (TUNE_P6600): New define.
4311 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
4312 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
4313 * config/mips/mips.md: Include p6600.md.
4314 (processor): Add p6600.
4315 * config/mips/p6600.md: New file.
4316 * doc/invoke.texi: Add p6600 to supported architectures.
4318 2018-06-13 Martin Sebor <msebor@redhat.com>
4320 PR tree-optimization/86114
4321 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
4323 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
4325 2018-06-13 Richard Biener <rguenther@suse.de>
4327 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
4328 Properly set vector type of the intermediate stmt.
4329 * tree-vect-stmts.c (vectorizable_operation): The destination
4330 var always has vectype_out type.
4332 2018-06-13 Jeff Law <law@redhat.com>
4334 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
4335 integer 0 for argument to print_rtl_with_bb.
4336 (rl78_reorg): Likewise.
4338 2018-06-13 David Malcolm <dmalcolm@redhat.com>
4340 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
4341 from rtx to rtx_insn *.
4342 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
4344 (add_sched_insns_for_speculation): Likewise for local "target",
4345 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
4346 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
4347 from rtx_insn ** to rtx_code_label **.
4348 (reorg_emit_nops): Likewise.
4349 (c6x_reorg): Likewise for local "call_labels".
4350 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
4352 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
4353 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
4354 the loops over LABEL_REFS.
4355 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
4357 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
4358 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
4359 (split_branches): Strengthen local "olabel" from rtx to
4360 rtx_insn *, adding a safe_as_a cast.
4361 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
4363 (add_insn_after): Likewise for first two params.
4364 (add_insn_before): Likewise.
4365 (remove_insn): Likewise for param.
4366 (emit_pattern_before_noloc): Likewise for second and third params.
4367 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
4368 (emit_call_insn_before_noloc): Likewise.
4369 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
4371 (emit_barrier_before): Likewise.
4372 (emit_label_before): Strengthen "label" param from "rtx" to
4373 "rtx_code_label *". Strengthen "before" param from "rtx" to
4375 (emit_insn_after_1): Strengthen "after" param from "rtx" to
4377 (emit_pattern_after_noloc): Likewise.
4378 (emit_insn_after_noloc): Likewise.
4379 (emit_jump_insn_after_noloc): Likewise.
4380 (emit_call_insn_after_noloc): Likewise.
4381 (emit_debug_insn_after_noloc): Likewise.
4382 (emit_barrier_after): Likewise.
4383 (emit_label_after): Likewise for both params.
4384 (emit_pattern_after_setloc): Likewise for "after" param. Convert
4385 "loc" param from "int" to "location_t".
4386 (emit_insn_after_setloc): Likewise.
4387 (emit_jump_insn_after_setloc): Likewise.
4388 (emit_call_insn_after_setloc): Likewise.
4389 (emit_debug_insn_after_setloc): Likewise.
4390 (emit_pattern_before_setloc): Likewise for "before" param. Convert
4391 "loc" param from "int" to "location_t".
4392 (emit_pattern_before): Convert NULL_RTX to NULL.
4393 (emit_insn_before_setloc): Convert "loc" param from "int" to
4395 (emit_jump_insn_before_setloc): Likewise.
4396 (emit_call_insn_before_setloc): Likewise.
4397 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
4398 rtx_insn *. Convert "loc" param from "int" to "location_t".
4399 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
4400 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
4401 Convert 3rd param from "int" to "location_t".
4402 (emit_barrier_before, emit_barrier_after, next_real_insn):
4403 Strengthen param from rtx to rtx_insn *.
4404 (emit_label_before): Strengthen 1st param from "rtx" to
4405 "rtx_code_label *". Strengthen 2nd param from "rtx" to
4407 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
4408 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
4409 Strengthen 2nd param from "rtx" to "rtx_insn *".
4410 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
4411 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
4412 Likewise. Convert 3rd param from "int" to "location_t".
4413 (emit_label_after): Strengthen 1st param from "rtx" to
4415 (next_real_insn, remove_insn): Strengthen param from "rtx" to
4417 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
4418 from "rtx" to "rtx_insn *".
4420 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
4422 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
4423 bodies streamed in with -Q.
4424 * dumpfile.c (dump_files): Add lto-stream-out dump file.
4425 * dumpfile.h (tree_dump_index): Add lto_stream_out.
4426 * gimple-streamer-out.c: Include gimple-pretty-print.h
4427 (output_bb): Dump stmts streamed.
4428 * lto-section-out.c: Include print-tree.h
4429 (lto_begin_section): Dump sections created.
4430 (lto_output_decl_index): Dump decl encoded.
4431 * lto-streamer-out.c: Include print-tree.h
4432 (create_output_block): Dump output block created.
4433 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
4434 (output_function): Dump function output.
4435 (output_constructor): Dump constructor streamed.
4436 (write_global_stream): Output indexes encoded.
4437 (produce_asm_for_decls): Dump streams encoded.
4438 * lto-streamer.c (streamer_dump_file): New global var.
4439 * lto-streamer.h (streamer_dump_file): Declare.
4440 * passes.c (ipa_write_summaries): Initialize streamer dump.
4441 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
4444 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
4447 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
4448 offsets for register save directives. Emit a second batch of save
4449 directives, if need be, when the function accesses prior frames.
4451 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4453 * config/arc/fpu.md (fmasf4): Force operand to register.
4454 (fnmasf4): Likewise.
4456 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4458 * config/arc/arc-protos.h (arc_pad_return): Remove.
4459 * config/arc/arc.c (machine_function): Remove force_short_suffix
4461 (arc_print_operand): Adjust printing of '&'.
4462 (arc_verify_short): Remove conditional printing of short suffix.
4463 (arc_final_prescan_insn): Remove reference to size_reason.
4464 (pad_return): New function.
4465 (arc_reorg): Call pad_return.
4466 (arc_pad_return): Remove.
4467 (arc_init_machine_status): Remove reference to force_short_suffix.
4468 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
4469 (attr length): When attribute iscompact is true force to 2
4470 regardless; in the case of maybe check if we want to force the
4471 instruction to have 4 bytes length.
4472 (nopv): Change it to generate 4 byte long nop as well.
4473 (blockage): New pattern.
4474 (simple_return): Remove call to arc_pad_return.
4475 (p_return_i): Likewise.
4477 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4479 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
4481 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4483 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
4486 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4488 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
4489 for ARC700 and ARCv2.
4491 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
4494 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
4495 operands[2] instead of operands[1].
4498 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
4500 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
4501 case, check whether the outer register overlaps an unallocatable
4502 register, not just whether it fits the required class.
4504 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
4506 * poly-int.h (can_div_trunc_p): Add new overload in which all values
4508 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
4509 (memrefs_conflict_p): Likewise.
4510 (init_alias_analysis): Likewise.
4511 * cfgexpand.c (expand_debug_expr): Likewise.
4512 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
4513 * cse.c (fold_rtx): Likewise.
4514 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
4515 * expr.c (emit_block_move_hints): Likewise.
4516 (clear_storage_hints, push_block, emit_push_insn): Likewise.
4517 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
4518 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
4519 (emit_group_store): Likewise.
4520 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
4521 to read the PRE/POST_MODIFY increment.
4522 * calls.c (store_one_arg): Use strip_offset.
4523 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
4525 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
4527 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
4528 (simplify_binary_operation_1): Extend CONST_INT handling to
4530 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
4531 than a HOST_WIDE_INT.
4532 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
4534 (adjust_mems, add_stores): Update accodingly.
4535 (vt_canonicalize_addr): Track polynomial offsets.
4536 (emit_note_insn_var_location): Likewise.
4537 (vt_add_function_parameter): Likewise.
4538 (vt_initialize): Likewise.
4540 2018-06-12 Jeff Law <law@redhat.com>
4542 * config.gcc (alpha*-*-freebsd*): Remove.
4543 * config/alpha/freebsd.h: Remove.
4545 2018-06-12 David Malcolm <dmalcolm@redhat.com>
4548 * spellcheck-tree.c (levenshtein_distance): Rename to...
4549 (get_edit_distance): ...this, and update for underlying renaming.
4550 * spellcheck-tree.h (levenshtein_distance): Rename to...
4551 (get_edit_distance): ...this.
4552 * spellcheck.c (levenshtein_distance): Rename to...
4553 (get_edit_distance): ...this. Convert from Levenshtein distance
4554 to Damerau-Levenshtein distance by supporting transpositions of
4555 adjacent characters. Rename "v1" to "v_next" and "v0" to
4557 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
4558 (selftest::test_edit_distance_unit_test_oneway): ...this, and
4559 update for underlying renaming.
4560 (selftest::levenshtein_distance_unit_test): Rename to...
4561 (selftest::test_get_edit_distance_unit): ...this, and update for
4562 underlying renaming.
4563 (selftest::test_find_closest_string): Add example from PR 69968
4564 where transposition helps
4565 (selftest::test_metric_conditions): Update for renaming.
4566 (selftest::test_metric_conditions): Likewise.
4567 (selftest::spellcheck_c_tests): Likewise.
4568 * spellcheck.h (levenshtein_distance): Rename both overloads to...
4569 (get_edit_distance): ...this.
4570 (best_match::consider): Update for renaming.
4572 2018-06-12 Martin Sebor <msebor@redhat.com>
4574 PR tree-optimization/85259
4575 * builtins.c (compute_objsize): Handle constant offsets.
4576 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
4577 true iff a warning has been issued.
4578 * gimple.h (gimple_nonartificial_location): New function.
4579 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
4580 gimple_nonartificial_location and handle -Wno-system-headers.
4581 (handle_builtin_stxncpy): Same.
4583 2018-06-12 Martin Sebor <msebor@redhat.com>
4586 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
4588 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
4590 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
4591 BUILTIN_VEC_XST entries for pointer to double and long long.
4593 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
4596 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
4598 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
4601 2018-06-12 Martin Liska <mliska@suse.cz>
4603 * doc/options.texi: Document IntegerRange.
4605 2018-06-12 Martin Liska <mliska@suse.cz>
4607 * config/i386/i386.opt: Make MPX-related options as Deprecated.
4608 * opt-functions.awk: Handle Deprecated flag.
4609 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
4611 (read_cmdline_option): Report warning for a deprecated option.
4612 * opts.h (struct cl_option): Add new field cl_deprecated.
4613 (CL_ERR_DEPRECATED): New.
4615 2018-06-12 Martin Liska <mliska@suse.cz>
4617 * doc/options.texi: Document Deprecated option flag.
4619 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4621 * config/arc/arc-arch.h (arc_extras): New enum.
4622 (arc_cpu_t):Add field extra.
4623 (arc_cpu_types): Consider the extras.
4624 * config/arc/arc-cpus.def: Add extras info.
4625 * config/arc/arc-opts.h (processor_type): Consider extra field.
4626 * config/arc/arc.c (arc_override_options): Handle extra field.
4628 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4630 * config/arc/arc-arch.h: Update ARC_OPTX macro.
4631 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
4633 * config/arc/arc.c (arc_init): Update pic warning.
4634 (irq_range): Update irq range parsing warnings.
4635 (arc_override_options): Update various warning messages.
4636 (arc_handle_aux_attribute): Likewise.
4638 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
4640 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
4642 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4644 * doc/sourcebuild.texi: Document usage of line number 0 in verify
4645 compiler messages directives.
4647 2018-06-12 Matthew Fortune <mfortune@gmail.com>
4649 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
4650 * config/mips/mips-tables.opt: Regenerate.
4651 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
4653 * doc/invoke.texi: Document -march=i6500.
4655 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
4657 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
4658 (i6400_gpmul): Add cpu_unit.
4659 (i6400_gpdiv): Likewise.
4660 (i6400_msa_add_d): Update reservations.
4661 (i6400_msa_int_add) Likewise.
4662 (i6400_msa_short_logic3) Likewise.
4663 (i6400_msa_short_logic2) Likewise.
4664 (i6400_msa_short_logic) Likewise.
4665 (i6400_msa_move) Likewise.
4666 (i6400_msa_cmp) Likewise.
4667 (i6400_msa_short_float2) Likewise.
4668 (i6400_msa_div_d) Likewise.
4669 (i6400_msa_long_logic1) Likewise.
4670 (i6400_msa_long_logic2) Likewise.
4671 (i6400_msa_mult) Likewise.
4672 (i6400_msa_long_float2) Likewise.
4673 (i6400_msa_long_float4) Likewise.
4674 (i6400_msa_long_float5) Likewise.
4675 (i6400_msa_long_float8) Likewise.
4676 (i6400_fpu_fadd): Include frint type.
4677 (i6400_fpu_store): New define_insn_reservation.
4678 (i6400_fpu_load): Likewise.
4679 (i6400_fpu_move): Likewise.
4680 (i6400_fpu_fcmp): Likewise.
4681 (i6400_fpu_fmadd): Likewise.
4682 (i6400_int_mult): Include imul3nc type and update reservation.
4683 (i6400_int_div): Include idiv3 type and update reservation.
4684 (i6400_int_load): Update to check type not move_type.
4685 (i6400_int_store): Likewise.
4686 (i6400_int_prefetch): Set zero latency.
4688 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
4690 * gcc.c: Document new %@{...} sequence.
4691 (LINK_COMMAND_SPEC): Use it for the -L switches.
4692 (cpp_unique_options): Use it for the -I switches.
4693 (at_file_argbuf): New global variable.
4694 (in_at_file): Likewise.
4695 (alloc_args): Create at_file_argbuf.
4696 (clear_args): Truncate at_file_argbuf.
4697 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
4698 (open_at_file): New function.
4699 (close_at_file): Likewise.
4700 (create_at_file): Delete.
4701 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
4704 (validate_switches_from_spec): Deal with %@{...} sequence.
4705 (validate_switches): Likewise.
4706 (driver::finalize): Call clear_args.
4708 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
4710 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
4712 2018-06-11 Martin Sebor <msebor@redhat.com>
4714 * doc/invoke.texi (-Wall): List -Wc++17-compat.
4715 (Wno-class-memaccess): Add @opindex.
4716 (Wno-templates, Wno-multiple-inheritance): Same.
4717 (Wno-virtual-inheritance, Wno-namespaces): Same.
4718 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
4719 (Wno-format-overflow, Wno-format-truncation): Same.
4720 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
4721 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
4722 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
4723 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
4724 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
4725 (Wno-misspelled-isr): Same.
4727 2018-06-11 Martin Sebor <msebor@redhat.com>
4729 * PR tree-optimization/86083
4730 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
4732 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
4734 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
4736 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
4739 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
4740 on the correct operand.
4741 (*movdi_internal64): Ditto.
4743 2018-06-11 Martin Liska <mliska@suse.cz>
4745 PR tree-optimization/86089
4746 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
4748 2018-06-11 Julia Koval <julia.koval@intel.com>
4750 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
4751 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
4752 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
4754 2018-06-11 Olivier Hainque <hainque@adacore.com>
4756 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
4757 for Ada with strict dwarf2.
4759 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
4762 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
4765 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
4767 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
4769 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
4771 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
4774 2018-06-08 Martin Liska <mliska@suse.cz>
4776 * tree-cfg.h (debug_function): Fix argument type to match
4779 2018-06-08 Martin Liska <mliska@suse.cz>
4781 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
4782 Remove usage of MPX-related (and removed) fields.
4783 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
4785 2018-06-08 David Malcolm <dmalcolm@redhat.com>
4787 * cfg.c (debug): Use TDF_NONE rather than 0.
4788 * cfghooks.c (debug): Likewise.
4789 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
4790 (struct dump_option_value_info): Convert to...
4791 (struct kv_pair): ...this template type.
4792 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
4794 (optinfo_verbosity_options): Likewise.
4795 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
4797 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
4798 than int for "optgroup_flags" param.
4799 (dump_generic_expr_loc): Use dump_flags_t rather than int for
4801 (dump_dec): Likewise.
4802 (dump_finish): Use TDF_NONE rather than 0.
4803 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
4804 rather than int for "optgroup_flags" param. Use TDF_NONE rather
4805 than 0. Update for change to option_ptr.
4806 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
4807 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
4808 0. Update for changes to optinfo_verbosity_options and
4810 (opt_info_switch_p): Convert optgroup_flags from int to
4812 (dump_basic_block): Use dump_flags_t rather than int
4813 for "dump_kind" param.
4814 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
4815 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
4816 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
4817 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
4818 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
4819 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
4820 TDF_NONE): Convert from macros to...
4821 (enum dump_flag): ...this new enum.
4822 (dump_flags_t): Update to use enum.
4823 (operator|, operator&, operator~, operator|=, operator&=):
4824 Implement for dump_flags_t.
4825 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
4826 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
4827 Convert from macros to...
4828 (enum optgroup_flag): ...this new enum.
4829 (optgroup_flags_t): New typedef.
4830 (operator|, operator|=): Implement for optgroup_flags_t.
4831 (struct dump_file_info): Convert field "alt_flags" to
4832 dump_flags_t. Convert field "optgroup_flags" to
4834 (dump_basic_block): Use dump_flags_t rather than int for param.
4835 (dump_generic_expr_loc): Likewise.
4836 (dump_dec): Likewise.
4837 (dump_register): Convert param "optgroup_flags" to
4839 (opt_info_enable_passes): Likewise.
4840 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
4842 * gimple-pretty-print.c (debug): Likewise.
4843 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
4844 (merged_store_group::apply_stores): Likewise.
4845 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
4846 * gimple.c (verify_gimple_pp): Likewise.
4847 * graphite-poly.c (print_pbb_body): Likewise.
4848 * passes.c (pass_manager::register_one_dump_file): Convert
4849 local "optgroup_flags" to optgroup_flags_t.
4850 * print-tree.c (print_node): Use TDF_NONE rather than 0.
4852 (debug_body): Likewise.
4853 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
4854 to optgroup_flags_t.
4855 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
4857 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
4858 (convert_mult_to_fma): Likewise.
4859 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
4860 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
4861 * tree-vect-data-refs.c (dump_lower_bound): Convert param
4862 "dump_kind" to dump_flags_t.
4864 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
4866 * config/rs6000/rs6000.c (min, max): Delete.
4868 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
4870 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
4871 -mabi=spe and -mabi=no-spe.
4873 2018-06-08 Martin Liska <mliska@suse.cz>
4875 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
4876 where we expect an existing summary.
4878 2018-06-08 Martin Liska <mliska@suse.cz>
4880 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
4881 * ipa-inline.h (estimate_edge_growth): Likewise.
4883 2018-06-08 Martin Liska <mliska@suse.cz>
4885 * cgraph.c (function_version_hasher::hash): Use
4886 cgraph_node::get_uid ().
4887 (function_version_hasher::equal):
4888 * cgraph.h (cgraph_node::get_uid): New method.
4889 * ipa-inline.c (update_caller_keys): Use
4890 cgraph_node::get_uid ().
4891 (update_callee_keys): Likewise.
4892 * ipa-utils.c (searchc): Likewise.
4893 (ipa_reduced_postorder): Likewise.
4894 * lto-cgraph.c (input_node): Likewise.
4895 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
4896 * symbol-summary.h (symtab_insertion): Likewise.
4897 (symtab_removal): Likewise.
4898 (symtab_duplication): Likewise.
4899 * tree-pretty-print.c (dump_function_header): Likewise.
4900 * tree-sra.c (convert_callers_for_node): Likewise.
4902 2018-06-08 Martin Liska <mliska@suse.cz>
4904 * cgraph.c (symbol_table::create_edge): Always assign a new
4906 (symbol_table::free_edge): Do not recycle numbers.
4907 * cgraph.h (cgraph_edge::get): New method.
4908 * symbol-summary.h (symtab_removal): Use it.
4909 (symtab_duplication): Likewise.
4910 (call_summary::hashable_uid): Remove.
4912 2018-06-08 Martin Liska <mliska@suse.cz>
4914 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
4915 (initialize_growth_caches): Remove.
4916 (free_growth_caches): Likewise.
4917 (do_estimate_edge_time): Use edge_growth_cache.
4918 (do_estimate_edge_size): Likewise.
4919 (do_estimate_edge_hints): Likewise.
4920 * ipa-inline.c (reset_edge_caches): Likewise.
4921 (recursive_inlining): Likewise.
4922 (inline_small_functions): Likewise.
4923 * ipa-inline.h (initialize_growth_caches): Remove.
4924 (estimate_edge_size): Likewise.
4925 (estimate_edge_time): Likewise.
4926 (estimate_edge_hints): Likewise.
4927 (reset_edge_growth_cache): Likewise.
4928 * symbol-summary.h (call_summary::remove): New method.
4930 2018-06-08 Martin Liska <mliska@suse.cz>
4932 * ipa-cp.c (class edge_clone_summary): New summary.
4933 (grow_edge_clone_vectors): Remove.
4934 (ipcp_edge_duplication_hook): Remove.
4935 (class edge_clone_summary_t): New call_summary class.
4936 (ipcp_edge_removal_hook): Remove.
4937 (edge_clone_summary_t::duplicate): New function.
4938 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
4939 (create_specialized_node): Likewise.
4940 (ipcp_driver): Initialize edge_clone_summaries and do not
4943 2018-06-08 Martin Liska <mliska@suse.cz>
4945 * symbol-summary.h (get): New function.
4946 (call_summary::m_initialize_when_cloning): New class member.
4948 2018-06-08 Martin Liska <mliska@suse.cz>
4950 * cgraph.c (cgraph_node::remove): Do not recycle uid.
4951 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
4952 (symbol_table::allocate_cgraph_symbol): Do not set uid.
4953 * passes.c (uid_hash_t): Record removed_nodes by their uids.
4954 (remove_cgraph_node_from_order): Use the removed_nodes set.
4955 (do_per_function_toporder): Likwise.
4956 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
4957 instead of summary_uid.
4958 (symtab_removal): Likewise.
4959 (symtab_duplication): Likewise.
4961 2018-06-08 Martin Liska <mliska@suse.cz>
4963 * ipa-cp.c (ipcp_store_bits_results): Use
4964 ipcp_transformation_sum.
4965 (ipcp_store_vr_results): Likewise.
4966 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
4968 (ipcp_transformation_initialize): ... this.
4969 (ipa_set_node_agg_value_chain):
4970 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
4971 (write_ipcp_transformation_info): Likewise.
4972 (read_ipcp_transformation_info): Likewise.
4973 (ipcp_update_bits): Likewise.
4974 (ipcp_update_vr): Likewise.
4975 (ipcp_transform_function): Likewise.
4976 * ipa-prop.h: Rename ipcp_transformation_summary to
4977 ipcp_transformation.
4978 (class ipcp_transformation_t): New function summary.
4979 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
4980 (ipa_get_agg_replacements_for_node): Likewise.
4982 2018-06-08 Martin Liska <mliska@suse.cz>
4984 * ipa-pure-const.c (struct funct_state_d): Do it class instead
4986 (class funct_state_summary_t): New function_summary class.
4987 (has_function_state): Remove.
4988 (get_function_state): Likewise.
4989 (set_function_state): Likewise.
4990 (add_new_function): Likewise.
4991 (funct_state_summary_t::insert): New function.
4992 (duplicate_node_data): Remove.
4993 (remove_node_data): Remove.
4994 (funct_state_summary_t::duplicate): New function.
4995 (register_hooks): Create new funct_state_summaries.
4996 (pure_const_generate_summary): Use it.
4997 (pure_const_write_summary): Likewise.
4998 (pure_const_read_summary): Likewise.
4999 (propagate_pure_const): Likewise.
5000 (propagate_nothrow): Likewise.
5001 (dump_malloc_lattice): Likewise.
5002 (propagate_malloc): Likewise.
5003 (execute): Do not register hooks, just remove summary
5005 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
5008 2018-06-08 Martin Liska <mliska@suse.cz>
5010 * ipa-reference.c (remove_node_data): Remove.
5011 (duplicate_node_data): Likewise.
5012 (class ipa_ref_var_info_summary_t): New class.
5013 (class ipa_ref_opt_summary_t): Likewise.
5014 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
5015 (get_reference_optimization_summary): Use
5016 ipa_ref_opt_sum_summaries.
5017 (set_reference_vars_info): Remove.
5018 (set_reference_optimization_summary): Likewise.
5019 (ipa_init): Create summaries.
5020 (init_function_info): Use function summary.
5021 (ipa_ref_opt_summary_t::duplicate): New function.
5022 (ipa_ref_opt_summary_t::remove): New function.
5023 (get_read_write_all_from_node): Fix GNU coding style.
5024 (propagate): Use function summary.
5025 (write_node_summary_p): Fix GNU coding style.
5026 (stream_out_bitmap): Likewise.
5027 (ipa_reference_read_optimization_summary): Use function summary.
5028 (ipa_reference_c_finalize): Do not release hooks.
5030 2018-06-08 Martin Liska <mliska@suse.cz>
5032 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
5033 (analyze_function_body): Extract multiple calls of get_create.
5034 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
5035 * ipa-inline.c (recursive_inlining): Use ::get method.
5036 * ipa-inline.h (estimate_edge_growth): Likewise.
5038 2018-06-08 Martin Liska <mliska@suse.cz>
5040 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
5042 (hsa_function_summary::hsa_function_summary): Use the new enum
5044 (hsa_gpu_implementation_p): Use hsa_summaries::get.
5045 * hsa-gen.c (hsa_get_host_function): Likewise.
5046 (get_brig_function_name): Likewise.
5047 * ipa-hsa.c (process_hsa_functions): Likewise.
5048 (ipa_hsa_write_summary): Likewise.
5049 * symbol-summary.h (symtab_duplication): Use ::get function/
5050 (get): New function.
5052 2018-06-08 Martin Liska <mliska@suse.cz>
5054 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
5056 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
5057 (hsa_register_kernel): Likewise.
5058 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
5059 * hsa-gen.c (hsa_get_host_function): Likewise.
5060 (get_brig_function_name): Likewise.
5061 (generate_hsa): Likewise.
5062 (pass_gen_hsail::execute): Likewise.
5063 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
5064 (devirtualization_time_bonus): Likewise.
5065 (ipcp_propagate_stage): Likewise.
5066 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
5067 (edge_set_predicate): Likewise.
5068 (evaluate_conditions_for_known_args): Likewise.
5069 (evaluate_properties_for_edge): Likewise.
5070 (ipa_fn_summary::reset): Likewise.
5071 (ipa_fn_summary_t::duplicate): Likewise.
5072 (dump_ipa_call_summary): Likewise.
5073 (ipa_dump_fn_summary): Likewise.
5074 (analyze_function_body): Likewise.
5075 (compute_fn_summary): Likewise.
5076 (estimate_edge_devirt_benefit): Likewise.
5077 (estimate_edge_size_and_time): Likewise.
5078 (estimate_calls_size_and_time): Likewise.
5079 (estimate_node_size_and_time): Likewise.
5080 (inline_update_callee_summaries): Likewise.
5081 (remap_edge_change_prob): Likewise.
5082 (remap_edge_summaries): Likewise.
5083 (ipa_merge_fn_summary_after_inlining): Likewise.
5084 (ipa_update_overall_fn_summary): Likewise.
5085 (read_ipa_call_summary): Likewise.
5086 (inline_read_section): Likewise.
5087 (write_ipa_call_summary): Likewise.
5088 (ipa_fn_summary_write): Likewise.
5089 (ipa_free_fn_summary): Likewise.
5090 * ipa-hsa.c (process_hsa_functions): Likewise.
5091 (ipa_hsa_write_summary): Likewise.
5092 (ipa_hsa_read_section): Likewise.
5093 * ipa-icf.c (sem_function::merge): Likewise.
5094 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
5095 (do_estimate_edge_time): Likewise.
5096 (estimate_size_after_inlining): Likewise.
5097 (estimate_growth): Likewise.
5098 (growth_likely_positive): Likewise.
5099 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
5100 (inline_call): Likewise.
5101 * ipa-inline.c (caller_growth_limits): Likewise.
5102 (can_inline_edge_p): Likewise.
5103 (can_inline_edge_by_limits_p): Likewise.
5104 (compute_uninlined_call_time): Likewise.
5105 (compute_inlined_call_time): Likewise.
5106 (want_inline_small_function_p): Likewise.
5107 (edge_badness): Likewise.
5108 (update_caller_keys): Likewise.
5109 (update_callee_keys): Likewise.
5110 (recursive_inlining): Likewise.
5111 (inline_small_functions): Likewise.
5112 (inline_to_all_callers_1): Likewise.
5113 (dump_overall_stats): Likewise.
5114 (early_inline_small_functions): Likewise.
5115 (early_inliner): Likewise.
5116 * ipa-inline.h (estimate_edge_growth): Likewise.
5117 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
5118 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5119 * ipa-prop.h (IPA_NODE_REF): Likewise.
5120 (IPA_EDGE_REF): Likewise.
5121 * ipa-pure-const.c (malloc_candidate_p): Likewise.
5122 (propagate_malloc): Likewise.
5123 * ipa-split.c (execute_split_functions): Likewise.
5124 * symbol-summary.h: Rename get to get_create.
5126 (get_create): Likewise.
5127 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
5129 2018-06-08 Martin Liska <mliska@suse.cz>
5131 * symbol-summary.h (release): Move definition out of class
5133 (symtab_removal): Likewise.
5134 (symtab_duplication): Likewise.
5136 2018-06-08 Martin Liska <mliska@suse.cz>
5138 * symbol-summary.h (function_summary): Move constructor
5139 implementation out of class declaration.
5140 (release): Likewise.
5141 (symtab_insertion): Likewise.
5142 (symtab_removal): Likewise.
5143 (symtab_duplication): Likewise.
5146 2018-06-08 Martin Liska <mliska@suse.cz>
5148 * Makefile.in: Remove support for MPX (macros, related functions,
5149 fields in cgraph_node, ...).
5150 * builtin-types.def (BT_BND): Likewise.
5151 (BT_FN_BND_CONST_PTR): Likewise.
5152 (BT_FN_CONST_PTR_BND): Likewise.
5153 (BT_FN_VOID_PTR_BND): Likewise.
5154 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
5155 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
5156 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
5157 (expand_builtin_mempcpy_with_bounds): Likewise.
5158 (expand_builtin_memset_with_bounds): Likewise.
5159 (expand_builtin_memset_args): Likewise.
5160 (std_expand_builtin_va_start): Likewise.
5161 (expand_builtin): Likewise.
5162 (expand_builtin_with_bounds): Likewise.
5163 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
5164 (DEF_LIB_BUILTIN_CHKP): Likewise.
5165 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
5166 (DEF_CHKP_BUILTIN): Likewise.
5167 (BUILT_IN_MEMCPY): Likewise.
5168 (BUILT_IN_MEMMOVE): Likewise.
5169 (BUILT_IN_MEMPCPY): Likewise.
5170 (BUILT_IN_MEMSET): Likewise.
5171 (BUILT_IN_STPCPY): Likewise.
5172 (BUILT_IN_STRCAT): Likewise.
5173 (BUILT_IN_STRCHR): Likewise.
5174 (BUILT_IN_STRCPY): Likewise.
5175 (BUILT_IN_STRLEN): Likewise.
5176 (BUILT_IN_MEMCPY_CHK): Likewise.
5177 (BUILT_IN_MEMMOVE_CHK): Likewise.
5178 (BUILT_IN_MEMPCPY_CHK): Likewise.
5179 (BUILT_IN_MEMSET_CHK): Likewise.
5180 (BUILT_IN_STPCPY_CHK): Likewise.
5181 (BUILT_IN_STRCAT_CHK): Likewise.
5182 (BUILT_IN_STRCPY_CHK): Likewise.
5183 * calls.c (store_bounds): Likewise.
5184 (emit_call_1): Likewise.
5185 (special_function_p): Likewise.
5186 (maybe_warn_nonstring_arg): Likewise.
5187 (initialize_argument_information): Likewise.
5188 (finalize_must_preallocate): Likewise.
5189 (compute_argument_addresses): Likewise.
5190 (expand_call): Likewise.
5191 * cfgexpand.c (expand_call_stmt): Likewise.
5192 (expand_return): Likewise.
5193 (expand_gimple_stmt_1): Likewise.
5194 (pass_expand::execute): Likewise.
5195 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
5196 (cgraph_node::remove): Likewise.
5197 (cgraph_node::dump): Likewise.
5198 (cgraph_node::verify_node): Likewise.
5199 * cgraph.h (chkp_function_instrumented_p): Likewise.
5200 (symtab_node::get_alias_target): Likewise.
5201 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
5202 (cgraph_local_p): Likewise.
5203 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
5204 (cgraph_edge::rebuild_references): Likewise.
5205 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
5206 (walk_polymorphic_call_targets): Likewise.
5207 (cgraph_node::expand_thunk): Likewise.
5208 (symbol_table::output_weakrefs): Likewise.
5209 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
5210 (ix86_handle_option): Likewise.
5211 * config/i386/constraints.md: Likewise.
5212 * config/i386/i386-builtin-types.def (BND): Likewise.
5216 * config/i386/i386-builtin.def (BDESC_END): Likewise.
5217 (BDESC_FIRST): Likewise.
5219 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
5220 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
5221 * config/i386/i386.c (enum reg_class): Likewise.
5222 (ix86_target_string): Likewise.
5223 (ix86_option_override_internal): Likewise.
5224 (ix86_conditional_register_usage): Likewise.
5225 (ix86_valid_target_attribute_inner_p): Likewise.
5226 (ix86_set_indirect_branch_type): Likewise.
5227 (ix86_set_current_function): Likewise.
5228 (ix86_function_arg_regno_p): Likewise.
5229 (init_cumulative_args): Likewise.
5230 (ix86_function_arg_advance): Likewise.
5231 (ix86_function_arg): Likewise.
5232 (ix86_pass_by_reference): Likewise.
5233 (ix86_function_value_regno_p): Likewise.
5234 (ix86_function_value_1): Likewise.
5235 (ix86_function_value_bounds): Likewise.
5236 (ix86_return_in_memory): Likewise.
5237 (ix86_setup_incoming_vararg_bounds): Likewise.
5238 (ix86_va_start): Likewise.
5239 (indirect_thunk_need_prefix): Likewise.
5240 (print_reg): Likewise.
5241 (ix86_print_operand): Likewise.
5242 (ix86_expand_call): Likewise.
5243 (ix86_output_function_return): Likewise.
5244 (reg_encoded_number): Likewise.
5245 (BDESC_VERIFYS): Likewise.
5246 (ix86_init_mpx_builtins): Likewise.
5247 (ix86_init_builtins): Likewise.
5248 (ix86_emit_cmove): Likewise.
5249 (ix86_emit_move_max): Likewise.
5250 (ix86_expand_builtin): Likewise.
5251 (ix86_builtin_mpx_function): Likewise.
5252 (ix86_get_arg_address_for_bt): Likewise.
5253 (ix86_load_bounds): Likewise.
5254 (ix86_store_bounds): Likewise.
5255 (ix86_load_returned_bounds): Likewise.
5256 (ix86_store_returned_bounds): Likewise.
5257 (ix86_class_likely_spilled_p): Likewise.
5258 (ix86_hard_regno_mode_ok): Likewise.
5259 (x86_order_regs_for_local_alloc): Likewise.
5260 (ix86_mitigate_rop): Likewise.
5261 (ix86_bnd_prefixed_insn_p): Likewise.
5262 (ix86_mpx_bound_mode): Likewise.
5263 (ix86_make_bounds_constant): Likewise.
5264 (ix86_initialize_bounds): Likewise.
5265 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
5266 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
5267 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
5268 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
5269 (TARGET_CHKP_BOUND_MODE): Likewise.
5270 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
5271 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
5272 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
5273 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
5274 * config/i386/i386.h (TARGET_MPX): Likewise.
5275 (TARGET_MPX_P): Likewise.
5276 (VALID_BND_REG_MODE): Likewise.
5277 (FIRST_BND_REG): Likewise.
5278 (LAST_BND_REG): Likewise.
5279 (enum reg_class): Likewise.
5280 (BND_REG_P): Likewise.
5281 (BND_REGNO_P): Likewise.
5282 (BNDmode): Likewise.
5283 (ADJUST_INSN_LENGTH): Likewise.
5284 * config/i386/i386.md: Likewise.
5285 * config/i386/i386.opt: Likewise.
5286 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
5287 (defined): Likewise.
5288 (LINK_MPX): Likewise.
5289 (MPX_SPEC): Likewise.
5290 (LIBMPX_SPEC): Likewise.
5291 (LIBMPXWRAPPERS_SPEC): Likewise.
5292 (CHKP_SPEC): Likewise.
5293 * config/i386/predicates.md: Likewise.
5294 * dbxout.c (dbxout_type): Likewise.
5295 * doc/extend.texi: Likewise.
5296 * doc/invoke.texi: Likewise.
5297 * doc/md.texi: Likewise.
5298 * doc/tm.texi: Likewise.
5299 * doc/tm.texi.in: Likewise.
5300 * dwarf2out.c (is_base_type): Likewise.
5301 (gen_formal_types_die): Likewise.
5302 (gen_subprogram_die): Likewise.
5303 (gen_type_die_with_usage): Likewise.
5304 (gen_decl_die): Likewise.
5305 (dwarf2out_late_global_decl): Likewise.
5306 * expr.c (expand_assignment): Likewise.
5307 (emit_storent_insn): Likewise.
5308 (store_expr_with_bounds): Likewise.
5309 (store_expr): Likewise.
5310 (expand_expr_real_1): Likewise.
5311 * expr.h (store_expr_with_bounds): Likewise.
5312 * function.c (use_register_for_decl): Likewise.
5313 (struct bounds_parm_data): Likewise.
5314 (assign_parms_augmented_arg_list): Likewise.
5315 (assign_parm_find_entry_rtl): Likewise.
5316 (assign_parm_is_stack_parm): Likewise.
5317 (assign_parm_load_bounds): Likewise.
5318 (assign_bounds): Likewise.
5319 (assign_parms): Likewise.
5320 (expand_function_start): Likewise.
5321 * gcc.c (CHKP_SPEC): Likewise.
5322 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
5323 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
5324 (wrestrict_dom_walker::check_call): Likewise.
5325 * gimple.c (gimple_build_call_from_tree): Likewise.
5326 * gimple.h (enum gf_mask): Likewise.
5327 (gimple_call_with_bounds_p): Likewise.
5328 (gimple_call_set_with_bounds): Likewise.
5329 * gimplify.c (gimplify_init_constructor): Likewise.
5330 * ipa-cp.c (initialize_node_lattices): Likewise.
5331 (propagate_constants_across_call): Likewise.
5332 (find_more_scalar_values_for_callers_subset): Likewise.
5333 * ipa-hsa.c (process_hsa_functions): Likewise.
5334 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
5335 * ipa-icf.c (sem_function::merge): Likewise.
5336 * ipa-inline.c (early_inliner): Likewise.
5337 * ipa-pure-const.c (warn_function_noreturn): Likewise.
5338 (warn_function_cold): Likewise.
5339 (propagate_pure_const): Likewise.
5340 * ipa-ref.h (enum GTY): Likewise.
5341 * ipa-split.c (find_retbnd): Likewise.
5342 (consider_split): Likewise.
5343 (split_function): Likewise.
5344 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
5345 * ipa.c (walk_polymorphic_call_targets): Likewise.
5346 (symbol_table::remove_unreachable_nodes): Likewise.
5347 (process_references): Likewise.
5348 (cgraph_build_static_cdtor_1): Likewise.
5349 * lto-cgraph.c (lto_output_node): Likewise.
5350 (output_refs): Likewise.
5351 (compute_ltrans_boundary): Likewise.
5352 (input_overwrite_node): Likewise.
5353 (input_node): Likewise.
5354 (input_cgraph_1): Likewise.
5355 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
5356 * passes.c (pass_manager::execute_early_local_passes): Likewise.
5357 (class pass_chkp_instrumentation_passes): Likewise.
5358 (make_pass_chkp_instrumentation_passes): Likewise.
5359 * passes.def: Likewise.
5360 * rtl.h (struct GTY): Likewise.
5361 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
5362 * stor-layout.c (layout_type): Likewise.
5363 * symtab.c: Likewise.
5364 * target.def: Likewise.
5365 * targhooks.c (default_chkp_bound_type): Likewise.
5366 (default_chkp_bound_mode): Likewise.
5367 (default_builtin_chkp_function): Likewise.
5368 (default_chkp_function_value_bounds): Likewise.
5369 (default_chkp_make_bounds_constant): Likewise.
5370 (default_chkp_initialize_bounds): Likewise.
5371 * targhooks.h (default_chkp_bound_type): Likewise.
5372 (default_chkp_bound_mode): Likewise.
5373 (default_builtin_chkp_function): Likewise.
5374 (default_chkp_function_value_bounds): Likewise.
5375 (default_chkp_make_bounds_constant): Likewise.
5376 (default_chkp_initialize_bounds): Likewise.
5377 * toplev.c (compile_file): Likewise.
5378 (process_options): Likewise.
5379 * tree-core.h (DEF_BUILTIN): Likewise.
5380 (DEF_BUILTIN_CHKP): Likewise.
5381 * tree-inline.c (declare_return_variable): Likewise.
5382 (remap_gimple_stmt): Likewise.
5383 (copy_bb): Likewise.
5384 (initialize_inlined_parameters): Likewise.
5385 (expand_call_inline): Likewise.
5386 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
5387 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
5388 (make_pass_ipa_chkp_produce_thunks): Likewise.
5389 (make_pass_chkp): Likewise.
5390 (make_pass_chkp_opt): Likewise.
5391 (make_pass_chkp_instrumentation_passes): Likewise.
5392 * tree-pretty-print.c (dump_generic_node): Likewise.
5393 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
5394 * tree-ssa-dce.c (propagate_necessity): Likewise.
5395 (eliminate_unnecessary_stmts): Likewise.
5396 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
5397 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
5398 * tree-ssa-sccvn.h: Likewise.
5399 * tree-ssa-strlen.c (get_string_length): Likewise.
5400 (valid_builtin_call): Likewise.
5401 (adjust_last_stmt): Likewise.
5402 (handle_builtin_strchr): Likewise.
5403 (handle_builtin_strcpy): Likewise.
5404 (handle_builtin_stxncpy): Likewise.
5405 (handle_builtin_memcpy): Likewise.
5406 (handle_builtin_strcat): Likewise.
5407 (strlen_check_and_optimize_stmt): Likewise.
5408 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
5409 * tree-streamer-in.c: Likewise.
5410 * tree-streamer.c (record_common_node): Likewise.
5411 * tree.c (tree_code_size): Likewise.
5412 (wide_int_to_tree_1): Likewise.
5413 (type_contains_placeholder_1): Likewise.
5414 (build_common_tree_nodes): Likewise.
5415 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
5416 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
5417 (POINTER_BOUNDS_P): Likewise.
5418 (BOUNDED_TYPE_P): Likewise.
5419 (BOUNDED_P): Likewise.
5420 (CALL_WITH_BOUNDS_P): Likewise.
5421 (pointer_bounds_type_node): Likewise.
5422 * value-prof.c (gimple_ic): Likewise.
5423 * var-tracking.c (vt_add_function_parameters): Likewise.
5424 * varasm.c (make_decl_rtl): Likewise.
5425 (assemble_start_function): Likewise.
5426 (output_constant): Likewise.
5427 (maybe_assemble_visibility): Likewise.
5428 * varpool.c (ctor_for_folding): Likewise.
5429 * chkp-builtins.def: Remove.
5430 * ipa-chkp.c: Remove.
5431 * ipa-chkp.h: Remove.
5432 * rtl-chkp.c: Remove.
5433 * rtl-chkp.h: Remove.
5434 * tree-chkp-opt.c: Remove.
5435 * tree-chkp.c: Remove.
5436 * tree-chkp.h: Remove.
5438 2018-06-07 Carl Love <cel@us.ibm.com>
5440 * config/rs6000/vsx.md (vextract_fp_from_shorth,
5441 vextract_fp_from_shortl): Add BE support.
5443 2018-06-07 Paul Koning <ni1d@arrl.net>
5445 * compare-elim.c (try_merge_compare): Don't merge compare if
5446 address contains a side effect.
5447 (try_eliminate_compare): Likewise.
5449 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
5451 * config.gcc: Support "tremont".
5452 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
5453 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5455 * config/i386/i386.c (m_TREMONT): Define.
5456 (processor_target_table): Add "tremont".
5457 (PTA_TREMONT): Define.
5458 (ix86_lea_outperforms): Add TARGET_TREMONT.
5459 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
5460 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
5461 and M_INTEL_GOLDMONT_PLUS.
5462 (fold_builtin_cpu): Add "tremont".
5463 (ix86_add_stmt_cost): Add TARGET_TREMONT.
5464 (ix86_option_override_internal): Add "tremont".
5465 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
5466 (processor_type): Add PROCESSOR_TREMONT.
5467 * config/i386/x86-tune.def: Add m_TREMONT.
5468 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
5470 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5472 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
5473 symbol defined for msp430i* devices to be lower case.
5475 2018-06-07 Richard Biener <rguenther@suse.de>
5477 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
5478 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
5479 Properly wrap signed arithmetic if overflow wraps.
5481 2018-06-07 Jakub Jelinek <jakub@redhat.com>
5483 PR tree-optimization/69615
5484 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
5485 of a cast from a same precision integral SSA_NAME in a bb dominated
5486 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
5487 cast to utype if rhs2 has already a compatible type.
5489 2018-06-07 Richard Biener <rguenther@suse.de>
5491 PR tree-optimization/85935
5492 * graphite-scop-detection.c (find_params_in_bb): Analyze
5493 condition operands with respect to the correct loop. Assert
5494 the analysis doesn't fail.
5496 2018-06-04 Carl Love <cel@us.ibm.com>
5498 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
5499 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
5500 as it is slightly cheaper.
5501 (first_match_or_eos_index_<mode>):
5502 Calculate index using natural element order.
5503 (first_match_index_<mode>):
5504 Calculate index using natural element order.
5505 (first_match_or_eos_index_<mode>):
5506 Calculate index using natural order.
5507 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
5508 for BE and LE modes.
5509 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
5510 P9V_BUILTIN_VCLZLSBB_V16QI.
5511 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
5514 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
5516 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
5517 indentation and line wrap for many prototypes. Add missing
5518 @smallexample directives around block of prototypes for vec_xl and
5521 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
5523 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
5524 track if we pass or return IEEE 128-bit floating point.
5525 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
5526 C++ mangling that is compatible with GCC 8.1.
5527 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
5528 (init_cumulative_args): Note if we pass or return IEEE 128-bit
5529 floating point types.
5530 (rs6000_function_arg_advance_1): Likewise.
5531 (rs6000_mangle_type): Optionally generate mangled names that match
5532 what GCC 8.1 generated for IEEE 128-bit floating point types.
5533 (rs6000_globalize_decl_name): If we have an external function that
5534 passes or returns IEEE 128-bit types, generate a weak reference
5535 from the mangled name used in GCC 8.1 to the current mangled
5537 (rs6000_init_builtins): Make __ibm128 use the long double type if
5538 long double is IBM extended double. Make __float128 use the long
5539 double type if long double is IEEE 128-bit.
5542 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
5543 macro for __ibm128 built-in functions.
5544 (PACK_IF): Add __ibm128 pack/unpack functions.
5545 (UNPACK_IF): Likewise.
5546 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
5547 enable long double built-in functions if long double is IEEE
5548 128-bit floating point.
5549 (rs6000_invalid_builtin): Update long double built-in function
5551 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
5552 functions, adjust the built-in function to use the long double
5553 built-in function if __ibm128 and long double are the same type.
5554 * doc/extend.texi (PowerPC builtins): Update documention for
5555 __builtin_{,un}pack_longdouble. Add documentation for
5556 __builtin_{,un}pack_ibm128.
5558 2018-06-06 Jim Wilson <jimw@sifive.com>
5560 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
5561 (struct machine_function): New field interrupt_mode.
5562 (riscv_handle_type_attribute): New function. Add forward declaration.
5563 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
5564 (riscv_expand_epilogue): Check interrupt_mode field.
5565 (riscv_set_current_function): Check interrupt attribute args and
5566 set interrupt_mode field.
5567 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
5568 (riscv_sret, riscv_uret): New.
5569 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
5570 new arguments to interrupt attribute.
5572 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
5575 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
5576 Don't handle -mcpu=power8 if -mpower9-vector is also used.
5578 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
5580 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
5581 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
5582 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5583 several redundant entries.
5585 2018-06-06 David Malcolm <dmalcolm@redhat.com>
5587 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
5588 type from "rtx" to "rtx_insn *".
5589 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
5590 for local "call_insn", removing cast.
5591 (ix86_expand_call): Likewise, introducing a "call_insn" local.
5593 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
5595 PR tree-optimization/86066
5596 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
5597 for BIT_INSERT_EXPR stores.
5599 2018-06-06 Richard Biener <rguenther@suse.de>
5601 PR tree-optimization/86062
5602 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5603 component refs ontop
5604 of to be offsetted base.
5606 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5608 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
5609 to be static and remove check on interrupt attribute name.
5611 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
5613 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5614 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
5616 2018-06-05 Steve Ellcey <sellcey@cavium.com>
5619 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
5621 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
5622 Remove second argument, change how error is called.
5623 (aarch64_layout_arg): Remove second argument from
5624 aarch64_err_no_fpadvsimd call.
5625 (aarch64_init_cumulative_args): Ditto.
5626 (aarch64_gimplify_va_arg_expr): Ditto.
5627 * config/aarch64/aarch64.md (mov<mode>): Ditto.
5629 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
5631 * config/i386/i386.md (simple_return_indirect_internal): New expander.
5632 (*simple_return_indirect_internal<mode>): Rename from
5633 simple_return_indirect_internal. Use W mode iterator.
5634 (rstorssp): New expander.
5635 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
5636 (clrssbsy): New expander.
5637 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
5639 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
5641 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
5643 (cmse_check_pointed_object): Likewise.
5645 2018-06-05 Martin Liska <mliska@suse.cz>
5647 PR gcov-profile/47618
5648 * doc/invoke.texi: Document how -fprofile-dir format
5651 2018-06-05 Richard Biener <rguenther@suse.de>
5653 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
5654 removal pretend DOM info isn't available so we do not update
5655 it and only remove edges, not dominated blocks. Actually free
5656 DOM info in case we removed something. Remove unreachable blocks.
5657 (mfb_keep_latches): Work with either DOM info or marked backedges.
5658 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
5659 first. Mark backedges if DOM info isn't available.
5660 (Re-)compute DOM info after cleanup_control_flow_pre.
5662 2018-06-05 Richard Biener <rguenther@suse.de>
5664 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
5665 (locus_discrim_hasher::hash): Adjust.
5666 (locus_discrim_hasher::equal): Likewise.
5667 (next_discriminator_for_locus): Work on line directly.
5668 (same_line_p): Pass in expanded locus1 as well.
5669 (assign_discriminators): Avoid redundant location expansions.
5671 2018-06-05 Richard Biener <rguenther@suse.de>
5673 PR tree-optimization/86046
5674 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
5675 if required after clearing TREE_ADDRESSABLE.
5677 2018-06-05 Richard Biener <rguenther@suse.de>
5679 PR tree-optimization/86047
5680 * tree-ssa-loop.c (for_each_index): Glob handling of all
5681 decls and constants and really handle all of them.
5683 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5686 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
5687 qualifier_void_pointer and qualifier_const_void_pointer.
5688 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
5689 (arm_init_builtins): Handle the above.
5690 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
5691 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
5692 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
5695 2018-06-05 Martin Liska <mliska@suse.cz>
5697 * auto-profile.c (read_autofdo_file): Do not use
5698 gcov_ctr_summary struct.
5699 (afdo_callsite_hot_enough_for_early_inline): Likewise.
5700 * coverage.c (struct counts_entry): Likewise.
5701 (read_counts_file): Read just single summary entry.
5702 (get_coverage_counts): Use gcov_summary struct.
5703 * coverage.h (get_coverage_counts): Likewise.
5704 * gcov-dump.c (dump_working_sets): Likewise.
5705 (tag_summary): Dump just single summary.
5706 * gcov-io.c (gcov_write_summary): Write just histogram
5708 (gcov_read_summary): Read just single summary.
5709 (compute_working_sets): Use gcov_summary struct.
5710 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
5711 of GCOV_COUNTERS_SUMMABLE.
5712 (GCOV_COUNTERS_SUMMABLE): Remove.
5713 (GCOV_FIRST_VALUE_COUNTER): Replace with
5714 GCOV_COUNTER_V_INTERVAL.
5715 (struct gcov_ctr_summary): Remove.
5716 (struct gcov_summary): Directly use fields of former
5718 (compute_working_sets): Use gcov_summary struct.
5719 * gcov.c (read_count_file): Do not use ctrs fields.
5720 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
5722 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
5724 * profile.c: Likewise.
5725 * profile.h: Likewise.
5727 2018-06-05 Martin Liska <mliska@suse.cz>
5729 PR gcov-profile/84846
5730 * gcov.c (output_lines): Print working directory only
5731 in intermediate format.
5733 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
5735 * config/s390/s390-builtin-types.def: Add void function type.
5736 * config/s390/s390-builtins.def: Use the function type for the
5739 2018-06-04 Jim Wilson <jimw@sifive.com>
5741 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
5743 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
5744 and maybe_eh_return. Change regno to unsigned int. Use new args to
5745 handle EH_RETURN_DATA_REGNO registers properly.
5746 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
5747 (riscv_expand_epilogue): Update comment. Change argument name and
5748 type. Update code to use new name and type. Pass new args to
5749 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
5751 * config/riscv/riscv.md (NORMAL_RETURN): New.
5752 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
5753 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
5754 (eh_return): Call gen_eh_return_internal and emit barrier.
5755 (eh_return_internal): Call riscv_expand_epilogue.
5757 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
5759 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
5760 bit_insertion field and declare can_be_merged_into method.
5761 (merged_store_group::can_be_merged_into): New method.
5762 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
5763 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
5764 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
5766 2018-06-04 Richard Biener <rguenther@suse.de>
5768 PR tree-optimization/85955
5769 * builtins.c (fold_builtin_sincos): Convert pointers to
5770 destination to appropriate type before dereferencing.
5772 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
5774 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
5776 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
5778 * expr.c (expand_expr_real_1): Force the operand into memory if
5779 its TYPE_MODE is BLKmode and if there is no integer mode for
5780 the number of bits being extracted.
5782 2018-06-04 Jakub Jelinek <jakub@redhat.com>
5786 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5787 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
5789 2018-06-04 Richard Biener <rguenther@suse.de>
5791 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
5792 (cleanup_tree_cfg_noloop): ... single caller. Do
5793 start_recording_case_labels later.
5795 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
5797 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
5798 to _IMMINTRIN_H_INCLUDED.
5799 * config/i386/pconfigintrin.h: Ditto.
5800 * config/i386/waitpkgintrin.h: Ditto.
5801 * config/i386/immintrin.h: Add includes for sgxintrin.h,
5802 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
5803 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
5804 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
5805 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
5806 waitpkgintrin.h and cldemoteintrin.h.
5808 2018-06-04 Richard Biener <rguenther@suse.de>
5810 PR tree-optimization/86038
5811 * tracer.c (find_best_successor): Check probability for
5812 being initialized, bail out if not.
5814 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
5817 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
5818 of bits to ignore when comparing architectures.
5820 2018-06-04 Jakub Jelinek <jakub@redhat.com>
5822 PR tree-optimization/69615
5823 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
5824 maximum or minimum of the type, try to merge it also as if
5825 range1 is + [-, x - 1] or + [x + 1, -].
5828 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
5830 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
5832 PR tree-optimization/86034
5833 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
5834 the unsigned bitfield type in a bit insertion sequence if it does not
5835 have a larger precision than the bitfield size.
5836 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
5838 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
5840 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
5842 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
5844 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
5845 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
5846 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
5847 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
5849 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
5851 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
5852 Disable -fdelete-null-pointer-checks for ELF toolchain.
5854 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
5855 Kito Cheng <kito.cheng@gmail.com>
5857 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
5858 (nds32le-*-*, nds32be-*-*): Integrate checking process.
5859 (nds32*-*-*): Add glibc and uclibc conditions.
5860 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
5861 (TARGET_EXCEPT_UNWIND_INFO): Define.
5862 * config/nds32/elf.h: New file.
5863 * config/nds32/linux.h: New file.
5864 * config/nds32/nds32-elf.opt: New file.
5865 * config/nds32/nds32-linux.opt: New file.
5866 * config/nds32/nds32-fp-as-gp.c
5867 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
5868 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
5870 (nds32_asm_file_end): Ditto.
5871 (nds32_print_operand): Ditto.
5872 (nds32_insert_attributes): Ditto.
5873 (nds32_init_libfuncs): New function.
5874 (TARGET_HAVE_TLS): Define.
5875 (TARGET_INIT_LIBFUNCS): Define.
5876 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
5878 (TARGET_ELF): Apply different mcmodel setting.
5879 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
5880 been migrated into elf.h and linux.h files.
5881 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
5882 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
5883 (mcmodel): The content has been migrated into nds32-elf.opt and
5884 nds32-linux.opt files.
5885 * config/nds32/t-elf: New file.
5886 * config/nds32/t-linux: New file.
5888 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
5889 Shiva Chen <shiva0217@gmail.com>
5891 * config/nds32/constants.md (unspec_volatile_element): Add
5892 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
5893 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
5895 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
5896 (make_pass_nds32_fp_as_gp): Declare.
5897 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
5899 (nds32_asm_function_end_prologue): Remove unused asm output.
5900 (nds32_asm_function_begin_epilogue): Remove unused asm output.
5901 (nds32_asm_file_start): Output necessary fp_as_gp information.
5902 (nds32_option_override): Adjust register usage.
5903 (nds32_expand_prologue): Consider fp_as_gp situation.
5904 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
5905 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
5908 (simple_return): Ditto.
5909 (omit_fp_begin): Output special directive for fp_as_gp.
5910 (omit_fp_end): Output special directive for fp_as_gp.
5911 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
5912 mforbid-fp-as-gp): New options.
5914 2018-06-01 Mark Wielaard <mark@klomp.org>
5916 * dwarf2out.c (dwarf2out_finish): Remove generation of
5917 DW_AT_loclists_base.
5919 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
5921 * gimple-ssa-store-merging.c: Include gimple-fold.h.
5922 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
5923 (struct merged_store_group): Add bit_insertion field.
5924 (dump_char_array): Use standard hexadecimal format.
5925 (merged_store_group::merged_store_group): Set bit_insertion to false.
5926 (merged_store_group::apply_stores): Use optimal buffer size. Deal
5927 with BIT_INSERT_EXPR stores. Move up code updating the mask and
5928 also print the mask in the dump file.
5929 (pass_store_merging::gate): Minor tweak.
5930 (imm_store_chain_info::coalesce_immediate): Fix wrong association
5931 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
5932 stores with INTEGER_CST stores.
5933 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
5934 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
5935 and use it throughout. Generate bit insertion sequences if need be.
5936 (pass_store_merging::process_store): Remove redundant condition.
5937 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
5939 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
5941 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
5942 the 128-bit floating point types. Fix function comment.
5944 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5946 * config/aarch64/aarch64-simd.md
5947 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
5949 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
5952 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
5954 PR tree-optimization/85989
5955 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
5957 (backprop::intersect_uses): Check it when deciding whether this
5958 is a backedge reference.
5959 (backprop::process_block): Add each phi to m_visited_phis
5960 after visiting it, then clear it at the end.
5962 2018-06-01 Richard Biener <rguenther@suse.de>
5964 * tree-vectorizer.h (vect_dr_stmt): New function.
5965 (vect_get_load_cost): Adjust.
5966 (vect_get_store_cost): Likewise.
5967 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
5968 Use vect_dr_stmt instead of DR_SMTT.
5969 (vect_record_base_alignments): Likewise.
5970 (vect_calculate_target_alignment): Likewise.
5971 (vect_compute_data_ref_alignment): Likewise and make static.
5972 (vect_update_misalignment_for_peel): Likewise.
5973 (vect_verify_datarefs_alignment): Likewise.
5974 (vector_alignment_reachable_p): Likewise.
5975 (vect_get_data_access_cost): Likewise. Pass down
5976 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
5977 (vect_get_peeling_costs_all_drs): Likewise.
5978 (vect_peeling_hash_get_lowest_cost): Likewise.
5979 (vect_enhance_data_refs_alignment): Likewise.
5980 (vect_find_same_alignment_drs): Likewise.
5981 (vect_analyze_data_refs_alignment): Likewise.
5982 (vect_analyze_group_access_1): Likewise.
5983 (vect_analyze_group_access): Likewise.
5984 (vect_analyze_data_ref_access): Likewise.
5985 (vect_analyze_data_ref_accesses): Likewise.
5986 (vect_vfa_segment_size): Likewise.
5987 (vect_small_gap_p): Likewise.
5988 (vectorizable_with_step_bound_p): Likewise.
5989 (vect_prune_runtime_alias_test_list): Likewise.
5990 (vect_analyze_data_refs): Likewise.
5991 (vect_supportable_dr_alignment): Likewise.
5992 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
5993 (vect_gen_prolog_loop_niters): Likewise.
5994 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5995 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
5997 (vect_recog_mask_conversion_pattern): Likewise.
5998 (vect_try_gather_scatter_pattern): Likewise.
5999 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
6000 to vect_get_store_cost.
6001 (vect_get_store_cost): Get stmt_info instead of DR.
6002 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
6003 (vect_get_load_cost): Get stmt_info instead of DR.
6005 2018-06-01 Richard Biener <rguenther@suse.de>
6008 * gimple-fold.c (var_decl_component_p): Also allow offsetted
6009 vars wrapped in MEM_REFs.
6011 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
6013 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
6014 Fix subreg tests so that we only return a choice between
6015 GENERAL_REGS and FP_REGS if the original classes included both.
6017 2018-06-01 Richard Biener <rguenther@suse.de>
6020 * tree-ssa-structalias.c (get_function_part_constraint):
6021 Handle NULL fi->decl.
6022 (find_func_aliases_for_call): Properly handle indirect
6023 fi from direct call.
6024 (find_func_clobbers): Likewise.
6025 (ipa_pta_execute): Likewise.
6026 (create_variable_info_for): For functions that are ifunc_resolver
6027 resolve to a varinfo that contains the result of the resolver call.
6028 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
6031 2018-05-31 Michael Collison <michael.collison@arm.com>
6033 * config/aarch64/aarch64.md:
6034 (*fix_to_zero_extenddfdi2): New pattern.
6035 * gcc.target/aarch64/fix_extend1.c: New testcase.
6037 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
6041 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
6042 and BUILT_IN_STRNCMP_EQ.
6043 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
6044 BUILT_IN_STRNCMP_EQ.
6045 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
6046 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
6047 (gimple_fold_builtin): Likewise.
6048 * tree-ssa-strlen.c (compute_string_length): New function.
6049 (determine_min_obsize): New function.
6050 (handle_builtin_string_cmp): New function to handle calls to
6051 string compare functions.
6052 (strlen_optimize_stmt): Add handling to builtin string compare
6054 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
6055 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
6056 * tree.c (build_common_builtin_nodes): Add new defines of
6057 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
6059 2018-05-31 Jakub Jelinek <jakub@redhat.com>
6062 * bb-reorder.c (pass_partition_blocks::gate): Return false for
6063 functions with naked attribute.
6065 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
6067 * config/i386/sse.md (avx_vec_concat<mode>):
6068 Substitute concat_tg_mode mode attribute with xtg_mode.
6069 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
6070 (concat_tg_mode): Remove mode attribute.
6072 2018-05-31 Martin Sebor <msebor@redhat.com>
6075 * calls.c (alloc_max_size): Correct a logic error/typo.
6076 Treat excessive arguments as infinite. Warn for invalid arguments.
6077 * doc/invoke.texi (-Walloc-size-larger-than): Update.
6079 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
6082 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
6083 and movx for Haswell.
6085 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
6086 Cesar Philippidis <cesar@codesourcery.com>
6089 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
6090 when emitting error on private/firstprivate reductions.
6091 * omp-low.c (lower_omp_target): Avoid reference-type processing
6092 on pointers for firstprivate clause.
6094 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
6096 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
6099 * config/aarch64/aarch64-simd.md
6100 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
6101 (aarch64_ld1_x3_<mode>): Likewise
6102 (aarch64_st1x2<VALLDIF:mode>): Likewise
6103 (aarch64_st1_x2_<mode>): Likewise
6104 (aarch64_st1x3<VALLDIF:mode>): Likewise
6105 (aarch64_st1_x3_<mode>): Likewise
6106 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
6107 (vld1_s8_x3): Likewise.
6108 (vld1_u16_x3): Likewise.
6109 (vld1_s16_x3): Likewise.
6110 (vld1_u32_x3): Likewise.
6111 (vld1_s32_x3): Likewise.
6112 (vld1_u64_x3): Likewise.
6113 (vld1_s64_x3): Likewise.
6114 (vld1_f16_x3): Likewise.
6115 (vld1_f32_x3): Likewise.
6116 (vld1_f64_x3): Likewise.
6117 (vld1_p8_x3): Likewise.
6118 (vld1_p16_x3): Likewise.
6119 (vld1_p64_x3): Likewise.
6120 (vld1q_u8_x3): Likewise.
6121 (vld1q_s8_x3): Likewise.
6122 (vld1q_u16_x3): Likewise.
6123 (vld1q_s16_x3): Likewise.
6124 (vld1q_u32_x3): Likewise.
6125 (vld1q_s32_x3): Likewise.
6126 (vld1q_u64_x3): Likewise.
6127 (vld1q_s64_x3): Likewise.
6128 (vld1q_f16_x3): Likewise.
6129 (vld1q_f32_x3): Likewise.
6130 (vld1q_f64_x3): Likewise.
6131 (vld1q_p8_x3): Likewise.
6132 (vld1q_p16_x3): Likewise.
6133 (vld1q_p64_x3): Likewise.
6134 (vst1_s64_x2): Likewise.
6135 (vst1_u64_x2): Likewise.
6136 (vst1_f64_x2): Likewise.
6137 (vst1_s8_x2): Likewise.
6138 (vst1_p8_x2): Likewise.
6139 (vst1_s16_x2): Likewise.
6140 (vst1_p16_x2): Likewise.
6141 (vst1_s32_x2): Likewise.
6142 (vst1_u8_x2): Likewise.
6143 (vst1_u16_x2): Likewise.
6144 (vst1_u32_x2): Likewise.
6145 (vst1_f16_x2): Likewise.
6146 (vst1_f32_x2): Likewise.
6147 (vst1_p64_x2): Likewise.
6148 (vst1q_s8_x2): Likewise.
6149 (vst1q_p8_x2): Likewise.
6150 (vst1q_s16_x2): Likewise.
6151 (vst1q_p16_x2): Likewise.
6152 (vst1q_s32_x2): Likewise.
6153 (vst1q_s64_x2): Likewise.
6154 (vst1q_u8_x2): Likewise.
6155 (vst1q_u16_x2): Likewise.
6156 (vst1q_u32_x2): Likewise.
6157 (vst1q_u64_x2): Likewise.
6158 (vst1q_f16_x2): Likewise.
6159 (vst1q_f32_x2): Likewise.
6160 (vst1q_f64_x2): Likewise.
6161 (vst1q_p64_x2): Likewise.
6162 (vst1_s64_x3): Likewise.
6163 (vst1_u64_x3): Likewise.
6164 (vst1_f64_x3): Likewise.
6165 (vst1_s8_x3): Likewise.
6166 (vst1_p8_x3): Likewise.
6167 (vst1_s16_x3): Likewise.
6168 (vst1_p16_x3): Likewise.
6169 (vst1_s32_x3): Likewise.
6170 (vst1_u8_x3): Likewise.
6171 (vst1_u16_x3): Likewise.
6172 (vst1_u32_x3): Likewise.
6173 (vst1_f16_x3): Likewise.
6174 (vst1_f32_x3): Likewise.
6175 (vst1_p64_x3): Likewise.
6176 (vst1q_s8_x3): Likewise.
6177 (vst1q_p8_x3): Likewise.
6178 (vst1q_s16_x3): Likewise.
6179 (vst1q_p16_x3): Likewise.
6180 (vst1q_s32_x3): Likewise.
6181 (vst1q_s64_x3): Likewise.
6182 (vst1q_u8_x3): Likewise.
6183 (vst1q_u16_x3): Likewise.
6184 (vst1q_u32_x3): Likewise.
6185 (vst1q_u64_x3): Likewise.
6186 (vst1q_f16_x3): Likewise.
6187 (vst1q_f32_x3): Likewise.
6188 (vst1q_f64_x3): Likewise.
6189 (vst1q_p64_x3): Likewise.
6191 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6193 * config/msp430/msp430.c (msp430_output_labelref): Prepend
6194 user_label_prefix to name.
6196 * tree-core.h: Update comment about the format of NAME string
6197 passed to handler in attribute_spec.
6199 * config/msp430/msp430.md: Remove erroneous subreg expression from
6200 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
6201 zero_extend{q,h}isi2.
6203 2018-05-30 Borislav Petkov <bp@suse.de>
6205 * doc/extend.texi: Document some architecture specific
6206 constraints and sort entries.
6208 2018-05-30 Martin Sebor <msebor@redhat.com>
6211 * builtins.c (expand_builtin_stpcpy_1): New function.
6212 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
6213 only if the former succeeds.
6215 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
6217 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
6220 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6222 * doc/invoke.texi (-flinker-output): Document
6224 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6226 * passes.c (ipa_write_summaries): Only modify statements if body
6228 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
6229 incrementally linking.
6230 (ipa_passes): Likewise.
6231 * lto-cgraph.c (lto_output_node): When incrementally linking do not
6232 pass down resolution info.
6233 * common.opt (flag_incremental_link): Update info.
6234 * gcc.c (plugin specs): Turn flinker-output=* to
6235 -plugin-opt=-linker-output-known
6236 * toplev.c (compile_file): Also cut compilation when doing incremental
6238 * flag-types. (enum lto_partition_model): Add
6239 LTO_LINKER_OUTPUT_NOLTOREL.
6240 (invoke.texi): Add -flinker-output docs.
6241 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
6242 link same way as WPA; do not stream in dead initializers.
6244 * dwarf2out.c (dwarf2out_die_ref_for_decl,
6245 darf2out_register_external_decl): Support incremental link.
6247 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6249 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
6251 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6253 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
6254 it down to simple_object_copy_lto_debug_sections.
6255 (run_gcc): Determine incremental LTO link time and configure
6256 lto1 into non-wpa mode, disable renaming of debug sections.
6258 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
6260 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
6261 descriptions of various incorrectly documented functions.
6263 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
6266 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6267 address check not strict.
6269 2018-05-30 Richard Biener <rguenther@suse.de>
6271 PR tree-optimization/85964
6272 * tracer.c (better_p): Drop initialized count check, we only
6273 call the function with initialized counts now.
6274 (find_best_successor): Do find a best edge if one
6275 has uninitialized count.
6276 (find_best_predecessor): Likewise. Do BB frequency check only
6277 if count is initialized.
6279 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
6281 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
6282 (aarch64_ldrstr_offset_compare): New.
6283 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
6284 load/store orderings.
6285 (aarch64_gen_adjusted_ldpstp): Likewise.
6287 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
6289 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
6290 Check for subset of GENERAL_REGS and FP_REGS.
6291 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
6294 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
6296 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
6297 and wi::to_poly_offset. Add the current offset and then check
6298 whether the sum fits, rather than using an unchecked addition of
6299 a checked term. Check for a shwi rather than a uhwi.
6300 * expr.c (get_bit_range): Use tree_to_poly_uint64.
6301 (store_constructor): Use poly_int_tree_p.
6302 (expand_expr_real_1): Likewise.
6303 * function.c (assign_temp): Likewise.
6304 * fold-const.c (const_binop): Use poly_int_tree_p and
6306 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
6308 * ipa-icf-gimple.c (func_checker::compare_operand): Use
6309 to_poly_offset for MEM offsets.
6310 * ipa-icf.c (sem_variable::equals): Likewise.
6311 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
6312 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
6313 wi::to_poly_offset for BIT_FIELD_REF offsets.
6314 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
6316 * var-tracking.c (emit_note_insn_var_location): Use
6317 tree_to_poly_uint64.
6319 2018-05-29 Jim Wilson <jimw@sifive.com>
6321 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
6323 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
6326 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
6327 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
6329 (sse4_1_round<mode>2): Use nonimmediate_operand
6330 for operand 1 predicate.
6332 2018-05-29 Martin Sebor <msebor@redhat.com>
6333 Richard Biener <rguenther@suse.de>
6336 * calls.c (get_size_range): Call determine_value_range instead
6337 of get_value_range..
6338 * tree-vrp.h (determine_value_range): Declared new function.
6339 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
6341 2018-05-29 Richard Biener <rguenther@suse.de>
6343 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
6344 sure to use non-pattern stmts for get_earlier_stmt arguments.
6345 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
6346 called on pattern stmts.
6347 (get_later_stmt): Likewise.
6349 2018-05-29 Martin Liska <mliska@suse.cz>
6351 PR gcov-profile/85759
6352 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
6355 2018-05-29 Jakub Jelinek <jakub@redhat.com>
6357 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
6359 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
6363 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
6364 VEC_PACK_FLOAT_EXPR): New tree codes.
6365 * tree-pretty-print.c (op_code_prio): Handle
6366 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
6367 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
6368 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
6369 * tree-inline.c (estimate_operator_cost): Likewise.
6370 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
6371 * fold-const.c (const_binop): Likewise.
6372 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
6373 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
6374 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6375 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
6376 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
6377 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
6378 * expr.c (expand_expr_real_2): Likewise.
6379 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
6380 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
6381 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
6383 * optabs.c (expand_widen_pattern_expr): For
6384 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
6385 sign from result type rather than operand's type.
6386 (expand_binop_directly): For vec_packu_float_optab and
6387 vec_packs_float_optab allow result type to be different from operand's
6389 * optabs-tree.c (optab_for_tree_code): Handle
6390 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
6391 VEC_PACK_FLOAT_EXPR. Formatting fixes.
6392 * tree-vect-generic.c (expand_vector_operations_1): Handle
6393 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
6394 VEC_PACK_FLOAT_EXPR.
6395 * tree-vect-stmts.c (supportable_widening_operation): Handle
6397 (supportable_narrowing_operation): Handle FLOAT_EXPR.
6398 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
6399 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
6400 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
6401 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
6403 (vec_pack<floatprefix>_float_<mode>): New expander.
6404 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
6406 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
6407 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
6408 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
6409 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
6410 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
6412 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
6413 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
6414 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
6415 VEC_PACK_FLOAT_EXPR): Document.
6417 2018-05-29 Richard Biener <rguenther@suse.de>
6419 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
6421 (stmt_vec_info_vec): Make pointer.
6422 (init_stmt_vec_info_vec): Remove.
6423 (free_stmt_vec_info_vec): Likewise.
6424 (set_stmt_vec_info_vec): New function.
6425 (free_stmt_vec_infos): Likewise.
6426 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
6427 (set_vinfo_for_stmt): Likewise.
6428 (get_earlier_stmt): Likewise.
6429 (get_later_stmt): Likewise.
6430 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
6431 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
6432 (vec_info::~vec_info): Free stmt_vec_infos.
6433 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
6434 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
6435 (pass_slp_vectorize::execute): Likewise.
6436 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
6437 (free_stmt_vec_info_vec): Likewise.
6438 (set_stmt_vec_info_vec): New function.
6439 (free_stmt_vec_infos): Likewise.
6440 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
6441 the global stmt_vec_info_vec.
6442 * tree-parloops.c (gather_scalar_reductions): Use
6443 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
6446 2018-05-29 Richard Biener <rguenther@suse.de>
6448 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
6450 2018-05-29 Martin Liska <mliska@suse.cz>
6451 David Malcolm <dmalcolm@redhat.com>
6453 * vec.c (test_reverse): New.
6454 (vec_c_tests): Add new test.
6455 * vec.h (vl_ptr>::reverse): New function.
6457 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
6459 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
6461 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
6464 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
6466 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
6468 2018-05-28 Richard Biener <rguenther@suse.de>
6470 PR tree-optimization/85933
6471 * tree-vect-data-refs.c (vect_record_base_alignments): Only
6472 look at stmts marked as vectorizable.
6474 2018-05-28 Richard Biener <rguenther@suse.de>
6476 PR tree-optimization/85934
6477 * tree-vect-generic.c (expand_vector_operations_1): Hoist
6478 vector boolean check before scalar optimization.
6480 2018-05-28 Jakub Jelinek <jakub@redhat.com>
6482 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
6485 2018-05-28 Mark Wielaard <mark@klomp.org>
6487 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
6488 if it is an expression containing a minus sign.
6490 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
6492 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
6494 2018-05-27 Paul Koning <ni1d@arrl.net>
6496 * config/pdp11/pdp11.md (truncsihi2): Remove.
6498 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
6499 Chung-Ju Wu <jasonwucj@gmail.com>
6501 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
6503 (unaligned_store_dw): Ditto.
6504 * config/nds32/nds32-memory-manipulation.c
6505 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
6506 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
6507 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
6508 (emit_setmem_word_loop): Rename to ...
6509 (emit_setmem_doubleword_loop): ... this.
6510 (nds32_gen_dup_4_byte_to_word_value): New function.
6511 (nds32_gen_dup_8_byte_to_double_word_value): New function.
6512 (nds32_expand_setmem_loop): Refine implementation.
6513 (nds32_expand_setmem_loop_v3m): Ditto.
6514 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
6517 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
6519 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
6521 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
6523 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
6524 (nds32_init_machine_status): Initialize machine->attr_naked_p and
6525 machine->attr_no_prologue_p.
6526 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
6527 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
6528 (nds32_expand_epilogue): Consider attr_naked_p.
6529 (nds32_expand_epilogue_v3pop): Likewise.
6530 (nds32_can_use_return_insn): Likewise.
6531 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
6532 attr_no_prologue_p fields.
6533 * config/nds32/nds32.opt (mret-in-naked-func): New option.
6535 2018-05-27 Jakub Jelinek <jakub@redhat.com>
6538 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
6540 * config/i386/sse.md
6541 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
6543 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
6545 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
6547 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
6549 (*<floatsuffix>floatv2div2sf2): Rename to ...
6550 (*float<floatunssuffix>v2div2sf2): ... this.
6551 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
6552 (float<floatunssuffix>v2div2sf2_mask): ... this.
6553 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
6554 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
6555 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
6557 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
6559 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
6561 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
6563 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
6565 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
6567 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
6568 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
6569 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
6570 gen_ufix_truncv8dfv8si2.
6571 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
6572 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
6573 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
6574 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
6575 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
6576 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
6577 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
6578 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
6580 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
6584 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
6586 2018-05-25 Jim Wilson <jimw@sifive.com>
6588 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
6589 * config/riscv/riscv.c (struct machine_function): Add
6590 interrupt_handler_p and attribute_checked_p fields.
6591 (riscv_attribute_table): Add interrupt.
6592 (riscv_interrupt_type_p): New.
6593 (riscv_save_reg_p): Save extra regs for interrupt handler.
6594 (riscv_use_save_libcall): Return false for interrupt handler.
6595 (riscv_first_stack_step): Add forward declaration.
6596 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
6597 for interrupt handler with large frame. Use it for saved reg list.
6598 (riscv_expand_prologue): Move flag_stack_usage_info support to
6599 eliminate duplication.
6600 (riscv_expand_epilogue): Generate mret for interrupt handler.
6601 (riscv_epilogue_uses): New.
6602 (riscv_can_use_return_insn): Return false for interrupt handler.
6603 (riscv_function_ok_for_sibcall): Likewise.
6604 (riscv_set_current_function): Add interrupt handler support.
6605 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
6606 * config/riscv/riscv.md (UNSPECV_MRET): New.
6608 (riscv_frflags, riscv_fsflags): Use tab after opcode.
6610 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
6612 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
6614 PR tree-optimization/85712
6615 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
6616 this candidate has already been replaced in-situ by a copy.
6618 2018-05-25 Jason Merrill <jason@redhat.com>
6620 PR c++/80485 - inline function non-zero address.
6621 * symtab.c (nonzero_address): Check DECL_COMDAT.
6623 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
6626 * config/alpha/alpha.md (ashlsi3): New insn pattern.
6627 (*ashlsi_se): Rename from *ashldi_se. Define as sign
6628 extension of SImode operation. Use const123_operand predicate.
6629 (*saddsi_1): Remove.
6630 (*saddl_se_1): Ditto.
6632 (*ssubl_se_1): Ditto.
6633 * config/alpha/predicates.md (const123_operand): New predicate.
6634 * config/alpha/constraints.md (P): Use IN_RANGE.
6636 2018-05-25 Richard Biener <rguenther@suse.de>
6638 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
6640 (ref_maybe_used_by_stmt_p): Likewise.
6641 (stmt_may_clobber_ref_p): Likewise.
6642 (stmt_may_clobber_ref_p_1): Likewise.
6643 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
6645 (ref_maybe_used_by_stmt_p): Likewise.
6646 (stmt_may_clobber_ref_p): Likewise.
6647 (stmt_may_clobber_ref_p_1): Likewise.
6648 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
6649 the alias oracle to disambiguate DRs with stmts DR analysis
6651 (vect_analyze_data_refs): Do not give up on not analyzable
6652 DRs for BB vectorization. Remove code truncating the dataref
6655 2018-05-25 Jakub Jelinek <jakub@redhat.com>
6658 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
6659 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
6660 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
6662 2018-05-25 Richard Biener <rguenther@suse.de>
6664 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
6665 function, combining stmt data ref gathering and fatal analysis
6667 (vect_analyze_data_refs): Remove now redudnant code and simplify.
6668 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
6669 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
6670 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
6671 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
6673 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6675 PR tree-optimization/85720
6676 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
6677 SCC if all partitions are builtins.
6678 (version_loop_by_alias_check): New parameter. Generate cancelable
6679 runtime alias check if all partitions are builtins.
6680 (distribute_loop): Update call to above function.
6682 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6684 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
6685 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
6686 (parm_default_def_partition_arg): Ditto.
6687 (set_parm_default_def_partition): Ditto.
6688 (get_parm_default_def_partitions): Ditto and make it static.
6689 (get_undefined_value_partitions): Ditto and make it static.
6690 (remove_ssa_form): Refactor call to init_var_map here.
6691 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
6692 computation for loop region.
6693 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
6694 (register_default_def): Delete.
6695 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
6696 (parm_default_def_partition_arg): Ditto.
6697 (set_parm_default_def_partition): Ditto.
6698 (get_parm_default_def_partitions): Ditto and make it static.
6699 (get_undefined_value_partitions): Ditto and make it static.
6700 (coalesce_with_default, coalesce_with_default): Update comment.
6701 (create_coalesce_list_for_region): New func factored out from
6702 create_outofssa_var_map.
6703 (populate_coalesce_list_for_outofssa): New func factored out from
6704 create_outofssa_var_map and coalesce_ssa_name.
6705 (create_outofssa_var_map): Delete.
6706 (coalesce_ssa_name): Refactor to support live range computation.
6707 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
6708 (get_parm_default_def_partitions): Delete.
6709 (get_undefined_value_partitions): Ditto.
6710 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
6711 computation for loop region.
6712 (new_tree_live_info, loe_visit_block): Ditto.
6713 (live_worklist, set_var_live_on_entry): Ditto.
6714 (calculate_live_on_exit, verify_live_on_entry): Ditto.
6715 * tree-ssa-live.h (struct _var_map): New fields.
6716 (init_var_map): Change decl.
6717 (region_contains_p): New.
6719 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6721 * tree-ssa-live.h (live_merge_and_clear): Delete.
6723 2018-05-25 Richard Biener <rguenther@suse.de>
6726 * tree-dump.c (dequeue_and_dump): Remove access to removed
6727 operand 2 of a SWITCH_EXPR.
6729 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6731 * doc/sourcebuild.texi (vect_double_cond_arith): Include
6732 multiplication and division.
6733 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
6734 (cond_udiv@var{m}, cond_umod@var{m}): Document.
6735 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
6736 (cond_udiv_optab, cond_umod_optab): New optabs.
6737 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
6738 (IFN_COND_RDIV): New internal functions.
6739 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
6740 TRUNC_MOD_EXPR and RDIV_EXPR.
6741 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
6742 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
6744 (SVE_INT_BINARY): Include mult.
6745 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
6746 (optab, sve_int_op): Handle mult.
6747 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
6749 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
6750 for SVE_INT_BINARY_SD.
6752 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6754 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
6755 (optab, sve_int_op): Handle div and udiv.
6756 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
6757 for SVE_INT_BINARY_SD.
6758 (*<optab><mode>3): New insn for the same.
6760 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6762 * tree-vect-patterns.c: Include predict.h.
6763 (vect_recog_divmod_pattern): Restrict check for division support
6764 to when optimizing for size.
6766 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6768 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
6769 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
6770 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
6771 (gimple_match_op::set_op): Likewise.
6772 (gimple_resimplify4): Declare.
6773 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
6774 (expr::gen_transform): Likewise.
6775 (decision_tree::gen): Generate a simplification routine for 4 operands.
6776 * gimple-match-head.c (gimple_simplify): Add an overload for
6777 4 operands. In the top-level function, handle up to 4 call
6778 arguments and call gimple_resimplify4.
6779 (gimple_resimplify4): New function.
6780 (build_call_internal): Pass a fourth operand.
6781 (maybe_push_to_seq): Likewise.
6782 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
6783 Fold VEC_COND_EXPRs of an operation and a default value into
6784 an IFN_COND_* function if possible.
6785 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
6787 (SVE_COND_FP_BINARY): Include them.
6788 (optab, sve_fp_op): Handle them.
6789 (SVE_INT_BINARY_REV): New code iterator.
6790 (SVE_COND_FP_BINARY_REV): New int iterator.
6791 (commutative): New int attribute.
6792 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
6794 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
6796 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
6797 (*cond_<optab><mode>): New patterns for reversed operands.
6799 2018-05-25 Richard Biener <rguenther@suse.de>
6801 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
6802 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
6803 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
6804 (STMT_VINFO_GROUPED_ACCESS): Adjust.
6805 * tree-vect-data-refs.c (everywhere): Adjust users.
6806 * tree-vect-loop.c (everywhere): Likewise.
6807 * tree-vect-slp.c (everywhere): Likewise.
6808 * tree-vect-stmts.c (everywhere): Likewise.
6809 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
6811 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6813 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
6815 (gcc_cv_as_section_exclude): ... this.
6816 Try Solaris as #exclude syntax.
6817 * configure: Regenerate.
6818 * config.in: Regenerate.
6819 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
6821 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
6822 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
6824 * varasm.c (default_elf_asm_named_section): Don't check if
6825 HAVE_GAS_SECTION_EXCLUDE is defined.
6827 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6829 * doc/md.texi: Update the documentation of the cond_* optabs
6830 to mention the new final operand. Fix GET_MODE_NUNITS call.
6831 Describe the scalar case too.
6832 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
6833 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
6835 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
6836 (get_conditional_internal_fn): Update comment.
6837 * tree-vect-loop.c (vectorizable_reduction): Pass the original
6838 accumulator value as a final argument to conditional functions.
6839 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
6840 a define_expand and add an "else" operand. Assert for now that
6841 the else operand is equal to operand 2. Use SVE_INT_BINARY and
6842 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
6843 (*cond_<optab><mode>): New patterns.
6844 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
6845 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6846 (UNSPEC_COND_EOR): Delete.
6847 (optab): Remove associated mappings.
6848 (SVE_INT_BINARY): New code iterator.
6849 (sve_int_op): Remove int attribute and add "minus" to the code
6851 (SVE_COND_INT_OP): Delete.
6852 (SVE_COND_FP_OP): Rename to...
6853 (SVE_COND_FP_BINARY): ...this.
6855 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6857 * optabs.c (can_reuse_operands_p): New function.
6858 (maybe_legitimize_operands): Try to reuse the results for
6861 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
6863 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
6864 Add {q} suffix to insn mnemonic.
6866 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6868 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
6869 (msp430_warn_func_return): New.
6871 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
6873 * fold-const.c (tree_nonzero_bits): New function.
6874 * fold-const.h (tree_nonzero_bits): Likewise.
6875 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
6876 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
6878 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
6882 * varasm.c (assemble_alias): Check ifunc_resolver only on
6885 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
6888 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
6889 when memory input operand is handled.
6891 2018-05-24 Luis Machado <luis.machado@linaro.org>
6893 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
6895 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
6897 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
6899 * match.pd: Delay FMA folds until after vectorization.
6901 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
6904 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6905 address check not strict.
6907 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
6909 * gimple-match.h (gimple_match_op): New class.
6910 (mprts_hook): Replace parameters with a gimple_match_op *.
6911 (maybe_build_generic_op): Likewise.
6912 (gimple_simplified_result_is_gimple_val): Replace parameters with
6913 a const gimple_match_op *.
6914 (gimple_simplify): Replace code_helper * and tree * parameters with
6915 a gimple_match_op * parameter.
6916 (gimple_resimplify1): Replace code_helper *, tree and tree *
6917 parameters with a gimple_match_op * parameter.
6918 (gimple_resimplify2): Likewise.
6919 (gimple_resimplify3): Likewise.
6920 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
6921 parameters with a gimple_match_op * parameter.
6922 * gimple-match-head.c (gimple_simplify): Change prototypes of
6923 auto-generated functions to take a gimple_match_op * instead of
6924 separate code_helper * and tree * parameters. Make the same
6925 change in the top-level overload and update calls to the
6926 gimple_resimplify routines. Update calls to the auto-generated
6927 functions and to maybe_push_res_to_seq in the publicly-facing
6928 operation-specific gimple_simplify overloads.
6929 (gimple_match_op::MAX_NUM_OPS): Define.
6930 (gimple_resimplify1): Replace rcode and ops with a single res_op
6931 parameter. Update call to gimple_simplify.
6932 (gimple_resimplify2): Likewise.
6933 (gimple_resimplify3): Likewise.
6934 (mprts_hook): Replace parameters with a gimple_match_op *.
6935 (maybe_build_generic_op): Likewise.
6936 (build_call_internal): Replace type, nargs and ops with
6937 a gimple_match_op *.
6938 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
6939 with a single gimple_match_op *. Update calls to mprts_hook,
6940 build_call_internal and gimple_simplified_result_is_gimple_val.
6941 Factor out code that is common to the tree_code and combined_fn cases.
6942 * genmatch.c (expr::gen_transform): Replace tem_code and
6943 tem_ops with a gimple_match_op called tem_op. Update calls
6944 to the gimple_resimplify functions and maybe_push_res_to_seq.
6945 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
6946 res_ops. Update call to the gimple_resimplify functions.
6947 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
6948 (decision_tree::gen): Make the functions take a gimple_match_op *
6949 called res_op instead of separate res_code and res_ops parameters.
6950 Update call accordingly.
6951 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
6952 and ops with a single res_op parameter. Update calls to
6953 maybe_build_generic_op and maybe_push_res_to_seq.
6954 (fold_stmt_1): Update calls to gimple_simplify and
6955 replace_stmt_with_simplification.
6956 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
6957 and gimple_simplified_result_is_gimple_val.
6958 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
6960 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
6961 with a gimple_match_op *.
6962 (vn_nary_build_or_lookup): Likewise. Update call to
6963 vn_nary_build_or_lookup_1.
6964 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
6965 gimple_match_op *. Update calls to the gimple_resimplify routines
6966 and to gimple_simplified_result_is_gimple_val.
6967 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
6968 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
6969 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
6970 (visit_nary_op): Likewise.
6971 (visit_reference_op_load): Likewise.
6973 2018-05-23 Luis Machado <luis.machado@linaro.org>
6975 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
6976 modifier for printing the step amount.
6978 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
6981 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
6984 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
6986 * doc/sourcebuild.texi (Endianness): New subsubsection.
6988 2018-05-23 Luis Machado <luis.machado@linaro.org>
6990 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6991 <prefetch_dynamic_strides>: New const bool field.
6992 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6993 prefetch_dynamic_strides.
6994 (exynosm1_prefetch_tune): Likewise.
6995 (thunderxt88_prefetch_tune): Likewise.
6996 (thunderx_prefetch_tune): Likewise.
6997 (thunderx2t99_prefetch_tune): Likewise.
6998 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
7000 (aarch64_override_options_internal): Update to set
7001 PARAM_PREFETCH_DYNAMIC_STRIDES.
7002 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7003 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7004 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7005 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7006 prefetch-dynamic-strides setting.
7008 2018-05-23 Luis Machado <luis.machado@linaro.org>
7010 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7011 <minimum_stride>: New const int field.
7012 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7013 minimum_stride field defaulting to -1.
7014 (exynosm1_prefetch_tune): Likewise.
7015 (thunderxt88_prefetch_tune): Likewise.
7016 (thunderx_prefetch_tune): Likewise.
7017 (thunderx2t99_prefetch_tune): Likewise.
7018 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
7019 <default_opt_level>: Set to 3.
7020 (aarch64_override_options_internal): Update to set
7021 PARAM_PREFETCH_MINIMUM_STRIDE.
7022 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7023 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7024 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7025 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7026 stride is constant and is below the minimum stride threshold.
7028 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7030 * config/arm/arm-cpus.in (mode26): Delete.
7031 (armv4): Delete mode26 reference.
7032 * config/arm/arm.c (arm_configure_build_target): Delete use of
7035 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
7037 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
7039 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
7040 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
7042 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
7043 (floatunsdidf2): Ditto.
7045 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
7047 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
7048 (fixuns_trunc<mode>si2_avx512f): Ditto.
7049 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
7050 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
7051 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
7053 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
7055 PR rtl-optimization/79985
7056 * df-scan.c (df_insn_refs_collect): Remove special case for
7057 global registers and asm statements.
7059 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
7061 * extend.texi (Global Register Variables): Rewrite the bullet list.
7062 Note that the register is available for allocation. Note that access
7063 via inline asm must use constraints. Add note about async-signal
7064 handlers. Remove paragraph about automagic register selection.
7066 2018-05-23 Richard Biener <rguenther@suse.de>
7068 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
7069 of fixed offset from memset VN.
7071 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
7073 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
7075 (alloc_cand_and_find_basis): Initialize first_interp field.
7076 (slsr_process_mul): Modify first_interp field.
7077 (slsr_process_add): Likewise.
7078 (slsr_process_cast): Modify first_interp field for each new
7080 (slsr_process_copy): Likewise.
7081 (dump_candidate): Dump first_interp field.
7082 (replace_mult_candidate): Process all interpretations, not just
7084 (replace_rhs_if_not_dup): Likewise.
7085 (replace_one_candidate): Likewise.
7087 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
7089 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
7091 (aarch64_needs_frame_chain): New function.
7092 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
7094 2018-05-23 Sudakshina Das <sudi.das@arm.com>
7097 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
7098 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
7099 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
7100 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
7101 as true for strict-align.
7102 (aarch64_can_inline_p): Perform checks even when callee has no
7103 attributes to check for strict alignment.
7104 * doc/extend.texi (AArch64 Function Attributes): Document
7106 * doc/invoke.texi: (AArch64 Options): Likewise.
7108 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
7110 PR tree-optimization/85853
7111 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
7112 the handling of the root of the node to...
7113 (vect_slp_analyze_node_operations_1): ...this new function,
7114 and run the whole thing with the child nodes' def types
7115 set according to their SLP node's def type.
7117 2018-05-23 Richard Biener <rguenther@suse.de>
7120 * tree-data-ref.c (create_runtime_alias_checks): Defer
7121 and ignore overflow warnings.
7123 2018-05-23 Yury Gribov <tetra2005@gmail.com>
7125 PR tree-optimization/85822
7126 * tree-vrp.c (is_masked_range_test): Fix handling of negative
7129 2018-05-23 Richard Biener <rguenther@suse.de>
7131 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
7132 memset constants via native_interpret_expr.
7134 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
7137 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
7139 (cgraph_node::create_alias): Likewise.
7140 (cgraph_node::get_availability): Check ifunc_resolver instead
7141 of looking up ifunc attribute.
7142 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
7143 * varasm.c (do_assemble_alias): Likewise.
7144 (assemble_alias): Likewise.
7145 (default_binds_local_p_3): Likewise.
7146 * cgraph.h (cgraph_node): Add ifunc_resolver.
7147 (cgraph_node::only_called_directly_or_aliased_p): Return false
7149 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
7151 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
7152 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
7153 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
7154 instead of looking up ifunc attribute.
7156 2018-05-22 Luis Machado <luis.machado@linaro.org>
7158 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
7160 2018-05-22 Martin Sebor <msebor@redhat.com>
7163 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
7164 only when expasion succeeds.
7165 (expand_builtin_strcmp): Same.
7166 (expand_builtin_strncmp): Same.
7168 2018-05-22 Martin Sebor <msebor@redhat.com>
7170 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
7172 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
7173 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7175 * config/aarch64/aarch64-ldpstp.md: Replace uses of
7176 aarch64_mem_pair_operand with memory_operand and delete operand swapping
7178 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
7179 Add check for legitimate_address.
7180 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
7181 (aarch64_swap_ldrstr_operands): New.
7182 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
7185 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
7186 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7188 * config/aarch64/aarch64.md: New patterns to generate stp
7190 (store_pair_sw, store_pair_dw): New patterns to generate stp for
7191 single words and double words.
7192 (load_pair_sw, load_pair_dw): Likewise.
7193 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
7195 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
7197 * config/aarch64/aarch64-ldpstp.md: Modify peephole
7198 for different mode ldpstp and add peephole for merged zero stores.
7200 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
7202 (aarch64_gen_store_pair): Rename calls to match new patterns.
7203 (aarch64_gen_load_pair): Rename calls to match new patterns.
7204 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
7205 (load_pair<DREG:mode><DREG2:mode>): ... This.
7206 (store_pair<mode>): Rename to...
7207 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
7208 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
7210 (V_INT_EQUIV): Handle SImode.
7211 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
7214 2018-05-22 Martin Sebor <msebor@redhat.com>
7217 * calls.c (maybe_warn_nonstring_arg): Use string length to set
7218 or ajust the presumed bound on an operation to avoid unnecessary
7221 2018-05-22 Martin Sebor <msebor@redhat.com>
7223 PR tree-optimization/85826
7224 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
7225 assuming that a DECL necesarily has a constant size.
7227 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
7230 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
7232 2018-05-22 Richard Biener <rguenther@suse.de>
7234 PR tree-optimization/85834
7235 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
7236 non-constant and non-zero memset arguments.
7238 2018-05-22 Martin Liska <mliska@suse.cz>
7241 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
7243 2018-05-22 Richard Biener <rguenther@suse.de>
7245 PR tree-optimization/85863
7246 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
7247 comparisons when vectype is specified.
7248 (vectorizable_condition): Do not specify vectype for
7249 vect_is_simple_cond when SLP vectorizing.
7251 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
7254 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
7255 define __ibm128 as long double.
7256 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
7257 as a distinct type when IEEE 128-bit support is enabled.
7258 (init_float128_ieee): Fix up conversions between IFmode and IEEE
7259 128-bit types to use the correct functions.
7260 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
7261 convert between 128-bit floating point types that have different
7262 modes but the same representation, instead of using gen_lowpart to
7264 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
7266 (IFKF_reg): New attributes to give the register constraints for
7268 (extend<mode>tf2_internal): New insns to mark an explicit
7269 conversion between 128-bit floating point types that have a
7270 different mode but share the same representation.
7272 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
7274 PR tree-optimization/85814
7275 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
7276 a null return from get_strinfo when unsharing the next
7277 strinfo in the chain.
7279 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7282 * varasm.c (weak_finish): Clean up weak_decls.
7284 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7286 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
7287 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
7288 UNSPEC_UADALP values.
7289 * config/aarch64/iterators.md (ABAL): New int iterator.
7292 (sur): Add mappings for the above.
7293 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
7295 (aarch64_<sur>abal<mode>_4): Likewise.
7296 (aarch64_<sur>adalp<mode>_3): Likewise.
7297 (<sur>sadv16qi): New define_expand.
7299 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
7301 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
7302 (*movdf_internal): Ditto.
7303 (*rcpsf2_sse): Ditto.
7304 (*rsqrtsf2_sse): Ditto.
7305 (*sqrt<mode>2_sse): Ditto.
7307 2018-05-21 Tamar Christina <tamar.christina@arm.com>
7309 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
7311 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
7312 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
7313 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
7314 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
7316 * config/aarch64/arm_neon.h: Likewise.
7317 * config/aarch64/iterators.md (VQ_I): New.
7319 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
7321 * config.gcc: Add arc/t-multilib-linux to tmake_file for
7323 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
7326 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
7328 * config/nds32/constraints.md (S): New constraint.
7329 * config/nds32/nds32.md (call_internal): Use constraint S.
7330 (call_value_internal): Likewise.
7331 (sibcall_internal): Likewise.
7332 (sibcall_value_internal): Likewise.
7334 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
7335 Chung-Ju Wu <jasonwucj@gmail.com>
7337 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
7340 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
7341 Chung-Ju Wu <jasonwucj@gmail.com>
7343 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
7344 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
7345 (nds32_rtx_costs_impl): Simplify.
7346 (nds32_address_cost_impl): Simplify.
7347 (nds32_init_rtx_costs): New function.
7348 (nds32_rtx_costs_speed_prefer): Likewise.
7349 (nds32_rtx_costs_size_prefer): Likewise.
7350 (nds32_address_cost_speed_prefer): Likewise.
7351 (nds32_address_cost_speed_fwprop): Likewise.
7352 (nds32_address_cost_size_prefer): Likewise.
7353 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
7354 * config/nds32/nds32.c (nds32_option_override): Use
7355 nds32_init_rtx_costs function.
7357 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
7359 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
7360 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
7361 (TARGET_PIPELINE_N8): Likewise.
7362 (TARGET_PIPELINE_N10): Likewise.
7363 (TARGET_PIPELINE_N13): Likewise.
7364 (TARGET_PIPELINE_GRAYWOLF): Likewise.
7366 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
7368 * config/nds32/nds32-fpu.md: Update copyright year.
7370 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7372 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
7374 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7376 * config/nds32/nds32.c
7377 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
7378 * config/nds32/nds32.opt (minline-asm-r15): New option.
7380 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7382 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
7384 * config/nds32/nds32.md (abssi2): New pattern.
7386 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
7388 * config/i386/i386.md (rex64namesuffix): New mode attribute.
7389 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
7390 Merge insn pattern from sse_cvtsi2ss<round_name> and
7391 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
7392 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
7393 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
7394 using SWI48 mode iterator.
7395 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
7396 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
7397 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
7398 pattern from sse_cvttss2si<round_saeonly_name>
7399 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
7400 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
7401 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
7402 using SWI48 mode iterator.
7403 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
7404 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
7405 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
7406 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
7407 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
7408 using SWI48 mode iterator.
7409 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
7410 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
7411 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
7412 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
7413 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
7414 SWI48 mode iterator.
7415 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
7416 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
7417 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
7418 pattern from sse_cvttsd2si<round_saeonly_name>
7419 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
7421 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7423 * config/nds32/nds32-md-auxiliary.c
7424 (nds32_valid_smw_lwm_base_p): Refine.
7425 (nds32_output_smw_single_word): Refine.
7426 (nds32_output_smw_double_word): New.
7427 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
7429 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7431 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
7432 (nds32_output_stack_pop): Refine.
7433 (nds32_expand_unaligned_load): Refine.
7434 (nds32_expand_unaligned_store): Refine.
7436 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7437 Chung-Ju Wu <jasonwucj@gmail.com>
7439 * config/nds32/constants.md: Add TP_REGNUM constant.
7440 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
7441 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
7443 * config/nds32/nds32-doubleword.md: Consider flag_pic.
7444 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
7445 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
7446 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
7447 and PIC code generation.
7448 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
7450 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
7452 * config/nds32/nds32.md: Support TLS and PIC.
7453 * config/nds32/nds32.c: Support TLS and PIC.
7454 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
7455 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
7458 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7460 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
7461 mode with E_ prefix.
7463 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7464 Chung-Ju Wu <jasonwucj@gmail.com>
7466 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
7467 * config/nds32/nds32-md-auxiliary.c
7468 (symbolic_reference_mentioned_p): New.
7469 (nds32_legitimize_ict_address): New.
7470 (nds32_expand_ict_move): New.
7471 (nds32_indirect_call_referenced_p): New.
7472 (nds32_symbol_binds_local_p): Delete.
7473 (nds32_long_call_p): Modify.
7474 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
7475 * config/nds32/nds32-protos.h
7476 (symbolic_reference_mentioned_p): Declare.
7477 (nds32_legitimize_ict_address): Declare.
7478 (nds32_expand_ict_move): Declare.
7479 (nds32_indirect_call_referenced_p): Declare.
7480 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
7481 (nds32_relax_group): Use nds32_ict_const_p as condition.
7482 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
7483 (nds32_asm_file_start): Output ict_model directive in asm code.
7484 (nds32_legitimate_address_p): Consider indirect call.
7485 (nds32_print_operand): Consider indirect call.
7486 (nds32_print_operand_address): Consider indirect call.
7487 (nds32_insert_attributes): Handle "indirect_call" attribute.
7488 (TARGET_LEGITIMATE_ADDRESS_P): Define.
7489 (TARGET_LEGITIMATE_CONSTANT_P): Define.
7490 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
7491 (TARGET_DELEGITIMIZE_ADDRESS): Define.
7492 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
7493 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
7494 (TARGET_ICT_MODEL_SMALL): Define.
7495 (TARGET_ICT_MODEL_LARGE): Define.
7496 * config/nds32/nds32.md (movsi): Consider ict model.
7497 (call, call_value): Consider ict model.
7498 (sibcall, sibcall_value): Consider ict model.
7499 * config/nds32/nds32.opt (mict-model): New option.
7500 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
7503 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
7504 Monk Chiang <sh.chiang04@gmail.com>
7505 Jim Wilson <jimw@sifive.com>
7507 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
7508 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
7509 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
7510 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
7511 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
7512 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
7513 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
7514 compute save_libcall_adjustment properly.
7515 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
7516 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
7517 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
7518 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
7519 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
7520 (ABI_SPEC): Handle mabi=ilp32e.
7521 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
7522 (RVE): Add RVE mask.
7523 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
7524 <-march>: Add rv32e as an example.
7526 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
7529 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
7530 (intra_create_variable_infos): Handle C++ constructors.
7532 2018-05-18 Martin Liska <mliska@suse.cz>
7534 * passes.def: Remove a redundant pass.
7536 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
7539 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
7541 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7543 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
7545 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
7547 (armv2, armv2a, armv3, armv3m): Delete architectures.
7548 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
7549 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
7550 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
7552 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
7553 (*mulsidi3adddi): Likewise.
7554 (mulsidi3): Likewise.
7555 (*mulsidi3_nov6): Likewise.
7556 (umulsidi3): Likewise.
7557 (umulsidi3_nov6): Likewise.
7558 (umaddsidi4): Likewise.
7559 (*umulsidi3adddi): Likewise.
7560 (smulsi3_highpart): Likewise.
7561 (*smulsi3_highpart_nov6): Likewise.
7562 (umulsi3_highpart): Likewise.
7563 (*umulsi3_highpart_nov6): Likewise.
7564 * config/arm/arm.h (arm_arch3m): Delete.
7565 * config/arm/arm.c (arm_arch3m): Delete.
7566 (arm_option_override_internal): Update armv3-related comment.
7567 (arm_configure_build_target): Delete use of isa_bit_mode32.
7568 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
7569 (arm_rtx_costs_internal): Delete check of arm_arch3m.
7570 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
7572 (mulusa3): Likewise.
7573 * config/arm/arm-protos.h (arm_arch3m): Delete.
7574 * config/arm/arm-tables.opt: Regenerate.
7575 * config/arm/arm-tune.md: Likewise.
7576 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
7577 deleted architectures.
7579 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7581 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
7582 (armv5t, armv5te): New features.
7583 (ARMv5, ARMv5e): Delete fgroups.
7584 (ARMv5t, ARMv5te): Adjust for above changes.
7586 (armv5, armv5e): Delete arches.
7587 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
7589 (*call_reg_arm): Likewise.
7590 (*call_value_reg_armv5): Likewise.
7591 (*call_value_reg_arm): Likewise.
7592 (*call_symbol): Likewise.
7593 (*call_value_symbol): Likewise.
7594 (*sibcall_insn): Likewise.
7595 (*sibcall_value_insn): Likewise.
7597 (prefetch): Likewise.
7598 (define_split and define_peephole2 dependent on arm_arch5):
7600 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
7602 (TARGET_ARM_QBIT): Likewise.
7603 (TARGET_DSP_MULTIPLY): Likewise.
7604 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
7605 (arm_arch5, arm_arch5e): Delete.
7606 (arm_arch5t, arm_arch5te): Declare.
7607 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
7608 (arm_arch5t): Declare.
7609 (arm_option_reconfigure_globals): Update for the above.
7610 (arm_options_perform_arch_sanity_checks): Update comment, replace
7611 use of arm_arch5 with arm_arch5t.
7612 (use_return_insn): Likewise.
7613 (arm_emit_call_insn): Likewise.
7614 (output_return_instruction): Likewise.
7615 (arm_final_prescan_insn): Likewise.
7616 (arm_coproc_builtin_available): Likewise.
7617 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
7618 arm_arch5e with arm_arch5t and arm_arch5te.
7619 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
7620 (arm_arch5t, arm_arch5te): Declare.
7621 * config/arm/arm-tables.opt: Regenerate.
7622 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
7623 * config/arm/t-multilib: Likewise.
7624 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
7625 instead of arm_arch5.
7626 (*call_reg_thumb1): Likewise.
7627 (*call_value_reg_thumb1_v5): Likewise.
7628 (*call_value_reg_thumb1): Likewise.
7629 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
7631 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
7633 2018-05-18 Martin Liska <mliska@suse.cz>
7635 PR gcov-profile/84846
7636 * doc/gcov.texi: Document -t option of gcov tool.
7638 2018-05-18 Martin Liska <mliska@suse.cz>
7640 PR gcov-profile/84846
7641 * gcov.c (print_usage): Add new -t option.
7642 (process_args): Handle the option.
7643 (generate_results): Use stdout as output when requested by
7646 2018-05-18 Martin Liska <mliska@suse.cz>
7648 PR gcov-profile/84846
7649 * coverage.c (coverage_init): Write PWD to .gcno file.
7650 * doc/gcov.texi: Document how working directory is printed.
7651 * gcov-dump.c (dump_gcov_file): Print PWD.
7652 * gcov.c (output_intermediate_file): Likewise.
7653 (read_graph_file): Read PWD string.
7654 (output_lines): Print PWD.
7656 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7659 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
7660 for retval and return false if all args to phi are zero.
7662 2018-05-18 Richard Biener <rguenther@suse.de>
7664 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
7666 (evrp_dom_walker::before_dom_children): Call it.
7668 2018-05-18 Richard Biener <rguenther@suse.de>
7670 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
7671 results when processing array refs with variable index.
7673 2018-05-18 Toon Moene <toon@moene.org>
7675 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
7676 directly after that of -floop-interchange. Indicate that both
7677 options are enabled by default when specifying -O3.
7679 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7681 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
7682 iterator. Delete separate integer-mode vec_set<mode> expander.
7683 (aarch64_simd_vec_setv2di): Delete.
7684 (vec_setv2di): Delete.
7685 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
7686 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
7687 the "w, r" alternative.
7689 2018-05-18 Martin Liska <mliska@suse.cz>
7691 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
7692 * tree-pass.h (make_pass_lower_switch_O0): New function.
7693 * tree-switch-conversion.c (node_has_low_bound): Remove.
7694 (node_has_high_bound): Likewise.
7695 (node_is_bounded): Likewise.
7696 (class pass_lower_switch): Make it a template type and create
7698 (pass_lower_switch::execute): Add template argument.
7699 (make_pass_lower_switch): New function.
7700 (make_pass_lower_switch_O0): New function.
7701 (do_jump_if_equal): Remove.
7702 (emit_case_nodes): Simplify to just handle all 3 cases and leave
7703 all the hard work to tree optimization passes.
7705 2018-05-18 Martin Liska <mliska@suse.cz>
7707 * dbgcnt.c (limit_low): Renamed from limit.
7708 (limit_high): New variable.
7709 (dbg_cnt_is_enabled): Check for upper limit.
7710 (dbg_cnt): Adjust dumping.
7711 (dbg_cnt_set_limit_by_index): Add new argument for high
7713 (dbg_cnt_set_limit_by_name): Likewise.
7714 (dbg_cnt_process_single_pair): Parse new format.
7715 (dbg_cnt_process_opt): Use strtok.
7716 (dbg_cnt_list_all_counters): Remove 'value' and add
7718 * doc/invoke.texi: Document changes.
7720 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
7722 * doc/sourcebuild.texi (scalar_all_fma): Document.
7723 * tree.def (FMA_EXPR): Delete.
7724 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
7725 * internal-fn.c (ternary_direct): New macro.
7726 (expand_ternary_optab_fn): Likewise.
7727 (direct_ternary_optab_supported_p): Likewise.
7728 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
7729 * builtins.c (fold_builtin_fma): Delete.
7730 (fold_builtin_3): Don't call it.
7731 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
7732 * expr.c (expand_expr_real_2): Likewise.
7733 * fold-const.c (operand_equal_p): Likewise.
7734 (fold_ternary_loc): Likewise.
7735 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
7736 * gimple.c (DEFTREECODE): Likewise.
7737 * gimplify.c (gimplify_expr): Likewise.
7738 * optabs-tree.c (optab_for_tree_code): Likewise.
7739 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
7740 * tree-eh.c (operation_could_trap_p): Likewise.
7741 (stmt_could_throw_1_p): Likewise.
7742 * tree-inline.c (estimate_operator_cost): Likewise.
7743 * tree-pretty-print.c (dump_generic_node): Likewise.
7744 (op_code_prio): Likewise.
7745 * tree-ssa-loop-im.c (stmt_cost): Likewise.
7746 * tree-ssa-operands.c (get_expr_operands): Likewise.
7747 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
7748 * fold-const-call.h (fold_fma): Delete.
7749 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
7750 CFN_FNMA and CFN_FNMS.
7752 * genmatch.c (combined_fn): New enum.
7753 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
7754 (commutative_op): New function.
7755 (commutate): Use it. Handle more than 2 operands.
7756 (dt_operand::gen_gimple_expr): Use commutative_op.
7757 (parser::parse_expr): Allow :c to be used with non-binary
7758 operators if the commutative operand is known.
7759 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
7760 CFN_FMS, CFN_FNMA and CFN_FNMS.
7761 (backprop::process_assign_use): Remove FMA_EXPR handling.
7762 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
7763 (gen_hsa_fma): New function.
7764 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
7765 IFN_FNMA and IFN_FNMS.
7766 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
7767 * gimple-fold.h (follow_all_ssa_edges): Declare.
7768 * gimple-fold.c (follow_all_ssa_edges): New function.
7769 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
7770 gimple_build interface and use follow_all_ssa_edges to fold the result.
7771 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
7772 instead of checking for optabs directly.
7773 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
7774 rather than FMA_EXPRs.
7775 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
7776 call to IFN_FMA instead of an FMA_EXPR.
7778 2018-05-17 Jim Wilson <jimw@sifive.com>
7780 * expr.c (do_tablejump): When converting index to Pmode, if we have a
7781 sign extended promoted subreg, and the range does not have the sign bit
7782 set, then do a sign extend.
7784 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
7785 test, check for sign extended subreg and/or constant operands, and
7786 do a sign extend in that case.
7788 2018-05-17 Steve Ellcey <sellcey@cavium.com>
7790 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
7791 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
7793 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
7794 Change logics_shift_reg to logics_shift_imm.
7795 (thunderx2t99_fp_loadpair_basic): Delete.
7796 (thunderx2t99_fp_storepair_basic): Delete.
7797 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
7798 (thunderx2t99_asimd_polynomial): Delete.
7799 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
7800 and neon_fp_mul_d_scalar_q.
7801 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
7802 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
7803 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
7804 (thunderx2t99_asimd_lut): Add missing tbl types.
7805 (thunderx2t99_asimd_ext): Delete.
7806 (thunderx2t99_asimd_load1_1_mult): Delete.
7807 (thunderx2t99_asimd_load1_2_mult): Delete.
7808 (thunderx2t99_asimd_load1_ldp): New.
7809 (thunderx2t99_asimd_load1): New.
7810 (thunderx2t99_asimd_load2): Add missing *load2* types.
7811 (thunderx2t99_asimd_load3): New.
7812 (thunderx2t99_asimd_load4): New.
7813 (thunderx2t99_asimd_store1_1_mult): Delete.
7814 (thunderx2t99_asimd_store1_2_mult): Delete.
7815 (thunderx2t99_asimd_store2_mult): Delete.
7816 (thunderx2t99_asimd_store2_onelane): Delete.
7817 (thunderx2t99_asimd_store_stp): New.
7818 (thunderx2t99_asimd_store1): New.
7819 (thunderx2t99_asimd_store2): New.
7820 (thunderx2t99_asimd_store3): New.
7821 (thunderx2t99_asimd_store4): New.
7823 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
7825 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
7826 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
7828 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
7829 Segher Boessenkool <segher@kernel.crashing.org>
7832 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
7835 2018-05-17 Richard Biener <rguenther@suse.de>
7837 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
7838 for pruning loop and prune defs feeding only already visited PHIs.
7840 2018-05-17 Richard Biener <rguenther@suse.de>
7842 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
7844 2018-05-17 Bin Cheng <bin.cheng@arm.com>
7845 Richard Biener <rguenther@suse.de>
7847 PR tree-optimization/85793
7848 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
7849 for VMAT_ELEMENTWISE.
7851 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
7853 * internal-fn.h (lookup_internal_fn): Declare
7854 * internal-fn.c (lookup_internal_fn): New function.
7855 * gimple.c (gimple_build_call_from_tree): Handle calls to
7857 * gimple-pretty-print.c (dump_gimple_call): Print "." before
7858 internal function names.
7859 * tree-pretty-print.c (dump_generic_node): Likewise.
7860 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
7862 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
7864 * gimple-fold.h (gimple_build): Make the function forms take
7865 combined_fn rather than built_in_function.
7866 (gimple_simplify): Likewise.
7867 * gimple-match-head.c (gimple_simplify): Likewise.
7868 * gimple-fold.c (gimple_build): Likewise.
7869 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
7870 rather than gimple_build_call_internal.
7871 (get_initial_defs_for_reduction): Likewise.
7872 (vect_create_epilog_for_reduction): Likewise.
7873 (vectorizable_live_operation): Likewise.
7875 2018-05-17 Martin Liska <mliska@suse.cz>
7877 * gimple-ssa-sprintf.c (format_directive): Do not use
7878 space in between 'G_' and '('.
7880 2018-05-17 Jakub Jelinek <jakub@redhat.com>
7883 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
7884 even if the mask is not all ones.
7887 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
7889 (ix86_gimple_fold_builtin): Likewise.
7892 * config/i386/i386.c: Include tree-vector-builder.h.
7893 (ix86_vector_shift_count): New function.
7894 (ix86_fold_builtin): Fold shift builtins by scalar count.
7895 (ix86_gimple_fold_builtin): Likewise.
7897 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
7898 _mm512_setzero): New intrinsics.
7900 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
7901 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7903 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
7904 code generation for cases where splatting a value is not useful.
7905 * simplify-rtx.c (simplify_ternary_operation): Simplify
7906 vec_merge across a vec_duplicate and a paradoxical subreg forming
7907 a vector mode to a vec_concat.
7909 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
7911 * config.gcc: Support "goldmont-plus".
7912 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
7914 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7915 PROCESSOR_GOLDMONT_PLUS.
7916 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
7917 (processor_target_table): Add "goldmont-plus".
7918 (PTA_GOLDMONT_PLUS): Define.
7919 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
7920 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
7921 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
7922 (fold_builtin_cpu): Add "goldmont-plus".
7923 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
7924 (ix86_option_override_internal): Add "goldmont-plus".
7925 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
7926 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
7927 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
7928 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
7930 2018-05-17 Richard Biener <rguenther@suse.de>
7932 PR tree-optimization/85757
7933 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
7934 remove defs that only feed that PHI from further processing.
7936 2018-05-16 Jim Wilson <jimw@sifive.com>
7938 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
7940 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
7942 2018-05-16 Mark Wielaard <mark@klomp.org>
7944 * dwarf2out.c (count_index_strings): New function.
7945 (output_indirect_strings): Call count_index_strings and generate
7946 header for dwarf_version >= 5.
7948 2018-05-16 Mark Wielaard <mark@klomp.org>
7950 * dwarf2out.c (dwarf_FORM): New function.
7951 (set_indirect_string): Use dwarf_FORM.
7952 (reset_indirect_string): Likewise.
7953 (size_of_die): Likewise.
7954 (value_format): Likewise.
7955 (output_die): Likewise.
7956 (add_skeleton_AT_string): Likewise.
7957 (output_macinfo_op): Likewise.
7958 (index_string): Likewise.
7959 (output_index_string_offset): Likewise.
7960 (output_index_string): Likewise.
7961 (count_index_strings): Likewise.
7963 2018-05-16 Carl Love <cel@us.ibm.com>
7965 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7966 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
7968 2018-05-16 Martin Jambor <mjambor@suse.cz>
7970 * ipa-prop.c (ipa_free_all_edge_args): Remove.
7971 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
7973 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
7975 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
7976 (fnma<mode>4): Likewise.
7977 (fms<mode>4): Likewise.
7978 (fnms<mode>4): Likewise.
7979 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
7980 (aarch64_fnma<mode>4): Likewise.
7981 (aarch64_fms<mode>4): Likewise.
7982 (aarch64_fnms<mode>4): Likewise.
7983 (aarch64_fnmadd<mode>4): Likewise.
7985 2018-05-16 Jason Merrill <jason@redhat.com>
7987 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
7989 2018-05-16 Richard Biener <rguenther@suse.de>
7991 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
7992 (dump_stmt_cost): Declare.
7993 (add_stmt_cost): Dump cost we add.
7994 (add_stmt_costs): New function.
7995 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
7997 (vect_analyze_stmt): Adjust prototype.
7998 (vectorizable_condition): Likewise.
7999 (vectorizable_live_operation): Likewise.
8000 (vectorizable_reduction): Likewise.
8001 (vectorizable_induction): Likewise.
8002 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
8003 cost vector to pass to vectorizable_ and record afterwards.
8004 (vect_model_reduction_cost): Take cost vector argument and adjust.
8005 (vect_model_induction_cost): Likewise.
8006 (vectorizable_reduction): Likewise.
8007 (vectorizable_induction): Likewise.
8008 (vectorizable_live_operation): Likewise.
8009 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
8010 SLP_TREE_NUMBER_OF_VEC_STMTS.
8011 (vect_analyze_slp_cost_1): Remove.
8012 (vect_analyze_slp_cost): Likewise.
8013 (vect_slp_analyze_node_operations): Take visited args and
8014 a target cost vector. Avoid processing already visited stmt sets.
8015 (vect_slp_analyze_operations): Use a local cost vector to gather
8016 costs and register those of non-discarded instances.
8017 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
8018 (vect_schedule_slp_instance): Remove copying of
8019 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
8021 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
8022 adding cost. Record cost entry location.
8023 (vect_prologue_cost_for_slp_op): Function to compute cost of
8024 a constant or invariant generated for SLP vect in the prologue,
8025 split out from vect_analyze_slp_cost_1.
8026 (vect_model_simple_cost): Make static. Adjust for SLP costing.
8027 (vect_model_promotion_demotion_cost): Likewise.
8028 (vect_model_store_cost): Likewise, make static.
8029 (vect_model_load_cost): Likewise.
8030 (vectorizable_bswap): Add cost vector arg and adjust.
8031 (vectorizable_call): Likewise.
8032 (vectorizable_simd_clone_call): Likewise.
8033 (vectorizable_conversion): Likewise.
8034 (vectorizable_assignment): Likewise.
8035 (vectorizable_shift): Likewise.
8036 (vectorizable_operation): Likewise.
8037 (vectorizable_store): Likewise.
8038 (vectorizable_load): Likewise.
8039 (vectorizable_condition): Likewise.
8040 (vectorizable_comparison): Likewise.
8041 (can_vectorize_live_stmts): Likewise.
8042 (vect_analyze_stmt): Likewise.
8043 (vect_transform_stmt): Adjust calls to vectorizable_*.
8044 * tree-vectorizer.c: Include gimple-pretty-print.h.
8045 (dump_stmt_cost): New function.
8047 2018-05-16 Richard Biener <rguenther@suse.de>
8049 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
8050 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
8051 * tree-ssa-dse.c: Include tree-ssa-loop.h.
8052 (check_name): New callback.
8053 (dse_classify_store): Track cycles via a visited bitmap of PHI
8054 defs and simplify handling of in-loop and across loop dead stores
8055 and properly fail for loop-variant refs. Handle byte-tracking with
8056 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
8059 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
8061 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
8062 (vect_get_mask_type_for_stmt): Likewise.
8063 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
8065 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
8066 to determine the statement's vector type and the vector type that
8067 should be used for calculating nunits. Deal with cases in which
8068 the type has to be deferred.
8069 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
8070 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
8071 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
8072 (vect_determine_vf_for_stmt): New functions, split out from...
8073 (vect_determine_vectorization_factor): ...here.
8074 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
8075 (vect_get_mask_type_for_stmt): New functions, split out from
8076 vect_determine_vectorization_factor.
8078 2018-05-16 Richard Biener <rguenther@suse.de>
8080 * tree-cfg.c (verify_gimple_assign_ternary): Properly
8081 verify the [VEC_]COND_EXPR embedded comparison.
8083 2018-05-15 Martin Sebor <msebor@redhat.com>
8085 PR tree-optimization/85753
8086 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
8087 RECORD_TYPE in addition to ARRAY_TYPE.
8089 2018-05-15 Martin Sebor <msebor@redhat.com>
8092 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
8094 2018-05-15 Richard Biener <rguenther@suse.de>
8096 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
8097 add by_clobber_p one. Change algorithm to collect all defs
8098 representing uses we need to walk and try reducing them to
8099 a single one before failing.
8100 (dse_dom_walker::dse_optimize_stmt): Adjust.
8102 2018-05-13 Mark Wielaard <mark@klomp.org>
8104 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
8105 (size_of_loc_descr): Likewise.
8106 (output_loc_operands): Likewise.
8107 (output_loc_operands_raw): Likewise.
8108 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
8109 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
8110 (hash_loc_operands): Likewise.
8111 (compare_loc_operands): Likewise.
8113 2018-05-14 Mark Wielaard <mark@klomp.org>
8115 * dwarf2out.c (count_index_addrs): New function.
8116 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
8118 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8120 PR tree-optimization/83648
8121 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
8122 return value as malloc candidate.
8124 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8127 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
8128 param as true in call to suggest_attribute.
8130 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
8132 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
8133 -mreadonly-in-sdata.
8135 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8137 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
8139 (aarch64_crypto_aesd_fused): Likewise.
8141 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
8143 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
8144 (movsi_aarch64): Likewise.
8145 (load_pairsi): Likewise.
8146 (load_pairdi): Likewise.
8147 (store_pairsi): Likewise.
8148 (store_pairdi): Likewise.
8149 (load_pairsf): Likewise.
8150 (load_pairdf): Likewise.
8151 (store_pairsf): Likewise.
8152 (store_pairdf): Likewise.
8153 (zero_extend): Likewise.
8154 (trunc): Swap alternatives.
8155 (fcvt_target): Add '?' to prefer w over r.
8157 2018-05-14 Jakub Jelinek <jakub@redhat.com>
8160 * config/i386/i386.md: Disallow non-commutative arithmetics in
8161 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
8162 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
8163 in the peephole2 before it.
8165 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
8167 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
8168 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
8169 (ix86_handle_option): Handle -mcldemote.
8170 * config.gcc: New header.
8171 * config/i386/cldemoteintrin.h: New file.
8172 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
8173 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
8175 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8176 OPTION_MASK_ISA_CLDEMOTE.
8177 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
8178 (ix86_valid_target_attribute_inner_p): Ditto.
8179 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
8180 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
8181 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
8182 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
8183 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
8185 * config/i386/i386.opt: Add -mcldemote.
8186 * config/i386/x86intrin.h: New header.
8187 * doc/invoke.texi: Add -mcldemote.
8189 2018-05-14 Richard Biener <rguenther@suse.de>
8191 * doc/match-and-simplify.texi: Adjust :s documentation.
8193 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
8195 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
8196 intended memcpy size.
8197 (REORDER_45): Likewise.
8199 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
8201 * sort.cc: New file.
8202 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
8203 * vec.c (qsort_chk): Use gcc_qsort.
8204 * Makefile.in (OBJS-libcommon): Add sort.o.
8205 (build/sort.o): New target. Use it...
8206 (BUILD_RTL): ... here, and...
8207 (build/gencfn-macros): ... here, and...
8208 (build/genmatch): ... here.
8210 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8211 Chung-Ju Wu <jasonwucj@gmail.com>
8213 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
8214 * config/nds32/nds32-graywolf.md: New file.
8215 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
8216 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
8218 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
8219 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
8220 * config/nds32/nds32.md (pipeline_model): Add graywolf.
8221 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
8222 * config/nds32/pipelines.md: Include n15 settings.
8224 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8225 Chung-Ju Wu <jasonwucj@gmail.com>
8227 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
8228 * config/nds32/nds32-n13.md: New file.
8229 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
8230 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
8232 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
8233 * config/nds32/nds32.md (pipeline_model): Add n13.
8234 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
8235 * config/nds32/pipelines.md: Include n13 settings.
8237 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8238 Chung-Ju Wu <jasonwucj@gmail.com>
8240 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
8241 * config/nds32/nds32-n10.md: New file.
8242 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
8243 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
8245 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
8246 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
8247 * config/nds32/nds32.md (pipeline_model): Add n10.
8248 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
8249 * config/nds32/pipelines.md: Include n10 settings.
8251 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
8252 Kito Cheng <kito.cheng@gmail.com>
8253 Chung-Ju Wu <jasonwucj@gmail.com>
8255 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
8256 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8257 Add enum values for DSP extension instructions.
8258 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
8260 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
8261 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
8263 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
8264 * config/nds32/nds32-dspext.md: New file for DSP implementation.
8265 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
8266 * config/nds32/nds32-intrinsic.md: Likewise.
8267 * config/nds32/nds32_intrinsic.h: Likewise.
8268 * config/nds32/nds32-md-auxiliary.c: Likewise.
8269 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
8270 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
8271 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
8272 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
8273 * config/nds32/nds32-protos.h: New declarations for DSP extension.
8274 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
8275 TYPE_DMAC in switch statement.
8276 * config/nds32/nds32.c: New checking and implementation for DSP
8277 extension instructions.
8278 * config/nds32/nds32.h: Likewise.
8279 * config/nds32/nds32.md: Likewise.
8280 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
8281 * config/nds32/predicates.md: Implement new predicates for DSP
8284 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8286 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
8287 Reformat alternatives and attributes so it is easier to identify
8288 which constraints/attributes go with which instruction.
8289 (mov<mode>_hardfloat32, FMOVE64): Likewise.
8290 (mov<mode>_softfloat32, FMOVE64): Likewise.
8291 (mov<mode>_hardfloat64, FMOVE64): Likewise.
8292 (mov<mode>_softfloat64, FMOVE64): Likewise.
8294 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
8296 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8298 (Basic PowerPC Built-in Functions): The new name of the
8299 subsection previously known as "PowerPC Built-in Functions".
8300 (Basic PowerPC Built-in Functions Available on all Configurations):
8302 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
8303 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
8304 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
8305 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
8307 2018-05-11 Martin Jambor <mjambor@suse.cz>
8310 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
8313 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
8316 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
8318 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
8320 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
8321 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
8322 (ix86_handle_option): Handle -mwaitpkg.
8323 * config.gcc: New header.
8324 * config/i386/cpuid.h (bit_WAITPKG): New bit.
8325 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
8326 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
8328 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8329 OPTION_MASK_ISA_WAITPKG.
8330 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
8331 (ix86_option_override_internal): Add PTA_WAITPKG.
8332 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
8333 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
8334 IX86_BUILTIN_TPAUSE.
8335 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
8336 __builtin_ia32_umwait and __builtin_ia32_tpause.
8337 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
8338 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
8339 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
8340 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
8341 UNSPECV_TPAUSE): New.
8342 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
8343 * config/i386/i386.opt: Add -mwaitpkg.
8344 * config/i386/waitpkgintrin.h: New file.
8345 * config/i386/x86intrin.h: New header.
8346 * doc/invoke.texi: Add -mwaitpkg.
8348 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
8351 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
8353 (cortex-m0): Use armv6s-m isa.
8354 (cortex-m0plus): Likewise.
8355 (cortex-m1): Likewise.
8356 (cortex-m0.small-multiply): Likewise.
8357 (cortex-m0plus.small-multiply): Likewise.
8358 (cortex-m1.small-multiply): Likewise.
8360 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
8361 Jakub Jelinek <jakub@redhat.com>
8363 PR tree-optimization/85692
8364 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
8365 source permute as well.
8367 2018-05-11 Martin Liska <mliska@suse.cz>
8370 * doc/extend.texi: Document LLVM style format for no_sanitize
8373 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8375 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
8376 mode_supports_vsx_dform_quad to mode_supports_dq_form.
8377 (mode_supports_vsx_dform_quad): Likewise.
8378 (mode_supports_vmx_dform): Move these functions to be next to the
8379 other mode_supports functions.
8380 (mode_supports_dq_form): Likewise.
8381 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
8382 mode_supports_dq_form.
8383 (reg_offset_addressing_ok_p): Likewise.
8384 (offsettable_ok_by_alignment): Likewise.
8385 (rs6000_legitimate_offset_address_p): Likewise.
8386 (legitimate_lo_sum_address_p): Likewise.
8387 (rs6000_legitimize_address): Likewise.
8388 (rs6000_legitimize_reload_address): Likewise.
8389 (rs6000_secondary_reload_inner): Likewise.
8390 (rs6000_preferred_reload_class): Likewise.
8391 (rs6000_output_move_128bit): Likewise.
8393 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
8395 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
8396 Generate SImode target register for null target.
8397 <case IX86_BUILTIN_XGETBV>: Ditto.
8398 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
8399 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
8401 2018-05-10 Carl Love <cel@us.ibm.com>
8403 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
8404 dcbtt and dcbtstt if operands[2] is 0.
8406 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
8409 * config/i386/sse.md (usadv64qi): New expander.
8411 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
8413 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
8414 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
8415 -maltivec=be support.
8416 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
8417 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
8418 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
8419 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
8420 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
8421 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
8422 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
8423 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
8424 altivec_vsumsws): Adjust.
8425 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
8426 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
8427 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
8429 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
8430 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
8431 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
8432 (altivec_lve<VI_char>x): Delete expand.
8433 (*altivec_lve<VI_char>x_internal): Rename to...
8434 (altivec_lve<VI_char>x): ... this.
8435 (altivec_lvxl_<mode>): Delete expand.
8436 (*altivec_lvxl_<mode>_internal): Rename to ...
8437 (altivec_lvxl_<mode>): ... this.
8438 (altivec_stvxl_<mode>): Delete expand.
8439 (*altivec_stvxl_<mode>_internal): Rename to ...
8440 (altivec_stvxl_<mode>): ... this.
8441 (altivec_stve<VI_char>x): Delete expand.
8442 (*altivec_stve<VI_char>x_internal): Rename to ...
8443 (altivec_stve<VI_char>x): ... this.
8444 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
8445 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
8446 reduc_plus_scal_<mode>): Adjust.
8447 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
8449 (rs6000_cpu_cpp_builtins): Adjust.
8450 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
8451 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
8452 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
8453 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
8454 -maltivec=be support.
8455 (rs6000_split_vec_extract_var): Adjust.
8456 (rs6000_split_v4si_init): Adjust.
8457 (swap_selector_for_mode): Delete.
8458 (altivec_expand_lvx_be, altivec_expand_stvx_be,
8459 altivec_expand_stvex_be): Delete.
8460 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
8461 -maltivec=be support.
8462 (rs6000_gimple_fold_builtin): Ditto.
8463 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
8465 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
8466 (TARGET_DIRECT_MOVE_64BIT): Adjust.
8467 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
8468 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
8469 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
8470 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
8471 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
8472 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
8473 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
8474 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
8475 anonymous split): Adjust.
8476 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
8477 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
8479 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
8481 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
8482 when --with-gxx-include-dir is also specified.
8483 * configure: Regenerate.
8485 2018-05-09 Jim Wilson <jimw@sifive.com>
8488 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
8489 * config/riscv/t-withmultilib: New.
8490 * config/riscv/withmultilib.h: New.
8491 * doc/install.texi: Document RISC-V --with-multilib-list support.
8493 2018-05-09 Richard Biener <rguenther@suse.de>
8495 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
8497 (vect_bb_vectorization_profitable_p): Adjust. Compute
8498 actual scalar cost using the cost vector and the add_stmt_cost
8501 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8503 PR rtl-optimization/85645
8504 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
8505 in the REG_CFA_REGISTER note for LR, don't leave it empty.
8507 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8509 PR rtl-optimization/85645
8510 * shrink-wrap.c (spread_components): Return a boolean saying if
8511 anything was changed.
8512 (try_shrink_wrapping_separate): Iterate spread_components until
8513 nothing changes anymore.
8515 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8517 PR rtl-optimization/85645
8518 * regrename.c (build_def_use): Also kill the chains that include the
8519 destination of a REG_CFA_REGISTER note.
8521 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8523 PR rtl-optimization/85645
8524 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
8525 insn that has a REG_CFA_REGISTER note.
8527 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
8529 * cfgexpand.c (expand_clobber): New function.
8530 (expand_gimple_stmt_1): Use it.
8531 * tree-vect-stmts.c (vect_clobber_variable): New function,
8533 (vectorizable_simd_clone_call): ...here.
8534 (vectorizable_store): Emit a clobber either side of an
8535 IFN_STORE_LANES sequence.
8536 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
8538 2018-05-09 Tom de Vries <tom@codesourcery.com>
8541 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
8542 (define_insn "trap_if_false"): Add exit after trap.
8544 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
8546 PR rtl-optimization/85638
8547 * bb-reorder.c: Include common/common-target.h.
8548 (create_forwarder_block): New function extracted from...
8549 (fix_up_crossing_landing_pad): ...here. Rename into...
8550 (dw2_fix_up_crossing_landing_pad): ...this.
8551 (sjlj_fix_up_crossing_landing_pad): New function.
8552 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
8553 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
8554 from both partitions and exit the loop after one iteration.
8556 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
8559 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8561 (Basic PowerPC Built-in Functions): The new name of the
8562 subsection previously known as "PowerPC Built-in Functions".
8563 (Basic PowerPC Built-in Functions Available on all Configurations):
8565 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
8567 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
8569 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
8571 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
8574 2018-05-08 Jim Wilson <jimw@sifive.com>
8576 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
8577 (LD_EMUL_SUFFIX): New.
8578 (LINK_SPEC): Use it.
8580 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
8582 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8584 (Basic PowerPC Built-in Functions): The new name of the
8585 subsection previously known as "PowerPC Built-in Functions".
8586 (Basic PowerPC Built-in Functions Available on all Configurations):
8588 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
8590 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
8592 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
8594 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
8597 2018-05-08 Jakub Jelinek <jakub@redhat.com>
8600 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
8601 after cmpelim optimization.
8603 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
8605 * config.gcc: Support "goldmont".
8606 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
8607 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8609 * config/i386/i386.c (m_GOLDMONT): Define.
8610 (processor_target_table): Add "goldmont".
8611 (PTA_GOLDMONT): Define.
8612 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
8613 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
8614 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
8615 (fold_builtin_cpu): Add "goldmont".
8616 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
8617 (ix86_option_override_internal): Add "goldmont".
8618 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
8619 (processor_type): Add PROCESSOR_GOLDMONT.
8620 * config/i386/i386.md: Add CPU "glm".
8621 * config/i386/glm.md: New file.
8622 * config/i386/x86-tune.def: Add m_GOLDMONT.
8623 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
8625 2018-05-08 Jakub Jelinek <jakub@redhat.com>
8628 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
8630 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
8631 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
8632 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
8635 * config/i386/i386.c (ix86_fold_builtin): Handle
8636 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
8639 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
8640 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
8642 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
8645 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
8646 (check_arch): Likewise.
8647 (check_fpu): Return the result rather than printing it.
8648 (end arch): Fix operator precedence.
8649 (end cpu): Likewise.
8650 (END): Print the result from check_fpu.
8652 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
8653 Alan Hayward <alan.hayward@arm.com>
8654 David Sherwood <david.sherwood@arm.com>
8656 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
8657 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
8658 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
8659 (*fcmuo<mode>_and): New patterns.
8661 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
8663 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
8664 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
8665 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
8666 (cmp_op, sve_imm_con): New code attributes.
8667 (SVE_COND_INT_CMP, imm_con): Delete.
8668 (cmp_op): Remove above unspecs from int attribute.
8669 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
8671 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
8672 comparison-specific unspecs.
8673 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
8674 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
8675 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
8676 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
8677 (*vec_fcm<cmp_op><mode>): Rename to...
8678 (*fcm<cmp_op><mode>): ...this and adjust likewise.
8679 (*vec_fcmuo<mode>): Rename to...
8680 (*fcmuo<mode>): ...this and adjust likewise.
8681 (*pred_fcm<cmp_op><mode>): New pattern.
8682 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
8683 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
8685 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
8687 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
8688 (aarch64_emit_sve_predicated_cond): New function.
8689 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
8690 (aarch64_emit_unspec_cond_or): Replace with...
8691 (aarch64_emit_sve_or_conds): ...this new function. Use
8692 aarch64_emit_sve_ptrue_op for the individual comparisons and
8693 aarch64_emit_binop to OR them together.
8694 (aarch64_emit_inverted_unspec_cond): Replace with...
8695 (aarch64_emit_sve_inverted_cond): ...this new function. Use
8696 aarch64_emit_sve_ptrue_op for the comparison and
8697 aarch64_emit_unop to invert the result.
8698 (aarch64_expand_sve_vec_cmp_float): Update after the above
8699 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
8701 2018-05-07 Nathan Sidwell <nathan@acm.org>
8703 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
8704 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
8705 (Backwards Compatibility): Likewise.
8707 2018-05-07 Luis Machado <luis.machado@linaro.org>
8711 2018-05-07 Luis Machado <luis.machado@linaro.org>
8713 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8714 <prefetch_dynamic_strides>: New const bool field.
8715 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8716 prefetch_dynamic_strides.
8717 (exynosm1_prefetch_tune): Likewise.
8718 (thunderxt88_prefetch_tune): Likewise.
8719 (thunderx_prefetch_tune): Likewise.
8720 (thunderx2t99_prefetch_tune): Likewise.
8721 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
8723 (aarch64_override_options_internal): Update to set
8724 PARAM_PREFETCH_DYNAMIC_STRIDES.
8725 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
8726 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
8727 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
8728 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
8729 prefetch-dynamic-strides setting.
8731 2018-05-07 Luis Machado <luis.machado@linaro.org>
8733 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8734 <minimum_stride>: New const int field.
8735 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8736 minimum_stride field.
8737 (exynosm1_prefetch_tune): Likewise.
8738 (thunderxt88_prefetch_tune): Likewise.
8739 (thunderx_prefetch_tune): Likewise.
8740 (thunderx2t99_prefetch_tune): Likewise.
8741 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
8742 (aarch64_override_options_internal): Update to set
8743 PARAM_PREFETCH_MINIMUM_STRIDE.
8744 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
8745 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
8746 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
8747 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
8748 stride is constant and is below the minimum stride threshold.
8750 2018-05-07 Luis Machado <luis.machado@linaro.org>
8752 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
8755 2018-05-07 Luis Machado <luis.machado@linaro.org>
8757 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8758 <prefetch_dynamic_strides>: New const bool field.
8759 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8760 prefetch_dynamic_strides.
8761 (exynosm1_prefetch_tune): Likewise.
8762 (thunderxt88_prefetch_tune): Likewise.
8763 (thunderx_prefetch_tune): Likewise.
8764 (thunderx2t99_prefetch_tune): Likewise.
8765 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
8767 (aarch64_override_options_internal): Update to set
8768 PARAM_PREFETCH_DYNAMIC_STRIDES.
8769 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
8770 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
8771 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
8772 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
8773 prefetch-dynamic-strides setting.
8775 2018-05-07 Luis Machado <luis.machado@linaro.org>
8777 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8778 <minimum_stride>: New const int field.
8779 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8780 minimum_stride field.
8781 (exynosm1_prefetch_tune): Likewise.
8782 (thunderxt88_prefetch_tune): Likewise.
8783 (thunderx_prefetch_tune): Likewise.
8784 (thunderx2t99_prefetch_tune): Likewise.
8785 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
8786 (aarch64_override_options_internal): Update to set
8787 PARAM_PREFETCH_MINIMUM_STRIDE.
8788 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
8789 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
8790 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
8791 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
8792 stride is constant and is below the minimum stride threshold.
8794 2018-05-06 Jakub Jelinek <jakub@redhat.com>
8797 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
8798 the type is addressable. Don't force op into register if it has
8801 2018-05-05 Roland McGrath <mcgrathr@google.com>
8804 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
8805 any section for which we don't know a specific type it should have,
8806 regardless of name. Previously this was done only for the exact
8807 names ".init_array", ".fini_array", and ".preinit_array".
8808 (default_elf_asm_named_section): Add comment about
8809 relationship with default_section_type_flags and SECTION_NOTYPE.
8810 (get_section): Don't consider it a type conflict if one side has
8811 SECTION_NOTYPE and the other doesn't, as long as neither has the
8812 SECTION_BSS et al used in the default_section_type_flags logic.
8814 2018-05-05 Tom de Vries <tom@codesourcery.com>
8817 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
8818 (workaround_barsyncs): New function.
8819 (nvptx_reorg): Use workaround_barsyncs.
8820 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
8821 (define_expand "nvptx_membar_cta"): New define_expand.
8822 (define_insn "*nvptx_membar_cta"): New insn.
8824 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
8826 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
8827 To improve optimization opportunities.
8828 * builtin-types.def: The new needed builtin types for the above.
8830 2018-05-04 Richard Biener <rguenther@suse.de>
8832 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
8833 * gimple-ssa-store-merging.c
8834 (imm_store_chain_info::output_merged_store): Remove redundant create,
8835 release split_store vector contents on failure.
8836 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
8837 scalar stmt vector on cache hit.
8839 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
8841 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
8843 * config.gcc (powerpc-xilinx-eabi*): Remove.
8844 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
8846 (fusion_addis_mem_combo_load): Ditto.
8847 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
8849 (rs6000_cpu_cpp_builtins): Ditto.
8850 * config/rs6000/rs6000-linux.c
8851 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
8852 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
8853 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
8855 (rs6000_setup_reg_addr_masks): Ditto.
8856 (rs6000_init_hard_regno_mode_ok): Ditto.
8857 (rs6000_option_override_internal): Ditto.
8858 (legitimate_lo_sum_address_p): Ditto.
8859 (rs6000_legitimize_address): Ditto.
8860 (rs6000_legitimize_reload_address): Ditto.
8861 (rs6000_legitimate_address_p): Ditto.
8862 (abi_v4_pass_in_fpr): Ditto.
8863 (setup_incoming_varargs): Ditto.
8864 (rs6000_gimplify_va_arg): Ditto.
8865 (rs6000_split_multireg_move): Ditto.
8866 (rs6000_savres_strategy): Ditto.
8867 (rs6000_emit_prologue_components): Ditto.
8868 (rs6000_emit_epilogue_components): Ditto.
8869 (rs6000_emit_prologue): Ditto.
8870 (rs6000_emit_epilogue): Ditto.
8871 (rs6000_elf_file_end): Ditto.
8872 (rs6000_function_value): Ditto.
8873 (rs6000_libcall_value): Ditto.
8874 * config/rs6000/rs6000.h: Ditto.
8875 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
8876 (TARGET_MINMAX): ... this. New.
8877 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
8878 * config/rs6000/rs6000.md: Remove Xilinx FP support.
8879 (*movsi_internal1_single): Delete.
8880 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
8881 mfpu=, mxilinx-fpu): Delete.
8882 * config/rs6000/singlefp.h: Delete.
8883 * config/rs6000/sysv4.h: Remove Xilinx FP support.
8884 * config/rs6000/t-rs6000: Ditto.
8885 * config/rs6000/t-xilinx: Delete.
8886 * config/rs6000/titan.md: Adjust for fp_type removal.
8887 * config/rs6000/vsx.md: Remove Xilinx FP support.
8888 (VStype_simple): Delete.
8889 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
8890 * config/rs6000/xfpu.h: Delete.
8891 * config/rs6000/xfpu.md: Delete.
8892 * config/rs6000/xilinx.h: Delete.
8893 * config/rs6000/xilinx.opt: Delete.
8894 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
8895 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
8897 2018-05-04 Tom de Vries <tom@codesourcery.com>
8900 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
8903 2018-05-04 Richard Biener <rguenther@suse.de>
8906 * tree-complex.c (update_complex_assignment): We are always in SSA form.
8907 (expand_complex_div_wide): Likewise.
8908 (expand_complex_operations_1): Likewise.
8909 (expand_complex_libcall): Preserve EH info of the original stmt.
8910 (tree_lower_complex): Handle removed blocks.
8911 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
8912 on complex multiplication and division libcall builtins.
8914 2018-05-04 Richard Biener <rguenther@suse.de>
8917 * fold-const.c (negate_expr_p): Restrict negation of operand
8918 zero of a division to when we know that can happen without
8920 (fold_negate_expr_1): Likewise.
8922 2018-05-04 Jakub Jelinek <jakub@redhat.com>
8925 * real.h (real_nextafter): Declare.
8926 * real.c (real_nextafter): New function.
8927 * fold-const-call.c (fold_const_nextafter): New function.
8928 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
8929 CASE_CFN_NEXTTOWARD.
8930 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
8931 even when arg1_mode is different from arg0_mode.
8933 2018-05-03 Nathan Sidwell <nathan@acm.org>
8935 * doc/extend.texi (Deprecated Features): Remove
8937 (Backwards Compatibility): Likewise.
8938 * doc/invoke.texi (C++ Language Options): Likewise.
8939 (C++ Dialect Options): Likewise.
8941 2018-05-03 Jakub Jelinek <jakub@redhat.com>
8944 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
8945 _mm512_mask_mullox_epi64): New intrinsics.
8947 2018-05-03 Tom de Vries <tom@codesourcery.com>
8950 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8951 dump files): Add offload-tree.
8953 2018-05-03 Richard Biener <rguenther@suse.de>
8955 PR tree-optimization/85615
8956 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
8957 to loops not nested in BBs loop father to avoid creating multi-entry
8960 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8962 PR tree-optimization/70291
8963 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
8964 arguments. Change return type to tree. Emit libcall as a new
8965 statement rather than replacing existing one when inplace_p is true.
8966 (expand_complex_multiplication_components): New function.
8967 (expand_complex_multiplication): Expand floating-point complex
8968 multiplication using the above.
8969 (expand_complex_division): Rename inner_type parameter to type.
8970 Update expand_complex_libcall call-site.
8971 (expand_complex_operations_1): Update expand_complex_multiplication
8972 and expand_complex_division call-sites.
8974 2018-05-02 Jakub Jelinek <jakub@redhat.com>
8977 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
8978 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
8979 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
8980 the highest significant bit of the shift count mask is clear. In
8981 check whether and[sq]i3 is needed verify that all significant bits
8982 of the shift count other than the highest are set.
8984 2018-05-02 Tom de Vries <tom@codesourcery.com>
8987 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
8988 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
8989 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
8990 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
8991 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
8992 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
8993 __builtin_goacc_parlevel_size.
8995 2018-05-02 Richard Biener <rguenther@suse.de>
8997 PR tree-optimization/85597
8998 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
8999 do not use split vect_get_vec_defs call but call vect_get_slp_defs
9002 2018-05-02 Tom de Vries <tom@codesourcery.com>
9005 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
9006 dump files): Add ltrans-tree.
9008 2018-05-02 Tom de Vries <tom@codesourcery.com>
9011 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
9012 dump files): Add wpa-ipa.
9014 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
9016 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
9017 powerpc*-*-linux*paired* target.
9018 * config/rs6000/750cl.h: Delete.
9019 * config/rs6000/paired.h: Delete.
9020 * config/rs6000/paired.md: Delete.
9021 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
9023 * config/rs6000/rs6000-builtin.def: Remove paired float support.
9024 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
9025 comment. Remove paired float support.
9026 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
9027 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
9029 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
9030 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
9032 * config/rs6000/rs6000.c: Remove paired float support.
9033 (paired_expand_vector_init, paired_expand_vector_move,
9034 paired_emit_vector_compare, paired_emit_vector_cond_expr,
9035 (paired_expand_lv_builtin, paired_expand_stv_builtin,
9036 paired_expand_builtin, paired_expand_predicate_builtin,
9037 paired_init_builtins): Delete.
9038 * config/rs6000/rs6000.h: Remove paired float support.
9039 * config/rs6000/rs6000.md: Remove paired float support.
9040 (move_from_CR_ov_bit): Delete.
9041 * config/rs6000/rs6000.opt (mpaired): Delete.
9042 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
9043 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
9045 2018-05-02 Richard Biener <rguenther@suse.de>
9048 * gimplify.c (gimplify_save_expr): When in SSA form allow
9049 SAVE_EXPRs to compute to SSA vars.
9051 2018-05-02 Jakub Jelinek <jakub@redhat.com>
9054 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
9055 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
9056 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
9057 clobber operands[2], instead use a new pseudo. Formatting fixes.
9059 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
9061 PR tree-optimization/85586
9062 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
9063 exit early for statements in the same group if the accesses are
9066 2018-05-02 Tom de Vries <tom@codesourcery.com>
9069 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
9072 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
9074 PR tree-optimization/85143
9075 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
9077 2018-05-01 Tom de Vries <tom@codesourcery.com>
9080 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
9081 not found" error message.
9083 2018-05-01 Tom de Vries <tom@codesourcery.com>
9086 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
9087 * vec.c (test_ordered_remove_if): New function.
9088 (vec_c_tests): Call test_ordered_remove_if.
9089 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
9090 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
9091 * tree-vect-patterns.c (vect_pattern_recog_1): Use
9092 VEC_ORDERED_REMOVE_IF.
9094 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9096 PR tree-optimization/82665
9097 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
9098 pointer subtraction where arguments come from a memchr call.
9100 2018-05-01 Jakub Jelinek <jakub@redhat.com>
9102 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
9103 --push-state --as-needed and --pop-state instead of --as-needed and
9104 --no-as-needed if ld supports it.
9105 * configure: Regenerated.
9108 * doc/install.texi2html: Replace _002d with - and _002a with * in
9109 generated html files using sed.
9111 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9114 * gcc-rich-location.c (blank_line_before_p): New function.
9115 (use_new_line): New function.
9116 (gcc_rich_location::add_fixit_insert_formatted): New function.
9117 * gcc-rich-location.h
9118 (gcc_rich_location::add_fixit_insert_formatted): New function.
9120 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9122 * selftest.c (assert_streq): Rename "expected" and "actual" to
9123 "val1" and "val2". Extend NULL-handling to cover both inputs
9124 symmetrically, while still requiring both to be non-NULL for a pass.
9125 * selftest.h (assert_streq): Rename "expected" and "actual" to
9127 (ASSERT_EQ): Likewise.
9128 (ASSERT_EQ_AT): Likewise.
9129 (ASSERT_KNOWN_EQ): Likewise.
9130 (ASSERT_KNOWN_EQ_AT): Likewise.
9131 (ASSERT_NE): Likewise.
9132 (ASSERT_MAYBE_NE): Likewise.
9133 (ASSERT_MAYBE_NE_AT): Likewise.
9134 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
9135 the assertion to pass.
9136 (ASSERT_STREQ_AT): Likewise.
9138 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
9140 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
9141 interaction with -pie.
9143 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9145 * selftest.h: Fix alphabetization of per-source-file selftest
9148 2018-04-30 Jason Merrill <jason@redhat.com>
9150 PR c++/61982 - dead stores to destroyed objects.
9151 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
9154 2018-04-30 Jason Merrill <jason@redhat.com>
9156 * tree.c (build_clobber): New.
9157 * tree.h: Declare it.
9158 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
9160 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9162 * diagnostic-show-locus.c (layout::layout): Update for
9163 location_get_source_line returning a char_span.
9164 (struct char_span): Move to input.h.
9165 (struct correction): Update for fields in char_span becoming
9167 (struct source_line): Update for location_get_source_line
9168 returning a char_span.
9169 (layout::print_line): Likewise.
9170 * edit-context.c (edited_file::print_content): Likewise.
9171 (edited_file::print_diff_hunk): Likewise.
9172 (edited_file::print_run_of_changed_lines): Likewise.
9173 (edited_file::get_num_lines): Likewise.
9174 (edited_line::edited_line): Likewise.
9175 * final.c (asm_show_source): Likewise.
9176 * input.c (location_get_source_line): Convert return type
9177 from const char * to char_span, losing the final "line_len"
9179 (dump_location_info): Update for the above.
9180 (get_substring_ranges_for_loc): Likewise. Use a char_span
9181 when handling the literal within the line.
9182 (test_reading_source_line): Update for location_get_source_line
9183 returning a char_span.
9184 * input.h (class char_span): Move here from
9185 diagnostic-show-locus.c, converting from a struct to a class.
9186 Make data members private.
9187 (char_span::operator bool): New.
9188 (char_span::length): New.
9189 (char_span::get_buffer): New.
9190 (char_span::operator[]): New.
9191 (char_span::subspan): Make const.
9192 (char_span::xstrdup): New.
9193 (location_get_source_line): Convert return type from const char *
9194 to char_span, losing the final "line_size" param.
9196 2018-04-30 Jan Hubicka <jh@suse.cz>
9198 * lto-wrapper.c (ltrans_priorities): New static var.
9199 (cmp_priority): New.
9200 (run_gcc): Read priorities and if doing parallel build order
9201 the Makefile by them.
9203 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9205 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
9207 2018-04-30 Richard Biener <rguenther@suse.de>
9209 * tree-cfg.c (verify_address): Remove base argument, add
9210 flag whether to check TREE_ADDRESSABLE and do that.
9211 (verify_expr): Remove.
9212 (verify_types_in_gimple_reference): Add pieces from verify_expr.
9213 (verify_gimple_assign_single): Likewise.
9214 (verify_gimple_switch): Likewise.
9215 (verify_expr_location_1): Dereference tp once. Add (disabled)
9216 piece from verify_expr.
9217 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
9219 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9221 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
9223 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9225 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
9226 (small_data_pattern): Likewise.
9227 (arc_rewrite_small_data): Likewise.
9228 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
9229 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
9230 (get_symbol_alignment): New function.
9231 (legitimate_small_data_address_p): Likewise.
9232 (legitimate_scaled_address): Update, call
9233 legitimate_small_data_address_p.
9234 (output_sdata): New static variable.
9235 (arc_print_operand): Update how we handle small data operands.
9236 (arc_print_operand_address): Likewise.
9237 (arc_legitimate_address_p): Update, use
9238 legitimate_small_data_address_p.
9239 (arc_rewrite_small_data_p): Remove.
9240 (arc_rewrite_small_data_1): Likewise.
9241 (arc_rewrite_small_data): Likewise.
9242 (small_data_pattern): Likewise.
9243 (compact_sda_memory_operand): Update to use
9244 legitimate_small_data_address_p and get_symbol_alignment.
9245 (prepare_move_operands): Don't rewite sdata pattern.
9246 (prepare_extend_operands): Remove.
9247 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
9249 (zero_extendqisi2): Likewise.
9250 (zero_extendhisi2): Likewise.
9251 (extendqihi2): Likewise.
9252 (extendqisi2): Likewise.
9253 (extendhisi2): Likewise.
9257 * config/arc/constraints.md (Usd): Change it to memory constraint.
9259 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9261 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
9262 as source of std instructions.
9263 * config/arc/arc.md (movsi_insn): Update pattern predicate to
9264 allow 6-bit constants as source for store instructions.
9265 (movdi_insn): Update instruction pattern to allow 6-bit constants
9266 as source for store instructions.
9268 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
9270 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
9272 2018-04-30 Nathan Sidwell <nathan@acm.org>
9273 Sandra Loosemore <sandra@codesourcery.com>
9275 * dumpfile.c (dump_open): Allow '-' for stdout.
9276 * doc/invoke.texi (Developer Options): Document dump filename
9277 determination early. Document stdin/stdout selection.
9279 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
9281 Microblaze Target: PIC data text relative
9283 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
9284 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
9286 * config/microblaze/microblaze.h (microblaze_constant_address_p):
9287 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
9288 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
9289 New addressing mode for data-text relative position indepenedent code.
9290 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
9291 'ADDRESS_SYMBOLIC_TXT_REL'.
9292 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
9293 (microblaze_legitimate_pic_operand): Exclude function calls from
9294 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
9295 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
9297 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
9298 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
9299 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
9300 for 'address + offset'.
9301 (microblaze_expand_prologue): Add new function prologue call for
9303 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
9304 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
9305 table in case of TARGET_PIC_DATA_TEXT_REL.
9306 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
9307 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
9308 Add new macros 'UNSPEC_TEXT',
9309 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
9310 + exclude function calls from 'UNSPEC_PLT' in case of data text
9312 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
9313 new target hook for generating address diff vector tables in case of
9315 * doc/tm.texi : Regenerate.
9316 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
9317 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
9318 of addr diff vector generation.
9319 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
9320 target hook definition.
9321 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
9322 Add default function for generate_pic_addr_diff_vec -> flag_pic.
9323 * doc/invoke.texi (Add new pic option): Add new microblaze pic
9324 option for data text relative.
9326 2018-04-30 Richard Biener <rguenther@suse.de>
9328 * tree-chrec.h (evolution_function_is_constant_p): Remove
9330 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
9332 2018-04-30 Richard Biener <rguenther@suse.de>
9335 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
9337 2018-04-30 Richard Biener <rguenther@suse.de>
9339 PR tree-optimization/28364
9340 PR tree-optimization/85275
9341 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
9342 copying first exit test.
9344 2018-04-28 Mark Wielaard <mark@klomp.org>
9346 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
9348 (dwarf_AT): Handle DW_AT_addr_base.
9349 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
9351 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
9354 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
9355 (*ashl<dwi>3_doubleword_mask_1): Ditto.
9356 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
9357 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
9359 2018-04-28 Richard Biener <rguenther@suse.de>
9361 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
9362 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
9363 to reflect use. Only add interesting stmts.
9365 2018-04-27 Martin Jambor <mjambor@suse.cz>
9368 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
9369 the jump function allows for passing through aggregate values.
9371 2018-04-27 David Malcolm <dmalcolm@redhat.com>
9373 * input.h (in_system_header_at): Convert from macro to inline
9375 (from_macro_expansion_at): Likewise.
9376 (from_macro_definition_at): Likewise.
9378 2018-04-27 Jeff Law <law@redhat.com>
9380 * config.gcc: Mark tile* targets as deprecated/obsolete.
9382 2018-04-27 Richard Biener <rguenther@suse.de>
9384 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
9387 2018-04-27 Richard Biener <rguenther@suse.de>
9389 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
9391 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
9393 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
9394 with Yd constraint. Set "preferred_for_speed" attribute from
9395 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
9397 (*movdi_internal): Ditto.
9398 (movti_interunit splitters): Remove
9399 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
9400 (movdi_interunit splitters): Ditto.
9401 * config/i386/constraints.md (Ye): Remove.
9402 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
9404 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9407 * config/aarch64/constraints.md (Usg): Limit to 31.
9410 2018-04-27 Jakub Jelinek <jakub@redhat.com>
9412 PR tree-optimization/85529
9413 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
9414 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
9415 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
9416 zero extension or masking of the MSB bit.
9417 (optimize_range_tests): Add FIRST_BB argument, pass it through
9418 to optimize_range_tests_var_bound.
9419 (maybe_optimize_range_tests, reassociate_bb): Adjust
9420 optimize_range_tests callers.
9422 2018-04-26 Richard Biener <rguenther@suse.de>
9423 Jakub Jelinek <jakub@redhat.com>
9425 * cgraph.h (symbol_table): Just declare debug method here.
9426 * symtab.c (symbol_table::debug): Define.
9428 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
9430 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
9432 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
9434 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
9435 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
9436 (*movdi_internal): Substitute Yi and Yj constraint with x
9437 and Ym and Yn constraint with y constraint. Update "isa"
9438 attribute and set "preferred_for_speed" attribute from
9439 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
9440 (*movsi_internal): Ditto.
9441 (*movdf_internal): Ditto.
9442 (*movsf_internal): Ditto.
9443 (*zero_extendsidi2): Ditto.
9444 * config/i386/sse.md (vec_set<mode>_0): Ditto.
9445 (sse2_loadld): Ditto.
9446 (*vec_extract<ssevecmodelower>_0): Ditto.
9447 (*vec_extractv4si_0_zext_sse4): Ditto.
9448 (vec_concatv2di): Ditto.
9449 (*vec_dup<mode>): Ditto.
9450 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
9451 * config/i386/constraints.md (Yi): Remove.
9456 2018-04-26 Nathan Sidwell <nathan@acm.org>
9458 * dumpfile.c (dump_open): New.
9459 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
9460 (dump_finish): Detect stdio/stderr by value not name.
9462 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
9464 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
9466 2018-04-26 Tom de Vries <tom@codesourcery.com>
9469 * config/nvptx/nvptx.c (verify_neutering_jumps)
9470 (verify_neutering_labels): New function
9471 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
9473 2018-04-26 Tom de Vries <tom@codesourcery.com>
9476 * config/nvptx/nvptx.c (needs_neutering_p): New function.
9477 (nvptx_single): Use needs_neutering_p to skip over insns that do not
9480 2018-04-26 Richard Biener <rguenther@suse.de>
9481 Tom de Vries <tom@codesourcery.com>
9484 * lto-streamer-out.c (output_function): Fixup loops if required to match
9485 discovery done in the reader.
9487 2018-04-26 Richard Biener <rguenther@suse.de>
9489 PR tree-optimization/85116
9490 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
9491 have a loop exit from the single latch predecessor. Remove
9492 case of header with just condition.
9493 (ch_base::copy_headers): Exclude infinite loops from any
9495 (pass_ch::execute): Record exits.
9497 2018-04-26 Richard Biener <rguenther@suse.de>
9499 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
9500 prologue cost vector and pass it to vect_get_load_cost.
9501 (vect_get_peeling_costs_all_drs): Likewise.
9502 (vect_peeling_hash_get_lowest_cost): Likewise.
9503 (vect_enhance_data_refs_alignment): Likewise.
9505 2018-04-26 Richard Biener <rguenther@suse.de>
9508 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
9509 checking of integer<->pointer conversions.
9510 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
9511 sign-/zero-extending pointer types.
9512 (expand_omp_for_static_chunk): Likewise.
9514 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
9515 Jean Lee <xiaoyur347@gmail.com>
9517 * config/mips/mips.c (mips_asan_shadow_offset): New function.
9518 (TARGET_ASAN_SHADOW_OFFSET): Define.
9519 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
9520 true for -fsanitize=address.
9522 2018-04-25 Mark Wielaard <mark@klomp.org>
9524 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
9527 2018-04-25 Jakub Jelinek <jakub@redhat.com>
9529 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
9530 than "alu", remove explicit "memory" and "imm_disp" attributes.
9531 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
9534 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
9535 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
9536 gen_lowpart_no_emit.
9538 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
9541 * config/i386/i386.c (ix86_expand_builtin): Change memory
9542 operand to XI, extend p0 to Pmode.
9543 * config/i386/i386.md: Change unspec volatile and operand
9544 1 mode to XI, change operand 0 mode to P.
9546 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9548 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
9549 GET_MODE_MASK before any checking.
9550 (nds32_can_use_bset_p): Likewise.
9551 (nds32_can_use_btgl_p): Likewise.
9553 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9555 * config/nds32/nds32-doubleword.md: New define_split pattern for
9556 illegal register number.
9558 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9560 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
9562 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9564 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
9566 2018-04-25 Richard Biener <rguenther@suse.de>
9568 * lto-streamer.h (LTO_major_version): Bump to 8.
9570 2018-04-25 Jakub Jelinek <jakub@redhat.com>
9572 * BASE-VER: Set to 9.0.0.
9574 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
9576 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
9577 in __abskf2 and __powikf2.
9579 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9582 * config/aarch64/constraints.md (Usg, Usj): New constraints.
9583 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
9584 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
9585 Use the above on operand 2. Reindent.
9586 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
9588 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
9591 * common/config/i386/i386-common.c (ix86_handle_option): Don't
9593 * config/i386/i386.opt (mcet): Removed.
9594 * doc/install.texi: Remove -mcet documentation.
9595 * doc/invoke.texi: Likewise.
9597 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
9600 * doc/install.texi: Remove -mcet from bootstrap-cet.
9602 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9605 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
9606 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
9610 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
9611 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
9612 containing a CONST_VECTOR.
9614 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
9616 * doc/install.texi: Update newlib dependency for nvptx.
9618 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9621 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
9622 instead of INTVAL when shifting x left.
9624 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
9626 PR tree-optimization/85478
9627 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
9628 vect_grouped_store_supported for single element vectors.
9630 2018-04-24 Richard Biener <rguenther@suse.de>
9633 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
9634 load cost increase to the case of non-constant step.
9636 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9639 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
9640 destination if any_malformed_asm.
9642 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
9645 * expr.c (store_field): In the bitfield case, if the value comes from
9646 a function call and is returned in registers by means of a PARALLEL,
9647 do not change the mode of the temporary unless BLKmode and VOIDmode.
9649 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
9651 PR rtl-optimization/85423
9652 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
9653 dependencies to debug insns when the previous insn is non-debug.
9655 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
9657 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
9658 enums into a single definition.
9659 (fls): Fix predicates and printing.
9662 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
9664 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
9665 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
9666 and short u6 immediate.
9667 (check_if_valid_sleep_operand): Remove.
9668 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
9670 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9672 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
9673 flag_always_save_lp condition.
9674 * config/nds32/nds32.opt (malways-save-lp): New option.
9676 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9678 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
9679 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
9680 * config/nds32/nds32.h
9681 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
9682 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
9684 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9686 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
9687 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
9689 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9690 Chung-Ju Wu <jasonwucj@gmail.com>
9692 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
9694 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
9695 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
9697 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9699 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
9701 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9703 * config/nds32/nds32-protos.h (nds32_data_alignment,
9704 nds32_local_alignment): Declare.
9705 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
9706 nds32_local_alignment): New functions.
9707 (TARGET_CONSTANT_ALIGNMENT): Define.
9708 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
9710 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9712 * config/nds32/nds32.c
9713 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
9714 (TARGET_MODES_TIEABLE_P): Likewise.
9716 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9718 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
9721 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
9722 Chung-Ju Wu <jasonwucj@gmail.com>
9724 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9725 for unaligned access.
9726 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9728 * config/nds32/nds32-intrinsic.md: Likewise.
9729 * config/nds32/nds32_intrinsic.h: Likewise.
9730 * config/nds32/nds32.h (nds32_builtins): Likewise.
9731 * config/nds32/nds32.opt (munaligned-access): New option.
9732 * config/nds32/nds32.c (nds32_asm_file_start): Display
9733 flag_unaligned_access status.
9735 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
9737 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
9738 -mno-relax is present.
9739 * config/riscv/linux.h (LINK_SPEC): Ditto.
9741 2018-04-20 Martin Sebor <msebor@redhat.com>
9744 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
9746 (gimple_fold_builtin_stxcpy_chk): Same.
9747 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
9749 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
9752 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
9753 __powikf2 when long double is IEEE 128-bit.
9755 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
9757 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
9758 step to make sure stack always aligned.
9760 2018-04-20 Carl Love <cel@us.ibm.com>
9763 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
9764 size check for arg0.
9766 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
9767 Tom de Vries <tom@codesourcery.com>
9770 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
9771 Emit insns for calls too.
9772 (nvptx_find_par): Always look for worker-level predecessor insn.
9773 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
9775 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
9776 (nvptx_process_pars): Propagate frames for calls.
9778 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
9781 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
9783 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
9784 (ix86_handle_option): Don't handle OPT_mibt.
9785 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
9787 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
9789 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
9790 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
9791 (ix86_target_macros): Define __CET__ with flag_cf_protection
9792 for -fcf-protection.
9793 * config/i386/i386.c (isa2_opts): Remove -mibt.
9794 * config/i386/i386.h (TARGET_IBT): Removed.
9795 (TARGET_IBT_P): Likewise.
9796 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
9797 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
9798 * config/i386/i386.opt (mcet): Update help message.
9801 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
9802 -mcet as an alias for -mshstk.
9804 2018-04-20 Richard Biener <rguenther@suse.de>
9807 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
9808 complexity by forcing a single use of the multiply operand.
9810 2018-04-20 Martin Jambor <mjambor@suse.cz>
9813 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
9814 recursion dependency to only apply to non-clones.
9816 2018-04-20 Martin Jambor <mjambor@suse.cz>
9819 * ipa-cp.c (create_specialized_node): Check that clones of
9820 self-recursive edges exist during IPA-CP.
9822 2018-04-19 Toon Moene <toon@moene.org>
9824 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
9827 2018-04-19 Jakub Jelinek <jakub@redhat.com>
9829 PR tree-optimization/85467
9830 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
9831 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
9832 VECTOR_CST element to type.
9834 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9837 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
9838 * config/i386/i386.md (builtin_setjmp_setup): Removed.
9839 (builtin_longjmp): Likewise.
9840 (save_stack_nonlocal): New pattern.
9841 (restore_stack_nonlocal): Likewise.
9843 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9846 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
9847 Replace ASM_OUTPUT_LABEL with fprintf.
9849 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9852 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
9853 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
9854 * config/i386/i386-c.c (ix86_target_macros_internal): Also
9855 define __IBT__ and __SHSTK__ for -fcf-protection.
9856 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
9858 (ix86_trampoline_init): Likewise.
9859 (x86_output_mi_thunk): Likewise.
9860 (ix86_notrack_prefixed_insn_p): Likewise.
9861 (ix86_option_override_internal): Don't disallow -fcf-protection.
9862 * config/i386/i386.md (rdssp<mode>): Also enable for
9864 (incssp<mode>): Likewise.
9865 (nop_endbr): Likewise.
9866 * config/i386/i386.opt (mcet): Change help message to built-in
9870 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
9871 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
9872 enable CET built-in functions.
9874 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
9876 * common/config/i386/i386-common.c
9877 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
9878 OPTION_MASK_ISA_MOVDIRI_UNSET,
9879 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
9880 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
9881 * config.gcc (movdirintrin.h): New header.
9882 * config/i386/cpuid.h (bit_MOVDIRI,
9883 bit_MOVDIR64B): New bits.
9884 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
9886 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
9887 (VOID, PVOID, PCVOID)): New function types.
9888 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
9889 __builtin_ia32_directstoreu_u64,
9890 __builtin_ia32_movdir64b): New builtins.
9891 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
9892 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
9894 (ix86_valid_target_attribute_inner_p): Ditto.
9895 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
9896 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
9897 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
9898 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
9899 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
9900 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
9901 (movdiri<mode>, movdir64b_<mode>): New.
9902 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
9903 * config/i386/immintrin.h: Include movdirintrin.h.
9904 * config/i386/movdirintrin.h: New file.
9905 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
9907 2018-04-19 Richard Biener <rguenther@suse.de>
9910 * cfg.c (clear_bb_flags): When loop state says we have
9911 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
9913 2018-04-19 Richard Biener <rguenther@suse.de>
9915 PR tree-optimization/84737
9916 * tree-vect-data-refs.c (vect_copy_ref_info): New function
9917 copying restrict info.
9918 (vect_setup_realignment): Use it.
9919 * tree-vectorizer.h (vect_copy_ref_info): Declare.
9920 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
9921 the first DR to all generated stores.
9922 (vectorizable_load): Likewise for loads.
9924 2018-04-19 Jakub Jelinek <jakub@redhat.com>
9926 PR tree-optimization/85446
9927 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
9928 the integral and pointer types to have the same precision.
9930 * doc/install.texi: Document --disable-cet being the default and
9933 2018-04-18 Martin Liska <mliska@suse.cz>
9935 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
9938 2018-04-18 Martin Liska <mliska@suse.cz>
9941 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9944 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9945 arguments if they are comparable.
9947 2018-04-18 Martin Liska <mliska@suse.cz>
9950 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9953 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9956 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
9959 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
9960 ENDBR after calling __morestack.
9962 2018-04-18 David Malcolm <dmalcolm@redhat.com>
9965 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
9966 by using gcc_base_ver to generate a gcc_driver_version, and use
9967 it when generating GCC_DRIVER_NAME.
9968 * configure: Regenerate.
9970 2018-04-18 Jakub Jelinek <jakub@redhat.com>
9973 * config.gcc: Obsolete powerpc*-*-*spe*.
9975 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9978 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
9979 (stabstr_D): Change type of unum from unsigned int to
9980 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
9983 2018-04-17 Jim Wilson <jimw@sifive.com>
9986 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
9987 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
9988 Set arg_pointer_offset after using pretend_args_size.
9990 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9992 PR rtl-optimization/85431
9993 * dse.c (record_store): Ignore zero width stores.
9996 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
9997 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
9998 __builtin_stack_restore rather than after it.
9999 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
10000 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
10001 argument instead of virtual_dynamic_stack_rtx.
10003 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
10005 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
10007 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
10008 Add note to error message to explain internal mapping of overloaded
10009 built-in function name to non-overloaded built-in function name.
10010 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
10013 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10016 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
10017 where the inputs overlap with the output.
10019 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10021 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
10022 (=v, v) alternative and explicit "memory" attribute.
10023 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
10024 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10026 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
10027 "sselog1" type instead of "sselog".
10028 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
10029 "sselog". Remove explicit "memory" attribute.
10030 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
10031 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10033 (vec_extract_hi_v32hi): Merge all alternatives into one, use
10034 "sselog1" type instead of "sselog". Remove explicit "memory"
10036 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
10037 use "sselog1" type instead of "sselog". Remove explicit "memory"
10039 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
10040 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10042 (vec_extract_hi_v64qi): Merge all alternatives into one, use
10043 "sselog1" type instead of "sselog". Remove explicit "memory"
10045 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
10046 use "sselog1" type instead of "sselog". Remove explicit "memory"
10050 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
10052 PR middle-end/85414
10053 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
10056 2018-04-17 Martin Jambor <mjambor@suse.cz>
10059 * ipa-cp.c (create_specialized_node): Call
10060 expand_all_artificial_thunks if necessary.
10062 2018-04-17 Martin Liska <mliska@suse.cz>
10065 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
10066 in message, remote space in between '_G' and '('.
10068 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10071 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
10072 avx512f_vmcmp<mode>3<round_saeonly_name>,
10073 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
10074 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
10075 avx512f_rndscale<mode><round_saeonly_name>,
10076 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
10077 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
10078 Use %<iptr>2 instead of %2 for -masm=intel.
10079 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
10080 avx512f_vcvttss2usi<round_saeonly_name>,
10081 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
10083 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
10084 avx512f_vcvttsd2usi<round_saeonly_name>,
10085 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
10086 Use %q1 instead of %1 for -masm=intel.
10087 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
10088 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
10089 of %3 for -masm=intel.
10090 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
10092 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
10094 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
10096 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
10097 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
10099 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
10101 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
10102 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
10103 %g1 and one with %0 and %1.
10104 (avx512er_vmrcp28<mode><round_saeonly_name>,
10105 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
10106 %1 for -masm=intel.
10107 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
10108 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
10109 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
10110 of %0 and %{%4%} for -masm=intel.
10111 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
10112 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
10113 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
10114 order of %0 and %{%5%}%{z%} for -masm=intel.
10116 2018-04-17 Jan Hubicka <jh@suse.cz>
10119 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
10121 2018-04-17 Martin Liska <mliska@suse.cz>
10124 * multiple_target.c (create_dispatcher_calls): Set apostrophes
10125 for target_clone error message. Make default implementation
10126 clone to be a local declaration.
10127 (separate_attrs): Add new argument and check for an empty
10129 (expand_target_clones): Handle it.
10130 (ipa_target_clone): Make redirection just for target_clones
10133 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
10134 Tom de Vries <tom@codesourcery.com>
10136 PR middle-end/84955
10137 * omp-expand.c (expand_oacc_for): Add dummy false branch for
10138 tiled basic blocks without omp continue statements.
10140 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
10143 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
10144 vec_extract expression as having side effects to make sure it gets
10147 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
10150 * config/i386/i386.c (get_builtin_code_for_version): Check
10153 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
10156 * config.gcc: Support "skylake".
10157 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10159 * config/i386/i386.c (m_SKYLAKE): Define.
10160 (processor_target_table): Add "skylake".
10161 (ix86_option_override_internal): Add "skylake".
10162 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
10163 PROCESSOR_CANNONLAKE.
10164 (get_builtin_code_for_version): Fix priority for
10165 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
10166 PROCESSOR_SKYLAKE-AVX512.
10167 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
10168 (processor_type): Add PROCESSOR_SKYLAKE.
10170 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
10171 Jason Merrill <jason@redhat.com>
10174 * convert.c (convert_to_integer_1): Use direct recursion for
10175 enumeral types and types with a precision less than the number
10176 of bits in their mode.
10178 2018-04-16 Julia Koval <julia.koval@intel.com>
10181 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
10182 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
10184 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
10187 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
10188 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
10189 and -mno-direct-move.
10191 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
10194 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
10195 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
10196 Ensure negative shifts result in {0}.
10198 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
10200 PR rtl-optimization/79916
10201 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
10202 regs (if any) to define how to gnerate SD moves when LRA is in
10205 2018-04-13 Jakub Jelinek <jakub@redhat.com>
10207 PR rtl-optimization/85393
10208 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
10209 * except.c (expand_dw2_landing_pad_for_region): Make static.
10210 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
10211 a label and unconditional jump to old_bb, rather than
10212 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
10215 PR rtl-optimization/85376
10216 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
10217 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
10218 instead of a specific value.
10220 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
10221 Bin Cheng <bin.cheng@arm.com>
10223 PR tree-optimization/82965
10224 PR tree-optimization/83991
10225 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
10226 by_profile_only parameter.
10227 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
10228 information if the loop was predicted to iterate too many times.
10229 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
10231 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
10234 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
10237 2018-04-13 Martin Liska <mliska@suse.cz>
10238 Jakub Jelinek <jakub@redhat.com>
10240 PR middle-end/81657
10241 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
10242 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
10243 * builtins.c (expand_builtin_memory_copy_args): Use
10244 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
10245 handle dest_addr == pc_rtx.
10247 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
10250 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
10251 asked to not generate direct moves.
10252 (fix_trunc<mode>si2_stfiwx): Similar.
10253 (fix_trunc<mode>si2_internal): Similar.
10255 2018-04-12 Jakub Jelinek <jakub@redhat.com>
10258 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
10259 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
10260 lookup if dest in some wider mode is known to be const0_rtx and
10261 if so, record permanent equivalence for it to be ZERO_EXTEND of
10262 the narrower mode destination.
10264 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
10266 * lto-streamer-out.c (output_function): Revert 259346.
10267 * omp-expand.c (expand_oacc_for): Likewise.
10269 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
10271 PR rtl-optimization/85354
10272 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
10273 * sel-sched.c (sel_global_init): ... here.
10275 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
10278 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
10279 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
10280 mode for PE-COFF targets.
10281 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
10282 (i386_pe_asm_lto_end): Likewise.
10283 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
10284 (TARGET_ASM_LTO_END): Likewise.
10285 * config/i386/winnt.c (saved_debug_info_level): New static variable.
10286 (i386_pe_asm_lto_start): New function.
10287 (i386_pe_asm_lto_end): Likewise.
10289 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
10290 Richard Biener <rguenther@suse.de>
10292 PR middle-end/84955
10293 * lto-streamer-out.c (output_function): Fix CFG loop state before
10295 * omp-expand.c (expand_oacc_for): Handle calls to internal
10296 functions like regular functions.
10298 2018-04-12 Richard Biener <rguenther@suse.de>
10301 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
10302 for the early LTO debug to properly generate references to it
10303 during DIE emission. Do not re-use that for the skeleton for
10305 (dwarf2out_early_finish): Likewise.
10307 2018-04-12 Jakub Jelinek <jakub@redhat.com>
10310 * config/i386/sse.md
10311 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
10312 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
10313 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
10314 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
10315 and output is a reg, avoid creating invalid lowpart subreg, but
10316 instead split into a 512-bit move. Don't split if not AVX512VL,
10317 input is xmm16+ reg and output is a mem.
10318 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
10319 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
10320 xmm16+ reg and output is a mem.
10322 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10324 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
10325 also for flag_dwarf2_cfi_asm.
10327 2018-04-12 Jakub Jelinek <jakub@redhat.com>
10329 PR rtl-optimization/85342
10330 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
10331 a bool scalar var inside of the loop instead. Don't try to update
10332 recog_data.operand after failed apply_change_group.
10334 2018-04-12 Tom de Vries <tom@codesourcery.com>
10337 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
10338 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
10339 array with flexible array member as array without given dimension.
10340 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
10341 argument for undefined param to true.
10343 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
10346 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
10347 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
10348 from PowerPC section.
10349 * config/rs6000/sysv4.opt (mcall-): Improve help text.
10350 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
10351 help text that is too long.
10352 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
10353 help text that is too long.
10354 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
10355 help text that is too long.
10357 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
10359 * config/alpha/alpha.md (stack_probe_internal): Rename
10360 from "probe_stack". Update all callers.
10362 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10364 PR rtl-optimization/84566
10365 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
10366 sched_macro_fuse_insns.
10368 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10371 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
10372 (compute_block_dependences): ... from here.
10374 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10376 PR tree-optimization/85331
10377 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
10378 from int to HOST_WIDE_INT.
10380 2018-04-11 Martin Jambor <mjambor@suse.cz>
10383 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
10384 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
10385 not the same as the source val.
10386 (cgraph_edge_brings_value_p): New parameter.
10387 (gather_edges_for_value): Pass destination value to
10388 cgraph_edge_brings_value_p.
10389 (perhaps_add_new_callers): Likewise.
10390 (get_info_about_necessary_edges): Likewise and exclude values brought
10391 only by self-recursive edges.
10392 (create_specialized_node): Redirect only clones of self-calling edges.
10393 (+self_recursive_pass_through_p): New function.
10394 (find_more_scalar_values_for_callers_subset): Use it.
10395 (find_aggregate_values_for_callers_subset): Likewise.
10396 (known_aggs_to_agg_replacement_list): Removed.
10397 (decide_whether_version_node): Re-calculate known constants for all
10398 remaining context clones.
10400 2018-04-11 Richard Biener <rguenther@suse.de>
10403 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
10404 from early DWARF output.
10405 (dwarf2out_early_finish): Output line info unconditionally into
10406 early DWARF and add reference to it.
10408 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10411 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
10412 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
10413 other than V2DFmode using iptr mode attribute.
10414 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
10416 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10418 PR rtl-optimization/84659
10419 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
10421 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10424 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
10426 (output_loc_list): Pass address of a dummy size variable even in the
10427 locview handling loop.
10428 (index_location_lists): Add comment on why skip_loc_list_entry can't
10431 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
10434 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
10437 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
10440 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
10441 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
10442 and -mstring-compare-inline-limit.
10444 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10447 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
10448 for stack clash protection in a register whenever we need it to be in
10451 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10453 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
10454 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
10456 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10459 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
10461 (mlong-double-): Ditto.
10462 * config/rs6000/sysv4.opt (msdata=): Ditto.
10463 (mtls-size=): Ditto.
10465 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10467 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
10468 erroneous entries for
10469 "vector int vec_ldl (int, long int *)", and
10470 "vector unsigned int vec_ldl (int, unsigned long int *)".
10471 Add comments and entries for
10472 "vector bool char vec_ldl (int, bool char *)",
10473 "vector bool short vec_ldl (int, bool short *)",
10474 "vector bool int vec_ldl (int, bool int *)",
10475 "vector bool long long vec_ldl (int, bool long long *)",
10476 "vector pixel vec_ldl (int, pixel *)",
10477 "vector long long vec_ldl (int, long long *)",
10478 "vector unsigned long long vec_ldl (int, unsigned long long *)".
10479 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
10480 type tree bool_long_long_type_node and correct definition of
10481 bool_V2DI_type_node to make reference to this new type tree.
10482 (rs6000_mangle_type): Replace erroneous reference to
10483 bool_long_type_node with bool_long_long_type_node.
10484 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
10485 comments to emphasize sign distinctions for char and int types and
10486 replace RS6000_BTI_bool_long constant with
10487 RS6000_BTI_bool_long_long constant. Also add comment to restrict
10488 use of RS6000_BTI_pixel.
10489 (bool_long_type_node): Remove this macro definition.
10490 (bool_long_long_type_node): New macro definition
10492 2018-04-10 Jakub Jelinek <jakub@redhat.com>
10494 PR rtl-optimization/85300
10495 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
10496 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
10497 simplify_unary_operation fails.
10499 2018-04-10 Martin Liska <mliska@suse.cz>
10501 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
10502 cgraph_edge and ipa_ref.
10504 2018-04-10 Jakub Jelinek <jakub@redhat.com>
10508 * config/i386/sse.md
10509 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
10510 computation of the VEC_MERGE selector from mask.
10511 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
10512 Fix decoding of the VEC_MERGE selector into mask.
10514 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
10516 PR tree-optimization/85286
10517 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
10519 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
10521 * final.c (final_1): Set insn_last_address as well as
10522 insn_current_address.
10524 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10527 * explow.c (emit_stack_probe): Call validize_mem on memory location
10528 before passing it to gen_probe_stack. Create address operand and
10529 legitimize it for the probe_stack_address case.
10531 2018-04-09 Jan Hubicka <jh@suse.cz>
10534 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
10535 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
10536 * tree.c (free_lang_data_in_type): Fix handling of binfos;
10538 (free_lang_data): Rebuild type inheritance graph.
10540 2018-04-09 Martin Sebor <msebor@redhat.com>
10542 * invoke.texi (-finline-small-functions): Mention other optimization
10544 (-findirect-inlining, -fpartial-inlining): Same.
10545 (-finline-functions-called-once): Same.
10546 (-freorder-blocks-and-partition): Same.
10548 2018-04-09 Jan Hubicka <jh@suse.cz>
10551 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
10552 jumps; choose last target that matches the criteria (i.e.
10553 no partition changes for non-crossing jumps).
10554 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
10555 support for redirecting crossing jumps to non-crossing.
10557 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
10559 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
10560 also for naked functions.
10562 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
10564 * config/arc/arc.md (add_shift): New pattern.
10565 (add_shift2): Likewise.
10566 (sub_shift): Likewise.
10567 (sub_shift_cmp0_noout): Likewise.
10568 (compare_si_ashiftsi): Likewise.
10569 (xbfu_cmp0_noout): New combine pattern.
10570 (xbfu_cmp0"): Likewise.
10571 (movsi_set_cc_insn): Place the predicable variant first.
10572 (commutative_binary_cmp0_noout): Remove clobber.
10573 (commutative_binary_cmp0): New pattern.
10574 (noncommutative_binary_cmp0): Likewise.
10575 (noncommutative_binary_cmp0_noout): Likewise.
10576 (noncommutative_binary_comparison_result_used): Removed.
10577 (rsub_cmp0): New pattern.
10578 (rsub_cmp0_noout): Likewise.
10579 (extzvsi): Changed, keep only meaningful variants.
10580 (SQH, SEZ): New iterators.
10581 (SQH_postfix): New mode attribute.
10582 (SEZ_prefix): New code attribute.
10583 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
10584 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
10585 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
10586 of numerical value.
10587 (noncommutative_operator): Check the availability of barrel
10590 2018-04-09 Richard Biener <rguenther@suse.de>
10592 PR tree-optimization/85284
10593 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
10594 Only use the niter constraining form of simple_iv when the exit
10595 is always executed.
10597 2018-04-09 Tom de Vries <tom@codesourcery.com>
10600 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
10601 (define_expand "*memory_barrier"): New define_expand.
10602 (define_insn "memory_barrier"): New insn.
10604 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10606 PR rtl-optimization/80463
10607 PR rtl-optimization/83972
10608 PR rtl-optimization/83480
10610 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
10611 correct producer for the insn.
10612 (tidy_control_flow): Fixup seqnos in case of debug insns.
10614 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10616 PR rtl-optimization/83913
10618 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
10619 different sched-times when merging exprs.
10621 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10623 PR rtl-optimization/83962
10625 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
10626 tidy_fallthru_edge and tidy_control_flow.
10628 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10630 PR rtl-optimization/83530
10632 * sel-sched.c (force_next_insn): New global variable.
10633 (remove_insn_for_debug): When force_next_insn is true, also leave only
10634 next insn in the ready list.
10635 (sel_sched_region): When the region wasn't scheduled, make another pass
10636 over it with force_next_insn set to 1.
10638 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
10640 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
10642 * config/nds32/constants.md (unspec_volatile_element): Add enum values
10643 for interrupt control.
10644 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
10645 functions for interrupt control.
10646 * config/nds32/nds32-intrinsic.md: Likewise.
10647 * config/nds32/nds32_intrinsic.h: Likewise.
10648 * config/nds32/nds32.h (nds32_builtins): Likewise.
10650 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
10652 * config/nds32/nds32.c (nds32_init_machine_status,
10653 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
10654 strict_aligned_p field.
10655 (nds32_expand_to_rtl_hook): New function.
10656 (TARGET_EXPAND_TO_RTL_HOOK): Define.
10657 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
10659 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10660 Chung-Ju Wu <jasonwucj@gmail.com>
10662 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
10663 * config/nds32/nds32-n7.md: New file.
10664 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
10665 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
10667 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
10668 * config/nds32/nds32.md (pipeline_model): Add n7.
10669 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
10670 * config/nds32/pipelines.md: Include n7 settings.
10672 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10673 Chung-Ju Wu <jasonwucj@gmail.com>
10675 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
10676 * config/nds32/nds32-e8.md: New file.
10677 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
10678 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
10680 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
10681 * config/nds32/nds32.md (pipeline_model): Add e8.
10682 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
10683 * config/nds32/pipelines.md: Include e8 settings.
10685 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10686 Chung-Ju Wu <jasonwucj@gmail.com>
10688 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
10689 * config/nds32/nds32-n8.md: New file.
10690 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
10691 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
10693 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
10694 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
10695 * config/nds32/nds32.md (pipeline_model): Add n8.
10696 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
10697 * config/nds32/pipelines.md: Include n8 settings.
10699 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10700 Chung-Ju Wu <jasonwucj@gmail.com>
10702 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
10703 * config/nds32/nds32-n9-2r1w.md: New file.
10704 * config/nds32/nds32-n9-3r2w.md: New file.
10705 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
10706 nds32_register_ports): New or modify for cpu n9.
10707 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
10709 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
10710 * config/nds32/nds32-utils.c: New file.
10711 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
10712 TARGET_MUL_SLOW): Define.
10713 * config/nds32/nds32.md (pipeline_model): New attribute.
10714 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
10715 New options that support cpu n9.
10716 * config/nds32/pipelines.md: Include n9 settings.
10717 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
10719 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
10721 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
10722 information if necessary.
10723 (output_cond_branch_compare_zero): Likewise.
10724 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
10725 (nds32_target_alignment): Refine for alignment.
10726 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
10727 (FUNCTION_BOUNDARY): Modify.
10728 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
10730 * config/nds32/nds32.opt (malways-align, malign-functions): New.
10732 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
10734 * config/nds32/constants.md (unspec_volatile_element): Add values for
10735 TLB operation and data prefetch.
10736 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
10737 functions for TLB operation and data prefetch.
10738 * config/nds32/nds32-intrinsic.md: Likewise.
10739 * config/nds32/nds32_intrinsic.h: Likewise.
10740 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
10741 (nds32_print_operand): Likewise.
10742 * config/nds32/nds32.h (nds32_builtins): Likewise.
10744 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
10745 Andrew Pinski <pinsika@gcc.gnu.org>
10747 PR middle-end/82976
10748 * match.pd: Use constant_boolean_node of correct type instead of
10749 boolean_true_node or boolean_false_node for simplifying
10750 pointer comparisons to zero.
10752 2018-04-07 Jakub Jelinek <jakub@redhat.com>
10754 PR tree-optimization/80021
10755 * tree.c (verify_type_variant): Make error call in verify_variant_match
10756 translatable and remove final full stop.
10758 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10760 * config/nds32/constants.md (unspec_volatile_element): Add
10761 UNSPEC_VOLATILE_EH_RETURN.
10762 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
10763 nds32_output_stack_pop): Support dwarf exception handling process.
10764 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
10765 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
10766 exception handling process.
10767 (nds32_compute_stack_frame): Likewise.
10768 (nds32_return_addr_rtx): Likewise.
10769 (nds32_initial_elimination_offset): Likewise.
10770 (nds32_expand_prologue): Likewise.
10771 (nds32_expand_epilogue): Likewise.
10772 (nds32_dynamic_chain_address): New function.
10773 * config/nds32/nds32.h (machine_function): Add fields for dwarf
10774 exception handling.
10775 (DYNAMIC_CHAIN_ADDRESS): Define.
10776 (EH_RETURN_DATA_REGNO): Define.
10777 (EH_RETURN_STACKADJ_RTX): Define.
10778 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
10779 patterns for dwarf exception handling.
10781 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10783 * config/nds32/nds32.h: Clean up obsolete macros.
10785 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10787 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10788 Add enum values for particular instructions.
10789 * config/nds32/nds32-intrinsic.c: Implementation of expanding
10790 particular intrinsic functions.
10791 * config/nds32/nds32-intrinsic.md: Likewise.
10792 * config/nds32/nds32_intrinsic.h: Likewise.
10793 * config/nds32/nds32.h (nds32_builtins): Likewise.
10794 * config/nds32/nds32.md (type): Add pbsad and pbsada.
10795 (btst, ave): New patterns for particular instructions.
10797 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10799 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10800 Add enum values for atomic load/store and memory sync.
10801 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
10803 * config/nds32/nds32-intrinsic.md: Likewise.
10804 * config/nds32/nds32_intrinsic.h: Likewise.
10805 * config/nds32/nds32.h (nds32_builtins): Likewise.
10807 2018-04-07 Jakub Jelinek <jakub@redhat.com>
10809 PR tree-optimization/85257
10810 * fold-const.c (native_encode_vector): If not all elts could fit
10811 and off is -1, return 0 rather than offset.
10812 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
10813 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
10814 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
10815 adjust buffer in native_interpret_expr call.
10817 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10819 * config/nds32/constants.md (unspec_volatile_element): Add cache
10820 control enum values.
10821 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
10822 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
10823 * config/nds32/nds32.c (nds32_cctl_names): New.
10824 (nds32_print_operand): Handle cache control register names.
10825 * config/nds32/nds32.h (nds32_builtins): New enum values.
10826 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
10828 * config/nds32/nds32.md (type): Add mmu.
10829 * config/nds32/pipelines.md (simple_insn): Add mmu.
10831 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10833 * config/nds32/nds32.md (type): Remove call.
10834 * config/nds32/pipelines.md (simple_insn): Likewise.
10836 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10838 * config/nds32/constants.md (unspec_volatile_element): Add
10839 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
10840 UNSPEC_VOLATILE_FMFCFG.
10841 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
10842 description for fmfcfg and fmfcsr.
10843 (bdesc_1arg): Add fmtcsr.
10844 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
10845 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
10846 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
10847 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
10848 unspec_fmfcfg): New patterns.
10849 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
10850 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
10851 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
10852 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
10853 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
10854 __nds32__fmfcfg): Define.
10856 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10858 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
10859 intrinsic register names.
10860 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
10861 intrinsic register enum values and macros.
10863 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10865 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
10866 for load/store addressing form.
10867 (nds32_print_operand_address): Likewise.
10869 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
10872 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
10873 based on LABEL_REF. Remove useless assertion.
10874 (pic_address_needs_scratch): Fix formatting.
10875 (sparc_legitimize_pic_address): Minor tweaks.
10876 (sparc_delegitimize_address): Adjust assertion accordingly.
10877 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
10878 into symbolic_operand.
10879 (movsi_high_pic_label_ref): Likewise.
10880 (movsi_lo_sum_pic_label_ref): Likewise.
10881 (movdi_pic_label_ref): Likewise.
10882 (movdi_high_pic_label_ref): Likewise.
10883 (movdi_lo_sum_pic_label_ref): Likewise.
10885 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
10887 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
10888 custom LIB_SPEC setup.
10890 2018-04-06 Ruslan Bukin <br@bsdpad.com>
10891 Kito Cheng <kito.cheng@gmail.com>
10893 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
10894 * config/riscv/freebsd.h: New.
10896 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10898 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
10899 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
10902 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10903 Kito Cheng <kito.cheng@gmail.com>
10905 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
10906 nds32_output_call, nds32_symbol_binds_local_p): New functions.
10907 * config/nds32/nds32-protos.h (nds32_output_call,
10908 nds32_output_return): Declare.
10909 * config/nds32/nds32.md: Refine all the call and return patterns.
10911 2018-04-06 Jakub Jelinek <jakub@redhat.com>
10914 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
10915 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
10917 PR rtl-optimization/84872
10918 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
10919 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
10920 EDGE_CROSSING edge.
10922 2018-04-06 Tamar Christina <tamar.christina@arm.com>
10924 * expr.c (copy_blkmode_to_reg): Revert 254862.
10925 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
10927 2018-04-06 Richard Biener <rguenther@suse.de>
10929 PR middle-end/85244
10930 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
10931 after seeing a component reference with an adjacent field. Treat
10932 refs to arrays at struct end of external decls similar to
10933 refs to unconstrained commons.
10935 2018-04-06 Jakub Jelinek <jakub@redhat.com>
10938 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
10939 look through SAVE_EXPRs with non-side-effects argument. Adjust
10941 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
10944 2018-04-06 Richard Biener <rguenther@suse.de>
10946 PR middle-end/85180
10947 * alias.c (find_base_term): New wrapper around find_base_term
10948 unwinding CSELIB_VAL_PTR changes.
10949 (find_base_term): Do not restore CSELIB_VAL_PTR during the
10952 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10954 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
10956 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
10957 constant definitions.
10958 ("nop"): lr 0,0 -> nopr r0
10959 ("nop_lr0", "nop_lr1"): New insn definitions.
10961 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10963 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
10964 NDS32_V3PUSH_AVAILABLE_P macro.
10966 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
10967 Chung-Ju Wu <jasonwucj@gmail.com>
10969 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
10970 (nds32*-*-*): Add float and fpu_config into supported_defaults.
10971 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
10972 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
10973 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
10974 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
10975 * config/nds32/constraints.md: New constraints and checking for hard
10976 float configuration.
10977 * config/nds32/iterators.md: New mode iterator and attribute for hard
10978 float configuration.
10979 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
10981 * config/nds32/nds32-fpu.md: New file.
10982 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
10983 deal with hard float code generation.
10984 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
10986 (abi_type, float_reg_number): New enum type.
10987 * config/nds32/nds32-predicates.c: New predicates for hard float.
10988 * config/nds32/nds32-protos.h: Declare functions for hard float.
10989 * config/nds32/nds32.c: Implementation for hard float configuration.
10990 * config/nds32/nds32.h: Definitions for hard float configuration.
10991 * config/nds32/nds32.md: Include hard float machine description and
10992 modify patterns for hard float configuration.
10993 * config/nds32/nds32.opt: New options for hard float configuration.
10994 * config/nds32/predicates.md: New predicates for hard float
10997 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10999 * common/config/nds32/nds32-common.c
11000 (nds32_option_optimization_table): Enable -mreleax-hint by default.
11002 2018-04-05 Jakub Jelinek <jakub@redhat.com>
11004 PR middle-end/85195
11005 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
11006 CONSTRUCTOR_ELT (ctor, ...)->value.
11008 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
11011 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
11013 2018-04-05 Tom de Vries <tom@codesourcery.com>
11016 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
11019 2018-04-05 Shiva Chen <shiva0217@gmail.com>
11020 Kito Cheng <kito.cheng@gmail.com>
11022 * config/nds32/constraints.md (U33): Fine-tune checking condition.
11023 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
11024 * config/nds32/nds32.h (nds32_16bit_address_type): Add
11025 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
11027 2018-04-05 Shiva Chen <shiva0217@gmail.com>
11028 Kito Cheng <kito.cheng@gmail.com>
11030 * config/nds32/constraints.md (Ufe): New memory constraint.
11031 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
11032 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
11033 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
11035 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
11036 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
11038 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11040 * config/nds32/nds32.md: Use optimize_size in the condition for
11041 alu-shift instructions.
11043 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11045 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
11047 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11049 * config/nds32/nds32.md (negsi2): Refine pattern.
11051 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
11052 Chung-Ju Wu <jasonwucj@gmail.com>
11054 * config/nds32/iterators.md (shift_rotate): New code iterator.
11055 (shift): New code attribute.
11056 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
11057 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
11058 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
11059 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
11060 bit-wise operations.
11061 (andsi3, *andsi3): Ditto.
11062 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
11063 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
11064 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
11065 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
11066 nds32_ior_operand, nds32_xor_operand): New predicates.
11068 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11070 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
11071 (addsi3, subsi3): ... this.
11073 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11075 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
11077 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11079 * config/nds32/nds32.md: Adjust indention.
11081 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
11083 * config/nds32/nds32.md (feature): New attribute.
11085 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11087 * config/nds32/nds32.md (subtype): New attribute.
11089 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
11092 * config/arm/arm-builtins.c (arm_expand_builtin): Change
11093 expansion to perform a bitwise AND of the argument followed by a
11094 boolean negation of the result.
11096 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
11098 PR rtl-optimization/84878
11099 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
11100 the basic block. Assert the use reference is not artificial and that
11101 it has an associated insn.
11103 2018-04-04 Michael Matz <matz@suse.de>
11105 * builtins.c (compute_objsize): Pass correct operand
11106 to array_at_struct_end_p.
11108 2018-04-04 Richard Biener <rguenther@suse.de>
11111 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
11112 from contexts for DINFO_LEVEL_TERSE and below.
11114 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11116 * config/nds32/nds32-doubleword.md (move_<mode>): Require
11117 resiter_operand condition.
11118 * config/nds32/nds32.md (*move<mode>): Ditto.
11120 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11121 Monk Chiang <sh.chiang04@gmail.com>
11123 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
11125 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11127 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
11129 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11130 Kito Cheng <kito.cheng@gmail.com>
11132 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
11133 nds32_cond_code_str, output_cond_branch,
11134 output_cond_branch_compare_zero, nds32_expand_cbranch,
11135 nds32_expand_cstore, nds32_expand_movcc,
11136 nds32_output_cbranchsi4_equality_zero,
11137 nds32_output_cbranchsi4_equality_reg,
11138 nds32_output_cbranchsi4_equality_reg_or_const_int,
11139 nds32_output_cbranchsi4_greater_less_zero: New functions.
11140 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
11141 nds32_expand_cstore, nds32_expand_movcc,
11142 nds32_output_cbranchsi4_equality_zero,
11143 nds32_output_cbranchsi4_equality_reg,
11144 nds32_output_cbranchsi4_equality_reg_or_const_int,
11145 nds32_output_cbranchsi4_greater_less_zero): Declare.
11146 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
11147 nds32_rimm11s_operand): New predicates.
11148 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
11149 * config/nds32/nds32.md: Rewrite all the branch and conditional move
11152 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11154 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
11155 * config/nds32/nds32.md: Ditto.
11156 * config/nds32/pipelines.md: Ditto.
11158 2018-04-04 Richard Biener <rguenther@suse.de>
11160 PR tree-optimization/85168
11161 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
11162 propagating abnormals.
11164 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11166 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
11168 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11169 Kito Cheng <kito.cheng@gmail.com>
11171 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
11172 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
11173 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
11174 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
11175 * config/nds32/nds32.md (sibcall_internal): New.
11176 (sibcall_register): Remove.
11177 (sibcall_immediate): Remove.
11178 (sibcall_value_internal): New.
11179 (sibcall_value_register): Remove.
11180 (sibcall_value_immediate): Remove.
11181 * config/nds32/predicates.md (nds32_general_register_operand): New.
11182 (nds32_call_address_operand): New.
11184 2018-04-03 Jakub Jelinek <jakub@redhat.com>
11186 PR rtl-optimization/85167
11187 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
11188 bb_defs if *split_p, instead preinitialize it to NULL.
11190 PR tree-optimization/85156
11191 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
11192 evaluating the argument multiple times.
11194 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
11196 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
11198 (_mm_cvtpd_ps): Likewise.
11199 (_mm_cvttpd_epi32): Likewise.
11200 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
11201 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
11202 vector, pixel, and bool following altivec.h include.
11204 2018-04-03 Martin Sebor <msebor@redhat.com>
11206 * doc/extend.texi (Common Function Attributes): Clarify.
11207 (const attribute): Likewise.
11208 (pure attribute): Likewise.
11210 2018-04-03 Jakub Jelinek <jakub@redhat.com>
11213 * config/i386/i386.c (ix86_expand_vector_set): Use
11214 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
11216 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
11218 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
11219 instructions when changing rounding bits to preserve precision bits
11220 in the x87 control word.
11222 2018-04-03 Martin Liska <mliska@suse.cz>
11224 PR tree-optimization/82491
11225 * rtl.h (strip_offset_and_add): Replace += suboffset with
11226 poly_uint64 () + suboffset.
11228 2018-03-29 Martin Liska <mliska@suse.cz>
11229 Martin Jambor <mjambor@suse.cz>
11232 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
11233 param_type is not an integral or pointer type.
11235 2018-04-03 Richard Biener <rguenther@suse.de>
11237 * sese.h (recompute_all_dominators): Remove.
11239 2018-04-02 Martin Sebor <msebor@redhat.com>
11241 * doc/invoke.texi (-Wrestrict): Fix typos.
11243 2018-04-02 Jim Wilson <jimw@sifive.com>
11245 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
11246 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
11247 (<optab>di3, <optab>si3_extend): Likewise.
11248 (<optab>si3_mask, <optab>si3_mask_1): New.
11249 (<optab>di3_mask, <optab>di3_mask_1): New.
11250 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
11251 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
11252 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
11254 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
11256 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
11259 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
11261 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
11262 (nds32_canonicalize_comparison): New function.
11264 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
11265 Kito Cheng <kito.cheng@gmail.com>
11266 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11268 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
11269 * config/nds32/constants.md (unspec_volatile_element): Add
11270 UNSPEC_VOLATILE_RELAX_GROUP.
11271 * config/nds32/nds32-relax-opt.c: New file.
11272 * config/nds32/nds32-predicates.c
11273 (nds32_symbol_load_store_p): New function.
11274 * config/nds32/nds32-protos.h
11275 (nds32_symbol_load_store_p): Declare function.
11276 (make_pass_nds32_relax_opt): Declare new rtl pass function.
11277 * config/nds32/nds32.c
11278 (nds32_register_pass): New function to register pass.
11279 (nds32_register_passes): New function to register passes.
11280 * config/nds32/nds32.md (relax_group): New pattern.
11281 * config/nds32/nds32.opt (mrelax-hint): New option.
11282 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
11284 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
11286 * config/nds32/t-nds32: Modify files dependency.
11288 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
11290 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
11291 (PROFILE_HOOK): Define its implementation.
11293 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
11295 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
11296 type and 32-bit size.
11298 2018-04-01 Jakub Jelinek <jakub@redhat.com>
11300 PR middle-end/85090
11301 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
11302 (V_128_256): New mode iterator.
11303 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
11304 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
11305 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
11307 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
11308 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
11310 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
11313 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
11314 NaN inputs correctly.
11316 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
11319 * config/rs6000/vsx.md (??r): New mode attribute.
11320 (*vsx_mov<mode>_64bit): Use it.
11321 (*vsx_mov<mode>_32bit): Likewise.
11323 2018-03-30 Martin Sebor <msebor@redhat.com>
11325 PR tree-optimization/84818
11326 * builtins.c (check_access): Use warning_n.
11328 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11331 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
11333 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
11336 2018-03-30 Julia Koval <julia.koval@intel.com>
11339 * x86-tune.def (movx, partial_reg_dependency): Enable for
11342 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
11344 PR inline-asm/84985
11345 * lra-constraints.c (process_alt_operands): Move setting
11346 this_alternative_matches below.
11348 2018-03-29 Martin Liska <mliska@suse.cz>
11351 * doc/invoke.texi: Document how LTO works with debug info.
11352 Describe auto-load support of binutils. Mention 'x86-64'
11353 as valid option value of -march option.
11355 2018-03-29 Jakub Jelinek <jakub@redhat.com>
11357 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
11360 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
11361 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
11362 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
11365 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
11368 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
11369 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
11370 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
11371 for RS6000_BTM_POWERPC64.
11372 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
11373 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
11374 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
11377 (DIVDEU): Likewise.
11379 2018-03-28 Carl Love <cel@us.ibm.com>
11382 2017-09-27 Carl Love <cel@us.ibm.com>
11384 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
11385 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
11386 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
11389 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11391 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
11392 instead of __vector bool.
11393 (_mm_max_pu8): Likewise.
11394 (_mm_min_pi16): Likewise.
11396 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
11399 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
11400 (DIVWEUO): Likewise.
11401 (DIVDEO): Likewise.
11402 (DIVDEUO): Likewise.
11403 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
11404 DIVWEUO and DIVDEUO.
11405 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
11406 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
11407 (div_extend): Likewise.
11408 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
11410 (__builtin_divweuo): Likewise.
11411 (__builtin_divdeo): Likewise.
11412 (__builtin_divdeuo): Likewise.
11414 2018-03-28 Jakub Jelinek <jakub@redhat.com>
11417 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
11418 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
11420 PR tree-optimization/82004
11421 * gimple-match-head.c (optimize_pow_to_exp): New function.
11422 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
11423 Don't fold to exp if optimize_pow_to_exp is false.
11425 2018-03-28 Martin Liska <mliska@suse.cz>
11428 * calls.c (initialize_argument_information): Fix trailing space.
11429 * common.opt: Fix typo and provide better explanation for
11430 -fsanitize-coverage option.
11431 * config/i386/i386.opt: Fix typo.
11433 2018-03-28 Jakub Jelinek <jakub@redhat.com>
11434 Martin Liska <mliska@suse.cz>
11437 * gimplify.c (asan_poison_variable): Don't do the check for
11438 gimplify_omp_ctxp here.
11439 (gimplify_decl_expr): Do it here.
11440 (gimplify_target_expr): Likewise.
11442 2018-03-28 Martin Liska <mliska@suse.cz>
11445 * config/i386/i386.c (ix86_function_arg_advance): Do not call
11446 chkp_type_bounds_count if MPX is not enabled.
11448 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
11450 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
11452 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
11455 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
11456 function to create the function decl for complex long double
11457 multiply and divide for -mabi=ieeelongdouble.
11458 (init_float128_ieee): Call it.
11460 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
11463 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
11464 -fcf-protection=branch -mibt.
11465 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
11467 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11470 * config/arm/arm.c (arm_valid_symbolic_address): Handle
11471 arm_word_relocations.
11473 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
11476 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
11477 extern array declarations.
11479 2018-03-27 Richard Biener <rguenther@suse.de>
11481 PR middle-end/84067
11482 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
11483 explicit single_use checks.
11485 2018-03-27 Richard Biener <rguenther@suse.de>
11487 PR tree-optimization/85082
11488 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
11491 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11493 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
11494 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
11495 Turn on fasynchronous-unwind-tables and funwind-tables.
11497 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
11500 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
11501 (*bmi_blsr_<mode>_ccz): Ditto.
11503 2018-03-26 Tom de Vries <tom@codesourcery.com>
11505 PR tree-optimization/85063
11506 * omp-general.c (offloading_function_p): New function. Factor out
11508 * omp-offload.c (pass_omp_target_link::gate): ... here.
11509 * omp-general.h (offloading_function_p): Declare.
11510 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
11511 with attribute omp declare target for offloading functions.
11513 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
11515 PR tree-optimization/84005
11516 * tree-data-ref.h (get_base_for_alignment): Declare.
11517 * tree-data-ref.c (get_base_for_alignment_1): New function.
11518 (get_base_for_alignment): Likewise.
11519 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
11520 get_base_for_alignment to find a suitable base object, instead
11521 of always using drb->base_address.
11523 2018-03-23 Jakub Jelinek <jakub@redhat.com>
11525 PR inline-asm/85022
11526 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
11527 known size by default.
11529 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
11531 PR inline-asm/85030
11532 * lra-constraints.c (process_alt_operands): Don't match BLKmode
11533 and non BLKmode operands.
11535 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11538 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
11539 Clean up attributes.
11541 2018-03-23 Richard Biener <rguenther@suse.de>
11544 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
11545 we are going to emit early debug for LTO.
11547 2018-03-23 Jakub Jelinek <jakub@redhat.com>
11549 PR inline-asm/85034
11550 * function.c (match_asm_constraints_1): Don't optimize if input
11551 doesn't satisfy general_operand predicate for output's mode.
11553 PR inline-asm/85022
11554 * alias.c (write_dependence_p): Don't require for x_canonicalized
11555 non-VOIDmode if x has VOIDmode.
11558 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
11559 just don't try to optimize it rather than assert it never happens.
11561 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
11563 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
11564 macro expansions for definition of ST_INTERNAL_<mode> and
11565 LD_INTERNAL_<mode> builtins.
11566 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
11568 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
11570 (altivec_expand_st_builtin): Likewise.
11571 (altivec_expand_builtin): Remove calls to deleted functions.
11572 (rs6000_address_for_altivec): Delete this function.
11573 * config/rs6000/vector.md: Remove expands for
11574 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
11576 2018-03-22 Sudakshina Das <sudi.das@arm.com>
11579 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
11580 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
11581 re-computing once computed.
11582 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
11583 (arm_init_machine_status): Initialize
11584 machine->static_chain_stack_bytes.
11586 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
11589 * doc/extend.texi: Add four new prototypes for vec_ld.
11590 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
11591 definitions for more logical presentation.
11592 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
11593 entries for V1TI variants of __builtin_altivec_ld builtin.
11594 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
11595 handling of V1TI variant of LVX icode pattern.
11596 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
11597 (rs6000_gimple_fold_builtin): Likewise.
11598 (altivec_init_builtins): Add code to define
11599 __builtin_altivec_lvx_v1ti function.
11601 2018-03-22 Jakub Jelinek <jakub@redhat.com>
11603 PR inline-asm/84941
11604 * function.c (match_asm_constraints_1): Don't do the optimization
11605 if input isn't a REG, SUBREG, MEM or constant.
11607 2018-03-22 Tom de Vries <tom@codesourcery.com>
11609 PR tree-optimization/84956
11610 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
11611 bb_has_abnormal_pred.
11613 2018-03-22 Jakub Jelinek <jakub@redhat.com>
11616 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
11617 DECL_INITIAL (decl) to decl at the end.
11618 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
11619 adjust the comment.
11621 2018-03-21 Joseph Myers <joseph@codesourcery.com>
11623 * doc/extend.texi (__builtin_tgmath): Document when complex
11624 integer types are treated as _Complex _Float64.
11626 2018-03-21 Tom de Vries <tom@codesourcery.com>
11628 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
11630 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11632 PR tree-optimization/84960
11633 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
11634 if it is ENTRY block, move them into single succ of ENTRY in that case.
11636 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
11638 PR tree-optimization/84811
11639 * poly-int.h (poly_span_traits): Remove the T3 parameter and
11640 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
11641 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
11642 (known_subrange_p): Update accordingly. Cast each value involved
11643 in the size comparison, rather than casting the result of the
11646 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11648 PR tree-optimization/84982
11649 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
11650 by flipping the least significant bit rather than all bits from
11651 bitpos to bitpos + bitsize - 1.
11653 2018-03-21 Nathan Sidwell <nathan@acm.org>
11655 * doc/extend.texi (Deprecated Features): Remove mention of
11656 long-deleted deprecations.
11658 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11661 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
11662 * configure: Regenerate.
11664 2018-03-21 Tom de Vries <tom@codesourcery.com>
11666 PR tree-optimization/83126
11667 * tree-parloops.c (num_phis): New function.
11668 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
11670 2018-03-21 Nathan Sidwell <nathan@acm.org>
11672 * doc/extend.texi (Deprecated Features): Update deprecated flags,
11673 mention anon-struct/union members and trailing attributes.
11675 2018-03-21 Bin Cheng <bin.cheng@arm.com>
11677 PR tree-optimization/84969
11678 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
11679 builtin memset partitions if they set different rhs values.
11681 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11683 PR rtl-optimization/84989
11684 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
11685 VEC_DUPLICATE with scalar result mode.
11687 2018-03-21 Martin Liska <mliska@suse.cz>
11690 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
11691 not intended return statement.
11693 2018-03-21 Martin Liska <mliska@suse.cz>
11696 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
11697 (chkp_find_bound_slots_1): Limit number of iterations.
11699 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
11702 * Minor grammar fixes for x86 options.
11704 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11707 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
11708 holding REG_CFA_RESTORE notes, instead turn them into a USE.
11710 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
11713 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
11714 (altivec_lvx_<mode>_1op): Likewise.
11715 (altivec_stvx_<mode>_2op): Likewise.
11716 (altivec_stvx_<mode>_1op): Likewise.
11717 (altivec_lvx_<VM2:mode>): New define_expand.
11718 (altivec_stvx_<VM2:mode>): Likewise.
11719 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
11720 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
11721 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
11722 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
11723 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
11724 (rs6000_gen_lvx): Likewise.
11725 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
11726 (altivec_expand_stv_builtin): Likewise.
11727 (altivec_expand_builtin): Likewise.
11728 * config/rs6000/vector.md: Likewise.
11730 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11733 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
11736 2018-03-20 Richard Biener <rguenther@suse.de>
11739 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
11740 sign-conversions as zero, fall back to standard scalar_stmt
11743 2018-03-20 Martin Liska <mliska@suse.cz>
11746 * predict.c (rebuild_frequencies): Handle case when we have
11747 PROFILE_ABSENT, but flag_guess_branch_prob is false.
11749 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11752 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
11753 flag_section_anchors.
11754 * varasm.c (use_blocks_for_decl_p): Remove hack for
11755 dw2_force_const_mem.
11758 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
11760 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
11761 be created, use lowpart_subreg of operands[0] rather than operands[0]
11763 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
11764 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
11765 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
11766 and n constraint instead of aarch64_shift_imm_di and Usd.
11767 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
11768 (*aarch64_<optab>_reg_minus<mode>3): ... this.
11770 2018-03-20 Sudakshina Das <sudi.das@arm.com>
11773 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
11774 to favor GPR over NEON registers.
11775 (<shift>di3_neon): Likewise.
11777 2018-03-20 Tom de Vries <tom@codesourcery.com>
11780 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
11781 (nvptx_process_pars): Emit bar.sync asap and alap.
11783 2018-03-20 Tom de Vries <tom@codesourcery.com>
11786 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
11787 seen_label if seen_label is already set.
11789 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11792 * config/i386/i386.c (fold_builtin_cpu): For features above 31
11793 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
11794 Use 1U instead of 1. Formatting fixes.
11797 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
11798 instead of TREE_TYPE (s1) for the return value.
11800 2018-03-19 Jakub Jelinek <jakub@redhat.com>
11802 PR tree-optimization/84946
11803 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
11804 bitsize + bitsize in poly_uint64 rather than poly_int64.
11807 * dwarf2asm.c: Include fold-const.c.
11808 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
11809 of decl rather than decl itself.
11811 PR rtl-optimization/84643
11812 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
11814 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
11817 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
11818 calling assemble_variable.
11820 2018-03-19 Sudakshina Das <sudi.das@arm.com>
11823 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
11824 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
11826 2018-03-19 Jim Wilson <jimw@sifive.com>
11829 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
11830 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
11831 (riscv_first_stack_step): Likewise.
11832 (riscv_option_override): Use STACK_BOUNDARY instead of
11833 MIN_STACK_BOUNDARY.
11834 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
11835 MIN_STACK_BOUNDARY.
11836 (BIGGEST_ALIGNMENT): Set to 128.
11837 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
11838 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
11841 2018-03-19 Richard Biener <rguenther@suse.de>
11843 PR tree-optimization/84933
11844 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
11845 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
11847 2018-03-19 Richard Biener <rguenther@suse.de>
11849 PR tree-optimization/84859
11850 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
11851 (cond_if_else_store_replacement): Perform sinking operation on
11852 single-store BBs regardless of MAX_STORES_TO_SINK setting.
11853 Generalize what a BB with a single eligible store is.
11855 2018-03-19 Richard Biener <rguenther@suse.de>
11857 PR tree-optimization/84929
11858 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
11859 chrec_is_positive against non-chrec arg.
11861 2018-03-19 Tamar Christina <tamar.christina@arm.com>
11864 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
11866 2018-03-18 Martin Liska <mliska@suse.cz>
11868 PR rtl-optimization/84635
11869 * regrename.c (build_def_use): Use matches_mode only when
11872 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
11874 PR tree-optimization/84913
11875 * tree-vect-loop.c (vectorizable_reduction): Don't try to
11876 vectorize chains of COND_EXPRs.
11878 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11880 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
11882 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11884 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
11886 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11888 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
11890 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
11891 Kito Cheng <kito.cheng@gmail.com>
11893 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
11894 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
11895 (nds32_adjust_reg_alloc_order): New function.
11896 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
11898 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
11900 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
11901 nds32_print_operand, nds32_print_operand_address): Use
11902 HOST_WIDE_INT_PRINT_DEC instead.
11904 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
11906 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
11908 2018-03-17 Jakub Jelinek <jakub@redhat.com>
11911 * config/i386/i386.c (initial_ix86_tune_features,
11912 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
11913 unsigned long long.
11914 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
11915 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
11916 rather than 1u << ix86_tune. Formatting fix.
11917 (ix86_option_override_internal): Change ix86_arch_mask from
11918 unsigned int to unsigned HOST_WIDE_INT, initialize to
11919 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
11920 (ix86_function_specific_restore): Likewise.
11922 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11925 * postreload.c (reload_combine_recognize_pattern): Perform
11926 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
11927 truncate_int_for_mode the result for the destination's mode.
11930 * hsa-gen.c (mem_type_for_type): Fix comment typo.
11931 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
11933 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11936 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
11939 * lra-assigns.c (lra_split_hard_reg_for): Don't use
11940 regno_allocno_class_array and sorted_pseudos.
11941 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
11942 insns where regno is used.
11944 2018-03-16 Martin Liska <mliska@suse.cz>
11947 * multiple_target.c (create_dispatcher_calls): Redirect
11948 reference in the symbol table.
11950 2018-03-16 Martin Liska <mliska@suse.cz>
11953 * multiple_target.c (create_dispatcher_calls): Redirect also
11956 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11960 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
11961 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
11962 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
11964 2018-03-16 Julia Koval <julia.koval@intel.com>
11966 * doc/invoke.texi (Skylake Server): Add CLWB.
11967 Cannonlake): Remove CLWB.
11969 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11971 PR tree-optimization/84841
11972 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
11974 (FLOAT_ONE_CONST_TYPE): Define.
11975 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
11976 (sort_by_operand_rank): Put entries with higher constant_type last
11977 rather than first to match comments.
11979 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
11981 * config/nios2/nios2.md (movsi_internal): Fix thinko in
11984 2018-03-15 Jakub Jelinek <jakub@redhat.com>
11987 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
11988 check and use address of target always.
11990 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
11993 * config/i386/i386.c (indirect_thunk_needed): Update comments.
11994 (indirect_thunk_bnd_needed): Likewise.
11995 (indirect_thunks_used): Likewise.
11996 (indirect_thunks_bnd_used): Likewise.
11997 (indirect_return_needed): New.
11998 (indirect_return_bnd_needed): Likewise.
11999 (output_indirect_thunk_function): Add a bool argument for
12001 (output_indirect_thunk_function): Don't generate alias for
12002 function return thunk.
12003 (ix86_code_end): Call output_indirect_thunk_function to generate
12004 function return thunks.
12005 (ix86_output_function_return): Set indirect_return_bnd_needed
12006 and indirect_return_needed instead of indirect_thunk_bnd_needed
12007 and indirect_thunk_needed.
12009 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
12011 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
12012 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
12013 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
12015 2018-03-15 David Malcolm <dmalcolm@redhat.com>
12016 Paul Hua <paul.hua.gm@gmail.com>
12019 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
12021 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
12023 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
12024 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
12025 resp. SFmode cases.
12027 2018-03-15 Tamar Christina <tamar.christina@arm.com>
12030 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
12031 instead of GET_MODE_SIZE when comparing Units.
12033 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
12036 * varasm.c (hash_section): Return an unchangeble hash value
12037 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
12038 Return !aarch64_can_use_per_function_literal_pools_p ().
12040 2018-03-15 Jakub Jelinek <jakub@redhat.com>
12043 * optabs.c (emit_conditional_move): Pass address of cmode's copy
12044 rather than address of cmode as last argument to prepare_cmp_insn.
12046 2018-03-15 Julia Koval <julia.koval@intel.com>
12048 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
12049 F_AVX512VNNI, F_AVX512BITALG): New.
12051 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
12054 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
12055 insn for floating-point loads and stores.
12057 2018-03-14 Carl Love <cel@us.ibm.com>
12059 * config/rs6000/rs6000-c.c: Add macro definitions for
12060 ALTIVEC_BUILTIN_VEC_PERMXOR.
12061 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
12062 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
12063 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
12064 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
12066 * config/doc/extend.texi: Add prototypes for vec_permxor.
12068 2018-03-14 David Malcolm <dmalcolm@redhat.com>
12071 * diagnostic-show-locus.c (class layout_point): Convert m_line
12072 from int to linenum_type.
12073 (line_span::comparator): Use linenum "compare" function when
12074 comparing line numbers.
12075 (test_line_span): New function.
12076 (layout_range::contains_point): Convert param "row" from int to
12078 (layout_range::intersects_line_p): Likewise.
12079 (layout::will_show_line_p): Likewise.
12080 (layout::print_source_line): Likewise.
12081 (layout::should_print_annotation_line_p): Likewise.
12082 (layout::print_annotation_line): Likewise.
12083 (layout::print_leading_fixits): Likewise.
12084 (layout::annotation_line_showed_range_p): Likewise.
12085 (struct line_corrections): Likewise for field m_row.
12086 (line_corrections::line_corrections): Likewise for param "row".
12087 (layout::print_trailing_fixits): Likewise.
12088 (layout::get_state_at_point): Likewise.
12089 (layout::get_x_bound_for_row): Likewise.
12090 (layout::print_line): Likewise.
12091 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
12092 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
12093 * input.c (selftest::test_linenum_comparisons): New function.
12094 (selftest::input_c_tests): Call it.
12095 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
12096 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
12097 * selftest.h (ASSERT_GT): New macro.
12098 (ASSERT_GT_AT): New macro.
12099 (ASSERT_LT): New macro.
12100 (ASSERT_LT_AT): New macro.
12102 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
12104 PR rtl-optimization/84780
12105 * combine.c (distribute_links): Don't make a link based on pc_rtx.
12107 2018-03-14 Martin Liska <mliska@suse.cz>
12109 * tree.c (record_node_allocation_statistics): Use
12110 get_stats_node_kind.
12111 (get_stats_node_kind): New function extracted from
12112 record_node_allocation_statistics.
12113 (free_node): Use get_stats_node_kind.
12115 2018-03-14 Richard Biener <rguenther@suse.de>
12117 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
12118 that the value-set of ANTIC_IN doesn't grow.
12121 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
12123 (BB_VISITED_WITH_VISITED_SUCCS): New define.
12124 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
12126 2018-03-14 Julia Koval <julia.koval@intel.com>
12128 * config.gcc (icelake-client, icelake-server): New.
12130 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
12131 (initial_ix86_arch_features): Ditto.
12132 (PTA_SKYLAKE): Add SGX.
12133 (PTA_ICELAKE): Remove.
12134 (PTA_ICELAKE_CLIENT): New.
12135 (PTA_ICELAKE_SERVER): New.
12136 (ix86_option_override_internal): Split up icelake on icelake client and
12138 (get_builtin_code_for_version): Ditto.
12139 (fold_builtin_cpu): Ditto.
12140 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
12141 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
12142 * config/i386/i386.h (processor_type): Ditto.
12143 * doc/invoke.texi: Ditto.
12145 2018-03-14 Jakub Jelinek <jakub@redhat.com>
12148 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
12149 INTEGER_CST offset, add it together with bitpos / 8 and
12150 sign extend based on POINTER_SIZE.
12154 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
12157 * config/i386/constraints.md (Yc): New register constraint.
12158 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
12159 Use Yc constraint for alternative 2 of operand 0. Remove
12160 preferred_for_speed attribute.
12162 2018-03-14 Richard Biener <rguenther@suse.de>
12164 PR tree-optimization/84830
12165 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
12166 with the old one to avoid oscillations.
12168 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
12171 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
12173 (assign_by_spills): Return a flag of reload assignment failure.
12174 Do not process the reload assignment failures. Do not spill other
12175 reload pseudos if they has the same reg class. Update n if
12177 (lra_assign): Add a return arg. Set up from the result of
12178 assign_by_spills call.
12179 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12180 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12181 usage_insns if it is not NULL.
12182 (spill_hard_reg_in_range): New function.
12183 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12184 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12185 function prototypes.
12186 (lra_assign): Change prototype.
12187 * lra.c (lra): Add code to deal with fails by splitting hard reg
12190 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
12192 * config/riscv/riscv.opt (mrelax): New option.
12193 * config/riscv/riscv.c (riscv_file_start): Emit ".option
12194 "norelax" when riscv_mrelax is disabled.
12195 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
12197 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12200 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
12201 reassociation for int modes.
12203 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12205 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
12206 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
12208 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
12209 * config/aarch64/aarch64-sve.md
12210 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
12211 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
12212 (*extend<mode><Vwide>2): Rename to...
12213 (aarch64_sve_extend<mode><Vwide>2): ...this.
12214 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
12215 renaming the old pattern to...
12216 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
12218 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
12219 define_expand, renaming the old pattern to...
12220 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
12221 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
12222 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
12223 account when deciding which SVE instruction the optab should use.
12224 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
12226 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12228 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
12229 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
12230 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
12231 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
12232 (tlsdesc_small_<mode>): Turn a define_expand and use
12233 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
12234 (tlsdesc_small_advsimd_<mode>): ...this.
12235 (tlsdesc_small_sve_<mode>): New pattern.
12237 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12239 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
12240 (UNSPEC_UMUL_HIGHPART): New constants.
12241 (MUL_HIGHPART): New int iteraor.
12242 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
12243 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
12245 (*<su>mul<mode>3_highpart): New define_insn.
12247 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
12250 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
12253 2018-03-13 Martin Liska <mliska@suse.cz>
12256 * (sem_item_optimizer::sem_item_optimizer): Initialize new
12258 (sem_item_optimizer::~sem_item_optimizer): Release it.
12259 (sem_item_optimizer::merge_classes): Register variable aliases.
12260 (sem_item_optimizer::fixup_pt_set): New function.
12261 (sem_item_optimizer::fixup_points_to_sets): Likewise.
12262 * ipa-icf.h: Declare new variables and functions.
12264 2018-03-13 Jakub Jelinek <jakub@redhat.com>
12266 PR middle-end/84834
12267 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
12268 integer_pow2p@2 and test integer_pow2p in condition.
12269 (A < 0 ? C : 0): Similarly for @1.
12271 PR middle-end/84831
12272 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
12273 characters starting at p contain '\0' character, don't look beyond
12277 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
12278 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
12281 * reg-stack.c (change_stack): Change update_end var from int to
12282 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
12283 also call set_block_for_insn on the newly added insns and rescan.
12286 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
12287 on the last operand.
12290 * tree.c (stabilize_reference_1): Return save_expr (e) for
12291 STATEMENT_LIST even if it doesn't have side-effects.
12293 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
12295 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
12297 2018-03-12 Renlin Li <renlin.li@arm.com>
12299 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
12300 aarch64_output_scalar_simd_mov_immediate.
12302 2018-03-12 Martin Sebor <msebor@redhat.com>
12304 PR tree-optimization/83456
12305 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
12306 for perfectly overlapping calls to memcpy.
12307 (gimple_fold_builtin_memory_chk): Same.
12308 (gimple_fold_builtin_strcpy): Handle no-warning.
12309 (gimple_fold_builtin_stxcpy_chk): Same.
12310 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
12312 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
12314 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
12315 parameter. Use it for SFmode.
12316 (rs6000_function_arg_advance_1): Adjust.
12317 (rs6000_function_arg): Adjust.
12318 (rs6000_gimplify_va_arg): Pass false for that new parameter.
12320 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
12322 PR rtl-optimization/84169
12323 PR rtl-optimization/84780
12324 * combine.c (can_combine_p): Check for a 2-insn combination whether
12325 the destination register is used between the two insns, too.
12327 2018-03-12 Richard Biener <rguenther@suse.de>
12329 PR tree-optimization/84803
12330 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
12331 for refs DR analysis didn't process.
12333 2018-03-12 Richard Biener <rguenther@suse.de>
12335 PR tree-optimization/84777
12336 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
12337 force-vectorize loops ignore whether we are optimizing for size.
12339 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
12341 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
12342 (TARGET_MD_ASM_ADJUST): Define.
12344 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
12345 Kito Cheng <kito.cheng@gmail.com>
12346 Chung-Ju Wu <jasonwucj@gmail.com>
12348 * config/nds32/nds32.c (nds32_compute_stack_frame,
12349 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
12350 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
12351 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
12352 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
12353 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
12354 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
12355 * config/nds32/nds32.md (prologue, epilogue): Use macro
12356 NDS32_V3PUSH_AVAILABLE_P to do checking.
12358 2018-03-11 Jakub Jelinek <jakub@redhat.com>
12361 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
12362 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
12363 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
12364 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
12365 addition of most attributes on !orig_type_die or the attribute not
12366 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
12368 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12369 Chung-Ju Wu <jasonwucj@gmail.com>
12371 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
12372 __NDS32_VH__ macro.
12373 * config/nds32/nds32.opt (mvh): New option.
12375 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12376 Chung-Ju Wu <jasonwucj@gmail.com>
12378 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
12380 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
12381 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
12384 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12385 Chung-Ju Wu <jasonwucj@gmail.com>
12387 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
12389 * config/nds32/nds32-multiple.md (strlensi): New pattern.
12390 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
12392 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
12393 Kito Cheng <kito.cheng@gmail.com>
12394 Chung-Ju Wu <jasonwucj@gmail.com>
12396 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
12397 UNSPEC_FFMISM and UNSPEC_FLMISM.
12398 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
12399 for ffb, ffmism and flmism.
12400 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
12401 (unspec_ffmism): Ditto.
12402 (unspec_flmism): Ditto.
12403 (nds32_expand_builtin_impl): Check if string extension is available.
12404 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
12405 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
12407 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
12410 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
12413 * lra-assigns.c (assign_by_spills): Return a flag of reload
12414 assignment failure. Do not process the reload assignment
12415 failures. Do not spill other reload pseudos if they has the same
12417 (lra_assign): Add a return arg. Set up from the result of
12418 assign_by_spills call.
12419 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12420 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12421 usage_insns if it is not NULL.
12422 (spill_hard_reg_in_range): New function.
12423 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12424 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12425 function prototypes.
12426 (lra_assign): Change prototype.
12427 * lra.c (lra): Add code to deal with fails by splitting hard reg
12430 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
12433 * config/i386/i386.opt: Replace Enforcment with Enforcement.
12435 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
12438 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
12439 (dw_val_node): Add val_symbolic_view.
12440 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
12441 (symview_upper_bound): New.
12442 (new_line_info_table): Initialize symviews_since_reset.
12443 (dwarf2out_source_line): Count symviews_since_reset and set
12444 symview_upper_bound.
12445 (dw_val_equal_p): Handle symview.
12446 (add_AT_symview): New.
12447 (print_dw_val): Handle symview.
12448 (attr_checksum, attr_checksum_ordered): Likewise.
12449 (same_dw_val_p, size_of_die): Likewise.
12450 (value_format, output_die): Likewise.
12451 (add_high_low_attributes): Use add_AT_symview for entry_view.
12452 (dwarf2out_finish): Reset symview_upper_bound, clear
12455 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
12458 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
12459 Add strict argument and use it.
12460 (rs6000_split_multireg_move): Update for new strict argument.
12461 (mem_operand_gpr): Disallow all non-offsettable addresses.
12462 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
12464 2018-03-09 Jakub Jelinek <jakub@redhat.com>
12467 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
12468 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
12469 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
12472 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
12473 decl, use remap_type if we want to use the type.
12475 2018-03-09 Martin Sebor <msebor@redhat.com>
12477 PR tree-optimization/84526
12478 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
12480 (builtin_access::generic_overlap): Be prepared to handle non-array
12483 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
12485 PR rtl-optimization/84682
12486 * lra-constraints.c (process_address_1): Check is_address flag
12487 for address constraints.
12488 (process_alt_operands): Likewise.
12489 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
12490 preprocess_constraints.
12491 * recog.h (preprocess_constraints): Add oploc parameter.
12493 * recog.c (preprocess_constraints): Test address_operand for
12494 CT_ADDRESS constraints.
12496 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
12499 * lra-assigns.c (assign_by_spills): Return a flag of reload
12500 assignment failure. Do not process the reload assignment
12501 failures. Do not spill other reload pseudos if they has the same
12503 (lra_assign): Add a return arg. Set up from the result of
12504 assign_by_spills call.
12505 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12506 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12507 usage_insns if it is not NULL.
12508 (spill_hard_reg_in_range): New function.
12509 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12510 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12511 function prototypes.
12512 (lra_assign): Change prototype.
12513 * lra.c (lra): Add code to deal with fails by splitting hard reg
12516 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12519 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
12520 Accept complain bool parameter. Only emit errors if it is true.
12521 (arm_parse_cpu_option_name): Likewise.
12522 (arm_target_thumb_only): Adjust callers of the above.
12523 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
12524 prototype to take a default true bool parameter.
12525 (arm_parse_arch_option_name): Likewise.
12527 2018-03-09 David Malcolm <dmalcolm@redhat.com>
12528 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
12532 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
12533 * configure: Regenerate.
12534 * configure.ac ("linker --version-script option"): New.
12535 ("linker soname option"): New.
12537 2018-03-09 Richard Biener <rguenther@suse.de>
12539 PR tree-optimization/84775
12540 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
12541 immediate uses of predicate stmts and mark them modified.
12544 PR tree-optimization/84178
12545 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
12547 (version_loop_for_if_conversion): Delay update_ssa call.
12548 (tree_if_conversion): Delay update_ssa until after predicate
12551 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
12554 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
12555 when the function accesses prior frames.
12557 2018-03-08 Jakub Jelinek <jakub@redhat.com>
12560 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
12561 gen_llsym, otherwise call maybe_gen_llsym.
12563 PR inline-asm/84742
12564 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
12565 has ',' character inside of it.
12567 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12570 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
12571 as clobbering CC_REGNUM.
12573 2018-03-08 Richard Biener <rguenther@suse.de>
12575 PR middle-end/84552
12576 * tree-scalar-evolution.c: Include tree-into-ssa.h.
12577 (follow_copies_to_constant): Do not follow SSA names registered
12580 2018-03-08 Richard Biener <rguenther@suse.de>
12582 PR tree-optimization/84178
12583 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
12585 (version_loop_for_if_conversion): Delay update_ssa call.
12586 (tree_if_conversion): Delay update_ssa until after predicate
12589 2018-03-08 David Malcolm <dmalcolm@redhat.com>
12591 PR tree-optimization/84178
12592 * tree-if-conv.c (release_bb_predicate): Remove the
12593 the assertion that the stmts have NULL use_ops.
12594 Discard the statements, asserting that they haven't
12595 yet been added to a BB.
12597 2018-03-08 Richard Biener <rguenther@suse.de>
12599 PR tree-optimization/84746
12600 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
12601 (phi_translate): Pass in destination ANTIC_OUT set.
12602 (phi_translate_1): Likewise. For a simplified result lookup
12603 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
12604 (phi_translate_set): Adjust.
12605 (do_pre_regular_insertion): Likewise.
12606 (do_pre_partial_partial_insertion): Likewise.
12608 2018-03-08 Martin Liska <mliska@suse.cz>
12610 PR gcov-profile/84735
12611 * doc/gcov.texi: Document usage of profile files.
12612 * gcov-io.h: Document changes in the format.
12614 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
12618 * dwarf2out.c (struct dw_line_info_table): Update comments for
12620 (FORCE_RESET_NEXT_VIEW): New.
12621 (FORCE_RESETTING_VIEW_P): New.
12622 (RESETTING_VIEW_P): Check for -1 too.
12623 (ZERO_VIEW_P): Likewise.
12624 (new_line_info_table): Force-reset next view.
12625 (dwarf2out_begin_function): Likewise.
12626 (dwarf2out_source_line): Simplify zero_view_p initialization.
12627 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
12628 view directly. Omit view when omitting .loc at line 0.
12630 2018-03-08 Jakub Jelinek <jakub@redhat.com>
12632 PR tree-optimization/84740
12633 * tree-switch-conversion.c (process_switch): Call build_constructors
12634 only if info.phi_count is non-zero.
12636 PR tree-optimization/84739
12637 * tree-tailcall.c (find_tail_calls): Check call arguments against
12638 DECL_ARGUMENTS (current_function_decl) rather than
12639 DECL_ARGUMENTS (func) when checking for tail recursion.
12641 2018-03-07 Jakub Jelinek <jakub@redhat.com>
12643 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
12644 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
12645 Volker Reichelt's entry and add entries for people that perform
12646 GCC fuzzy testing and report numerous bugs.
12648 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
12651 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
12652 readonly data in sdata, if that is disabled.
12653 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
12654 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
12655 -mreadonly-in-sdata option.
12657 2018-03-07 Martin Sebor <msebor@redhat.com>
12659 PR tree-optimization/84468
12660 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
12661 basic block when looking for nul assignment.
12663 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
12666 * except.h (output_function_exception_table): Adjust prototype.
12667 * except.c (output_function_exception_table): Remove FNNAME parameter
12668 and add SECTION parameter. Ouput one part of the table at a time.
12669 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
12670 the first part of the exception table and emit unwind directives.
12671 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
12672 (i386_pe_seh_cold_init): Likewise.
12673 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
12674 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
12675 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
12676 (ix86_output_call_insn): Emit a nop in one more case for SEH.
12677 * config/i386/winnt.c: Include except.h.
12678 (struct seh_frame_state): Add reg_offset, after_prologue and
12679 in_cold_section fields.
12680 (i386_pe_seh_end_prologue): Set seh->after_prologue.
12681 (i386_pe_seh_cold_init): New function.
12682 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
12683 to seh->in_cold_section.
12684 (seh_emit_push): Record the offset of the push.
12685 (seh_emit_save): Record the offet of the save.
12686 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
12687 Test seh->after_prologue to disregard the epilogue.
12688 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
12689 (i386_pe_end_cold_function): New function.
12691 2018-03-07 Jakub Jelinek <jakub@redhat.com>
12694 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
12695 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
12698 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
12700 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
12701 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
12703 PR middle-end/84723
12704 * multiple_target.c: Include tree-inline.h and intl.h.
12705 (expand_target_clones): Diagnose and fail if node->definition and
12706 !tree_versionable_function_p (node->decl).
12708 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
12710 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
12712 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
12713 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
12714 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
12716 2018-03-06 Jakub Jelinek <jakub@redhat.com>
12719 * combine.c (try_combine): Use reg_or_subregno instead of handling
12720 just paradoxical SUBREGs and REGs.
12722 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
12724 * config/arc/arc.c (arc_finalize_pic): Remove function.
12725 (arc_must_save_register): We use single base PIC register, remove
12726 checks to save/restore the PIC register.
12727 (arc_expand_prologue): Likewise.
12728 * config/arc/arc-protos.h (arc_set_default_type_attributes):
12730 (arc_verify_short): Likewise.
12731 (arc_attr_type): Likewise.
12732 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
12733 (walk_stores): Likewise.
12734 (arc_address_cost): Make it static.
12735 (arc_verify_short): Likewise.
12736 (branch_dest): Likewise.
12737 (arc_attr_type): Likewise.
12738 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
12739 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
12740 (arc_final_prescan_insn): Remove inserting the nops due to
12741 hardware hazards. It is done in reorg step.
12742 (insn_length_variant_t): Remove.
12743 (insn_length_parameters_t): Likewise.
12744 (arc_insn_length_parameters): Likewise.
12745 (arc_get_insn_variants): Likewise.
12746 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
12748 2018-03-06 Jakub Jelinek <jakub@redhat.com>
12750 PR inline-asm/84683
12751 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
12754 PR tree-optimization/84687
12755 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
12757 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
12759 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12761 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
12762 Rename to ppc_speculation_barrier.
12763 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
12764 __builtin_ppc_speculation_barrier.
12766 2018-03-05 Jakub Jelinek <jakub@redhat.com>
12769 * combine.c (combine_simplify_rtx): Don't try to simplify if
12770 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
12773 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
12775 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
12776 to 32 bytes when compiling for POWER9.
12778 2018-03-05 Jakub Jelinek <jakub@redhat.com>
12781 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
12782 regparm >= 3 with no arg reg available also for calls with
12783 flag_force_indirect_call. Pass decl to ix86_function_regparm.
12786 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
12788 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
12790 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
12793 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
12795 2018-03-05 Richard Biener <rguenther@suse.de>
12797 PR tree-optimization/84486
12798 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
12799 When inserting a __builtin_assume_aligned call set the LHS
12800 SSA name alignment info accordingly.
12802 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
12804 PR tree-optimization/84114
12805 * config/aarch64/aarch64.c (aarch64_reassociation_width)
12806 Avoid reassociation of FLOAT_MODE addition.
12808 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
12810 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
12811 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
12812 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
12813 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
12814 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
12815 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
12816 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
12818 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
12819 __builtin_ia32_wbinvd): New builtins.
12820 (SPECIAL_ARGS2): New.
12821 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
12822 (SPECIAL_ARGS2): New.
12823 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
12824 (ix86_valid_target_attribute_inner_p): Ditto.
12825 (ix86_init_mmx_sse_builtins): Add special_args2.
12826 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
12827 TARGET_WBNOINVD_P): New.
12828 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
12829 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
12830 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
12831 * config/i386/immintrin.h (_wbinvd): New intrinsic.
12832 * config/i386/pconfigintrin.h: New file.
12833 * config/i386/wbnoinvdintrin.h: Ditto.
12834 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
12836 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
12838 2018-03-05 Richard Biener <rguenther@suse.de>
12840 PR tree-optimization/84670
12841 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
12843 (BB_VISITED_WITH_VISITED_SUCCS): New define.
12844 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
12845 (compute_antic_aux): Only assert the number of values in ANTIC_IN
12846 doesn't grow if all successors (recursively) were visited at least
12849 2018-03-05 Richard Biener <rguenther@suse.de>
12851 PR tree-optimization/84650
12852 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
12853 if executed in the loop pipeline.
12855 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
12857 * doc/configfiles.texi (Configuration Files): Move info about
12858 conditionalizing $target-protos.h to...
12859 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
12860 differs from $target-protos.h.
12862 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
12863 Chung-Ju Wu <jasonwucj@gmail.com>
12865 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
12866 * config/nds32/nds32-multiple.md (setmemsi): Define.
12867 * config/nds32/nds32-memory-manipulation.c
12868 (nds32_gen_dup_4_byte_to_word_value): New.
12869 (emit_setmem_word_loop): New.
12870 (emit_setmem_byte_loop): New.
12871 (nds32_expand_setmem_loop): New.
12872 (nds32_expand_setmem_loop_v3m): New.
12873 (nds32_expand_setmem_unroll): New.
12874 (nds32_expand_setmem): New.
12876 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12877 Chung-Ju Wu <jasonwucj@gmail.com>
12879 * config/nds32/nds32-memory-manipulation.c
12880 (nds32_emit_load_store): New.
12881 (nds32_emit_post_inc_load_store): New.
12882 (nds32_emit_mem_move): New.
12883 (nds32_emit_mem_move_block): New.
12884 (nds32_expand_movmemsi_loop_unknown_size): New.
12885 (nds32_expand_movmemsi_loop_known_size): New.
12886 (nds32_expand_movmemsi_loop): New.
12887 (nds32_expand_movmemsi_unroll): New.
12888 (nds32_expand_movmemqi): Rename ...
12889 (nds32_expand_movmemsi): ... to this.
12890 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
12891 (movmemsi): ... to this.
12892 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
12893 (nds32_expand_movmemsi): ... to this.
12895 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12896 Monk Chiang <sh.chiang04@gmail.com>
12897 Chung-Ju Wu <jasonwucj@gmail.com>
12899 * config/nds32/nds32-protos.h
12900 (nds32_expand_load_multiple): New arguments.
12901 (nds32_expand_store_multiple): Ditto.
12902 (nds32_valid_multiple_load_store): Rename ...
12903 (nds32_valid_multiple_load_store_p): ... to this.
12904 * config/nds32/nds32-memory-manipulation.c
12905 (nds32_expand_load_multiple): Refine implementation.
12906 (nds32_expand_store_multiple): Ditto.
12907 * config/nds32/nds32-multiple.md
12908 (load_multiple): Update nds32_expand_load_multiple interface.
12909 (store_multiple): Update nds32_expand_store_multiple interface.
12910 * config/nds32/nds32-predicates.c
12911 (nds32_valid_multiple_load_store): Rename ...
12912 (nds32_valid_multiple_load_store_p): ... to this and refine
12914 * config/nds32/predicates.md
12915 (nds32_load_multiple_and_update_address_operation): New predicate.
12916 (nds32_store_multiple_and_update_address_operation): New predicate.
12918 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12919 Chung-Ju Wu <jasonwucj@gmail.com>
12921 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
12922 (combo): New attribute.
12923 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
12925 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
12927 * config/nds32/nds32.opt: Change -mcmodel= default value.
12929 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
12930 Monk Chiang <sh.chiang04@gmail.com>
12931 Chung-Ju Wu <jasonwucj@gmail.com>
12933 * config/nds32/constants.md (unspec_element): New enum.
12934 * config/nds32/constraints.md (Umw): New constraint.
12935 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
12936 * config/nds32/nds32-intrinsic.md: Likewise.
12937 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
12938 (nds32_valid_smw_lwm_base_p): New.
12939 (nds32_output_smw_single_word): New.
12940 (nds32_output_lmw_single_word): New.
12941 (nds32_expand_unaligned_load): New.
12942 (nds32_expand_unaligned_store): New.
12943 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
12944 (nds32_output_smw_single_word): Declare.
12945 (nds32_output_lmw_single_word): Declare.
12946 (nds32_expand_unaligned_load): Declare.
12947 (nds32_expand_unaligned_store): Declare.
12948 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
12949 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
12950 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
12951 NDS32_BUILTIN_UASTORE_DW.
12952 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
12955 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
12956 Kito Cheng <kito.cheng@gmail.com>
12957 Chung-Ju Wu <jasonwucj@gmail.com>
12959 * config/nds32/nds32-intrinsic.c
12960 (nds32_expand_builtin_null_ftype_reg): Delete.
12961 (nds32_expand_builtin_reg_ftype_imm): Ditto.
12962 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
12963 (nds32_read_argument): New.
12964 (nds32_legitimize_target): Ditto.
12965 (nds32_legitimize_argument): Ditto.
12966 (nds32_check_constant_argument): Ditto.
12967 (nds32_expand_unop_builtin): Ditto.
12968 (nds32_expand_unopimm_builtin): Ditto.
12969 (nds32_expand_binop_builtin): Ditto.
12970 (nds32_builtin_decl_impl): Ditto.
12971 (builtin_description): Ditto.
12972 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
12973 (nds32_init_builtins_impl): Ditto.
12974 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
12975 (nds32_builtin_decl): New.
12976 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
12977 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
12979 2018-03-02 Jeff Law <law@redhat.com>
12981 * reorg.c (stop_search_p): Handle DEBUG_INSN.
12982 (redundant_insn, fill_simple_delay_slots): Likewise.
12983 (fill_slots_from_thread): Likewise.
12984 * resource.c (mark_referenced_resources): Likewise.
12985 (mark_set_resources, find_dead_or_set_registers): Likewise.
12987 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12989 * substring-locations.h (format_warning_va): Formatting fix for
12990 ATTRIBUTE_GCC_DIAG.
12991 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
12993 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
12994 * substring-locations.c: Include intl.h.
12995 (format_warning_va): Turned into small wrapper around
12996 format_warning_n_va, renamed to ...
12997 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
12998 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
13000 (format_warning_at_substring_n): New function.
13001 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
13002 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
13003 format_warning_at_substring with just a shorter name instead of
13004 const function pointer.
13005 (fmtwarn_n): New function.
13006 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
13007 appropriate, get rid of all the fmtstr temporaries, move conditionals
13008 with G_() wrapped string literals directly into fmtwarn arguments,
13009 cast dir.len to (int), formatting fixes.
13011 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
13013 * doc/invoke.texi: Remove "Cilk Plus" references.
13015 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13016 Richard Biener <rguenther@suse.de>
13019 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
13020 for error or warning attributes if CALL_FROM_THUNK_P is set.
13023 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13026 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
13027 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
13030 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
13031 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
13033 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
13034 instead of -1U in last predictors element's probability member.
13036 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
13039 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
13040 arguments if they are comparable.
13042 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
13044 PR tree-optimization/84634
13045 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
13046 masks and masked_loop_p with a single loop_masks, making sure it's
13047 null for bb vectorization.
13049 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
13051 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
13052 (vect_analyze_data_ref_access): Use loop->safe_len rather than
13053 loop->force_vectorize to check whether there is no alias.
13055 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13058 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
13060 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
13062 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
13063 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
13064 instead of a loop around prev_real_insn.
13065 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
13068 PR inline-asm/84625
13069 * config/i386/i386.c (ix86_print_operand): Use conditional
13070 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
13073 2018-03-02 Richard Biener <rguenther@suse.de>
13075 PR tree-optimization/84427
13076 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
13077 (bitmap_set_subtract_values): Rewrite to handle multiple
13080 (prune_clobbered_mems): Likewise.
13081 (phi_translate): Take edge instead of pred/phiblock.
13082 (phi_translate_1): Likewise.
13083 (phi_translate_set): Likewise. Insert all translated
13084 exprs for a value into the set, keeping possibly multiple
13085 expressions per value.
13086 (compute_antic_aux): Adjust for phi_translate changes.
13087 When intersecting union the expressions and prune those
13088 not in the final value set, keeping possibly multiple
13089 expressions per value. Do not use value-insertion
13090 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
13091 all expressions. Add verification that the value-sets
13092 only shrink during iteration.
13093 (compute_partial_antic_aux): Adjust for the phi_translate changes.
13094 (do_pre_regular_insertion): Likewise.
13095 (do_pre_partial_partial_insertion): Likewise.
13097 2018-03-02 Richard Biener <rguenther@suse.de>
13100 * config/darwin.c (saved_debug_info_level): New static global.
13101 (darwin_asm_lto_start): Disable debug info generation for LTO out.
13102 (darwin_asm_lto_end): Restore debug info generation settings.
13104 2018-03-01 Martin Liska <mliska@suse.cz>
13107 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13108 volatile arguments.
13110 2018-03-01 Richard Biener <rguenther@suse.de>
13113 * dwarf2out.c (gen_variable_die): Properly handle late VLA
13114 type annotation with LTO when debug was disabled at compile-time.
13116 2018-03-01 Matthew Fortune <mfortune@gmail.com>
13118 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
13120 (mips_final_postscan_insn): Likewise.
13122 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
13124 PR rtl-optimization/84528
13125 * alias.c (init_alias_target): Add commentary.
13126 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
13127 a unique base value if the frame pointer is not eliminated
13128 to the stack pointer.
13130 2018-03-01 Tom de Vries <tom@codesourcery.com>
13132 PR rtl-optimization/83327
13133 * lra-int.h (hard_regs_spilled_into): Declare.
13134 * lra.c (hard_regs_spilled_into): Define.
13135 (init_reg_info): Init hard_regs_spilled_into.
13136 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
13137 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
13138 (process_bb_lives): Handle hard_regs_spilled_into.
13139 (lra_create_live_ranges_1): Before doing liveness propagation, clear
13140 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
13142 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
13144 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
13145 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
13146 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
13147 * config/rs6000/aix72.h: New file.
13149 2018-02-28 Jakub Jelinek <jakub@redhat.com>
13151 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
13152 instead of warning_at with conditional singular and plural messages
13156 * stor-layout.c (update_alignment_for_field): For
13157 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
13158 && !DECL_PACKED (field), do the alignment update, just use
13159 only desired_align instead of MAX (type_align, desired_align)
13161 (place_field): Don't do known_align < desired_align handling
13162 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
13163 is non-NULL, instead do it after rli->prev_field handling and
13164 only if not within a bitfield word. For DECL_PACKED (field)
13165 use type_align of BITS_PER_UNIT.
13167 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
13169 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
13170 superfluous parentheses and trailing spaces.
13172 2018-02-28 Richard Biener <rguenther@suse.de>
13174 PR tree-optimization/84584
13175 * graphite-scop-detection.c (scop_detection::add_scop): Discard
13176 SCoPs with fake exit edge.
13178 2018-02-28 Martin Liska <mliska@suse.cz>
13181 * timevar.c (timer::print): Fix format to properly print 100%
13184 2018-02-28 Richard Biener <rguenther@suse.de>
13186 PR middle-end/84607
13187 * genmatch.c (capture_info::walk_match): Do not mark
13188 captured expressions without operands as expr_p given
13189 they act more like predicates and should be subject to
13190 "lost tail" side-effect preserving.
13192 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
13194 PR rtl-optimization/81611
13195 * auto-inc-dec.c (attempt_change): Move dead note from
13196 mem_insn if it's the next use of regno
13197 (find_address): Take address use of reg holding
13198 non-incremented value. Add parm to limit search to the named
13200 (merge_in_block): Attempt to use a mem insn that is the next
13201 use of the original regno.
13203 2018-02-27 Martin Sebor <msebor@redhat.com>
13206 * doc/invoke.texi (-Wmissing-attributes): New option.
13207 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
13209 2018-02-27 Martin Sebor <msebor@redhat.com>
13211 PR translation/84207
13212 * diagnostic-core.h (warning_n, error_n, inform_n): Change
13213 n argument to unsigned HOST_WIDE_INT.
13214 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
13215 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
13216 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
13217 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
13219 2018-02-27 Richard Biener <rguenther@suse.de>
13221 PR tree-optimization/84512
13222 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
13223 Do not use the estimate returned from record_stmt_cost for
13224 the scalar iteration cost but sum properly using add_stmt_cost.
13226 2018-02-27 Richard Biener <rguenther@suse.de>
13228 PR tree-optimization/84466
13229 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
13230 Adjust last change to less strictly validate use operands.
13232 2018-02-27 Martin Liska <mliska@suse.cz>
13234 PR gcov-profile/84548
13235 * gcov.c (process_file): Allow partial overlap and consider it
13236 also as group functions.
13237 (output_lines): Properly calculate range of lines for a group.
13239 2018-02-27 Martin Liska <mliska@suse.cz>
13241 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
13242 'ggc' suffixes. Change first column width.
13243 (timer::print): Fix formatting of the column.
13245 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
13247 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
13248 preserve inline entry blocks for the sake of debug inline
13249 entry point markers alone.
13250 (remove_unused_locals): Suggest in comments a better place to
13251 force the preservation of inline entry blocks that are
13252 otherwise unused, but do not preserve them.
13254 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
13256 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
13258 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
13261 * config/i386/constraints.md (Bs): Replace
13262 ix86_indirect_branch_register with
13263 TARGET_INDIRECT_BRANCH_REGISTER.
13265 * config/i386/i386.md (indirect_jump): Likewise.
13266 (tablejump): Likewise.
13267 (*sibcall_memory): Likewise.
13268 (*sibcall_value_memory): Likewise.
13269 Peepholes of indirect call and jump via memory: Likewise.
13270 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
13271 (*sibcall_value_GOT_32): Likewise.
13272 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13273 (GOT_memory_operand): Likewise.
13274 (call_insn_operand): Likewise.
13275 (sibcall_insn_operand): Likewise.
13276 (GOT32_symbol_operand): Likewise.
13277 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
13279 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
13281 PR rtl-optimization/83496
13282 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
13283 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
13284 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
13285 redundant insn, if any.
13286 (relax_delay_slots): Likewise.
13287 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
13289 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
13291 PR tree-optimization/83965
13292 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
13293 that grouped statements are part of a reduction chain. Return
13294 true if the statement is not marked as a reduction itself but
13295 is part of a group.
13296 (vect_recog_dot_prod_pattern): Don't check whether the statement
13297 is part of a group here.
13298 (vect_recog_sad_pattern): Likewise.
13299 (vect_recog_widen_sum_pattern): Likewise.
13301 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
13304 * final.c (rest_of_clean_state): Also look for calls inside sequences.
13306 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
13309 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
13311 (ix86_output_indirect_function_return): New prototype.
13312 (ix86_split_simple_return_pop_internal): Likewise.
13313 * config/i386/i386.c (indirect_return_via_cx): New.
13314 (indirect_return_via_cx_bnd): Likewise.
13315 (indirect_thunk_name): Handle return va CX_REG.
13316 (output_indirect_thunk_function): Create alias for
13317 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
13318 (ix86_output_indirect_jmp): Remove the bool argument.
13319 (ix86_output_indirect_function_return): New function.
13320 (ix86_split_simple_return_pop_internal): Likewise.
13321 * config/i386/i386.md (*indirect_jump): Don't pass false
13322 to ix86_output_indirect_jmp.
13323 (*tablejump_1): Likewise.
13324 (simple_return_pop_internal): Change it to define_insn_and_split.
13325 Call ix86_split_simple_return_pop_internal to split it for
13326 -mfunction-return=.
13327 (simple_return_indirect_internal): Call
13328 ix86_output_indirect_function_return instead of
13329 ix86_output_indirect_jmp.
13331 2018-02-26 Jakub Jelinek <jakub@redhat.com>
13334 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
13335 memset and value initialization afterwards.
13337 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
13339 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
13341 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
13344 * common/config/aarch64/aarch64-common.c
13345 (aarch_option_optimization_table[]): Switch
13346 off fomit-frame-pointer
13348 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
13349 Chung-Ju Wu <jasonwucj@gmail.com>
13351 * config/nds32/nds32-multiple.md (load_multiple): Disallow
13353 (store_multiple): Ditto.
13355 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
13357 * config.gcc: Add --with-cpu support for nds32 target.
13358 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
13359 * config/nds32/nds32.opt: Add -mcpu= option.
13361 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
13363 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
13364 isel=yes): Warn for these deprecated options.
13366 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
13368 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
13369 ISA_2_5_MASKS_EMBEDDED.
13371 2018-02-23 Jakub Jelinek <jakub@redhat.com>
13373 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
13374 p->max as pointers rather than using iterative_hash_expr.
13376 2018-02-23 Carl Love <cel@us.ibm.com>
13378 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
13379 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
13381 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
13382 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
13383 P8V_BUILTIN_VEC_VUNSIGNED2.
13385 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
13388 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
13389 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
13391 * lra-constraints.c (curr_insn_transform): Set up
13392 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
13394 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
13395 * lra-eliminations.c (spill_pseudos): Ditto.
13396 (process_insn_for_elimination): Ditto.
13397 * lra-lives.c (reg_early_clobber_p): Use the new macros.
13398 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
13399 LRA_NON_CLOBBERED_ALT.
13401 2018-02-22 Martin Sebor <msebor@redhat.com>
13403 PR tree-optimization/84480
13404 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
13405 to maybe_diag_stxncpy_trunc. Call it.
13406 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
13407 from gimple_fold_builtin_strcpy. Print inlining stack.
13408 (handle_builtin_stxncpy): Print inlining stack.
13409 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
13411 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
13414 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
13415 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
13416 and -fcheck-pointer-bounds are used together.
13417 (indirect_thunk_prefix): New enum.
13418 (indirect_thunk_need_prefix): New function.
13419 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
13420 "_nt" instead of "_bnd" for NOTRACK prefix.
13421 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
13422 (output_indirect_thunk_function): Likewise.
13424 (ix86_code_end): Update output_indirect_thunk_function calls.
13425 (ix86_output_indirect_branch_via_reg): Replace
13426 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
13427 (ix86_output_indirect_branch_via_push): Likewise.
13428 (ix86_output_function_return): Likewise.
13429 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
13430 incompatible with -fcf-protection=branch and
13431 -fcheck-pointer-bounds.
13433 2018-02-22 Steve Ellcey <sellcey@cavium.com>
13436 * config/aarch64/aarch64.c (aarch64_print_address_internal):
13437 Change gcc_assert call to output_operand_lossage.
13439 2018-02-22 Steve Ellcey <sellcey@cavium.com>
13441 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
13443 2018-02-22 DJ Delorie <dj@redhat.com>
13444 Sebastian Perta <sebastian.perta@renesas.com>
13445 Oleg Endo <olegendo@gcc.gnu.org>
13447 * config/rx/rx.c (rx_rtx_costs): New function.
13448 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
13450 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
13452 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
13454 2018-02-22 Martin Liska <mliska@suse.cz>
13457 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
13458 Add "native" as a possible value.
13460 2018-02-22 Martin Liska <mliska@suse.cz>
13463 * config/i386/i386.c (ix86_option_override_internal):
13464 Add "native" as a possible value for -march and -mtune.
13466 2018-02-22 Jakub Jelinek <jakub@redhat.com>
13469 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
13470 to all type variants.
13472 PR tree-optimization/84503
13473 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
13474 width as info->bitpos + info->bitsize - start.
13475 (merged_store_group::merge_overlapping): Simplify width computation.
13476 (check_no_overlap): New function.
13477 (imm_store_chain_info::try_coalesce_bswap): Compute expected
13478 start + width and last_order of the group, fail if check_no_overlap
13480 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
13481 to group if check_no_overlap fails.
13483 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
13485 * config/rs6000/altivec.md: Delete contraint arguments to
13486 define_expand, define_split, and define_peephole2, and in
13487 define_insn_and_split if always unused.
13488 * config/rs6000/darwin.md: Ditto.
13489 * config/rs6000/dfp.md: Ditto.
13490 * config/rs6000/rs6000.md: Ditto.
13491 * config/rs6000/sync.md: Ditto.
13492 * config/rs6000/vector.md: Ditto.
13493 * config/rs6000/vsx.md: Ditto.
13495 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
13497 * config/rs6000/altivec.md: Write output control strings as braced
13498 blocks instead of double-quoted strings.
13499 * config/rs6000/darwin.md: Ditto.
13500 * config/rs6000/rs6000.md: Ditto.
13501 * config/rs6000/vector.md: Ditto.
13502 * config/rs6000/vsx.md: Ditto.
13504 2018-02-21 Jason Merrill <jason@redhat.com>
13506 PR c++/84314 - ICE with templates and fastcall attribute.
13507 * attribs.c (build_type_attribute_qual_variant): Remove assert.
13509 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
13511 * ipa-cp.c (determine_versionability): Fix comment typos.
13513 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
13516 * ipa-cp.c (determine_versionability): Do not version functions caling
13519 2018-02-21 Martin Liska <mliska@suse.cz>
13522 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
13523 Add "native" as a possible value.
13524 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
13525 the macro when native cpu detection is available.
13527 2018-02-21 Martin Liska <mliska@suse.cz>
13530 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
13531 Add "native" as a possible value.
13532 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
13533 when native cpu detection is available.
13535 2018-02-21 Jakub Jelinek <jakub@redhat.com>
13536 Martin Sebor <msebor@redhat.com>
13538 PR tree-optimization/84478
13539 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
13541 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
13542 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
13543 support which is conservatively correct, for 2 only stay conservative
13544 for maxlen. Formatting and comment capitalization fixes. Add STRICT
13545 argument to the 2 argument get_range_strlen, adjust 6 arg
13546 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
13548 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
13549 (gimple_fold_builtin_strlen): Pass true as last argument to
13552 2018-02-20 Martin Sebor <msebor@redhat.com>
13554 PR middle-end/84095
13555 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
13556 (builtin_memref::set_base_and_offset): Same. Handle inner references.
13557 (builtin_memref::builtin_memref): Factor out parts into
13558 set_base_and_offset and call it.
13560 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
13562 PR middle-end/84406
13563 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
13564 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
13565 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
13566 search at the associated MODE_INT.
13568 2018-02-20 Jeff Law <law@redhat.com>
13570 PR middle-end/82123
13571 PR tree-optimization/81592
13572 PR middle-end/79257
13573 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
13574 for range data rather than using global data.
13575 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
13576 range data rather than using global data.
13577 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
13578 pass it to children as needed.
13579 (struct directive::fmtresult): Similarly.
13580 (struct directive::set_width): Similarly.
13581 (struct directive::set_precision): Similarly.
13582 (format_integer, format_directive, parse_directive): Similarly.
13583 (format_none): Accept unnamed vr_values parameter.
13584 (format_percent, format_floating, format_character): Similarly.
13585 (format_string, format_plain): Similarly.
13586 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
13587 the EVRP range analyzer for range data rather than using global data.
13588 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
13589 gimple-ssa-evrp-analyze.h
13590 (class sprintf_dom_walker): Add after_dom_children member function.
13591 Add evrp_range_analyzer member.
13592 (sprintf_dom_walker::before_dom_children): Call into the EVRP
13593 range analyzer as needed.
13594 (sprintf_dom_walker::after_dom_children): New member function.
13595 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
13597 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
13598 (evrp_range_analyzer::pop_to_marker): Likewise.
13600 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
13602 PR tree-optimization/84419
13603 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
13604 with the required type if its current type is compatible but
13607 2018-02-20 Jakub Jelinek <jakub@redhat.com>
13609 PR middle-end/82004
13610 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
13611 after vectorization.
13613 2018-02-20 Martin Liska <mliska@suse.cz>
13616 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
13617 possible values if we don't have a hint.
13619 2018-02-20 Martin Liska <mliska@suse.cz>
13623 * final.c (shorten_branches): Build align_tab array with one
13625 * opts.c (finish_options): Add alignment option limit check.
13626 (MAX_CODE_ALIGN): Likewise.
13627 (MAX_CODE_ALIGN_VALUE): Likewise.
13628 * doc/invoke.texi: Document maximum allowed option value for
13629 all -falign-* options.
13631 2018-02-19 Jakub Jelinek <jakub@redhat.com>
13634 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
13635 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
13636 * var-tracking.c (emit_note_insn_var_location): Remove all references
13637 to NOTE_INSN_CALL_ARG_LOCATION.
13638 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
13639 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
13640 Use copy_rtx_if_shared.
13641 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
13642 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
13643 (dwarf2out_var_location): Remove handling of
13644 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
13646 * final.c (final_scan_insn): Remove all references to
13647 NOTE_INSN_CALL_ARG_LOCATION.
13648 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
13649 before dumping final insns.
13650 * except.c (emit_note_eh_region_end): Remove all references to
13651 NOTE_INSN_CALL_ARG_LOCATION.
13652 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
13653 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
13654 * config/arc/arc.c (hwloop_optimize): Likewise.
13655 * config/arm/arm.c (create_fix_barrier): Likewise.
13656 * config/s390/s390.c (s390_chunkify_start): Likewise.
13657 * config/sh/sh.c (find_barrier): Likewise.
13658 * config/i386/i386.c (rest_of_insert_endbranch,
13659 ix86_seh_fixup_eh_fallthru): Likewise.
13660 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
13661 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
13662 * config/frv/frv.c (frv_function_prologue): Likewise.
13663 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
13665 (note_outside_basic_block_p): Remove all references to
13666 NOTE_INSN_CALL_ARG_LOCATION.
13667 * gengtype.c (adjust_field_rtx_def): Likewise.
13668 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
13670 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
13671 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
13674 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
13677 PR tree-optimization/84452
13678 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
13679 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
13682 2018-02-19 Martin Liska <mliska@suse.cz>
13685 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
13687 2018-02-19 Martin Liska <mliska@suse.cz>
13688 Richard Sandiford <richard.sandiford@linaro.org>
13690 PR tree-optimization/82491
13691 * gimple-fold.c (get_base_constructor): Make earlier bail out
13694 2018-02-19 Carl Love <cel@us.ibm.com>
13696 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
13697 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
13699 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
13700 P8V_BUILTIN_VEC_NEG.
13702 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
13704 * config/rl78/rl78.md (movdf): New define expand.
13706 2018-02-19 Martin Liska <mliska@suse.cz>
13709 * doc/invoke.texi: Fix typo.
13710 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
13712 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
13714 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
13715 handle rs6000_single_float and rs6000_double_float specially for
13718 2018-02-16 Jeff Law <law@redhat.com>
13720 * config/rx/rx.c (add_pop_cfi_notes): New function.;
13721 (pop_regs): Use it.
13723 2018-02-16 Jakub Jelinek <jakub@redhat.com>
13726 * ipa-inline.c (inline_small_functions): Fix a typo.
13728 2018-02-16 Nathan Sidwell <nathan@acm.org>
13730 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
13732 2018-02-16 Carl Love <cel@us.ibm.com>
13734 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
13735 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
13736 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
13737 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
13738 expansion to P8V_BUILTIN_VEC_FLOAT2.
13740 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
13742 PR rtl-optimization/70023
13743 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
13744 src_regno into account.
13746 2018-02-16 Carl Love <cel@us.ibm.com>
13748 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
13749 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
13750 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
13751 * config/rs6000/rs6000.c: Remove case statements for
13752 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
13753 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
13754 and P9V_BUILTIN_VEC_VINSERT4B.
13755 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
13756 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
13757 * config/rs6000/vsx.md:
13758 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
13761 2018-02-16 Carl Love <cel@us.ibm.com>
13763 * config/rs6000/altivec.h: Add builtin names vec_extract4b
13765 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
13767 * config/rs6000/rs6000-c.c: Add the definitions for
13768 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
13769 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
13770 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
13771 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
13772 definition for insert4b and define insn *insert3b_internal.
13773 * doc/extend.texi: Add documentation for vec_extract4b.
13775 2018-02-16 Nathan Sidwell <nathan@acm.org>
13777 * doc/extend.texi (Backwards Compatibility): Mention friend
13778 injection. Note for-scope is deprecated.
13779 * doc/invoke.texi (-ffriend-injection): Deprecate.
13781 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
13783 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
13784 that moved to I2, also allow destinations that are a paradoxical
13785 subreg (instead of a normal reg).
13787 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
13790 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
13793 2018-02-16 Richard Biener <rguenther@suse.de>
13795 PR tree-optimization/84037
13796 PR tree-optimization/84016
13798 * config/i386/i386.c (ix86_builtin_vectorization_cost):
13799 Adjust vec_construct for the fact we need additional higher latency
13800 128bit inserts for AVX256 and AVX512 vector builds.
13801 (ix86_add_stmt_cost): Scale vector construction cost for
13804 2018-02-16 Richard Biener <rguenther@suse.de>
13806 PR tree-optimization/84417
13807 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
13808 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
13809 (non_rewritable_lvalue_p): Likewise, use poly-ints.
13811 2018-02-16 Martin Liska <mliska@suse.cz>
13814 * internal-fn.def (ASAN_CHECK): Set proper flags.
13815 (ASAN_MARK): Likewise.
13817 2018-02-16 Julia Koval <julia.koval@intel.com>
13819 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
13820 from PTA_CANNONLAKE.
13822 2018-02-16 Jakub Jelinek <jakub@redhat.com>
13825 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
13826 Use ++iter rather than iter++ for std::list iterators.
13827 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
13828 defer deleting them until all nodes in the forest are processed. Do
13829 free even leaf nodes. Change to_process into auto_vec.
13832 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
13833 * vec.h (vec_default_construct): Use memset instead of placement new
13834 if BROKEN_VALUE_INITIALIZATION is defined.
13835 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
13836 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
13839 PR rtl-optimization/83723
13840 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
13841 * lra.c (lra_substitute_pseudo): Likewise. If true, use
13842 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
13844 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
13846 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
13848 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
13850 PR rtl-optimization/81443
13851 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
13852 from inner REGs to paradoxical SUBREGs.
13854 2018-02-16 Richard Biener <rguenther@suse.de>
13856 PR tree-optimization/84399
13857 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
13858 For operands we can analyze at their definition make sure we can
13859 analyze them at each use as well.
13861 2018-02-16 Richard Biener <rguenther@suse.de>
13863 PR tree-optimization/84190
13864 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
13865 volatile accesses if the decl isn't volatile.
13867 2018-02-15 Jason Merrill <jason@redhat.com>
13869 PR c++/84314 - ICE with templates and fastcall attribute.
13870 * attribs.c (build_type_attribute_qual_variant): Don't clobber
13871 TYPE_CANONICAL on an existing type.
13873 2018-02-15 Jakub Jelinek <jakub@redhat.com>
13875 PR tree-optimization/84383
13876 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
13877 dstoff nor call operand_equal_p if dstbase is NULL.
13879 PR tree-optimization/84334
13880 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
13881 also a CONSTANT_CLASS_P, punt.
13883 2018-02-14 Jim Wilson <jimw@sifive.com>
13885 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
13886 first SMALL_OPERAND check. New local min_second_step. Move assert
13887 to where locals are set. Add TARGET_RVC support.
13888 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
13890 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
13892 * doc/invoke.texi: Correct -Wformat-overflow code sample.
13894 2018-02-14 Martin Sebor <msebor@redhat.com>
13896 PR tree-optimization/83698
13897 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
13898 arrays constrain the offset range to their bounds.
13899 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
13900 (builtin_access::overlap): Avoid setting the size of overlap if it's
13902 (maybe_diag_overlap): Also consider arrays when deciding what values
13903 of offsets to include in diagnostics.
13905 2018-02-14 Martin Sebor <msebor@redhat.com>
13908 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
13909 that correspond to the kind of a declaration.
13911 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
13914 * config/pa/pa.md: Load address of PIC label using the linkage table
13915 if the label is nonlocal.
13917 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
13919 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
13920 warning message if user requests -maltivec=be.
13921 * doc/invoke.texi: Document deprecation of -maltivec=be.
13923 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
13926 * config/rs6000/rs6000-c.c: Update definitions for
13927 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
13928 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
13930 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13933 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
13934 add _get_ssp intrinsics. Remove argument from
13935 __builtin_ia32_rdssp[d|q].
13936 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
13937 * config/i386/i386-builtin.def: Remove argument from
13938 __builtin_ia32_rdssp[d|q].
13939 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
13940 ix86_expand_special_args_builtin for _rdssp[d|q].
13941 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
13942 Clear register before usage.
13943 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
13944 Add documentation for new _get_ssp and _inc_ssp intrinsics.
13946 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
13948 PR tree-optimization/84357
13949 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
13950 operand 1 of an ARRAY_REF too.
13952 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
13955 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
13956 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
13958 (set_of_reg): New struct.
13959 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
13960 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
13961 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
13963 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
13964 Split into bitclr, bitset, bitinvert patterns if appropriate.
13965 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
13966 use rx_fuse_in_memory_bitop.
13967 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
13968 to named insn, correct maximum insn length.
13970 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
13973 * machmode.def: Define a complex mode for PARTIAL_INT.
13974 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
13976 * doc/rtl.texi: Document CSPImode.
13977 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
13979 (msp430_hard_regno_nregs_with_padding): Likewise.
13981 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
13984 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
13986 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
13988 PR rtl-optimization/84169
13989 * combine.c (try_combine): New variable split_i2i3. Set it to true if
13990 we generated a parallel as new i3 and we split that to new i2 and i3
13991 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
13992 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
13993 those to i2, not i1. Partially rewrite this scan code.
13995 2018-02-13 Jakub Jelinek <jakub@redhat.com>
13998 * stor-layout.c (place_field): For variable length fields, adjust
13999 offset_align afterwards not just based on the field's alignment,
14000 but also on the size.
14002 PR middle-end/84309
14003 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
14004 of exps and logs in the use_exp2 case.
14006 2018-02-13 Jeff Law <law@redhat.com>
14008 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
14009 entry for "vector".
14011 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
14014 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
14018 * common.opt (gas-loc-support, gas-locview-support): New.
14019 (ginline-points, ginternal-reset-location-views): New.
14020 * doc/invoke.texi: Document them. Use @itemx where intended.
14021 (gvariable-location-views): Adjust.
14022 * target.def (reset_location_view): New.
14023 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
14024 (TARGET_RESET_LOCATION_VIEW): New.
14025 * doc/tm.texi: Rebuilt.
14026 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
14027 (dwarf2out_default_as_locview_support): New.
14028 (output_asm_line_debug_info): Use option variables.
14029 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
14030 (output_loc_list): Likewise.
14031 (add_high_low_attributes): Check option variables.
14032 Don't output entry view attribute in strict mode.
14033 (gen_inlined_subroutine_die): Check option variables.
14034 (dwarf2out_inline_entry): Likewise.
14035 (init_sections_and_labels): Likewise.
14036 (dwarf2out_early_finish): Likewise.
14037 (maybe_reset_location_view): New, from...
14038 (dwarf2out_var_location): ... here. Call it.
14039 * debug.h (dwarf2out_default_as_loc_support): Declare.
14040 (dwarf2out_default_as_locview_support): Declare.
14041 * hooks.c (hook_int_rtx_insn_0): New.
14042 * hooks.h (hook_int_rtx_insn_0): Declare.
14043 * toplev.c (process_options): Take -gas-loc-support and
14044 -gas-locview-support from dwarf2out. Enable
14045 -gvariable-location-views by default only with locview
14046 assembler support. Enable -ginternal-reset-location-views by
14047 default only if the target defines the corresponding hook.
14048 Enable -ginline-points by default if location views are
14049 enabled; force it disabled if statement frontiers are
14051 * tree-inline.c (expand_call_inline): Check option variables.
14052 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
14054 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
14056 PR tree-optimization/84321
14057 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
14058 handling. Also check whether the anti-range contains any values
14059 that satisfy the mask; switch to a VR_RANGE if not.
14061 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
14064 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
14066 2018-02-13 Martin Jambor <mjambor@suse.cz>
14069 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
14070 of call statements, also set location of a load to a temporary.
14072 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
14074 * config/rl78/rl78.c (add_vector_labels): New function.
14075 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
14076 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
14077 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
14078 which checks that no arguments are passed.
14079 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
14080 * doc/extend.texi: Documentation for the new attribute.
14082 2018-02-13 Andreas Schwab <schwab@suse.de>
14084 * config/riscv/linux.h (CPP_SPEC): Define.
14086 2018-02-13 Jakub Jelinek <jakub@redhat.com>
14089 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
14090 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
14091 OPTION_MASK_ISA_AES as first argument to def_builtin_const
14092 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
14093 instead of OPTION_MASK_ISA_PCLMUL as first argument to
14094 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
14095 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
14096 temporarily for AES and PCLMUL builtins.
14098 PR tree-optimization/84339
14099 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
14100 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
14103 PR middle-end/84309
14104 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
14105 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
14106 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
14108 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
14110 * omp-simd-clone.h: New file.
14111 * omp-simd-clone.c: Include omp-simd-clone.h.
14112 (expand_simd_clones): No longer static.
14113 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
14114 cgraph.h and omp-simd-clone.h.
14115 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
14116 (vect_recog_widen_shift_pattern): Formatting fix.
14117 (vect_pattern_recog_1): Don't check optab for calls.
14120 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
14121 operands[2] into a REG before using gen_lowpart on it.
14123 2018-02-12 Jeff Law <law@redhat.com>
14126 * config/sh/sh.c (find_barrier): Consider a sibling call
14129 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
14130 successfully back substituting a reg.
14132 2018-02-12 Richard Biener <rguenther@suse.de>
14134 PR tree-optimization/84037
14135 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
14136 parameter, move visited init to caller.
14137 (vect_slp_analyze_operations): Separate cost from validity
14138 check, initialize visited once for all instances.
14139 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
14141 * tree-vect-stmts.c (vect_model_simple_cost): Make early
14143 (vect_model_promotion_demotion_cost): Likewise.
14144 (vectorizable_bswap): Guard cost modeling with !slp_node
14145 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
14147 (vectorizable_call): Likewise.
14148 (vectorizable_conversion): Likewise.
14149 (vectorizable_assignment): Likewise.
14150 (vectorizable_shift): Likewise.
14151 (vectorizable_operation): Likewise.
14152 (vectorizable_store): Likewise.
14153 (vectorizable_load): Likewise.
14154 (vectorizable_condition): Likewise.
14155 (vectorizable_comparison): Likewise.
14157 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
14160 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
14161 (ASAN_MARK): Fix fnspec to account for return value, change pointer
14162 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
14164 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
14166 PR middle-end/83665
14167 * params.def (inline-min-speedup): Increase from 8 to 15.
14168 (max-inline-insns-auto): Decrease from 40 to 30.
14169 * ipa-split.c (consider_split): Add some buffer for function to
14170 be considered inlining candidate.
14171 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
14174 2018-02-12 Richard Biener <rguenther@suse.de>
14176 PR tree-optimization/84037
14177 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
14178 matched stmts if we cannot swap the non-matched ones.
14180 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
14182 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
14183 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
14184 _mm_maskz_scalef_round_ss): New intrinsics.
14185 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
14186 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
14187 __builtin_ia32_scalefss_round): Remove.
14188 (__builtin_ia32_scalefsd_mask_round,
14189 __builtin_ia32_scalefss_mask_round): New intrinsics.
14190 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
14191 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
14192 ((match_operand:VF_128 2 "<round_nimm_predicate>"
14193 "<round_constraint>")): Changed to ...
14194 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
14195 "<round_scalar_constraint>")): ... this.
14196 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
14197 %0, %1, %2<round_op3>}"): Changed to ...
14198 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
14199 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
14200 %2<round_scalar_mask_op3>}"): ... this.
14201 * config/i386/subst.md (round_scalar_nimm_predicate): New.
14203 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
14205 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
14206 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
14207 (_mm_maskz_sqrt_round_ss): New intrinsics.
14208 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
14209 (__builtin_ia32_sqrtsd_mask_round)
14210 (__builtin_ia32_sqrtss_mask_round): New builtins.
14211 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
14212 (__builtin_ia32_sqrtss_round): Remove.
14213 (__builtin_ia32_sqrtsd_mask_round)
14214 (__builtin_ia32_sqrtss_mask_round): New builtins.
14215 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
14216 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
14217 ((match_operand:VF_128 1 "vector_operand"
14218 "xBm,<round_constraint>")): Changed to ...
14219 ((match_operand:VF_128 1 "vector_operand"
14220 "xBm,<round_scalar_constraint>")): ... this.
14221 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
14222 %0, %2, %<iptr>1<round_op3>}): Changed to ...
14223 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
14224 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
14225 %<iptr>1<round_scalar_mask_op3>}): ... this.
14226 ((set_attr "prefix" "<round_prefix>")): Changed to ...
14227 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
14229 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
14232 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
14233 Cast vec_cmpeq result to correct type.
14234 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
14235 Cast vec_cmpgt result to correct type.
14237 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
14239 * final.c (final_scan_insn_1): Renamed from...
14240 (final_scan_insn): ... this. New wrapper, to recover
14241 seen from the outermost call in recursive ones.
14242 * config/sparc/sparc.c (output_return): Drop seen from call.
14243 (output_sibcall): Likewise.
14244 * config/visium/visium.c (output_branch): Likewise.
14246 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
14248 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
14251 2018-02-10 Alan Modra <amodra@gmail.com>
14254 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
14255 Specify LR as an input.
14257 2018-02-10 Jakub Jelinek <jakub@redhat.com>
14260 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
14261 remove_member_access_dummy_vars): New functions.
14262 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
14263 lower_omp_1, execute_lower_omp): Use them.
14265 PR rtl-optimization/84308
14266 * shrink-wrap.c (spread_components): Release todo vector.
14268 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
14270 PR rtl-optimization/57193
14271 * ira-color.c (struct allocno_color_data): Add member
14272 conflict_allocno_hard_prefs.
14273 (update_conflict_allocno_hard_prefs): New.
14274 (bucket_allocno_compare_func): Add a preference based on
14275 conflict_allocno_hard_prefs.
14276 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
14277 (color_allocnos): Remove a dead code. Initiate
14278 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
14280 2018-02-09 Jakub Jelinek <jakub@redhat.com>
14283 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
14284 constraint from =wa to wa. Avoid a subreg on the output operand,
14285 instead use a pseudo and subreg it in a move.
14286 (p9_xxbrd_<mode>): Changed to ...
14287 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
14288 (p9_xxbrd_v2df): New expander.
14289 (p9_xxbrw_<mode>): Changed to ...
14290 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
14291 (p9_xxbrw_v4sf): New expander.
14293 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
14295 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
14297 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
14300 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
14301 multiply in 32-bit mode.
14302 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
14303 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
14306 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
14308 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
14309 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
14310 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
14311 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
14313 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
14316 * dwarf2out.c (is_trivial_indirect_ref): New function.
14317 (dwarf2out_late_global_decl): Do not generate a location
14318 attribute for variables that have a non-trivial DECL_VALUE_EXPR
14319 and that are not defined in the current unit.
14321 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
14323 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
14324 instead of a libcall for UNORDERED.
14326 2018-02-09 Tamar Christina <tamar.christina@arm.com>
14329 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
14330 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
14332 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14335 * config/s390/s390.c (s390_set_current_function): Invoke
14336 s390_indirect_branch_settings also if fndecl didn't change.
14338 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
14340 * config/rs6000/rs6000.md (blockage): Set length to zero.
14342 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
14344 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
14346 2018-02-09 Jakub Jelinek <jakub@redhat.com>
14349 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
14350 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
14354 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
14355 PARALLEL incoming that failed vt_get_decl_and_offset check.
14357 PR middle-end/84237
14358 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
14359 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
14361 (get_variable_section): For decls in named .bss* sections pass true as
14362 second argument to bss_initializer_p.
14364 2018-02-09 Marek Polacek <polacek@redhat.com>
14365 Jakub Jelinek <jakub@redhat.com>
14368 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
14369 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
14370 Sync some changes from cxx_fold_indirect_ref.
14372 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
14374 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
14376 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
14377 (BLOCK_INLINE_ENTRY_LABEL): New.
14378 (dwarf2out_var_location): Disregard inline entry markers.
14379 (inline_entry_data): New struct.
14380 (inline_entry_data_hasher): New hashtable type.
14381 (inline_entry_data_hasher::hash): New.
14382 (inline_entry_data_hasher::equal): New.
14383 (inline_entry_data_table): New variable.
14384 (add_high_low_attributes): Add DW_AT_entry_pc and
14385 DW_AT_GNU_entry_view attributes if a pending entry is found
14386 in inline_entry_data_table. Add old entry_pc attribute only
14387 if debug nonbinding markers are disabled.
14388 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
14389 markers are enabled.
14390 (block_within_block_p, dwarf2out_inline_entry): New.
14391 (dwarf2out_finish): Check that no entries remained in
14392 inline_entry_data_table.
14393 * final.c (reemit_insn_block_notes): Handle inline entry notes.
14394 (final_scan_insn, notice_source_line): Likewise.
14395 (rest_of_clean_state): Skip inline entry markers.
14396 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
14398 * gimple.c (gimple_build_debug_inline_entry): New.
14399 * gimple.h (enum gimple_debug_subcode): Add
14400 GIMPLE_DEBUG_INLINE_ENTRY.
14401 (gimple_build_debug_inline_entry): Declare.
14402 (gimple_debug_inline_entry_p): New.
14403 (gimple_debug_nonbind_marker_p): Adjust.
14404 * insn-notes.def (INLINE_ENTRY): New.
14405 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
14406 inline entry marker notes.
14407 (print_insn): Likewise.
14408 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
14409 (INSN_DEBUG_MARKER_KIND): Likewise.
14410 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
14411 * tree-inline.c (expand_call_inline): Build and insert
14412 debug_inline_entry stmt.
14413 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
14414 inline entry blocks early, if nonbind markers are enabled.
14415 (dump_scope_block): Dump fragment info.
14416 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
14417 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
14418 (gimple_build_debug_inline_entry): New.
14419 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
14420 Enable/disable inline entry points too.
14421 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
14422 (DEBUG_INSN): Describe inline entry markers.
14424 * common.opt (gvariable-location-views): New.
14425 (gvariable-location-views=incompat5): New.
14426 * config.in: Rebuilt.
14427 * configure: Rebuilt.
14428 * configure.ac: Test assembler for view support.
14429 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
14430 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
14431 * dwarf2out.c (var_loc_view): New typedef.
14432 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
14433 (dwarf2out_locviews_in_attribute): New.
14434 (dwarf2out_locviews_in_loclist): New.
14435 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
14436 (enum dw_line_info_opcode): Add LI_adv_address.
14437 (struct dw_line_info_table): Add view.
14438 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
14439 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
14440 (zero_view_p): New variable.
14441 (ZERO_VIEW_P): New macro.
14442 (output_asm_line_debug_info): New.
14443 (struct var_loc_node): Add view.
14444 (add_AT_view_list, AT_loc_list): New.
14445 (add_var_loc_to_decl): Add view param. Test it against last.
14446 (new_loc_list): Add view params. Record them.
14447 (AT_loc_list_ptr): Handle loc and view lists.
14448 (view_list_to_loc_list_val_node): New.
14449 (print_dw_val): Handle dw_val_class_view_list.
14450 (size_of_die): Likewise.
14451 (value_format): Likewise.
14452 (loc_list_has_views): New.
14453 (gen_llsym): Set vl_symbol too.
14454 (maybe_gen_llsym, skip_loc_list_entry): New.
14455 (dwarf2out_maybe_output_loclist_view_pair): New.
14456 (output_loc_list): Output view list or entries too.
14457 (output_view_list_offset): New.
14458 (output_die): Handle dw_val_class_view_list.
14459 (output_dwarf_version): New.
14460 (output_compilation_unit_header): Use it.
14461 (output_skeleton_debug_sections): Likewise.
14462 (output_rnglists, output_line_info): Likewise.
14463 (output_pubnames, output_aranges): Update version comments.
14464 (output_one_line_info_table): Output view numbers in asm comments.
14465 (dw_loc_list): Determine current endview, pass it to new_loc_list.
14466 Call maybe_gen_llsym.
14467 (loc_list_from_tree_1): Adjust.
14468 (add_AT_location_description): Create view list attribute if
14469 needed, check it's absent otherwise.
14470 (convert_cfa_to_fb_loc_list): Adjust.
14471 (maybe_emit_file): Call output_asm_line_debug_info for test.
14472 (dwarf2out_var_location): Reset views as needed. Precompute
14473 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
14474 attribute. Set view.
14475 (new_line_info_table): Reset next view.
14476 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
14477 (dwarf2out_source_line): Likewise. Output view resets and labels to
14478 the assembler, or select appropriate line info opcodes.
14479 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
14480 (optimize_string_length): Catch it. Adjust.
14481 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
14482 dw_val_class_view_list, and remove it if no longer needed.
14483 (hash_loc_list): Hash view numbers.
14484 (loc_list_hasher::equal): Compare them.
14485 (optimize_location_lists): Check whether a view list symbol is
14486 needed, and whether the locview attribute is present, and
14487 whether they match. Remove the locview attribute if no longer
14489 (index_location_lists): Call skip_loc_list_entry for test.
14490 (dwarf2out_finish): Call output_asm_line_debug_info for test.
14491 Use output_dwarf_version.
14492 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
14493 (struct dw_val_node): Add val_view_list.
14494 * final.c (SEEN_NEXT_VIEW): New.
14495 (set_next_view_needed): New.
14496 (clear_next_view_needed): New.
14497 (maybe_output_next_view): New.
14498 (final_start_function): Rename to...
14499 (final_start_function_1): ... this. Take pointer to FIRST,
14500 add SEEN parameter. Emit param bindings in the initial view.
14501 (final_start_function): Reintroduce SEEN-less interface.
14502 (final): Rename to...
14503 (final_1): ... this. Take SEEN parameter. Output final pending
14504 next view at the end.
14505 (final): Reintroduce seen-less interface.
14506 (final_scan_insn): Output pending next view before switching
14507 sections or ending a block. Mark the next view as needed when
14508 outputting variable locations. Notify debug backend of section
14509 changes, and of location view changes.
14510 (rest_of_handle_final): Adjust.
14511 * toplev.c (process_options): Autodetect value for debug variable
14512 location views option. Warn on incompat5 without -gdwarf-5.
14513 * doc/invoke.texi (gvariable-location-views): New.
14514 (gvariable-location-views=incompat5): New.
14515 (gno-variable-location-views): New.
14517 2018-02-08 David Malcolm <dmalcolm@redhat.com>
14519 PR tree-optimization/84136
14520 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
14521 that the result of find_edge is non-NULL.
14523 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
14526 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
14527 storing integer register in SImode. Fix cost of 256 and 512
14528 byte aligned SSE register store.
14530 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
14532 * config/i386/i386.c (ix86_multiplication_cost): Fix
14533 multiplication cost for TARGET_AVX512DQ.
14535 2018-02-08 Marek Polacek <polacek@redhat.com>
14537 PR tree-optimization/84238
14538 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
14541 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14543 PR tree-optimization/84265
14544 * tree-vect-stmts.c (vectorizable_store): Don't treat
14545 VMAT_CONTIGUOUS accesses as grouped.
14546 (vectorizable_load): Likewise.
14548 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14550 PR tree-optimization/81635
14551 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
14552 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
14553 (test_round_for_mask): New functions.
14554 (wide_int_cc_tests): Call test_round_for_mask.
14555 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
14556 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
14557 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
14558 range returned by get_range_info.
14560 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
14563 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
14564 * symtab.c: Include builtins.h
14565 (symtab_node::output_to_lto_symbol_table_p): Move here
14566 from lto-streamer-out.c:output_symbol_p.
14567 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
14568 (output_symbol_p): Move all logic to symtab.c
14569 (produce_symtab): Update.
14571 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14573 * config/s390/s390-opts.h (enum indirect_branch): Define.
14574 * config/s390/s390-protos.h (s390_return_addr_from_memory)
14575 (s390_indirect_branch_via_thunk)
14576 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
14577 (enum s390_indirect_branch_type): Define.
14578 * config/s390/s390.c (struct s390_frame_layout, struct
14579 machine_function): Remove.
14580 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
14581 (indirect_branch_table_label_no, indirect_branch_table_name):
14583 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
14584 (enum s390_indirect_branch_option): Define.
14585 (s390_return_addr_from_memory): New function.
14586 (s390_handle_string_attribute): New function.
14587 (s390_attribute_table): Add new attribute handler.
14588 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
14589 (s390_indirect_branch_via_thunk): New function.
14590 (s390_indirect_branch_via_inline_thunk): New function.
14591 (s390_function_ok_for_sibcall): When jumping via thunk disallow
14592 sibling call optimization for non z10 compiles.
14593 (s390_emit_call): Force indirect branch target to be a single
14594 register. Add r1 clobber for non-z10 compiles.
14595 (s390_emit_epilogue): Emit return jump via return_use expander.
14596 (s390_reorg): Handle JUMP_INSNs as execute targets.
14597 (s390_option_override_internal): Perform validity checks for the
14598 new command line options.
14599 (s390_indirect_branch_attrvalue): New function.
14600 (s390_indirect_branch_settings): New function.
14601 (s390_set_current_function): Invoke s390_indirect_branch_settings.
14602 (s390_output_indirect_thunk_function): New function.
14603 (s390_code_end): Implement target hook.
14604 (s390_case_values_threshold): Implement target hook.
14605 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
14607 * config/s390/s390.h (struct s390_frame_layout)
14608 (struct machine_function): Move here from s390.c.
14609 (TARGET_INDIRECT_BRANCH_NOBP_RET)
14610 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
14611 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
14612 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
14613 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
14614 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
14615 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
14616 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
14617 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
14618 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
14619 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
14620 (mnemonic attribute): Add values which aren't recognized
14622 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
14623 pattern for branch conversion. Fix mnemonic attribute.
14624 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
14625 indirect branch via thunk if requested.
14626 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
14627 ("*indirect_jump"): Disable for branch conversion using out of
14629 ("indirect_jump_via_thunk<mode>_z10")
14630 ("indirect_jump_via_thunk<mode>")
14631 ("indirect_jump_via_inlinethunk<mode>_z10")
14632 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
14633 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
14634 ("casesi_jump_via_inlinethunk<mode>_z10")
14635 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
14636 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
14637 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
14638 ("*indirect2_jump"): Disable for branch conversion.
14639 ("casesi_jump"): Turn into expander and expand patterns for branch
14641 ("return_use"): New expander.
14642 ("*return"): Emit return via thunk and rename it to ...
14643 ("*return<mode>"): ... this one.
14644 * config/s390/s390.opt: Add new options and and enum for the
14647 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14649 * lra-constraints.c (match_reload): Unconditionally use
14650 gen_lowpart_SUBREG, rather than selecting between that
14651 and equivalent gen_rtx_SUBREG code.
14653 2018-02-08 Richard Biener <rguenther@suse.de>
14655 PR tree-optimization/84233
14656 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
14657 changed flag instead of boguously re-using phi_inserted.
14659 2018-02-08 Martin Jambor <mjambor@suse.cz>
14661 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
14662 static local variables.
14664 2018-02-08 Richard Biener <rguenther@suse.de>
14666 PR tree-optimization/84278
14667 * tree-vect-stmts.c (vectorizable_store): When looking for
14668 smaller vector types to perform grouped strided loads/stores
14669 make sure the mode is supported by the target.
14670 (vectorizable_load): Likewise.
14672 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
14674 * config/aarch64/aarch64.c (aarch64_components_for_bb):
14675 Increase LDP/STP opportunities by adding adjacent callee-saves.
14677 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
14679 PR rtl-optimization/84068
14680 PR rtl-optimization/83459
14681 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
14683 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
14685 PR tree-optimization/84224
14686 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
14687 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
14688 non-zero arguments.
14690 2018-02-07 Iain Sandoe <iain@codesourcery.com>
14693 * config/rs6000/altivec.md (*restore_world): Remove LR use.
14694 * config/rs6000/predicates.md (restore_world_operation): Adjust op
14695 count, remove one USE.
14697 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
14699 * doc/install.texi (Configuration): Document the
14700 --with-long-double-format={ibm,ieee} PowerPC configuration
14704 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
14705 Convert from define_expand to be define_insn_and_split. Rework
14706 float/double/_Float128 conversions to QI/HI/SImode to work with
14707 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
14708 conversions to QI/HImode types did a store and then a load to
14709 truncate the value. For conversions to VSX registers, don't split
14710 the insn, instead emit the code directly. Use the code iterator
14711 any_fix to combine signed and unsigned conversions.
14712 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
14713 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
14714 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
14715 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
14716 (fix_<mode>di2_hw): Likewise.
14717 (fixuns_<mode>di2_hw): Likewise.
14718 (fix_<mode>si2_hw): Likewise.
14719 (fixuns_<mode>si2_hw): Likewise.
14720 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
14721 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
14722 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
14723 fix<uns>_trunc<SFDF:mode>si2_p8.
14724 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
14725 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
14726 (fix<uns>_<mode>_mem): Likewise.
14727 (fctiw<u>z_<mode>_mem): Likewise.
14728 (fix<uns>_<mode>_mem): Likewise.
14729 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
14730 the register allocator from doing a direct move to the GPRs to do
14731 a store, and instead use the ISA 3.0 store byte/half-word from
14732 vector register instruction. For IEEE 128-bit floating point,
14733 also optimize stores of 32-bit ints.
14734 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
14736 2018-02-07 Alan Hayward <alan.hayward@arm.com>
14738 * genextract.c (push_pathstr_operand): New function to support
14740 (walk_rtx): Call push_pathstr_operand.
14741 (print_path): Support [a-zA-Z].
14743 2018-02-07 Richard Biener <rguenther@suse.de>
14745 PR tree-optimization/84037
14746 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
14747 (cse_and_gimplify_to_preheader): Declare.
14748 (vect_get_place_in_interleaving_chain): Likewise.
14749 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14751 (_loop_vec_info::~_loop_vec_info): Delete it.
14752 (cse_and_gimplify_to_preheader): New function.
14753 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
14754 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
14755 (vectorizable_load): Likewise. For grouped stores always base
14756 the IV on the first element.
14757 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
14758 condition before gimplifying.
14760 2018-02-07 Jakub Jelinek <jakub@redhat.com>
14762 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
14763 *DIV_EXPR and *MOD_EXPR.
14765 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
14768 * config/i386/i386.c (ix86_option_override_internal): Mask out
14769 the CF_SET bit when checking -fcf-protection.
14771 2018-02-07 Tom de Vries <tom@codesourcery.com>
14774 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
14777 2018-02-07 Richard Biener <rguenther@suse.de>
14779 PR tree-optimization/84204
14780 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
14783 PR tree-optimization/84205
14784 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
14785 special-case isl_ast_op_zdiv_r.
14787 PR tree-optimization/84223
14788 * graphite-scop-detection.c (gather_bbs::before_dom_children):
14789 Only add conditions from within the region.
14790 (gather_bbs::after_dom_children): Adjust.
14792 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
14795 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
14796 * config/avr/avr.md: Only post-reload split REG-REG moves if
14797 either register is GENERAL_REG_P.
14799 2018-02-07 Jakub Jelinek <jakub@redhat.com>
14801 PR tree-optimization/84235
14802 * tree-ssa-scopedtables.c
14803 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
14804 if the subtraction is performed in floating point type where NaNs are
14805 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
14806 build 1. Formatting fix.
14808 2018-02-06 Jakub Jelinek <jakub@redhat.com>
14811 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
14812 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
14813 and skip it regardless of bb boundaries. Use CALL_P macro,
14814 don't test INSN_P (insn) together with CALL_P or JUMP_P check
14815 unnecessarily, formatting fix.
14817 2018-02-06 Michael Collison <michael.collison@arm.com>
14819 * config/arm/thumb2.md:
14820 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
14821 (*thumb_mov_notscc): Ditto.
14823 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
14826 * config/rs6000/rs6000.md (su code attribute): Use "u" for
14827 unsigned_fix, not "s".
14829 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14831 * configure.ac (gcc_fn_eh_frame_ro): New function.
14832 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
14833 correct .eh_frame permissions.
14834 * configure: Regenerate.
14836 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
14838 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
14839 irrelevant options.
14841 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14843 * config/rs6000/rs6000.c (rs6000_option_override_internal):
14844 Display warning message for -mno-speculate-indirect-jumps.
14846 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
14848 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
14850 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
14852 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
14854 PR tree-optimization/84225
14855 * tree-eh.c (find_trapping_overflow): Only call
14856 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
14858 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14861 * config/i386/i386.c: Reimplement the check of possible options
14862 -mibt/-mshstk conbination. Change error messages.
14863 * doc/invoke.texi: Fix a typo: remove extra '='.
14865 2018-02-06 Marek Polacek <polacek@redhat.com>
14867 PR tree-optimization/84228
14868 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
14870 2018-02-06 Tamar Christina <tamar.christina@arm.com>
14873 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
14874 emitted arch directives.
14875 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
14876 __ARM_FEATURE_COPROC before changing architectures.
14878 2018-02-06 Richard Biener <rguenther@suse.de>
14880 * config/i386/i386.c (print_reg): Fix typo.
14881 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
14883 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
14885 * configure: Regenerate.
14887 2018-02-05 Martin Sebor <msebor@redhat.com>
14889 PR tree-optimization/83369
14890 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
14893 2018-02-05 Martin Liska <mliska@suse.cz>
14895 * doc/invoke.texi: Cherry-pick upstream r323995.
14897 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
14899 * ira.c (ira_init_register_move_cost): Adjust comment.
14901 2018-02-05 Martin Liska <mliska@suse.cz>
14903 PR gcov-profile/84137
14904 * doc/gcov.texi: Fix typo in documentation.
14906 2018-02-05 Martin Liska <mliska@suse.cz>
14908 PR gcov-profile/83879
14909 * doc/gcov.texi: Document necessity of --dynamic-list-data when
14910 using dlopen functionality.
14912 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
14914 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
14915 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
14916 _mm_maskz_range_ss, _mm_mask_range_round_ss,
14917 _mm_maskz_range_round_ss): New intrinsics.
14918 (__builtin_ia32_rangesd128_round)
14919 (__builtin_ia32_rangess128_round): Remove.
14920 (__builtin_ia32_rangesd128_mask_round,
14921 __builtin_ia32_rangess128_mask_round): New builtins.
14922 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
14923 __builtin_ia32_rangess128_round): Remove.
14924 (__builtin_ia32_rangesd128_mask_round,
14925 __builtin_ia32_rangess128_mask_round): New builtins.
14926 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
14927 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
14928 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
14929 "<round_saeonly_constraint>")): Changed to ...
14930 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
14931 "<round_saeonly_scalar_constraint>")): ... this.
14932 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
14933 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
14934 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
14935 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
14936 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
14938 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
14940 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
14942 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
14943 Remove all values except native, 8540 and 8548.
14945 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
14947 * config/i386/i386.c (ix86_output_function_return): Pass
14948 INVALID_REGNUM, instead of -1, as invalid register number to
14949 indirect_thunk_name and output_indirect_thunk.
14951 2018-02-02 Julia Koval <julia.koval@intel.com>
14953 * config.gcc: Add -march=icelake.
14954 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
14955 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
14956 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
14957 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
14958 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
14959 (processor_target_table): Add icelake.
14960 (ix86_option_override_internal): Handle new PTAs.
14961 (get_builtin_code_for_version): Handle icelake.
14962 (M_INTEL_COREI7_ICELAKE): New.
14963 (fold_builtin_cpu): Handle icelake.
14964 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
14965 * doc/invoke.texi: Add -march=icelake.
14967 2018-02-02 Julia Koval <julia.koval@intel.com>
14969 * config/i386/i386.c (ix86_option_override_internal): Change flags type
14970 to wide_int_bitmask.
14971 * wide-int-bitmask.h: New.
14973 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14976 * config/i386/i386.md: Replace Pmode with word_mode in
14977 builtin_setjmp_setup and builtin_longjmp to support x32.
14979 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
14983 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
14985 (rs6000_supported_cpu_names): New static variable.
14986 (linux_cpu_translation_table): Likewise.
14987 (elf_platform) <cpu>: Define new static variable and use it.
14988 Translate kernel AT_PLATFORM name to canonical name if needed.
14989 Error if platform name is unknown.
14991 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
14994 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
14996 2018-02-01 Jeff Law <law@redhat.com>
14999 * config/i386/i386.c (release_scratch_register_on_entry): Add new
15000 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
15001 the scratch if RELEASE_VIA_POP is false.
15002 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
15003 If we have to save a temporary register, decrement SIZE appropriately.
15004 Pass new arguments to release_scratch_register_on_entry.
15005 (ix86_adjust_stack_and_probe): Likewise.
15006 (ix86_emit_probe_stack_range): Pass new arguments to
15007 release_scratch_register_on_entry.
15009 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
15011 PR rtl-optimization/84157
15012 * combine.c (change_zero_ext): Use REG_P predicate in
15013 front of HARD_REGISTER_P predicate.
15015 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
15017 * config/avr/avr.c (avr_option_override): Move disabling of
15018 -fdelete-null-pointer-checks to...
15019 * common/config/avr/avr-common.c (avr_option_optimization_table):
15022 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15024 PR tree-optimization/81635
15025 * tree-data-ref.c (split_constant_offset_1): For types that
15026 wrap on overflow, try to use range info to prove that wrapping
15029 2018-02-01 Renlin Li <renlin.li@arm.com>
15032 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
15033 TAILCALL_ADDR_REGS.
15034 (aarch64_register_move_cost): Likewise.
15035 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
15036 TAILCALL_ADDR_REGS.
15037 (REG_CLASS_NAMES): Likewise.
15038 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
15039 TAILCALL_ADDR_REGS. Remove IP registers.
15040 * config/aarch64/aarch64.md (Ucs): Update register constraint.
15042 2018-02-01 Richard Biener <rguenther@suse.de>
15044 * domwalk.h (dom_walker::dom_walker): Add additional constructor
15045 for specifying RPO order and allow NULL for that.
15046 * domwalk.c (dom_walker::dom_walker): Likewise.
15047 (dom_walker::walk): Handle NULL RPO order.
15048 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
15050 (rewrite_update_dom_walker): Likewise.
15051 (mark_def_dom_walker): Likewise.
15053 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15055 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
15056 (aarch64_maybe_expand_sve_subreg_move): Declare.
15057 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
15058 * config/aarch64/predicates.md (aarch64_any_register_operand): New
15060 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
15061 that are semantically a reverse operation.
15062 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
15063 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
15064 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
15066 (aarch64_can_change_mode_class): For big-endian, forbid changes
15067 between two SVE modes if they have different element sizes.
15069 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15071 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
15072 the TImode handling for big-endian targets.
15074 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15076 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
15077 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
15079 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
15080 Remove BSWAP handing for big-endian targets and use the form of
15081 LD1RQ appropariate for the mode.
15083 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15085 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
15086 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
15087 duplicated element.
15089 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15092 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
15093 check for operands that need to go through aarch64_sve_reload_be.
15095 2018-02-01 Jakub Jelinek <jakub@redhat.com>
15097 PR tree-optimization/81661
15098 PR tree-optimization/84117
15099 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
15100 * tree-eh.c: Include gimplify.h.
15101 (find_trapping_overflow, replace_trapping_overflow,
15102 rewrite_to_non_trapping_overflow): New functions.
15103 * tree-vect-loop.c: Include tree-eh.h.
15104 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
15105 * tree-data-ref.c: Include tree-eh.h.
15106 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
15108 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
15110 PR rtl-optimization/84123
15111 * combine.c (change_zero_ext): Check if hard register satisfies
15112 can_change_dest_mode before calling gen_lowpart_SUBREG.
15114 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
15117 * ira.c (ira_init_register_move_cost): Remove assert.
15119 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15121 PR rtl-optimization/84071
15122 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
15123 * doc/tm.texi: Regenerate.
15125 2018-01-31 Richard Biener <rguenther@suse.de>
15127 PR tree-optimization/84132
15128 * tree-data-ref.c (analyze_miv_subscript): Properly
15129 check whether evolution_function_is_affine_multivariate_p
15130 before calling gcd_of_steps_may_divide_p.
15132 2018-01-31 Julia Koval <julia.koval@intel.com>
15135 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
15136 * config/i386/i386.md (rdpid_rex64) New.
15137 (rdpid): Make 32bit only.
15139 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
15142 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
15143 an IDENTIFIER_NODE for FUNCTION_TYPE's.
15145 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15148 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15150 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15152 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15154 PR rtl-optimization/84071
15155 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
15156 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
15158 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
15160 * config/arc/arc.c (arc_handle_aux_attribute): New function.
15161 (arc_attribute_table): Add 'aux' attribute.
15162 (arc_in_small_data_p): Consider aux like variables.
15163 (arc_is_aux_reg_p): New function.
15164 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
15165 (arc_get_aux_arg): New function.
15166 (prepare_move_operands): Handle aux-register access.
15167 (arc_handle_aux_attribute): New function.
15168 * doc/extend.texi (ARC Variable attributes): Add subsection.
15170 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
15172 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
15173 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
15174 (arc_attribute_table): Add 'uncached' attribute.
15175 (arc_print_operand): Print '.di' flag for uncached memory
15177 (arc_in_small_data_p): Do not consider for small data the uncached
15179 (arc_is_uncached_mem_p): New function.
15180 * config/arc/predicates.md (compact_store_memory_operand): Check
15181 for uncached memory accesses.
15182 (nonvol_nonimm_operand): Likewise.
15183 * doc/extend.texi (ARC Type Attribute): New subsection.
15185 2018-01-31 Jakub Jelinek <jakub@redhat.com>
15188 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
15189 falign-loops=): Add Optimization flag.
15191 2018-01-30 Jeff Law <law@redhat.com>
15194 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
15195 INT_REGISTERS_SAVED. Check it prior to calling
15196 get_scratch_register_on_entry.
15197 (ix86_adjust_stack_and_probe): Similarly.
15198 (ix86_emit_probe_stack_range): Similarly.
15199 (ix86_expand_prologue): Corresponding changes.
15201 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15204 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
15205 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
15207 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
15210 * lra-constraints.c (curr_insn_transform): Process AND in the
15213 2018-01-30 Jakub Jelinek <jakub@redhat.com>
15215 PR rtl-optimization/83986
15216 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
15217 dependence against last_pending_memory_flush in addition to
15218 pending_jump_insns.
15220 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
15222 PR tree-optimization/81611
15223 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
15226 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15229 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
15232 2018-01-30 Richard Biener <rguenther@suse.de>
15233 Jakub Jelinek <jakub@redhat.com>
15235 PR tree-optimization/84111
15236 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
15237 inner loops added during recursion, as they don't have up-to-date
15240 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
15243 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
15244 (can_inline_edge_by_limits_p): ... here.
15245 (can_early_inline_edge_p, check_callers,
15246 update_caller_keys, update_callee_keys, recursive_inlining,
15247 add_new_edges_to_heap, speculation_useful_p,
15248 inline_small_functions,
15249 inline_small_functions, flatten_function,
15250 inline_to_all_callers_1): Update.
15252 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
15254 * profile-count.c (profile_count::combine_with_ipa_count): Handle
15257 2018-01-30 Richard Biener <rguenther@suse.de>
15259 PR tree-optimization/83008
15260 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
15261 invariant and constant vector uses in stmts when they need
15262 more than one stmt.
15264 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15267 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
15268 * configure: Regenerate.
15270 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
15272 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
15274 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
15275 Use gen_rtx_REG rather than gen_lowpart.
15277 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
15279 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
15280 rather than 0 when creating partial subregs.
15282 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
15284 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
15287 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
15290 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
15291 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
15292 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
15293 flags. This restores the settings used before the 2017-07-24.
15294 Turning off pre increment/decrement/modify allows IVOPTS to
15295 optimize DF/SF loops where the index is an int.
15297 2018-01-29 Richard Biener <rguenther@suse.de>
15298 Kelvin Nilsen <kelvin@gcc.gnu.org>
15301 * tree-vect-stmts.c (vectorizable_call): Don't call
15302 targetm.vectorize_builtin_md_vectorized_function if callee is
15305 2018-01-22 Carl Love <cel@us.ibm.com>
15307 * doc/extend.tex: Fix typo in second arg in
15308 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
15310 2018-01-29 Richard Biener <rguenther@suse.de>
15312 PR tree-optimization/84086
15313 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
15314 (flush_ssaname_freelist): When SSA names were released reset
15315 the SCEV hash table.
15317 2018-01-29 Richard Biener <rguenther@suse.de>
15319 PR tree-optimization/84057
15320 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
15321 removed paths when removing edges.
15323 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
15325 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
15326 -mfunction-return=@var{choice}.
15328 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
15330 PR diagnostic/84034
15331 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
15332 Handle CR like TAB.
15333 (layout::print_source_line): Likewise.
15334 (test_get_line_width_without_trailing_whitespace): Add test cases.
15336 2018-01-27 Jakub Jelinek <jakub@redhat.com>
15338 PR middle-end/84040
15339 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
15342 2018-01-26 Jim Wilson <jimw@sifive.com>
15344 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
15346 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
15349 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15351 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
15352 and CMP + SUB-immediate -> SUBS.
15354 2018-01-26 Martin Sebor <msebor@redhat.com>
15356 PR tree-optimization/83896
15357 * tree-ssa-strlen.c (get_string_len): Rename...
15358 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
15359 Avoid assuming length is constant.
15360 (handle_char_store): Use HOST_WIDE_INT for string length.
15362 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
15365 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
15366 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
15368 2018-01-26 Richard Biener <rguenther@suse.de>
15370 PR rtl-optimization/84003
15371 * dse.c (record_store): Only record redundant stores when
15372 the earlier store aliases at least all accesses the later one does.
15374 2018-01-26 Jakub Jelinek <jakub@redhat.com>
15376 PR rtl-optimization/83985
15377 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
15378 REG_CFA_RESTORE insns.
15379 (delete_unmarked_insns): Don't ignore separate shrink wrapping
15380 REG_CFA_RESTORE insns here.
15383 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
15384 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
15386 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15388 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
15389 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
15390 (arc_init): Likewise.
15391 (arc_override_options): Likewise.
15392 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
15394 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
15396 * config/arc/arc.h (TARGET_DBNZ): Define.
15397 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
15398 properly set the tune attribute.
15399 (dbnz): Use TARGET_DBNZ guard.
15400 * config/arc/arc.opt (mtune): Add core3 option.
15402 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15404 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
15405 recognize new pic like addresses.
15406 (arc_delegitimize_address): Clean up.
15408 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15410 * config/arc/arc-arches.def: Option mrf16 valid for all
15412 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
15413 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
15414 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
15415 * config/arc/arc-tables.opt: Regenerate.
15416 * config/arc/arc.c (arc_conditional_register_usage): Handle
15417 reduced register file case.
15418 (arc_file_start): Set must have build attributes.
15419 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
15420 mrf16 option value.
15421 * config/arc/arc.opt (mrf16): Add new option.
15422 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
15423 * config/arc/genmultilib.awk: Handle new mrf16 option.
15424 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
15425 * config/arc/t-multilib: Regenerate.
15426 * doc/invoke.texi (ARC Options): Document mrf16 option.
15428 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15430 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
15431 * config/arc/arc.c (arc_handle_secure_attribute): New function.
15432 (arc_attribute_table): Add 'secure_call' attribute.
15433 (arc_print_operand): Print secure call operand.
15434 (arc_function_ok_for_sibcall): Don't optimize tail calls when
15436 (arc_is_secure_call_p): New function. * config/arc/arc.md
15437 (call_i): Add support for sjli instruction.
15438 (call_value_i): Likewise.
15439 * config/arc/constraints.md (Csc): New constraint.
15441 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15442 John Eric Martin <John.Martin@emmicro-us.com>
15444 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
15445 * config/arc/arc.c (_arc_jli_section): New struct.
15446 (arc_jli_section): New type.
15447 (rc_jli_sections): New static variable.
15448 (arc_handle_jli_attribute): New function.
15449 (arc_attribute_table): Add jli_always and jli_fixed attribute.
15450 (arc_file_end): New function.
15451 (TARGET_ASM_FILE_END): Define.
15452 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
15453 (arc_add_jli_section): New function.
15454 (jli_call_scan): Likewise.
15455 (arc_reorg): Call jli_call_scan.
15456 (arc_output_addsi): Remove 'S' from printing asm operand.
15457 (arc_is_jli_call_p): New function.
15458 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
15460 (movhi_insn): Likewise.
15461 (movsi_insn): Likewise.
15462 (movsi_set_cc_insn): Likewise.
15463 (loadqi_update): Likewise.
15464 (load_zeroextendqisi_update): Likewise.
15465 (load_signextendqisi_update): Likewise.
15466 (loadhi_update): Likewise.
15467 (load_zeroextendhisi_update): Likewise.
15468 (load_signextendhisi_update): Likewise.
15469 (loadsi_update): Likewise.
15470 (loadsf_update): Likewise.
15471 (movsicc_insn): Likewise.
15472 (bset_insn): Likewise.
15473 (bxor_insn): Likewise.
15474 (bclr_insn): Likewise.
15475 (bmsk_insn): Likewise.
15476 (bicsi3_insn): Likewise.
15477 (cmpsi_cc_c_insn): Likewise.
15478 (movsi_ne): Likewise.
15479 (movsi_cond_exec): Likewise.
15480 (clrsbsi2): Likewise.
15481 (norm_f): Likewise.
15491 (call_i): Remove 'S' asm letter, add jli instruction.
15492 (call_value_i): Likewise.
15493 * config/arc/arc.op (mjli-always): New option.
15494 * config/arc/constraints.md (Cji): New constraint.
15495 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
15497 (subsf3_fpx): Likewise.
15498 (mulsf3_fpx): Likewise.
15499 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
15501 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
15502 function attrbutes.
15503 * doc/invoke.texi (ARC): Document mjli-always option.
15505 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
15507 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
15508 avoid addition with 0 and use incw and decw where possible.
15510 2018-01-26 Richard Biener <rguenther@suse.de>
15512 PR tree-optimization/81082
15513 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
15514 association if it requires casting to unsigned.
15515 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
15516 from fold_plusminus_mult_expr to catch important cases late when
15517 range info is available.
15519 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15521 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
15522 * configure.ac (hidden_linkonce): New test.
15523 * configure: Regenerate.
15524 * config.in: Regenerate.
15526 2018-01-26 Julia Koval <julia.koval@intel.com>
15528 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
15529 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
15530 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
15531 _mm_mask_bitshuffle_epi64_mask): Fix type.
15532 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
15533 USI_FTYPE_V4DI_V4DI_USI): Remove.
15534 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
15535 __builtin_ia32_vpshufbitqmb256_mask,
15536 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
15537 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
15538 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
15540 2018-01-26 Alan Modra <amodra@gmail.com>
15543 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
15544 UNSPEC_VBPERMQ. Sort other unspecs.
15546 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
15548 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
15550 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
15552 PR middle-end/83055
15553 * predict.c (drop_profile): Do not push/pop cfun; update also
15555 (handle_missing_profiles): Fix logic looking for zero profiles.
15557 2018-01-25 Jakub Jelinek <jakub@redhat.com>
15559 PR middle-end/83977
15560 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
15561 on functions with #pragma omp declare simd or functions with simd
15563 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
15564 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
15565 Remove trailing \n from warning_at calls.
15567 2018-01-25 Tom de Vries <tom@codesourcery.com>
15570 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
15571 for neutered workers.
15573 2018-01-24 Joseph Myers <joseph@codesourcery.com>
15576 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
15577 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
15579 2018-01-24 Jeff Law <law@redhat.com>
15582 * i386.c (get_probe_interval): Move to earlier point.
15583 (ix86_compute_frame_layout): If -fstack-clash-protection and
15584 the frame is larger than the probe interval, then use pushes
15585 to save registers rather than reg->mem moves.
15586 (ix86_expand_prologue): Remove conditional for int_registers_saved
15589 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
15592 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
15593 min/max for never referenced object.
15595 2018-01-24 Jakub Jelinek <jakub@redhat.com>
15597 PR middle-end/83977
15598 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
15600 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
15601 attributes from DECL_ATTRIBUTES (decl) without affecting
15602 DECL_ATTRIBUTES (current_function_decl).
15603 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
15604 functions with non-NULL DECL_ABSTRACT_ORIGIN.
15606 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
15608 PR tree-optimization/83979
15609 * fold-const.c (fold_comparison): Use constant_boolean_node
15610 instead of boolean_{true,false}_node.
15612 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
15614 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
15617 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15619 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15620 Simplify the clause that sets the length attribute.
15621 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15622 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
15623 clause that sets the length attribute.
15624 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15626 2018-01-24 Tom de Vries <tom@codesourcery.com>
15629 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
15630 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
15631 Add strict parameter.
15632 (prevent_branch_around_nothing): Insert dummy insn between branch to
15633 label and label with no ptx insn inbetween.
15634 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
15636 2018-01-24 Tom de Vries <tom@codesourcery.com>
15639 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
15640 for neutered threads in warp.
15641 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
15643 2018-01-24 Richard Biener <rguenther@suse.de>
15645 PR tree-optimization/83176
15646 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
15649 2018-01-24 Richard Biener <rguenther@suse.de>
15651 PR tree-optimization/82819
15652 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
15653 code generating pluses that are no-ops in the target precision.
15655 2018-01-24 Richard Biener <rguenther@suse.de>
15657 PR middle-end/84000
15658 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
15660 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15662 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
15663 to merge probabilities.
15664 * predict.c (probably_never_executed): Also mark as cold functions
15665 with global 0 profile and guessed local profile.
15666 * profile-count.c (profile_probability::combine_with_count): New
15668 * profile-count.h (profile_probability::operator*,
15669 profile_probability::operator*=, profile_probability::operator/,
15670 profile_probability::operator/=): Reduce precision to adjusted
15671 and set value to guessed on contradictory divisions.
15672 (profile_probability::combine_with_freq): Remove.
15673 (profile_probability::combine_wiht_count): Declare.
15674 (profile_count::force_nonzero):: Set to adjusted.
15675 (profile_count::probability_in):: Set quality to adjusted.
15676 * tree-ssa-tail-merge.c (replace_block_by): Use
15677 combine_with_count.
15679 2018-01-23 Andrew Waterman <andrew@sifive.com>
15680 Jim Wilson <jimw@sifive.com>
15682 * config/riscv/riscv.c (riscv_stack_boundary): New.
15683 (riscv_option_override): Set riscv_stack_boundary. Handle
15684 riscv_preferred_stack_boundary_arg.
15685 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
15686 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
15687 (STACK_BOUNDARY): Set to riscv_stack_boundary.
15688 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
15689 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
15690 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
15692 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
15695 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
15696 of struct ix86_frame.
15697 (ix86_expand_epilogue): Likewise. Add a local variable for
15698 the reg_save_offset field in struct ix86_frame.
15700 2018-01-23 Bin Cheng <bin.cheng@arm.com>
15702 PR tree-optimization/82604
15703 * tree-loop-distribution.c (enum partition_kind): New enum item
15704 PKIND_PARTIAL_MEMSET.
15705 (partition_builtin_p): Support above new enum item.
15706 (generate_code_for_partition): Ditto.
15707 (compute_access_range): Differentiate cases that equality can be
15708 proven at all loops, the innermost loops or no loops.
15709 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
15710 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
15711 (finalize_partitions, distribute_loop): Don't fuse partition of
15712 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
15713 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
15714 parloop is enabled.
15716 2018-01-23 Martin Liska <mliska@suse.cz>
15718 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
15719 order to ignore the predictor.
15720 (PRED_POLYMORPHIC_CALL): Likewise.
15721 (PRED_RECURSIVE_CALL): Likewise.
15723 2018-01-23 Martin Liska <mliska@suse.cz>
15725 * tree-profile.c (tree_profiling): Print function header to
15726 aware reader which function we are working on.
15727 * value-prof.c (gimple_find_values_to_profile): Do not print
15728 not interesting value histograms.
15730 2018-01-23 Martin Liska <mliska@suse.cz>
15732 * profile-count.h (enum profile_quality): Add
15733 profile_uninitialized as the first value. Do not number values
15734 as they are zero based.
15735 (profile_count::verify): Update sanity check.
15736 (profile_probability::verify): Likewise.
15738 2018-01-23 Nathan Sidwell <nathan@acm.org>
15740 * doc/invoke.texi (ffor-scope): Deprecate.
15742 2018-01-23 David Malcolm <dmalcolm@redhat.com>
15744 PR tree-optimization/83510
15745 * domwalk.c (set_all_edges_as_executable): New function.
15746 (dom_walker::dom_walker): Convert bool param
15747 "skip_unreachable_blocks" to enum reachability. Move setup of
15748 edge flags to set_all_edges_as_executable and only do it when
15749 reachability is REACHABLE_BLOCKS.
15750 * domwalk.h (enum dom_walker::reachability): New enum.
15751 (dom_walker::dom_walker): Convert bool param
15752 "skip_unreachable_blocks" to enum reachability.
15753 (set_all_edges_as_executable): New decl.
15754 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
15755 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
15757 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
15758 but converting true to REACHABLE_BLOCKS.
15759 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
15761 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
15762 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
15763 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
15765 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
15766 if check_all_array_refs will be called.
15768 2018-01-23 David Malcolm <dmalcolm@redhat.com>
15770 * tree.c (selftest::test_location_wrappers): Add more test
15773 2018-01-23 David Malcolm <dmalcolm@redhat.com>
15775 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
15776 (selftest::test_bit_in_range): Likewise.
15778 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
15781 * doc/sourcebuild.texi (vect_float): Say that the selector
15782 only describes the situation when -funsafe-math-optimizations is on.
15783 (vect_float_strict): Document.
15785 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
15787 PR tree-optimization/83965
15788 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
15789 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
15790 instead of checking only for a reduction.
15791 (vect_recog_widen_sum_pattern): Likewise.
15793 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15795 * predict.c (probably_never_executed): Only use precise profile info.
15796 (compute_function_frequency): Skip after inlining hack since we now
15797 have quality checking.
15799 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15801 * profile-count.h (profile_probability::very_unlikely,
15802 profile_probability::unlikely, profile_probability::even): Set
15803 precision to guessed.
15805 2018-01-23 Richard Biener <rguenther@suse.de>
15807 PR tree-optimization/83963
15808 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
15809 Properly terminate dominator walk when crossing the exit edge not
15810 when visiting its source block.
15812 2018-01-23 Jakub Jelinek <jakub@redhat.com>
15815 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
15816 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
15818 2018-01-22 Jakub Jelinek <jakub@redhat.com>
15820 PR tree-optimization/83957
15821 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
15822 semicolon after for body surrounded by braces.
15824 PR tree-optimization/83081
15825 * profile-count.h (profile_probability::split): New method.
15826 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
15827 Use profile_probability::split.
15828 (do_compare_rtx_and_jump): Fix adjustment of probabilities
15829 when splitting a single conditional jump into 2.
15831 2018-01-22 David Malcolm <dmalcolm@redhat.com>
15833 PR tree-optimization/69452
15834 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
15837 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15839 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
15840 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
15841 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
15843 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15845 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
15847 * config/rl78/rl78.md (movdi): New define_expand.
15848 * config/rl78/rl78.c (rl78_split_movdi): New function.
15850 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
15853 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
15855 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
15856 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
15857 128-bit to produce an UNSPEC move to get the double word with the
15858 signbit and then a shift directly to do signbit.
15859 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
15860 implementation with a new version that just does either a direct
15861 move or a regular move. Move memory interface to separate insns.
15862 Move insns so they are next to the expander.
15863 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
15864 with signbit move. Split big and little endian case.
15865 (signbit<mode>2_dm_mem_le): Likewise.
15866 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
15867 (signbit<mode>2_dm2): Likewise.
15869 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15871 * config/rl78/rl78.md (anddi3): New define_expand.
15873 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15875 * config/rl78/rl78.md (umindi3): New define_expand.
15877 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15879 * config/rl78/rl78.md (smindi3): New define_expand.
15881 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15883 * config/rl78/rl78.md (smaxdi3): New define_expand.
15885 2018-01-22 Carl Love <cel@us.ibm.com>
15887 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
15888 LVX_V1TI): Add macro expansion.
15889 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
15890 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
15891 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
15892 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
15893 Change check to determine if the instruction is a byte reversing
15894 entry. Fix typo in comment.
15895 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
15896 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
15897 Add def_builtin calls for new builtins.
15898 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
15899 Add define_insn expansion.
15901 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15903 * config/rl78/rl78.md (umaxdi3): New define_expand.
15905 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15907 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
15908 for non-QImode registers.
15910 2018-01-22 Richard Biener <rguenther@suse.de>
15912 PR tree-optimization/83963
15913 * graphite-scop-detection.c (scop_detection::get_sese): Delay
15914 including the loop exit block.
15915 (scop_detection::merge_sese): Likewise.
15916 (scop_detection::add_scop): Do it here instead.
15918 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15920 * doc/sourcebuild.texi (arm_softfloat): Document.
15922 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
15925 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
15926 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
15927 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
15929 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15930 David Edelsohn <dje.gcc@gmail.com>
15933 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15934 Change "crset eq" to "crset 2".
15935 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15936 (*call_indirect_aix<mode>_nospec): Likewise.
15937 (*call_value_indirect_aix<mode>_nospec): Likewise.
15938 (*call_indirect_elfv2<mode>_nospec): Likewise.
15939 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
15940 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
15941 change assembly output from . to $.
15942 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15943 (indirect_jump<mode>_nospec): Change assembly output from . to $.
15944 (*tablejump<mode>_internal1_nospec): Likewise.
15946 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
15949 * config/sh/sh_optimize_sett_clrt.cc:
15950 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
15952 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
15954 PR tree-optimization/83940
15955 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
15956 offset_dt to vect_constant_def rather than vect_unknown_def_type.
15957 (vect_check_load_store_mask): Add a mask_dt_out parameter and
15958 use it to pass back the definition type.
15959 (vect_check_store_rhs): Likewise rhs_dt_out.
15960 (vect_build_gather_load_calls): Add a mask_dt argument and use
15961 it instead of a call to vect_is_simple_use.
15962 (vectorizable_store): Update calls to vect_check_load_store_mask
15963 and vect_check_store_rhs. Use the dt returned by the latter instead
15964 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
15965 instead of calls to vect_is_simple_use. Pass the scalar rather
15966 than the vector operand to vect_is_simple_use when handling
15967 second and subsequent copies of an rhs value.
15968 (vectorizable_load): Update calls to vect_check_load_store_mask
15969 and vect_build_gather_load_calls. Use the cached mask_dt and
15970 gs_info.offset_dt instead of calls to vect_is_simple_use.
15972 2018-01-20 Jakub Jelinek <jakub@redhat.com>
15974 PR middle-end/83945
15975 * tree-emutls.c: Include gimplify.h.
15976 (lower_emutls_2): New function.
15977 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
15978 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
15979 it before further processing.
15982 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
15983 UINTVAL (trueop1) instead of INTVAL (op1).
15985 2018-01-19 Jakub Jelinek <jakub@redhat.com>
15989 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
15990 INCOMING_FRAME_SP_OFFSET if not defined.
15991 (scan_trace): Add ENTRY argument. If true and
15992 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
15993 emit a note to adjust the CFA offset.
15994 (create_cfi_notes): Adjust scan_trace callers.
15995 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
15996 INCOMING_FRAME_SP_OFFSET in the CIE.
15997 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
15998 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
16000 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
16001 * doc/tm.texi: Regenerated.
16003 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16005 PR rtl-optimization/83147
16006 * lra-constraints.c (remove_inheritance_pseudos): Use
16007 lra_substitute_pseudo_within_insn.
16009 2018-01-19 Tom de Vries <tom@codesourcery.com>
16010 Cesar Philippidis <cesar@codesourcery.com>
16013 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
16015 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
16018 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
16019 spaces for function labels.
16021 2018-01-19 Martin Liska <mliska@suse.cz>
16023 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
16024 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
16025 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
16026 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
16027 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
16028 (PRED_CONST_RETURN): Change from 69 to 65.
16029 (PRED_NULL_RETURN): Change from 91 to 71.
16030 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
16031 (PRED_LOOP_GUARD): Change from 66 to 73.
16033 2018-01-19 Martin Liska <mliska@suse.cz>
16035 * predict.c (predict_insn_def): Add new assert.
16036 (struct branch_predictor): Change type to signed integer.
16037 (test_prediction_value_range): Amend test to cover
16038 PROB_UNINITIALIZED.
16039 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
16040 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
16041 (PRED_LOOP_ITERATIONS_MAX): Likewise.
16042 (PRED_LOOP_IV_COMPARE): Likewise.
16043 * predict.h (PROB_UNINITIALIZED): Define new constant.
16045 2018-01-19 Martin Liska <mliska@suse.cz>
16047 * predict.c (dump_prediction): Add new format for
16048 analyze_brprob.py script which is enabled with -details
16050 * profile-count.h (precise_p): New function.
16052 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
16054 PR tree-optimization/83922
16055 * tree-vect-loop.c (vect_verify_full_masking): Return false if
16056 there are no statements that need masking.
16057 (vect_active_double_reduction_p): New function.
16058 (vect_analyze_loop_operations): Use it when handling phis that
16059 are not in the loop header.
16061 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
16063 PR tree-optimization/83914
16064 * tree-vect-loop.c (vectorizable_induction): Don't convert
16065 init_expr or apply the peeling adjustment for inductions
16066 that are nested within the vectorized loop.
16068 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16070 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
16073 2018-01-18 Jakub Jelinek <jakub@redhat.com>
16077 * function.h (gimplify_parameters): Add gimple_seq * argument.
16078 * function.c: Include gimple.h and options.h.
16079 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
16080 for the added local temporaries if needed.
16081 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
16082 if there are any parameter cleanups, wrap whole body into a
16083 try/finally with the cleanups.
16085 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
16088 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
16089 Use GET_MODE_CLASS for scalar floating point.
16091 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
16095 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
16096 Fix call of call_cgraph_insertion_hooks.
16098 2018-01-18 Martin Sebor <msebor@redhat.com>
16100 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
16102 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
16105 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
16108 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
16111 * common.opt: (-ffile-prefix-map): New option.
16112 * opts.c (common_handle_option): Defer it.
16113 * opts-global.c (handle_common_deferred_options): Handle it.
16114 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
16115 * file-prefix-map.h: New file.
16116 (remap_debug_filename, add_debug_prefix_map): ...here.
16117 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
16118 * final.c (debug_prefix_map, add_debug_prefix_map
16119 remap_debug_filename): Move to...
16120 * file-prefix-map.c: New file.
16121 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
16122 generalize, get rid of alloca(), use strrchr() instead of strchr().
16123 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
16124 Implement in terms of add_prefix_map().
16125 (remap_macro_filename, remap_debug_filename): Implement in term of
16127 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
16128 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
16129 * dbxout.c: Include file-prefix-map.h.
16130 * varasm.c: Likewise.
16131 * vmsdbgout.c: Likewise.
16132 * xcoffout.c: Likewise.
16133 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
16134 * doc/cppopts.texi (-fmacro-prefix-map): Document.
16135 * doc/invoke.texi (-ffile-prefix-map): Document.
16136 (-fdebug-prefix-map): Update description.
16138 2018-01-18 Martin Liska <mliska@suse.cz>
16140 * config/i386/i386.c (indirect_thunk_name): Document that also
16142 (output_indirect_thunk): Document why both instructions
16143 (pause and lfence) are generated.
16145 2018-01-18 Richard Biener <rguenther@suse.de>
16147 PR tree-optimization/83887
16148 * graphite-scop-detection.c
16149 (scop_detection::get_nearest_dom_with_single_entry): Remove.
16150 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
16151 (scop_detection::merge_sese): Re-implement with a flood-fill
16152 algorithm that properly finds a SESE region if it exists.
16154 2018-01-18 Jakub Jelinek <jakub@redhat.com>
16157 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
16158 pointer_diff optimizations use view_convert instead of convert.
16160 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16162 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
16163 Generate different code for -mno-speculate-indirect-jumps.
16164 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
16165 (*call_indirect_aix<mode>): Disable for
16166 -mno-speculate-indirect-jumps.
16167 (*call_indirect_aix<mode>_nospec): New define_insn.
16168 (*call_value_indirect_aix<mode>): Disable for
16169 -mno-speculate-indirect-jumps.
16170 (*call_value_indirect_aix<mode>_nospec): New define_insn.
16171 (*sibcall_nonlocal_sysv<mode>): Generate different code for
16172 -mno-speculate-indirect-jumps.
16173 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
16175 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
16177 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
16178 long double type, set the flags for noting the default long double
16179 type, even if we don't pass or return a long double type.
16181 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
16184 * ipa-inline.c (flatten_function): Do not overwrite final inlining
16187 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
16189 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
16190 support for merge[hl].
16191 (fold_mergehl_helper): New helper function.
16192 (tree-vector-builder.h): New #include for tree_vector_builder usage.
16193 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
16194 (altivec_vmrglw_direct): Add xxmrglw insn.
16196 2018-01-17 Andrew Waterman <andrew@sifive.com>
16198 * config/riscv/riscv.c (riscv_conditional_register_usage): If
16199 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
16201 2018-01-17 David Malcolm <dmalcolm@redhat.com>
16204 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
16205 call the lto_location_cache before reading the
16206 DECL_SOURCE_LOCATION of the types.
16208 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
16209 Richard Sandiford <richard.sandiford@linaro.org>
16211 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
16212 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
16213 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
16214 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
16215 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
16217 * config/aarch64/constraints.md (aarch64_movti_operand):
16219 * config/aarch64/predicates.md (Uti): Add new constraint.
16221 2018-01-17 Carl Love <cel@us.ibm.com>
16223 * config/rs6000/vsx.md (define_expand xl_len_r,
16224 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
16225 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
16227 (define_expand, define_insn): Move the shift left from the
16228 define_insn to the define_expand for lxvl and stxvl instructions.
16229 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
16230 and XL_LEN_R definitions to PURE.
16232 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
16234 * config/i386/i386.c (indirect_thunk_name): Declare regno
16235 as unsigned int. Compare regno with INVALID_REGNUM.
16236 (output_indirect_thunk): Ditto.
16237 (output_indirect_thunk_function): Ditto.
16238 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
16239 in the call to output_indirect_thunk_function.
16241 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
16243 PR middle-end/83884
16244 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
16245 rather than the size of inner_type to determine the stack slot size
16246 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
16248 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
16251 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
16254 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
16256 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
16257 endian Linux systems to optionally enable multilibs for selecting
16258 the long double type if the user configured an explicit type.
16259 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
16260 have no long double multilibs if not defined.
16261 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
16262 warn if the user used -mabi={ieee,ibm}longdouble and we built
16263 multilibs for long double.
16264 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
16265 appropriate multilib option.
16266 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
16268 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
16269 for building long double multilibs.
16270 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
16272 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
16274 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
16277 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
16279 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
16282 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
16285 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
16288 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
16290 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
16291 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
16293 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
16295 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
16296 different rtl trees depending on TARGET_64BIT.
16297 (rs6000_gen_lvx): Likewise.
16299 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
16301 * config/visium/visium.md (nop): Tweak comment.
16302 (hazard_nop): Likewise.
16304 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16306 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
16307 -mspeculate-indirect-jumps.
16308 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
16309 for -mno-speculate-indirect-jumps.
16310 (*call_indirect_elfv2<mode>_nospec): New define_insn.
16311 (*call_value_indirect_elfv2<mode>): Disable for
16312 -mno-speculate-indirect-jumps.
16313 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
16314 (indirect_jump): Emit different RTL for
16315 -mno-speculate-indirect-jumps.
16316 (*indirect_jump<mode>): Disable for
16317 -mno-speculate-indirect-jumps.
16318 (*indirect_jump<mode>_nospec): New define_insn.
16319 (tablejump): Emit different RTL for
16320 -mno-speculate-indirect-jumps.
16321 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
16322 (tablejumpsi_nospec): New define_expand.
16323 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
16324 (tablejumpdi_nospec): New define_expand.
16325 (*tablejump<mode>_internal1): Disable for
16326 -mno-speculate-indirect-jumps.
16327 (*tablejump<mode>_internal1_nospec): New define_insn.
16328 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
16331 2018-01-16 Artyom Skrobov tyomitch@gmail.com
16333 * caller-save.c (insert_save): Drop unnecessary parameter. All
16336 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16337 Richard Biener <rguenth@suse.de>
16340 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
16341 return early, inline manually is_gimple_sizepos. Make sure if we
16342 call gimplify_expr we don't end up with a gimple constant.
16343 * tree.c (variably_modified_type_p): Don't return true for
16344 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
16345 * gimplify.h (is_gimple_sizepos): Remove.
16347 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16349 PR tree-optimization/83857
16350 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
16351 vectorizable_live_operation for pure SLP statements.
16352 (vectorizable_live_operation): Handle PHIs.
16354 2018-01-16 Richard Biener <rguenther@suse.de>
16356 PR tree-optimization/83867
16357 * tree-vect-stmts.c (vect_transform_stmt): Precompute
16358 nested_in_vect_loop_p since the scalar stmt may get invalidated.
16360 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16363 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
16364 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
16365 If off is not INTEGER_CST, issue a may not be aligned warning
16366 rather than isn't aligned. Use isn%'t rather than isn't.
16367 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
16369 <case MULT_EXPR>: Improve the case when bottom and one of the
16370 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
16371 operand, in that case check if the other operand is multiple of
16372 bottom divided by the INTEGER_CST operand.
16374 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16377 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
16378 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
16379 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
16380 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
16381 * config/pa/pa.c (pa_function_arg_advance): Likewise.
16382 (pa_function_arg, pa_arg_partial_bytes): Likewise.
16383 (pa_function_arg_size): New function.
16385 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16387 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
16388 in a separate statement.
16390 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16392 PR tree-optimization/83847
16393 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
16394 group gathers and scatters.
16396 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16398 PR rtl-optimization/86620
16399 * params.def (max-sched-ready-insns): Bump minimum value to 1.
16401 PR rtl-optimization/83213
16402 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
16403 to last if both are JUMP_INSNs.
16405 PR tree-optimization/83843
16406 * gimple-ssa-store-merging.c
16407 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
16408 store_immediate_info for bswap/nop orig_stores.
16410 2018-01-15 Andrew Waterman <andrew@sifive.com>
16412 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
16414 <UDIV>: Increase cost if !TARGET_DIV.
16416 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
16418 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
16419 (define_attr "cr_logical_3op"): New.
16420 (cceq_ior_compare): Adjust.
16421 (cceq_ior_compare_complement): Adjust.
16422 (*cceq_rev_compare): Adjust.
16423 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
16424 (is_cracked_insn): Adjust.
16425 (insn_must_be_first_in_group): Adjust.
16426 * config/rs6000/40x.md: Adjust.
16427 * config/rs6000/440.md: Adjust.
16428 * config/rs6000/476.md: Adjust.
16429 * config/rs6000/601.md: Adjust.
16430 * config/rs6000/603.md: Adjust.
16431 * config/rs6000/6xx.md: Adjust.
16432 * config/rs6000/7450.md: Adjust.
16433 * config/rs6000/7xx.md: Adjust.
16434 * config/rs6000/8540.md: Adjust.
16435 * config/rs6000/cell.md: Adjust.
16436 * config/rs6000/e300c2c3.md: Adjust.
16437 * config/rs6000/e500mc.md: Adjust.
16438 * config/rs6000/e500mc64.md: Adjust.
16439 * config/rs6000/e5500.md: Adjust.
16440 * config/rs6000/e6500.md: Adjust.
16441 * config/rs6000/mpc.md: Adjust.
16442 * config/rs6000/power4.md: Adjust.
16443 * config/rs6000/power5.md: Adjust.
16444 * config/rs6000/power6.md: Adjust.
16445 * config/rs6000/power7.md: Adjust.
16446 * config/rs6000/power8.md: Adjust.
16447 * config/rs6000/power9.md: Adjust.
16448 * config/rs6000/rs64.md: Adjust.
16449 * config/rs6000/titan.md: Adjust.
16451 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16453 * config/i386/predicates.md (indirect_branch_operand): Rewrite
16454 ix86_indirect_branch_register logic.
16456 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16458 * config/i386/constraints.md (Bs): Update
16459 ix86_indirect_branch_register check. Don't check
16460 ix86_indirect_branch_register with GOT_memory_operand.
16462 * config/i386/predicates.md (GOT_memory_operand): Don't check
16463 ix86_indirect_branch_register here.
16464 (GOT32_symbol_operand): Likewise.
16466 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16468 * config/i386/predicates.md (constant_call_address_operand):
16469 Rewrite ix86_indirect_branch_register logic.
16470 (sibcall_insn_operand): Likewise.
16472 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16474 * config/i386/constraints.md (Bs): Replace
16475 ix86_indirect_branch_thunk_register with
16476 ix86_indirect_branch_register.
16478 * config/i386/i386.md (indirect_jump): Likewise.
16479 (tablejump): Likewise.
16480 (*sibcall_memory): Likewise.
16481 (*sibcall_value_memory): Likewise.
16482 Peepholes of indirect call and jump via memory: Likewise.
16483 * config/i386/i386.opt: Likewise.
16484 * config/i386/predicates.md (indirect_branch_operand): Likewise.
16485 (GOT_memory_operand): Likewise.
16486 (call_insn_operand): Likewise.
16487 (sibcall_insn_operand): Likewise.
16488 (GOT32_symbol_operand): Likewise.
16490 2018-01-15 Jakub Jelinek <jakub@redhat.com>
16492 PR middle-end/83837
16493 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
16494 type rather than type addr's type points to.
16495 (expand_omp_atomic_mutex): Likewise.
16496 (expand_omp_atomic): Likewise.
16498 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16501 * config/i386/i386.c (output_indirect_thunk_function): Use
16502 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
16503 for __x86_return_thunk.
16505 2018-01-15 Richard Biener <rguenther@suse.de>
16507 PR middle-end/83850
16508 * expmed.c (extract_bit_field_1): Fix typo.
16510 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16513 * config/arm/iterators.md (VF): New mode iterator.
16514 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
16515 Remove integer-related logic from pattern.
16516 (neon_vabd<mode>_3): Likewise.
16518 2018-01-15 Jakub Jelinek <jakub@redhat.com>
16520 PR middle-end/82694
16521 * common.opt (fstrict-overflow): No longer an alias.
16522 (fwrapv-pointer): New option.
16523 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
16524 also for pointer types based on flag_wrapv_pointer.
16525 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
16526 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
16527 opts->x_flag_wrapv got set.
16528 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
16529 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
16530 POINTER_TYPE_OVERFLOW_UNDEFINED.
16531 * match.pd: Likewise in address comparison pattern.
16532 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
16534 2018-01-15 Richard Biener <rguenther@suse.de>
16537 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
16538 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
16539 Reset type names to their identifier if their TYPE_DECL doesn't
16540 have linkage (and thus is used for ODR and devirt).
16541 (save_debug_info_for_decl): Remove.
16542 (save_debug_info_for_type): Likewise.
16543 (add_tree_to_fld_list): Adjust.
16544 * tree-pretty-print.c (dump_generic_node): Make dumping of
16545 type names more robust.
16547 2018-01-15 Richard Biener <rguenther@suse.de>
16549 * BASE-VER: Bump to 8.0.1.
16551 2018-01-14 Martin Sebor <msebor@redhat.com>
16554 * builtins.c (check_access): Avoid warning when the no-warning bit
16557 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
16559 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
16560 * ira-color (allocno_hard_regs_compare): Likewise.
16562 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
16565 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
16566 Use .pushsection/.popsection.
16568 2018-01-14 Martin Sebor <msebor@redhat.com>
16571 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
16573 2018-01-14 Jakub Jelinek <jakub@redhat.com>
16575 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
16576 entry from extra_headers.
16577 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
16578 extra_headers, make the list bitwise identical to the i?86-*-* one.
16580 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16582 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
16583 -mcmodel=large with -mindirect-branch=thunk,
16584 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
16585 -mfunction-return=thunk-extern.
16586 * doc/invoke.texi: Document -mcmodel=large is incompatible with
16587 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
16588 -mfunction-return=thunk and -mfunction-return=thunk-extern.
16590 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16592 * config/i386/i386.c (print_reg): Print the name of the full
16593 integer register without '%'.
16594 (ix86_print_operand): Handle 'V'.
16595 * doc/extend.texi: Document 'V' modifier.
16597 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16599 * config/i386/constraints.md (Bs): Disallow memory operand for
16600 -mindirect-branch-register.
16602 * config/i386/predicates.md (indirect_branch_operand): Likewise.
16603 (GOT_memory_operand): Likewise.
16604 (call_insn_operand): Likewise.
16605 (sibcall_insn_operand): Likewise.
16606 (GOT32_symbol_operand): Likewise.
16607 * config/i386/i386.md (indirect_jump): Call convert_memory_address
16608 for -mindirect-branch-register.
16609 (tablejump): Likewise.
16610 (*sibcall_memory): Likewise.
16611 (*sibcall_value_memory): Likewise.
16612 Disallow peepholes of indirect call and jump via memory for
16613 -mindirect-branch-register.
16614 (*call_pop): Replace m with Bw.
16615 (*call_value_pop): Likewise.
16616 (*sibcall_pop_memory): Replace m with Bs.
16617 * config/i386/i386.opt (mindirect-branch-register): New option.
16618 * doc/invoke.texi: Document -mindirect-branch-register option.
16620 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16622 * config/i386/i386-protos.h (ix86_output_function_return): New.
16623 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
16624 set function_return_type.
16625 (indirect_thunk_name): Add ret_p to indicate thunk for function
16627 (output_indirect_thunk_function): Pass false to
16628 indirect_thunk_name.
16629 (ix86_output_indirect_branch_via_reg): Likewise.
16630 (ix86_output_indirect_branch_via_push): Likewise.
16631 (output_indirect_thunk_function): Create alias for function
16632 return thunk if regno < 0.
16633 (ix86_output_function_return): New function.
16634 (ix86_handle_fndecl_attribute): Handle function_return.
16635 (ix86_attribute_table): Add function_return.
16636 * config/i386/i386.h (machine_function): Add
16637 function_return_type.
16638 * config/i386/i386.md (simple_return_internal): Use
16639 ix86_output_function_return.
16640 (simple_return_internal_long): Likewise.
16641 * config/i386/i386.opt (mfunction-return=): New option.
16642 (indirect_branch): Mention -mfunction-return=.
16643 * doc/extend.texi: Document function_return function attribute.
16644 * doc/invoke.texi: Document -mfunction-return= option.
16646 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16648 * config/i386/i386-opts.h (indirect_branch): New.
16649 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
16650 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
16651 with local indirect jump when converting indirect call and jump.
16652 (ix86_set_indirect_branch_type): New.
16653 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
16654 (indirectlabelno): New.
16655 (indirect_thunk_needed): Likewise.
16656 (indirect_thunk_bnd_needed): Likewise.
16657 (indirect_thunks_used): Likewise.
16658 (indirect_thunks_bnd_used): Likewise.
16659 (INDIRECT_LABEL): Likewise.
16660 (indirect_thunk_name): Likewise.
16661 (output_indirect_thunk): Likewise.
16662 (output_indirect_thunk_function): Likewise.
16663 (ix86_output_indirect_branch_via_reg): Likewise.
16664 (ix86_output_indirect_branch_via_push): Likewise.
16665 (ix86_output_indirect_branch): Likewise.
16666 (ix86_output_indirect_jmp): Likewise.
16667 (ix86_code_end): Call output_indirect_thunk_function if needed.
16668 (ix86_output_call_insn): Call ix86_output_indirect_branch if
16670 (ix86_handle_fndecl_attribute): Handle indirect_branch.
16671 (ix86_attribute_table): Add indirect_branch.
16672 * config/i386/i386.h (machine_function): Add indirect_branch_type
16673 and has_local_indirect_jump.
16674 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
16676 (tablejump): Likewise.
16677 (*indirect_jump): Use ix86_output_indirect_jmp.
16678 (*tablejump_1): Likewise.
16679 (simple_return_indirect_internal): Likewise.
16680 * config/i386/i386.opt (mindirect-branch=): New option.
16681 (indirect_branch): New.
16684 (thunk-inline): Likewise.
16685 (thunk-extern): Likewise.
16686 * doc/extend.texi: Document indirect_branch function attribute.
16687 * doc/invoke.texi: Document -mindirect-branch= option.
16689 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
16692 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
16694 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
16696 * ipa-inline.c (want_inline_small_function_p): Return false if
16697 inlining has already failed with CIF_FINAL_ERROR.
16698 (update_caller_keys): Call want_inline_small_function_p before
16700 (update_callee_keys): Likewise.
16702 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
16704 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
16706 (rs6000_quadword_masked_address_p): Likewise.
16707 (quad_aligned_load_p): Likewise.
16708 (quad_aligned_store_p): Likewise.
16709 (const_load_sequence_p): Add comment to describe the outer-most loop.
16710 (mimic_memory_attributes_and_flags): New function.
16711 (rs6000_gen_stvx): Likewise.
16712 (replace_swapped_aligned_store): Likewise.
16713 (rs6000_gen_lvx): Likewise.
16714 (replace_swapped_aligned_load): Likewise.
16715 (replace_swapped_load_constant): Capitalize argument name in
16716 comment describing this function.
16717 (rs6000_analyze_swaps): Add a third pass to search for vector loads
16718 and stores that access quad-word aligned addresses and replace
16719 with stvx or lvx instructions when appropriate.
16720 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
16721 New function prototype.
16722 (rs6000_quadword_masked_address_p): Likewise.
16723 (rs6000_gen_lvx): Likewise.
16724 (rs6000_gen_stvx): Likewise.
16725 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
16726 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
16727 when memory address is aligned.
16728 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
16729 this split to select lvx instruction when memory address is aligned.
16730 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
16731 instruction when memory address is aligned.
16732 (*vsx_le_perm_load_v16qi): Likewise.
16733 (four unnamed splitters): Modify to select the stvx instruction
16734 when memory is aligned.
16736 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
16738 * predict.c (determine_unlikely_bbs): Handle correctly BBs
16739 which appears in the queue multiple times.
16741 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16742 Alan Hayward <alan.hayward@arm.com>
16743 David Sherwood <david.sherwood@arm.com>
16745 * tree-vectorizer.h (vec_lower_bound): New structure.
16746 (_loop_vec_info): Add check_nonzero and lower_bounds.
16747 (LOOP_VINFO_CHECK_NONZERO): New macro.
16748 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
16749 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
16750 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
16751 fields. Make seg_len the distance travelled, not including the
16753 (dr_direction_indicator): Declare.
16754 (dr_zero_step_indicator): Likewise.
16755 (dr_known_forward_stride_p): Likewise.
16756 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
16758 (runtime_alias_check_p): Allow runtime alias checks with
16760 (operator ==): Compare access_size and align.
16761 (prune_runtime_alias_test_list): Rework for new distinction between
16762 the access_size and seg_len.
16763 (create_intersect_range_checks_index): Likewise. Cope with polynomial
16765 (get_segment_min_max): New function.
16766 (create_intersect_range_checks): Use it.
16767 (dr_step_indicator): New function.
16768 (dr_direction_indicator): Likewise.
16769 (dr_zero_step_indicator): Likewise.
16770 (dr_known_forward_stride_p): Likewise.
16771 * tree-loop-distribution.c (data_ref_segment_size): Return
16772 DR_STEP * (niters - 1).
16773 (compute_alias_check_pairs): Update call to the dr_with_seg_len
16775 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
16776 (vect_preserves_scalar_order_p): New function, split out from...
16777 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
16778 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
16779 (vect_vfa_access_size): New function.
16780 (vect_vfa_align): Likewise.
16781 (vect_compile_time_alias): Take access_size_a and access_b arguments.
16782 (dump_lower_bound): New function.
16783 (vect_check_lower_bound): Likewise.
16784 (vect_small_gap_p): Likewise.
16785 (vectorizable_with_step_bound_p): Likewise.
16786 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
16787 depencies if the vectorization factor is 1. Convert the checks
16788 for nonzero steps into checks on the bounds of DR_STEP. Try using
16789 a bunds check for variable steps if the minimum required step is
16790 relatively small. Update calls to the dr_with_seg_len
16791 constructor and to vect_compile_time_alias.
16792 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
16794 (vect_loop_versioning): Call it.
16795 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
16797 (vect_estimate_min_profitable_iters): Account for any bounds checks.
16799 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16800 Alan Hayward <alan.hayward@arm.com>
16801 David Sherwood <david.sherwood@arm.com>
16803 * doc/sourcebuild.texi (vect_scatter_store): Document.
16804 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
16806 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
16808 * genopinit.c (main): Add supports_vec_scatter_store and
16809 supports_vec_scatter_store_cached to target_optabs.
16810 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
16811 IFN_MASK_SCATTER_STORE.
16812 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
16814 * internal-fn.h (internal_store_fn_p): Declare.
16815 (internal_fn_stored_value_index): Likewise.
16816 * internal-fn.c (scatter_store_direct): New macro.
16817 (expand_scatter_store_optab_fn): New function.
16818 (direct_scatter_store_optab_supported_p): New macro.
16819 (internal_store_fn_p): New function.
16820 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
16821 IFN_MASK_SCATTER_STORE.
16822 (internal_fn_mask_index): Likewise.
16823 (internal_fn_stored_value_index): New function.
16824 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
16825 for scatter stores.
16826 * optabs-query.h (supports_vec_scatter_store_p): Declare.
16827 * optabs-query.c (supports_vec_scatter_store_p): New function.
16828 * tree-vectorizer.h (vect_get_store_rhs): Declare.
16829 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
16830 true for scatter stores.
16831 (vect_gather_scatter_fn_p): Handle scatter stores too.
16832 (vect_check_gather_scatter): Consider using scatter stores if
16833 supports_vec_scatter_store_p.
16834 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
16835 scatter stores too.
16836 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
16837 internal_fn_stored_value_index.
16838 (check_load_store_masking): Handle scatter stores too.
16839 (vect_get_store_rhs): Make public.
16840 (vectorizable_call): Use internal_store_fn_p.
16841 (vectorizable_store): Handle scatter store internal functions.
16842 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
16843 when deciding whether the end of the group has been reached.
16844 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
16845 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
16846 (mask_scatter_store<mode>): New insns.
16848 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16849 Alan Hayward <alan.hayward@arm.com>
16850 David Sherwood <david.sherwood@arm.com>
16852 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
16853 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
16854 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
16856 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
16857 Use vect_truncate_gather_scatter_offset if we can't treat the
16858 operation as a normal gather load or scatter store.
16859 (get_group_load_store_type): Take the gather_scatter_info
16860 as argument. Try using a gather load or scatter store for
16861 single-element groups.
16862 (get_load_store_type): Update calls to get_group_load_store_type
16863 and vect_use_strided_gather_scatters_p.
16865 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16866 Alan Hayward <alan.hayward@arm.com>
16867 David Sherwood <david.sherwood@arm.com>
16869 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
16870 optional tree argument.
16871 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
16873 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
16874 but continue to use the current value as a fallback.
16875 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
16876 to compare the updates.
16877 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
16878 (get_load_store_type): Use it when handling a strided access.
16879 (vect_get_strided_load_store_ops): New function.
16880 (vect_get_data_ptr_increment): Likewise.
16881 (vectorizable_load): Handle strided gather loads. Always pass
16882 a step to vect_create_data_ref_ptr and bump_vector_ptr.
16884 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16885 Alan Hayward <alan.hayward@arm.com>
16886 David Sherwood <david.sherwood@arm.com>
16888 * doc/md.texi (gather_load@var{m}): Document.
16889 (mask_gather_load@var{m}): Likewise.
16890 * genopinit.c (main): Add supports_vec_gather_load and
16891 supports_vec_gather_load_cached to target_optabs.
16892 * optabs-tree.c (init_tree_optimization_optabs): Use
16893 ggc_cleared_alloc to allocate target_optabs.
16894 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
16895 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
16897 * internal-fn.h (internal_load_fn_p): Declare.
16898 (internal_gather_scatter_fn_p): Likewise.
16899 (internal_fn_mask_index): Likewise.
16900 (internal_gather_scatter_fn_supported_p): Likewise.
16901 * internal-fn.c (gather_load_direct): New macro.
16902 (expand_gather_load_optab_fn): New function.
16903 (direct_gather_load_optab_supported_p): New macro.
16904 (direct_internal_fn_optab): New function.
16905 (internal_load_fn_p): Likewise.
16906 (internal_gather_scatter_fn_p): Likewise.
16907 (internal_fn_mask_index): Likewise.
16908 (internal_gather_scatter_fn_supported_p): Likewise.
16909 * optabs-query.c (supports_at_least_one_mode_p): New function.
16910 (supports_vec_gather_load_p): Likewise.
16911 * optabs-query.h (supports_vec_gather_load_p): Declare.
16912 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
16913 and memory_type field.
16914 (NUM_PATTERNS): Bump to 15.
16915 * tree-vect-data-refs.c: Include internal-fn.h.
16916 (vect_gather_scatter_fn_p): New function.
16917 (vect_describe_gather_scatter_call): Likewise.
16918 (vect_check_gather_scatter): Try using internal functions for
16919 gather loads. Recognize existing calls to a gather load function.
16920 (vect_analyze_data_refs): Consider using gather loads if
16921 supports_vec_gather_load_p.
16922 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
16923 (vect_get_gather_scatter_offset_type): Likewise.
16924 (vect_convert_mask_for_vectype): Likewise.
16925 (vect_add_conversion_to_patterm): Likewise.
16926 (vect_try_gather_scatter_pattern): Likewise.
16927 (vect_recog_gather_scatter_pattern): New pattern recognizer.
16928 (vect_vect_recog_func_ptrs): Add it.
16929 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
16930 internal_fn_mask_index and internal_gather_scatter_fn_p.
16931 (check_load_store_masking): Take the gather_scatter_info as an
16932 argument and handle gather loads.
16933 (vect_get_gather_scatter_ops): New function.
16934 (vectorizable_call): Check internal_load_fn_p.
16935 (vectorizable_load): Likewise. Handle gather load internal
16937 (vectorizable_store): Update call to check_load_store_masking.
16938 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
16939 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
16940 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
16941 (aarch64_gather_scale_operand_d): New predicates.
16942 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
16943 (mask_gather_load<mode>): New insns.
16945 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16946 Alan Hayward <alan.hayward@arm.com>
16947 David Sherwood <david.sherwood@arm.com>
16949 * optabs.def (fold_left_plus_optab): New optab.
16950 * doc/md.texi (fold_left_plus_@var{m}): Document.
16951 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
16952 * internal-fn.c (fold_left_direct): Define.
16953 (expand_fold_left_optab_fn): Likewise.
16954 (direct_fold_left_optab_supported_p): Likewise.
16955 * fold-const-call.c (fold_const_fold_left): New function.
16956 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
16957 * tree-parloops.c (valid_reduction_p): New function.
16958 (gather_scalar_reductions): Use it.
16959 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
16960 (vect_finish_replace_stmt): Declare.
16961 * tree-vect-loop.c (fold_left_reduction_fn): New function.
16962 (needs_fold_left_reduction_p): New function, split out from...
16963 (vect_is_simple_reduction): ...here. Accept reductions that
16964 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
16965 (vect_force_simple_reduction): Also store the reduction type in
16966 the assignment's STMT_VINFO_REDUC_TYPE.
16967 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
16968 (merge_with_identity): New function.
16969 (vect_expand_fold_left): Likewise.
16970 (vectorize_fold_left_reduction): Likewise.
16971 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
16972 scalar phi in place for it. Check for target support and reject
16973 cases that would reassociate the operation. Defer the transform
16974 phase to vectorize_fold_left_reduction.
16975 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
16976 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
16977 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
16979 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16981 * tree-if-conv.c (predicate_mem_writes): Remove redundant
16982 call to ifc_temp_var.
16984 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16985 Alan Hayward <alan.hayward@arm.com>
16986 David Sherwood <david.sherwood@arm.com>
16988 * target.def (legitimize_address_displacement): Take the original
16989 offset as a poly_int.
16990 * targhooks.h (default_legitimize_address_displacement): Update
16992 * targhooks.c (default_legitimize_address_displacement): Likewise.
16993 * doc/tm.texi: Regenerate.
16994 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
16995 as an argument, moving assert of ad->disp == ad->disp_term to...
16996 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
16997 Try calling targetm.legitimize_address_displacement before expanding
16998 the address rather than afterwards, and adjust for the new interface.
16999 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
17000 Match the new hook interface. Handle SVE addresses.
17001 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
17002 new hook interface.
17004 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17006 * Makefile.in (OBJS): Add early-remat.o.
17007 * target.def (select_early_remat_modes): New hook.
17008 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
17009 * doc/tm.texi: Regenerate.
17010 * targhooks.h (default_select_early_remat_modes): Declare.
17011 * targhooks.c (default_select_early_remat_modes): New function.
17012 * timevar.def (TV_EARLY_REMAT): New timevar.
17013 * passes.def (pass_early_remat): New pass.
17014 * tree-pass.h (make_pass_early_remat): Declare.
17015 * early-remat.c: New file.
17016 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
17018 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
17020 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17021 Alan Hayward <alan.hayward@arm.com>
17022 David Sherwood <david.sherwood@arm.com>
17024 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
17025 vfm1 with a bound_epilog parameter.
17026 (vect_do_peeling): Update calls accordingly, and move the prologue
17027 call earlier in the function. Treat the base bound_epilog as 0 for
17028 fully-masked loops and retain vf - 1 for other loops. Add 1 to
17029 this base when peeling for gaps.
17030 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
17031 with fully-masked loops.
17032 (vect_estimate_min_profitable_iters): Handle the single peeled
17033 iteration in that case.
17035 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17036 Alan Hayward <alan.hayward@arm.com>
17037 David Sherwood <david.sherwood@arm.com>
17039 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
17040 single-element interleaving even if the size is not a power of 2.
17041 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
17042 accesses for single-element interleaving if the group size is
17045 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17046 Alan Hayward <alan.hayward@arm.com>
17047 David Sherwood <david.sherwood@arm.com>
17049 * doc/md.texi (fold_extract_last_@var{m}): Document.
17050 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
17051 * optabs.def (fold_extract_last_optab): New optab.
17052 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
17053 * internal-fn.c (fold_extract_direct): New macro.
17054 (expand_fold_extract_optab_fn): Likewise.
17055 (direct_fold_extract_optab_supported_p): Likewise.
17056 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
17057 * tree-vect-loop.c (vect_model_reduction_cost): Handle
17058 EXTRACT_LAST_REDUCTION.
17059 (get_initial_def_for_reduction): Do not create an initial vector
17060 for EXTRACT_LAST_REDUCTION reductions.
17061 (vectorizable_reduction): Leave the scalar phi in place for
17062 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
17063 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
17064 epilogue code for EXTRACT_LAST_REDUCTION and defer the
17065 transform phase to vectorizable_condition.
17066 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
17068 (vect_finish_stmt_generation): ...here.
17069 (vect_finish_replace_stmt): New function.
17070 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
17071 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
17073 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
17075 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17076 Alan Hayward <alan.hayward@arm.com>
17077 David Sherwood <david.sherwood@arm.com>
17079 * doc/md.texi (extract_last_@var{m}): Document.
17080 * optabs.def (extract_last_optab): New optab.
17081 * internal-fn.def (EXTRACT_LAST): New internal function.
17082 * internal-fn.c (cond_unary_direct): New macro.
17083 (expand_cond_unary_optab_fn): Likewise.
17084 (direct_cond_unary_optab_supported_p): Likewise.
17085 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
17086 loops using EXTRACT_LAST.
17087 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
17088 (extract_last_<mode>): ...this optab.
17089 (vec_extract<mode><Vel>): Update accordingly.
17091 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17092 Alan Hayward <alan.hayward@arm.com>
17093 David Sherwood <david.sherwood@arm.com>
17095 * target.def (empty_mask_is_expensive): New hook.
17096 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
17097 * doc/tm.texi: Regenerate.
17098 * targhooks.h (default_empty_mask_is_expensive): Declare.
17099 * targhooks.c (default_empty_mask_is_expensive): New function.
17100 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
17101 if the target says that empty masks are expensive.
17102 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
17104 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
17106 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17107 Alan Hayward <alan.hayward@arm.com>
17108 David Sherwood <david.sherwood@arm.com>
17110 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
17111 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
17112 (vect_use_loop_mask_for_alignment_p): New function.
17113 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
17114 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
17115 niters_skip argument. Make sure that the first niters_skip elements
17116 of the first iteration are inactive.
17117 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
17118 Update call to vect_set_loop_masks_directly.
17119 (get_misalign_in_elems): New function, split out from...
17120 (vect_gen_prolog_loop_niters): ...here.
17121 (vect_update_init_of_dr): Take a code argument that specifies whether
17122 the adjustment should be added or subtracted.
17123 (vect_update_init_of_drs): Likewise.
17124 (vect_prepare_for_masked_peels): New function.
17125 (vect_do_peeling): Skip prologue peeling if we're using a mask
17126 instead. Update call to vect_update_inits_of_drs.
17127 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17129 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
17130 alignment. Do not include the number of peeled iterations in
17131 the minimum threshold in that case.
17132 (vectorizable_induction): Adjust the start value down by
17133 LOOP_VINFO_MASK_SKIP_NITERS iterations.
17134 (vect_transform_loop): Call vect_prepare_for_masked_peels.
17135 Take the number of skipped iterations into account when calculating
17137 * tree-vect-stmts.c (vect_gen_while_not): New function.
17139 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17140 Alan Hayward <alan.hayward@arm.com>
17141 David Sherwood <david.sherwood@arm.com>
17143 * doc/sourcebuild.texi (vect_fully_masked): Document.
17144 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
17145 default value to 0.
17146 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
17148 (vect_analyze_loop_2): ...here. Don't check the vectorization
17149 factor against the number of loop iterations if the loop is
17152 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17153 Alan Hayward <alan.hayward@arm.com>
17154 David Sherwood <david.sherwood@arm.com>
17156 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
17157 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
17158 (dump_groups): Update accordingly.
17159 (iv_use::mem_type): New member variable.
17160 (address_p): New function.
17161 (record_use): Add a mem_type argument and initialize the new
17163 (record_group_use): Add a mem_type argument. Use address_p.
17164 Remove obsolete null checks of base_object. Update call to record_use.
17165 (find_interesting_uses_op): Update call to record_group_use.
17166 (find_interesting_uses_cond): Likewise.
17167 (find_interesting_uses_address): Likewise.
17168 (get_mem_type_for_internal_fn): New function.
17169 (find_address_like_use): Likewise.
17170 (find_interesting_uses_stmt): Try find_address_like_use before
17171 calling find_interesting_uses_op.
17172 (addr_offset_valid_p): Use the iv mem_type field as the type
17173 of the addressed memory.
17174 (add_autoinc_candidates): Likewise.
17175 (get_address_cost): Likewise.
17176 (split_small_address_groups_p): Use address_p.
17177 (split_address_groups): Likewise.
17178 (add_iv_candidate_for_use): Likewise.
17179 (autoinc_possible_for_pair): Likewise.
17180 (rewrite_groups): Likewise.
17181 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
17182 (determine_group_iv_cost): Update after split of USE_ADDRESS.
17183 (get_alias_ptr_type_for_ptr_address): New function.
17184 (rewrite_use_address): Rewrite address uses in calls that were
17185 identified by find_address_like_use.
17187 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17188 Alan Hayward <alan.hayward@arm.com>
17189 David Sherwood <david.sherwood@arm.com>
17191 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
17193 * gimple-expr.h (is_gimple_addressable: Likewise.
17194 * gimple-expr.c (is_gimple_address): Likewise.
17195 * internal-fn.c (expand_call_mem_ref): New function.
17196 (expand_mask_load_optab_fn): Use it.
17197 (expand_mask_store_optab_fn): Likewise.
17199 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17200 Alan Hayward <alan.hayward@arm.com>
17201 David Sherwood <david.sherwood@arm.com>
17203 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
17204 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
17205 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
17206 (cond_umax@var{mode}): Document.
17207 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
17208 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
17209 (cond_umin_optab, cond_umax_optab): New optabs.
17210 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
17211 (COND_IOR, COND_XOR): New internal functions.
17212 * internal-fn.h (get_conditional_internal_fn): Declare.
17213 * internal-fn.c (cond_binary_direct): New macro.
17214 (expand_cond_binary_optab_fn): Likewise.
17215 (direct_cond_binary_optab_supported_p): Likewise.
17216 (get_conditional_internal_fn): New function.
17217 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
17218 Cope with reduction statements that are vectorized as calls rather
17220 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
17221 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
17222 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
17223 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
17224 (UNSPEC_COND_EOR): New unspecs.
17225 (optab): Add mappings for them.
17226 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
17227 (sve_int_op, sve_fp_op): New int attributes.
17229 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17230 Alan Hayward <alan.hayward@arm.com>
17231 David Sherwood <david.sherwood@arm.com>
17233 * optabs.def (while_ult_optab): New optab.
17234 * doc/md.texi (while_ult@var{m}@var{n}): Document.
17235 * internal-fn.def (WHILE_ULT): New internal function.
17236 * internal-fn.h (direct_internal_fn_supported_p): New override
17237 that takes two types as argument.
17238 * internal-fn.c (while_direct): New macro.
17239 (expand_while_optab_fn): New function.
17240 (convert_optab_supported_p): Likewise.
17241 (direct_while_optab_supported_p): New macro.
17242 * wide-int.h (wi::udiv_ceil): New function.
17243 * tree-vectorizer.h (rgroup_masks): New structure.
17244 (vec_loop_masks): New typedef.
17245 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
17246 and fully_masked_p.
17247 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
17248 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
17249 (vect_max_vf): New function.
17250 (slpeel_make_loop_iterate_ntimes): Delete.
17251 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
17252 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
17253 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
17254 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
17255 internal-fn.h, stor-layout.h and optabs-query.h.
17256 (vect_set_loop_mask): New function.
17257 (add_preheader_seq): Likewise.
17258 (add_header_seq): Likewise.
17259 (interleave_supported_p): Likewise.
17260 (vect_maybe_permute_loop_masks): Likewise.
17261 (vect_set_loop_masks_directly): Likewise.
17262 (vect_set_loop_condition_masked): Likewise.
17263 (vect_set_loop_condition_unmasked): New function, split out from
17264 slpeel_make_loop_iterate_ntimes.
17265 (slpeel_make_loop_iterate_ntimes): Rename to..
17266 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
17267 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
17268 (vect_do_peeling): Update call accordingly.
17269 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
17271 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17272 mask_compare_type, can_fully_mask_p and fully_masked_p.
17273 (release_vec_loop_masks): New function.
17274 (_loop_vec_info): Use it to free the loop masks.
17275 (can_produce_all_loop_masks_p): New function.
17276 (vect_get_max_nscalars_per_iter): Likewise.
17277 (vect_verify_full_masking): Likewise.
17278 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
17279 retries, and free the mask rgroups before retrying. Check loop-wide
17280 reasons for disallowing fully-masked loops. Make the final decision
17281 about whether use a fully-masked loop or not.
17282 (vect_estimate_min_profitable_iters): Do not assume that peeling
17283 for the number of iterations will be needed for fully-masked loops.
17284 (vectorizable_reduction): Disable fully-masked loops.
17285 (vectorizable_live_operation): Likewise.
17286 (vect_halve_mask_nunits): New function.
17287 (vect_double_mask_nunits): Likewise.
17288 (vect_record_loop_mask): Likewise.
17289 (vect_get_loop_mask): Likewise.
17290 (vect_transform_loop): Handle the case in which the final loop
17291 iteration might handle a partial vector. Call vect_set_loop_condition
17292 instead of slpeel_make_loop_iterate_ntimes.
17293 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
17294 (check_load_store_masking): New function.
17295 (prepare_load_store_mask): Likewise.
17296 (vectorizable_store): Handle fully-masked loops.
17297 (vectorizable_load): Likewise.
17298 (supportable_widening_operation): Use vect_halve_mask_nunits for
17300 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
17301 (vect_gen_while): New function.
17302 * config/aarch64/aarch64.md (umax<mode>3): New expander.
17303 (aarch64_uqdec<mode>): New insn.
17305 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17306 Alan Hayward <alan.hayward@arm.com>
17307 David Sherwood <david.sherwood@arm.com>
17309 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
17310 (reduc_xor_scal_optab): New optabs.
17311 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
17312 (reduc_xor_scal_@var{m}): Document.
17313 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
17314 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
17315 internal functions.
17316 * fold-const-call.c (fold_const_call): Handle them.
17317 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
17318 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
17319 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
17320 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
17321 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
17322 (UNSPEC_XORV): New unspecs.
17323 (optab): Add entries for them.
17324 (BITWISEV): New int iterator.
17325 (bit_reduc_op): New int attributes.
17327 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17328 Alan Hayward <alan.hayward@arm.com>
17329 David Sherwood <david.sherwood@arm.com>
17331 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
17332 * internal-fn.def (VEC_SHL_INSERT): New internal function.
17333 * optabs.def (vec_shl_insert_optab): New optab.
17334 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
17335 (duplicate_and_interleave): Likewise.
17336 * tree-vect-loop.c: Include internal-fn.h.
17337 (neutral_op_for_slp_reduction): New function, split out from
17338 get_initial_defs_for_reduction.
17339 (get_initial_def_for_reduction): Handle option 2 for variable-length
17340 vectors by loading the neutral value into a vector and then shifting
17341 the initial value into element 0.
17342 (get_initial_defs_for_reduction): Replace the code argument with
17343 the neutral value calculated by neutral_op_for_slp_reduction.
17344 Use gimple_build_vector for constant-length vectors.
17345 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
17346 but the first group_size elements have a neutral value.
17347 Use duplicate_and_interleave otherwise.
17348 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
17349 Update call to get_initial_defs_for_reduction. Handle SLP
17350 reductions for variable-length vectors by creating one vector
17351 result for each scalar result, with the elements associated
17352 with other scalar results stubbed out with the neutral value.
17353 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
17354 Require IFN_VEC_SHL_INSERT for double reductions on
17355 variable-length vectors, or SLP reductions that have
17356 a neutral value. Require can_duplicate_and_interleave_p
17357 support for variable-length unchained SLP reductions if there
17358 is no neutral value, such as for MIN/MAX reductions. Also require
17359 the number of vector elements to be a multiple of the number of
17360 SLP statements when doing variable-length unchained SLP reductions.
17361 Update call to vect_create_epilog_for_reduction.
17362 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
17363 and remove initial values.
17364 (duplicate_and_interleave): Make public.
17365 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
17366 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
17368 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17369 Alan Hayward <alan.hayward@arm.com>
17370 David Sherwood <david.sherwood@arm.com>
17372 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
17373 (can_duplicate_and_interleave_p): New function.
17374 (vect_get_and_check_slp_defs): Take the vector of statements
17375 rather than just the current one. Remove excess parentheses.
17376 Restriction rejectinon of vect_constant_def and vect_external_def
17377 for variable-length vectors to boolean types, or types for which
17378 can_duplicate_and_interleave_p is false.
17379 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
17380 (duplicate_and_interleave): New function.
17381 (vect_get_constant_vectors): Use gimple_build_vector for
17382 constant-length vectors and suitable variable-length constant
17383 vectors. Use duplicate_and_interleave for other variable-length
17384 vectors. Don't defer the update when inserting new statements.
17386 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17387 Alan Hayward <alan.hayward@arm.com>
17388 David Sherwood <david.sherwood@arm.com>
17390 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
17391 min_profitable_iters doesn't go negative.
17393 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17394 Alan Hayward <alan.hayward@arm.com>
17395 David Sherwood <david.sherwood@arm.com>
17397 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
17398 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
17399 * optabs.def (vec_mask_load_lanes_optab): New optab.
17400 (vec_mask_store_lanes_optab): Likewise.
17401 * internal-fn.def (MASK_LOAD_LANES): New internal function.
17402 (MASK_STORE_LANES): Likewise.
17403 * internal-fn.c (mask_load_lanes_direct): New macro.
17404 (mask_store_lanes_direct): Likewise.
17405 (expand_mask_load_optab_fn): Handle masked operations.
17406 (expand_mask_load_lanes_optab_fn): New macro.
17407 (expand_mask_store_optab_fn): Handle masked operations.
17408 (expand_mask_store_lanes_optab_fn): New macro.
17409 (direct_mask_load_lanes_optab_supported_p): Likewise.
17410 (direct_mask_store_lanes_optab_supported_p): Likewise.
17411 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
17413 (vect_load_lanes_supported): Likewise.
17414 * tree-vect-data-refs.c (strip_conversion): New function.
17415 (can_group_stmts_p): Likewise.
17416 (vect_analyze_data_ref_accesses): Use it instead of checking
17417 for a pair of assignments.
17418 (vect_store_lanes_supported): Take a masked_p parameter.
17419 (vect_load_lanes_supported): Likewise.
17420 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
17421 vect_store_lanes_supported and vect_load_lanes_supported.
17422 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
17423 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
17424 parameter. Don't allow gaps for masked accesses.
17425 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
17426 and vect_load_lanes_supported.
17427 (get_load_store_type): Take a masked_p parameter and update
17428 call to get_group_load_store_type.
17429 (vectorizable_store): Update call to get_load_store_type.
17430 Handle IFN_MASK_STORE_LANES.
17431 (vectorizable_load): Update call to get_load_store_type.
17432 Handle IFN_MASK_LOAD_LANES.
17434 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17435 Alan Hayward <alan.hayward@arm.com>
17436 David Sherwood <david.sherwood@arm.com>
17438 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
17440 * config/aarch64/aarch64-protos.h
17441 (aarch64_sve_struct_memory_operand_p): Declare.
17442 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
17443 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
17444 (VPRED, vpred): Handle SVE structure modes.
17445 * config/aarch64/constraints.md (Utx): New constraint.
17446 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
17447 (aarch64_sve_struct_nonimmediate_operand): New predicates.
17448 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
17449 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
17450 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
17451 structure modes. Split into pieces after RA.
17452 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
17453 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
17455 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
17456 SVE structure modes.
17457 (aarch64_classify_address): Likewise.
17458 (sizetochar): Move earlier in file.
17459 (aarch64_print_operand): Handle SVE register lists.
17460 (aarch64_array_mode): New function.
17461 (aarch64_sve_struct_memory_operand_p): Likewise.
17462 (TARGET_ARRAY_MODE): Redefine.
17464 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17465 Alan Hayward <alan.hayward@arm.com>
17466 David Sherwood <david.sherwood@arm.com>
17468 * target.def (array_mode): New target hook.
17469 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
17470 * doc/tm.texi: Regenerate.
17471 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
17472 * hooks.c (hook_optmode_mode_uhwi_none): New function.
17473 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
17474 targetm.array_mode.
17475 * stor-layout.c (mode_for_array): Likewise. Support polynomial
17478 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17479 Alan Hayward <alan.hayward@arm.com>
17480 David Sherwood <david.sherwood@arm.com>
17482 * fold-const.c (fold_binary_loc): Check the argument types
17483 rather than the result type when testing for a vector operation.
17485 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17487 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
17488 * doc/tm.texi: Regenerate.
17490 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17491 Alan Hayward <alan.hayward@arm.com>
17492 David Sherwood <david.sherwood@arm.com>
17494 * doc/invoke.texi (-msve-vector-bits=): Document new option.
17495 (sve): Document new AArch64 extension.
17496 * doc/md.texi (w): Extend the description of the AArch64
17497 constraint to include SVE vectors.
17498 (Upl, Upa): Document new AArch64 predicate constraints.
17499 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
17501 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
17502 (msve-vector-bits=): New option.
17503 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
17504 SVE when these are disabled.
17505 (sve): New extension.
17506 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
17507 modes. Adjust their number of units based on aarch64_sve_vg.
17508 (MAX_BITSIZE_MODE_ANY_MODE): Define.
17509 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
17510 aarch64_addr_query_type.
17511 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
17512 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
17513 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
17514 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
17515 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
17516 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
17517 (aarch64_simd_imm_zero_p): Delete.
17518 (aarch64_check_zero_based_sve_index_immediate): Declare.
17519 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
17520 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
17521 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
17522 (aarch64_sve_float_mul_immediate_p): Likewise.
17523 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
17524 rather than an rtx.
17525 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
17526 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
17527 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
17528 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
17529 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
17530 (aarch64_regmode_natural_size): Likewise.
17531 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
17532 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
17534 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
17535 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
17536 for VG and the SVE predicate registers.
17537 (V_ALIASES): Add a "z"-prefixed alias.
17538 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
17539 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
17540 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
17541 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
17542 (REG_CLASS_NAMES): Add entries for them.
17543 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
17544 and the predicate registers.
17545 (aarch64_sve_vg): Declare.
17546 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
17547 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
17548 (REGMODE_NATURAL_SIZE): Define.
17549 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
17551 * config/aarch64/aarch64.c: Include cfgrtl.h.
17552 (simd_immediate_info): Add a constructor for series vectors,
17553 and an associated step field.
17554 (aarch64_sve_vg): New variable.
17555 (aarch64_dbx_register_number): Handle VG and the predicate registers.
17556 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
17557 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
17558 (VEC_ANY_DATA, VEC_STRUCT): New constants.
17559 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
17560 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
17561 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
17562 (aarch64_get_mask_mode): New functions.
17563 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
17564 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
17565 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
17566 predicate modes and predicate registers. Explicitly restrict
17567 GPRs to modes of 16 bytes or smaller. Only allow FP registers
17568 to store a vector mode if it is recognized by
17569 aarch64_classify_vector_mode.
17570 (aarch64_regmode_natural_size): New function.
17571 (aarch64_hard_regno_caller_save_mode): Return the original mode
17573 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
17574 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
17575 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
17576 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
17578 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
17579 does not overlap dest if the function is frame-related. Handle
17581 (aarch64_split_add_offset): New function.
17582 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
17583 them aarch64_add_offset.
17584 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
17585 and update call to aarch64_sub_sp.
17586 (aarch64_add_cfa_expression): New function.
17587 (aarch64_expand_prologue): Pass extra temporary registers to the
17588 functions above. Handle the case in which we need to emit new
17589 DW_CFA_expressions for registers that were originally saved
17590 relative to the stack pointer, but now have to be expressed
17591 relative to the frame pointer.
17592 (aarch64_output_mi_thunk): Pass extra temporary registers to the
17594 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
17595 IP0 and IP1 values for SVE frames.
17596 (aarch64_expand_vec_series): New function.
17597 (aarch64_expand_sve_widened_duplicate): Likewise.
17598 (aarch64_expand_sve_const_vector): Likewise.
17599 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
17600 Handle SVE constants. Use emit_move_insn to move a force_const_mem
17601 into the register, rather than emitting a SET directly.
17602 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
17603 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
17604 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
17605 (offset_9bit_signed_scaled_p): New functions.
17606 (aarch64_replicate_bitmask_imm): New function.
17607 (aarch64_bitmask_imm): Use it.
17608 (aarch64_cannot_force_const_mem): Reject expressions involving
17609 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
17610 (aarch64_classify_index): Handle SVE indices, by requiring
17611 a plain register index with a scale that matches the element size.
17612 (aarch64_classify_address): Handle SVE addresses. Assert that
17613 the mode of the address is VOIDmode or an integer mode.
17614 Update call to aarch64_classify_symbol.
17615 (aarch64_classify_symbolic_expression): Update call to
17616 aarch64_classify_symbol.
17617 (aarch64_const_vec_all_in_range_p): New function.
17618 (aarch64_print_vector_float_operand): Likewise.
17619 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
17620 "vN" for FP registers with SVE modes. Handle (const ...) vectors
17621 and the FP immediates 1.0 and 0.5.
17622 (aarch64_print_address_internal): Handle SVE addresses.
17623 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
17624 (aarch64_regno_regclass): Handle predicate registers.
17625 (aarch64_secondary_reload): Handle big-endian reloads of SVE
17627 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
17628 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
17629 (aarch64_convert_sve_vector_bits): New function.
17630 (aarch64_override_options): Use it to handle -msve-vector-bits=.
17631 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
17632 rather than an rtx.
17633 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
17634 Handle SVE vector and predicate modes. Accept VL-based constants
17635 that need only one temporary register, and VL offsets that require
17636 no temporary registers.
17637 (aarch64_conditional_register_usage): Mark the predicate registers
17638 as fixed if SVE isn't available.
17639 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
17640 Return true for SVE vector and predicate modes.
17641 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
17642 rather than an unsigned int. Handle SVE modes.
17643 (aarch64_preferred_simd_mode): Update call accordingly. Handle
17645 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
17647 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
17648 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
17649 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
17650 (aarch64_sve_float_mul_immediate_p): New functions.
17651 (aarch64_sve_valid_immediate): New function.
17652 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
17653 Explicitly reject structure modes. Check for INDEX constants.
17654 Handle PTRUE and PFALSE constants.
17655 (aarch64_check_zero_based_sve_index_immediate): New function.
17656 (aarch64_simd_imm_zero_p): Delete.
17657 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
17658 vector modes. Accept constants in the range of CNT[BHWD].
17659 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
17660 ask for an Advanced SIMD mode.
17661 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
17662 (aarch64_simd_vector_alignment): Handle SVE predicates.
17663 (aarch64_vectorize_preferred_vector_alignment): New function.
17664 (aarch64_simd_vector_alignment_reachable): Use it instead of
17666 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
17667 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
17669 (MAX_VECT_LEN): Delete.
17670 (expand_vec_perm_d): Add a vec_flags field.
17671 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
17672 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
17673 (aarch64_evpc_ext): Don't apply a big-endian lane correction
17675 (aarch64_evpc_rev): Rename to...
17676 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
17677 (aarch64_evpc_rev_global): New function.
17678 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
17679 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
17681 (aarch64_evpc_sve_tbl): New function.
17682 (aarch64_expand_vec_perm_const_1): Update after rename of
17683 aarch64_evpc_rev. Handle SVE permutes too, trying
17684 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
17685 than aarch64_evpc_tbl.
17686 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
17687 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
17688 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
17689 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
17690 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
17691 (aarch64_expand_sve_vcond): New functions.
17692 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
17693 of aarch64_vector_mode_p.
17694 (aarch64_dwarf_poly_indeterminate_value): New function.
17695 (aarch64_compute_pressure_classes): Likewise.
17696 (aarch64_can_change_mode_class): Likewise.
17697 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
17698 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
17699 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
17700 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
17701 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
17702 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
17703 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
17704 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
17706 (Dn, Dl, Dr): Accept const as well as const_vector.
17707 (Dz): Likewise. Compare against CONST0_RTX.
17708 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
17709 of "vector" where appropriate.
17710 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
17711 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
17712 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
17713 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
17714 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
17715 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
17716 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
17717 (v_int_equiv): Extend to SVE modes.
17718 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
17720 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
17721 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
17722 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
17723 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
17724 (SVE_COND_FP_CMP): New int iterators.
17725 (perm_hilo): Handle the new unpack unspecs.
17726 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
17728 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
17729 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
17730 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
17731 (aarch64_equality_operator, aarch64_constant_vector_operand)
17732 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
17733 (aarch64_sve_nonimmediate_operand): Likewise.
17734 (aarch64_sve_general_operand): Likewise.
17735 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
17736 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
17737 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
17738 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
17739 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
17740 (aarch64_sve_float_arith_immediate): Likewise.
17741 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
17742 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
17743 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
17744 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
17745 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
17746 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
17747 (aarch64_sve_float_arith_operand): Likewise.
17748 (aarch64_sve_float_arith_with_sub_operand): Likewise.
17749 (aarch64_sve_float_mul_operand): Likewise.
17750 (aarch64_sve_vec_perm_operand): Likewise.
17751 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
17752 (aarch64_mov_operand): Accept const_poly_int and const_vector.
17753 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
17754 as well as const_vector.
17755 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
17756 in file. Use CONST0_RTX and CONSTM1_RTX.
17757 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
17758 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
17759 Use aarch64_simd_imm_zero.
17760 * config/aarch64/aarch64-sve.md: New file.
17761 * config/aarch64/aarch64.md: Include it.
17762 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
17763 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
17764 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
17765 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
17766 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
17767 (sve): New attribute.
17768 (enabled): Disable instructions with the sve attribute unless
17770 (movqi, movhi): Pass CONST_POLY_INT operaneds through
17771 aarch64_expand_mov_immediate.
17772 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
17773 CNT[BHSD] immediates.
17774 (movti): Split CONST_POLY_INT moves into two halves.
17775 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
17776 Split additions that need a temporary here if the destination
17777 is the stack pointer.
17778 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
17779 (*add<mode>3_poly_1): New instruction.
17780 (set_clobber_cc): New expander.
17782 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17784 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
17785 parameter and use it instead of GET_MODE_SIZE (innermode). Use
17786 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
17787 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
17788 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
17789 Change innermode from fixed_mode_size to machine_mode.
17790 (simplify_subreg): Update call accordingly. Handle a constant-sized
17791 subreg of a variable-length CONST_VECTOR.
17793 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17794 Alan Hayward <alan.hayward@arm.com>
17795 David Sherwood <david.sherwood@arm.com>
17797 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
17798 (add_offset_to_base): New function, split out from...
17799 (create_mem_ref): ...here. When handling a scale other than 1,
17800 check first whether the address is valid without the offset.
17801 Add it into the base if so, leaving the index and scale as-is.
17803 2018-01-12 Jakub Jelinek <jakub@redhat.com>
17806 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
17807 fold_for_warn before checking if arg2 is INTEGER_CST.
17809 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
17811 * config/rs6000/predicates.md (load_multiple_operation): Delete.
17812 (store_multiple_operation): Delete.
17813 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
17814 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
17815 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
17816 guarded by TARGET_STRING.
17817 (rs6000_output_load_multiple): Delete.
17818 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
17819 OPTION_MASK_STRING / TARGET_STRING handling.
17820 (print_operand) <'N', 'O'>: Add comment that these are unused now.
17821 (const rs6000_opt_masks) <"string">: Change mask to 0.
17822 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
17823 (MASK_STRING): Delete.
17824 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
17826 (load_multiple): Delete.
17833 (store_multiple): Delete.
17840 (movmemsi_8reg): Delete.
17841 (corresponding unnamed define_insn): Delete.
17842 (movmemsi_6reg): Delete.
17843 (corresponding unnamed define_insn): Delete.
17844 (movmemsi_4reg): Delete.
17845 (corresponding unnamed define_insn): Delete.
17846 (movmemsi_2reg): Delete.
17847 (corresponding unnamed define_insn): Delete.
17848 (movmemsi_1reg): Delete.
17849 (corresponding unnamed define_insn): Delete.
17850 * config/rs6000/rs6000.opt (mno-string): New.
17851 (mstring): Replace by deprecation warning stub.
17852 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
17854 2018-01-12 Jakub Jelinek <jakub@redhat.com>
17856 * regrename.c (regrename_do_replace): If replacing the same
17857 reg multiple times, try to reuse last created gen_raw_REG.
17860 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
17861 main to workaround a bug in GDB.
17863 2018-01-12 Tom de Vries <tom@codesourcery.com>
17866 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
17868 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
17870 PR rtl-optimization/80481
17871 * ira-color.c (get_cap_member): New function.
17872 (allocnos_conflict_by_live_ranges_p): Use it.
17873 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
17874 (setup_slot_coalesced_allocno_live_ranges): Ditto.
17876 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
17879 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
17880 (*saddl_se_1): Ditto.
17881 (*ssubsi_1): Ditto.
17882 (*ssubl_se_1): Ditto.
17884 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17886 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
17887 rather than wi::to_widest for DR_INITs.
17888 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
17889 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
17890 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
17892 (vect_analyze_group_access_1): Note that here.
17894 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17896 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
17897 polynomial type sizes.
17899 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17901 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
17902 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
17903 (gimple_add_tmp_var): Likewise.
17905 2018-01-12 Martin Liska <mliska@suse.cz>
17907 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
17908 (gimple_alloc_sizes): Likewise.
17909 (dump_gimple_statistics): Use PRIu64 in printf format.
17910 * gimple.h: Change uint64_t to int.
17912 2018-01-12 Martin Liska <mliska@suse.cz>
17914 * tree-core.h: Use uint64_t instead of int.
17915 * tree.c (tree_node_counts): Likewise.
17916 (tree_node_sizes): Likewise.
17917 (dump_tree_statistics): Use PRIu64 in printf format.
17919 2018-01-12 Martin Liska <mliska@suse.cz>
17921 * Makefile.in: As qsort_chk is implemented in vec.c, add
17922 vec.o to linkage of gencfn-macros.
17923 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
17924 passing the info to record_node_allocation_statistics.
17925 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
17927 * ggc-common.c (struct ggc_usage): Add operator== and use
17928 it in operator< and compare function.
17929 * mem-stats.h (struct mem_usage): Likewise.
17930 * vec.c (struct vec_usage): Remove operator< and compare
17931 function. Can be simply inherited.
17933 2018-01-12 Martin Jambor <mjambor@suse.cz>
17936 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
17937 * tree-ssa-math-opts.c: Include domwalk.h.
17938 (convert_mult_to_fma_1): New function.
17939 (fma_transformation_info): New type.
17940 (fma_deferring_state): Likewise.
17941 (cancel_fma_deferring): New function.
17942 (result_of_phi): Likewise.
17943 (last_fma_candidate_feeds_initial_phi): Likewise.
17944 (convert_mult_to_fma): Added deferring logic, split actual
17945 transformation to convert_mult_to_fma_1.
17946 (math_opts_dom_walker): New type.
17947 (math_opts_dom_walker::after_dom_children): New method, body moved
17948 here from pass_optimize_widening_mul::execute, added deferring logic
17950 (pass_optimize_widening_mul::execute): Moved most of code to
17951 math_opts_dom_walker::after_dom_children.
17952 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
17953 * config/i386/i386.c (ix86_option_override_internal): Added
17954 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
17956 2018-01-12 Richard Biener <rguenther@suse.de>
17959 * dwarf2out.c (gen_variable_die): Do not reset old_die for
17960 inline instance vars.
17962 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
17965 * config/rx/rx.c (rx_is_restricted_memory_address):
17966 Handle SUBREG case.
17968 2018-01-12 Richard Biener <rguenther@suse.de>
17970 PR tree-optimization/80846
17971 * target.def (split_reduction): New target hook.
17972 * targhooks.c (default_split_reduction): New function.
17973 * targhooks.h (default_split_reduction): Declare.
17974 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
17975 target requests first reduce vectors by combining low and high
17977 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
17978 (get_vectype_for_scalar_type_and_size): Export.
17979 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
17980 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
17981 * doc/tm.texi: Regenerate.
17982 * config/i386/i386.c (ix86_split_reduction): Implement
17983 TARGET_VECTORIZE_SPLIT_REDUCTION.
17985 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17988 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
17989 in PIC mode except for TARGET_VXWORKS_RTP.
17990 * config/sparc/sparc.c: Include cfgrtl.h.
17991 (TARGET_INIT_PIC_REG): Define.
17992 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
17993 (sparc_pic_register_p): New predicate.
17994 (sparc_legitimate_address_p): Use it.
17995 (sparc_legitimize_pic_address): Likewise.
17996 (sparc_delegitimize_address): Likewise.
17997 (sparc_mode_dependent_address_p): Likewise.
17998 (gen_load_pcrel_sym): Remove 4th parameter.
17999 (load_got_register): Adjust call to above. Remove obsolete stuff.
18000 (sparc_expand_prologue): Do not call load_got_register here.
18001 (sparc_flat_expand_prologue): Likewise.
18002 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
18003 (sparc_use_pseudo_pic_reg): New function.
18004 (sparc_init_pic_reg): Likewise.
18005 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
18006 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
18008 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
18010 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
18011 Add item for branch_cost.
18013 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
18015 PR rtl-optimization/83565
18016 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
18017 not extend the result to a larger mode for rotate operations.
18018 (num_sign_bit_copies1): Likewise.
18020 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18023 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
18025 Use values-Xc.o for -pedantic.
18026 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
18028 2018-01-12 Martin Liska <mliska@suse.cz>
18031 * ipa-devirt.c (final_warning_record::grow_type_warnings):
18033 (possible_polymorphic_call_targets): Use it.
18034 (ipa_devirt): Likewise.
18036 2018-01-12 Martin Liska <mliska@suse.cz>
18038 * profile-count.h (enum profile_quality): Use 0 as invalid
18039 enum value of profile_quality.
18041 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
18043 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
18044 -mext-string options.
18046 2018-01-12 Richard Biener <rguenther@suse.de>
18048 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
18049 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
18050 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
18052 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
18054 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
18056 * configure.ac (--with-long-double-format): Add support for the
18057 configuration option to change the default long double format on
18059 * config.gcc (powerpc*-linux*-*): Likewise.
18060 * configure: Regenerate.
18061 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
18062 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
18063 used without modification.
18065 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18067 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
18068 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
18069 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
18070 MISC_BUILTIN_SPEC_BARRIER.
18071 (rs6000_init_builtins): Likewise.
18072 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
18074 (speculation_barrier): New define_insn.
18075 * doc/extend.texi: Document __builtin_speculation_barrier.
18077 2018-01-11 Jakub Jelinek <jakub@redhat.com>
18080 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
18081 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
18082 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
18084 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
18085 integral modes instead of "ss" and "sd".
18086 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
18087 vectors with 32-bit and 64-bit elements.
18088 (vecdupssescalarmodesuffix): New mode attribute.
18089 (vec_dup<mode>): Use it.
18091 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
18094 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
18095 frame if argument is passed on stack.
18097 2018-01-11 Jakub Jelinek <jakub@redhat.com>
18100 * ree.c (combine_reaching_defs): Optimize also
18101 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
18102 reg2=any_extend(exp); reg1=reg2;, formatting fix.
18104 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
18106 PR middle-end/83189
18107 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
18109 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
18111 PR middle-end/83718
18112 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
18113 after they are computed.
18115 2018-01-11 Bin Cheng <bin.cheng@arm.com>
18117 PR tree-optimization/83695
18118 * gimple-loop-linterchange.cc
18119 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
18120 reset cached scev information after interchange.
18121 (pass_linterchange::execute): Remove call to scev_reset_htab.
18123 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18125 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
18126 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
18127 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
18128 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
18129 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
18130 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
18131 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
18132 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
18133 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
18134 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
18135 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
18136 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
18137 (V_lane_reg): Likewise.
18138 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
18140 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
18141 (vfmal_lane_low<mode>_intrinsic,
18142 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
18143 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
18144 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
18145 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
18146 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
18147 vfmsl_lane_high<mode>_intrinsic): New define_insns.
18149 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18151 * config/arm/arm-cpus.in (fp16fml): New feature.
18152 (ALL_SIMD): Add fp16fml.
18153 (armv8.2-a): Add fp16fml as an option.
18154 (armv8.3-a): Likewise.
18155 (armv8.4-a): Add fp16fml as part of fp16.
18156 * config/arm/arm.h (TARGET_FP16FML): Define.
18157 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
18159 * config/arm/arm-modes.def (V2HF): Define.
18160 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
18161 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
18162 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
18163 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
18164 vfmsl_low, vfmsl_high): New set of builtins.
18165 * config/arm/iterators.md (PLUSMINUS): New code iterator.
18166 (vfml_op): New code attribute.
18167 (VFMLHALVES): New int iterator.
18168 (VFML, VFMLSEL): New mode attributes.
18169 (V_reg): Define mapping for V2HF.
18170 (V_hi, V_lo): New mode attributes.
18171 (VF_constraint): Likewise.
18172 (vfml_half, vfml_half_selector): New int attributes.
18173 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
18175 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
18176 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
18178 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
18179 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
18180 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
18181 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
18183 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
18184 Document new effective target and option set.
18186 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18188 * config/arm/arm-cpus.in (armv8_4): New feature.
18189 (ARMv8_4a): New fgroup.
18190 (armv8.4-a): New arch.
18191 * config/arm/arm-tables.opt: Regenerate.
18192 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
18193 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
18194 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
18195 Add matching rules for -march=armv8.4-a and extensions.
18196 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
18198 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
18201 * config/rx/rx.md (BW): New mode attribute.
18202 (sync_lock_test_and_setsi): Add mode suffix to insn output.
18204 2018-01-11 Richard Biener <rguenther@suse.de>
18206 PR tree-optimization/83435
18207 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
18208 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
18209 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
18211 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18212 Alan Hayward <alan.hayward@arm.com>
18213 David Sherwood <david.sherwood@arm.com>
18215 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
18217 (aarch64_classify_address): Initialize it. Track polynomial offsets.
18218 (aarch64_print_address_internal): Use it to check for a zero offset.
18220 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18221 Alan Hayward <alan.hayward@arm.com>
18222 David Sherwood <david.sherwood@arm.com>
18224 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
18225 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
18226 Return a poly_int64 rather than a HOST_WIDE_INT.
18227 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
18228 rather than a HOST_WIDE_INT.
18229 * config/aarch64/aarch64.h (aarch64_frame): Protect with
18230 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
18231 hard_fp_offset, frame_size, initial_adjust, callee_offset and
18232 final_offset from HOST_WIDE_INT to poly_int64.
18233 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
18234 to_constant when getting the number of units in an Advanced SIMD
18236 (aarch64_builtin_vectorized_function): Check for a constant number
18238 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
18240 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
18241 attribute instead of GET_MODE_NUNITS.
18242 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
18243 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
18244 GET_MODE_SIZE for fixed-size registers.
18245 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
18246 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
18247 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
18248 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
18249 (aarch64_print_operand, aarch64_print_address_internal)
18250 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
18251 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
18252 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
18253 Handle polynomial GET_MODE_SIZE.
18254 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
18255 wider than SImode without modification.
18256 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
18257 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
18258 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
18259 passing and returning SVE modes.
18260 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
18261 rather than GEN_INT.
18262 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
18263 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
18264 (aarch64_allocate_and_probe_stack_space): Likewise.
18265 (aarch64_layout_frame): Cope with polynomial offsets.
18266 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
18267 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
18268 polynomial offsets.
18269 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
18270 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
18271 poly_int64 rather than a HOST_WIDE_INT.
18272 (aarch64_get_separate_components, aarch64_process_components)
18273 (aarch64_expand_prologue, aarch64_expand_epilogue)
18274 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
18275 (aarch64_anchor_offset): New function, split out from...
18276 (aarch64_legitimize_address): ...here.
18277 (aarch64_builtin_vectorization_cost): Handle polynomial
18278 TYPE_VECTOR_SUBPARTS.
18279 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
18281 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
18282 number of elements from the PARALLEL rather than the mode.
18283 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
18284 rather than GET_MODE_BITSIZE.
18285 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
18286 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
18287 (aarch64_expand_vec_perm_const_1): Handle polynomial
18288 d->perm.length () and d->perm elements.
18289 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
18290 Apply to_constant to d->perm elements.
18291 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
18292 polynomial CONST_VECTOR_NUNITS.
18293 (aarch64_move_pointer): Take amount as a poly_int64 rather
18295 (aarch64_progress_pointer): Avoid temporary variable.
18296 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
18297 the mode attribute instead of GET_MODE.
18299 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18300 Alan Hayward <alan.hayward@arm.com>
18301 David Sherwood <david.sherwood@arm.com>
18303 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
18304 x exists before using it.
18305 (aarch64_add_constant_internal): Rename to...
18306 (aarch64_add_offset_1): ...this. Replace regnum with separate
18307 src and dest rtxes. Handle the case in which they're different,
18308 including when the offset is zero. Replace scratchreg with an rtx.
18309 Use 2 additions if there is no spare register into which we can
18310 move a 16-bit constant.
18311 (aarch64_add_constant): Delete.
18312 (aarch64_add_offset): Replace reg with separate src and dest
18313 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
18314 Use aarch64_add_offset_1.
18315 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
18316 an rtx rather than an int. Take the delta as a poly_int64
18317 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
18318 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
18319 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
18320 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
18321 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
18322 and aarch64_add_sp.
18323 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
18324 aarch64_add_constant.
18326 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18328 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
18329 Use scalar_float_mode.
18331 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18333 * config/aarch64/aarch64-simd.md
18334 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
18335 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
18336 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
18337 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
18338 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
18339 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
18340 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
18341 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
18342 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
18343 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
18345 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18348 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
18349 targ_options->x_arm_arch_string is non NULL.
18351 2018-01-11 Tamar Christina <tamar.christina@arm.com>
18353 * config/aarch64/aarch64.h
18354 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
18356 2018-01-11 Sudakshina Das <sudi.das@arm.com>
18359 * expmed.c (emit_store_flag_force): Swap if const op0
18360 and change VOIDmode to mode of op0.
18362 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18364 PR rtl-optimization/83761
18365 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
18366 than bytes to mode_for_size.
18368 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18370 PR middle-end/83189
18371 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
18372 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
18375 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18377 PR middle-end/83575
18378 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
18379 when in layout mode.
18380 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
18381 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
18384 2018-01-10 Michael Collison <michael.collison@arm.com>
18386 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
18387 * config/aarch64/aarch64-option-extension.def: Add
18388 AARCH64_OPT_EXTENSION of 'fp16fml'.
18389 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18390 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
18391 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
18392 * config/aarch64/constraints.md (Ui7): New constraint.
18393 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
18394 (VFMLA_SEL_W): Ditto.
18397 (VFMLA16_LOW): New int iterator.
18398 (VFMLA16_HIGH): Ditto.
18399 (UNSPEC_FMLAL): New unspec.
18400 (UNSPEC_FMLSL): Ditto.
18401 (UNSPEC_FMLAL2): Ditto.
18402 (UNSPEC_FMLSL2): Ditto.
18403 (f16mac): New code attribute.
18404 * config/aarch64/aarch64-simd-builtins.def
18405 (aarch64_fmlal_lowv2sf): Ditto.
18406 (aarch64_fmlsl_lowv2sf): Ditto.
18407 (aarch64_fmlalq_lowv4sf): Ditto.
18408 (aarch64_fmlslq_lowv4sf): Ditto.
18409 (aarch64_fmlal_highv2sf): Ditto.
18410 (aarch64_fmlsl_highv2sf): Ditto.
18411 (aarch64_fmlalq_highv4sf): Ditto.
18412 (aarch64_fmlslq_highv4sf): Ditto.
18413 (aarch64_fmlal_lane_lowv2sf): Ditto.
18414 (aarch64_fmlsl_lane_lowv2sf): Ditto.
18415 (aarch64_fmlal_laneq_lowv2sf): Ditto.
18416 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
18417 (aarch64_fmlalq_lane_lowv4sf): Ditto.
18418 (aarch64_fmlsl_lane_lowv4sf): Ditto.
18419 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
18420 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
18421 (aarch64_fmlal_lane_highv2sf): Ditto.
18422 (aarch64_fmlsl_lane_highv2sf): Ditto.
18423 (aarch64_fmlal_laneq_highv2sf): Ditto.
18424 (aarch64_fmlsl_laneq_highv2sf): Ditto.
18425 (aarch64_fmlalq_lane_highv4sf): Ditto.
18426 (aarch64_fmlsl_lane_highv4sf): Ditto.
18427 (aarch64_fmlalq_laneq_highv4sf): Ditto.
18428 (aarch64_fmlsl_laneq_highv4sf): Ditto.
18429 * config/aarch64/aarch64-simd.md:
18430 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
18431 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
18432 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
18433 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
18434 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
18435 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
18436 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
18437 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
18438 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
18439 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
18440 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
18441 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
18442 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
18443 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
18444 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
18445 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
18446 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
18447 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
18448 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
18449 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
18450 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
18451 (vfmlsl_low_u32): Ditto.
18452 (vfmlalq_low_u32): Ditto.
18453 (vfmlslq_low_u32): Ditto.
18454 (vfmlal_high_u32): Ditto.
18455 (vfmlsl_high_u32): Ditto.
18456 (vfmlalq_high_u32): Ditto.
18457 (vfmlslq_high_u32): Ditto.
18458 (vfmlal_lane_low_u32): Ditto.
18459 (vfmlsl_lane_low_u32): Ditto.
18460 (vfmlal_laneq_low_u32): Ditto.
18461 (vfmlsl_laneq_low_u32): Ditto.
18462 (vfmlalq_lane_low_u32): Ditto.
18463 (vfmlslq_lane_low_u32): Ditto.
18464 (vfmlalq_laneq_low_u32): Ditto.
18465 (vfmlslq_laneq_low_u32): Ditto.
18466 (vfmlal_lane_high_u32): Ditto.
18467 (vfmlsl_lane_high_u32): Ditto.
18468 (vfmlal_laneq_high_u32): Ditto.
18469 (vfmlsl_laneq_high_u32): Ditto.
18470 (vfmlalq_lane_high_u32): Ditto.
18471 (vfmlslq_lane_high_u32): Ditto.
18472 (vfmlalq_laneq_high_u32): Ditto.
18473 (vfmlslq_laneq_high_u32): Ditto.
18474 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
18475 (AARCH64_FL_FOR_ARCH8_4): New.
18476 (AARCH64_ISA_F16FML): New ISA flag.
18477 (TARGET_F16FML): New feature flag for fp16fml.
18478 (doc/invoke.texi): Document new fp16fml option.
18480 2018-01-10 Michael Collison <michael.collison@arm.com>
18482 * config/aarch64/aarch64-builtins.c:
18483 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
18484 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18485 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
18486 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
18487 (AARCH64_ISA_SHA3): New ISA flag.
18488 (TARGET_SHA3): New feature flag for sha3.
18489 * config/aarch64/iterators.md (sha512_op): New int attribute.
18490 (CRYPTO_SHA512): New int iterator.
18491 (UNSPEC_SHA512H): New unspec.
18492 (UNSPEC_SHA512H2): Ditto.
18493 (UNSPEC_SHA512SU0): Ditto.
18494 (UNSPEC_SHA512SU1): Ditto.
18495 * config/aarch64/aarch64-simd-builtins.def
18496 (aarch64_crypto_sha512hqv2di): New builtin.
18497 (aarch64_crypto_sha512h2qv2di): Ditto.
18498 (aarch64_crypto_sha512su0qv2di): Ditto.
18499 (aarch64_crypto_sha512su1qv2di): Ditto.
18500 (aarch64_eor3qv8hi): Ditto.
18501 (aarch64_rax1qv2di): Ditto.
18502 (aarch64_xarqv2di): Ditto.
18503 (aarch64_bcaxqv8hi): Ditto.
18504 * config/aarch64/aarch64-simd.md:
18505 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
18506 (aarch64_crypto_sha512su0qv2di): Ditto.
18507 (aarch64_crypto_sha512su1qv2di): Ditto.
18508 (aarch64_eor3qv8hi): Ditto.
18509 (aarch64_rax1qv2di): Ditto.
18510 (aarch64_xarqv2di): Ditto.
18511 (aarch64_bcaxqv8hi): Ditto.
18512 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
18513 (vsha512h2q_u64): Ditto.
18514 (vsha512su0q_u64): Ditto.
18515 (vsha512su1q_u64): Ditto.
18516 (veor3q_u16): Ditto.
18517 (vrax1q_u64): Ditto.
18518 (vxarq_u64): Ditto.
18519 (vbcaxq_u16): Ditto.
18520 * config/arm/types.md (crypto_sha512): New type attribute.
18521 (crypto_sha3): Ditto.
18522 (doc/invoke.texi): Document new sha3 option.
18524 2018-01-10 Michael Collison <michael.collison@arm.com>
18526 * config/aarch64/aarch64-builtins.c:
18527 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
18528 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18529 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
18530 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
18531 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
18532 (AARCH64_ISA_SM4): New ISA flag.
18533 (TARGET_SM4): New feature flag for sm4.
18534 * config/aarch64/aarch64-simd-builtins.def
18535 (aarch64_sm3ss1qv4si): Ditto.
18536 (aarch64_sm3tt1aq4si): Ditto.
18537 (aarch64_sm3tt1bq4si): Ditto.
18538 (aarch64_sm3tt2aq4si): Ditto.
18539 (aarch64_sm3tt2bq4si): Ditto.
18540 (aarch64_sm3partw1qv4si): Ditto.
18541 (aarch64_sm3partw2qv4si): Ditto.
18542 (aarch64_sm4eqv4si): Ditto.
18543 (aarch64_sm4ekeyqv4si): Ditto.
18544 * config/aarch64/aarch64-simd.md:
18545 (aarch64_sm3ss1qv4si): Ditto.
18546 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
18547 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
18548 (aarch64_sm4eqv4si): Ditto.
18549 (aarch64_sm4ekeyqv4si): Ditto.
18550 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
18551 (sm3part_op): Ditto.
18552 (CRYPTO_SM3TT): Ditto.
18553 (CRYPTO_SM3PART): Ditto.
18554 (UNSPEC_SM3SS1): New unspec.
18555 (UNSPEC_SM3TT1A): Ditto.
18556 (UNSPEC_SM3TT1B): Ditto.
18557 (UNSPEC_SM3TT2A): Ditto.
18558 (UNSPEC_SM3TT2B): Ditto.
18559 (UNSPEC_SM3PARTW1): Ditto.
18560 (UNSPEC_SM3PARTW2): Ditto.
18561 (UNSPEC_SM4E): Ditto.
18562 (UNSPEC_SM4EKEY): Ditto.
18563 * config/aarch64/constraints.md (Ui2): New constraint.
18564 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
18565 * config/arm/types.md (crypto_sm3): New type attribute.
18566 (crypto_sm4): Ditto.
18567 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
18568 (vsm3tt1aq_u32): Ditto.
18569 (vsm3tt1bq_u32): Ditto.
18570 (vsm3tt2aq_u32): Ditto.
18571 (vsm3tt2bq_u32): Ditto.
18572 (vsm3partw1q_u32): Ditto.
18573 (vsm3partw2q_u32): Ditto.
18574 (vsm4eq_u32): Ditto.
18575 (vsm4ekeyq_u32): Ditto.
18576 (doc/invoke.texi): Document new sm4 option.
18578 2018-01-10 Michael Collison <michael.collison@arm.com>
18580 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
18581 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
18582 (AARCH64_FL_FOR_ARCH8_4): New.
18583 (AARCH64_FL_V8_4): New flag.
18584 (doc/invoke.texi): Document new armv8.4-a option.
18586 2018-01-10 Michael Collison <michael.collison@arm.com>
18588 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18589 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
18590 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
18591 * config/aarch64/aarch64-option-extension.def: Add
18592 AARCH64_OPT_EXTENSION of 'sha2'.
18593 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
18594 (crypto): Disable sha2 and aes if crypto disabled.
18595 (crypto): Enable aes and sha2 if enabled.
18596 (simd): Disable sha2 and aes if simd disabled.
18597 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
18599 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
18600 (TARGET_SHA2): New feature flag for sha2.
18601 (TARGET_AES): New feature flag for aes.
18602 * config/aarch64/aarch64-simd.md:
18603 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
18604 conditional on TARGET_AES.
18605 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
18606 (aarch64_crypto_sha1hsi): Make pattern conditional
18608 (aarch64_crypto_sha1hv4si): Ditto.
18609 (aarch64_be_crypto_sha1hv4si): Ditto.
18610 (aarch64_crypto_sha1su1v4si): Ditto.
18611 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
18612 (aarch64_crypto_sha1su0v4si): Ditto.
18613 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
18614 (aarch64_crypto_sha256su0v4si): Ditto.
18615 (aarch64_crypto_sha256su1v4si): Ditto.
18616 (doc/invoke.texi): Document new aes and sha2 options.
18618 2018-01-10 Martin Sebor <msebor@redhat.com>
18620 PR tree-optimization/83781
18621 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
18624 2018-01-11 Martin Sebor <msebor@gmail.com>
18625 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18627 PR tree-optimization/83501
18628 PR tree-optimization/81703
18630 * tree-ssa-strlen.c (get_string_cst): Rename...
18631 (get_string_len): ...to this. Handle global constants.
18632 (handle_char_store): Adjust.
18634 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
18635 Jim Wilson <jimw@sifive.com>
18637 * config/riscv/riscv-protos.h (riscv_output_return): New.
18638 * config/riscv/riscv.c (struct machine_function): New naked_p field.
18639 (riscv_attribute_table, riscv_output_return),
18640 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
18641 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
18642 (riscv_compute_frame_info): Only compute frame->mask if not a naked
18644 (riscv_expand_prologue): Add early return for naked function.
18645 (riscv_expand_epilogue): Likewise.
18646 (riscv_function_ok_for_sibcall): Return false for naked function.
18647 (riscv_set_current_function): New.
18648 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
18649 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
18650 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
18651 * doc/extend.texi (RISC-V Function Attributes): New.
18653 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
18655 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
18656 check for 128-bit long double before checking TCmode.
18657 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
18658 128-bit long doubles before checking TFmode or TCmode.
18659 (FLOAT128_IBM_P): Likewise.
18661 2018-01-10 Martin Sebor <msebor@redhat.com>
18663 PR tree-optimization/83671
18664 * builtins.c (c_strlen): Unconditionally return zero for the empty
18666 Use -Warray-bounds for warnings.
18667 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
18668 for non-constant array indices with COMPONENT_REF, arrays of
18669 arrays, and pointers to arrays.
18670 (gimple_fold_builtin_strlen): Determine and set length range for
18671 non-constant character arrays.
18673 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
18675 PR middle-end/81897
18676 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
18679 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
18681 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
18683 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
18686 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
18687 VECTOR_MEM_ALTIVEC_OR_VSX_P.
18688 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
18689 indexed_or_indirect_operand predicate.
18690 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
18691 (*vsx_le_perm_load_v8hi): Likewise.
18692 (*vsx_le_perm_load_v16qi): Likewise.
18693 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
18694 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
18695 (*vsx_le_perm_store_v8hi): Likewise.
18696 (*vsx_le_perm_store_v16qi): Likewise.
18697 (eight unnamed splitters): Likewise.
18699 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
18701 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
18702 * config/rs6000/emmintrin.h: Likewise.
18703 * config/rs6000/mmintrin.h: Likewise.
18704 * config/rs6000/xmmintrin.h: Likewise.
18706 2018-01-10 David Malcolm <dmalcolm@redhat.com>
18709 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
18711 * tree.c (tree_nop_conversion): Return true for location wrapper
18713 (maybe_wrap_with_location): New function.
18714 (selftest::check_strip_nops): New function.
18715 (selftest::test_location_wrappers): New function.
18716 (selftest::tree_c_tests): Call it.
18717 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
18718 (maybe_wrap_with_location): New decl.
18719 (EXPR_LOCATION_WRAPPER_P): New macro.
18720 (location_wrapper_p): New inline function.
18721 (tree_strip_any_location_wrapper): New inline function.
18723 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
18726 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
18727 stack_realign_offset for the largest alignment of stack slot
18729 (ix86_find_max_used_stack_alignment): New function.
18730 (ix86_finalize_stack_frame_flags): Use it. Set
18731 max_used_stack_alignment if we don't realign stack.
18732 * config/i386/i386.h (machine_function): Add
18733 max_used_stack_alignment.
18735 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
18737 * config/arm/arm.opt (-mbranch-cost): New option.
18738 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
18741 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
18744 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
18745 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
18747 2018-01-10 Richard Biener <rguenther@suse.de>
18750 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
18751 early out so it also covers the case where we have a non-NULL
18754 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
18756 PR tree-optimization/83753
18757 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
18758 for non-strided grouped accesses if the number of elements is 1.
18760 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18763 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
18764 * i386.h (TARGET_USE_GATHER): Define.
18765 * x86-tune.def (X86_TUNE_USE_GATHER): New.
18767 2018-01-10 Martin Liska <mliska@suse.cz>
18770 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
18771 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
18773 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
18774 CLEANUP_NO_PARTITIONING is not set.
18776 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
18778 * doc/rtl.texi: Remove documentation of (const ...) wrappers
18779 for vectors, as a partial revert of r254296.
18780 * rtl.h (const_vec_p): Delete.
18781 (const_vec_duplicate_p): Don't test for vector CONSTs.
18782 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
18783 * expmed.c (make_tree): Likewise.
18786 * common.md (E, F): Use CONSTANT_P instead of checking for
18788 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
18789 checking for CONST_VECTOR.
18791 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18793 PR middle-end/83575
18794 * predict.c (force_edge_cold): Handle in more sane way edges
18795 with no prediction.
18797 2018-01-09 Carl Love <cel@us.ibm.com>
18799 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
18801 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
18802 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
18803 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
18804 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
18805 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
18806 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
18807 * config/rs6000/rs6000-protos.h: Add extern defition for
18808 rs6000_generate_float2_double_code.
18809 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
18811 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
18812 (float2_v2df): Add define_expand.
18814 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
18817 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
18818 op_mode in the force_to_mode call.
18820 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18822 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
18823 instead of checking each element individually.
18824 (aarch64_evpc_uzp): Likewise.
18825 (aarch64_evpc_zip): Likewise.
18826 (aarch64_evpc_ext): Likewise.
18827 (aarch64_evpc_rev): Likewise.
18828 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
18829 instead of checking each element individually. Return true without
18831 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
18832 whether all selected elements come from the same input, instead of
18833 checking each element individually. Remove calls to gen_rtx_REG,
18834 start_sequence and end_sequence and instead assert that no rtl is
18837 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18839 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
18840 order of HIGH and CONST checks.
18842 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18844 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
18845 if the destination isn't an SSA_NAME.
18847 2018-01-09 Richard Biener <rguenther@suse.de>
18849 PR tree-optimization/83668
18850 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
18852 (canonicalize_loop_form): ... here, renamed from ...
18853 (canonicalize_loop_closed_ssa_form): ... this and amended to
18854 swap successor edges for loop exit blocks to make us use
18855 the RPO order we need for initial schedule generation.
18857 2018-01-09 Joseph Myers <joseph@codesourcery.com>
18859 PR tree-optimization/64811
18860 * match.pd: When optimizing comparisons with Inf, avoid
18861 introducing or losing exceptions from comparisons with NaN.
18863 2018-01-09 Martin Liska <mliska@suse.cz>
18866 * asan.c (shadow_mem_size): Add gcc_assert.
18868 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
18870 Don't save registers in main().
18873 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
18874 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
18875 * config/avr/avr.c (avr_set_current_function): Don't error if
18876 naked, OS_task or OS_main are specified at the same time.
18877 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
18879 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
18881 * common/config/avr/avr-common.c (avr_option_optimization_table):
18882 Switch on -mmain-is-OS_task for optimizing compilations.
18884 2018-01-09 Richard Biener <rguenther@suse.de>
18886 PR tree-optimization/83572
18887 * graphite.c: Include cfganal.h.
18888 (graphite_transform_loops): Connect infinite loops to exit
18889 and remove fake edges at the end.
18891 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18893 * ipa-inline.c (edge_badness): Revert accidental checkin.
18895 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18898 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
18899 symbols; not inline clones.
18901 2018-01-09 Jakub Jelinek <jakub@redhat.com>
18904 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
18905 hard registers. Formatting fixes.
18907 PR preprocessor/83722
18908 * gcc.c (try_generate_repro): Pass
18909 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
18910 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
18913 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
18914 Kito Cheng <kito.cheng@gmail.com>
18916 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
18917 (riscv_leaf_function_p): Delete.
18918 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
18920 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18922 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
18924 (do_ifelse): New function.
18925 (do_isel): New function.
18926 (do_sub3): New function.
18927 (do_add3): New function.
18928 (do_load_mask_compare): New function.
18929 (do_overlap_load_compare): New function.
18930 (expand_compare_loop): New function.
18931 (expand_block_compare): Call expand_compare_loop() when appropriate.
18932 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
18933 option description.
18934 (-mblock-compare-inline-loop-limit): New option.
18936 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18939 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
18940 Reverse order of second and third operands in first alternative.
18941 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
18942 of first and second elements in UNSPEC_VPERMR vector.
18943 (altivec_expand_vec_perm_le): Likewise.
18945 2018-01-08 Jeff Law <law@redhat.com>
18947 PR rtl-optimizatin/81308
18948 * tree-switch-conversion.c (cfg_altered): New file scoped static.
18949 (process_switch): If group_case_labels makes a change, then set
18951 (pass_convert_switch::execute): If a switch is converted, then
18952 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
18954 PR rtl-optimization/81308
18955 * recog.c (split_all_insns): Conditionally cleanup the CFG after
18958 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
18960 PR target/83663 - Revert r255946
18961 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
18962 generation for cases where splatting a value is not useful.
18963 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
18964 across a vec_duplicate and a paradoxical subreg forming a vector
18965 mode to a vec_concat.
18967 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18969 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
18970 -march=armv8.3-a variants.
18971 * config/arm/t-multilib: Likewise.
18972 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
18974 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18976 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
18978 (cceq_ior_compare_complement): Give it a name so I can use it, and
18979 change boolean_or_operator predicate to boolean_operator so it can
18980 be used to generate a crand.
18981 (eqne): New code iterator.
18982 (bd/bd_neg): New code_attrs.
18983 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
18984 a single define_insn.
18985 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
18986 decrement (bdnzt/bdnzf/bdzt/bdzf).
18987 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
18988 with the new names of the branch decrement patterns, and added the
18989 names of the branch decrement conditional patterns.
18991 2018-01-08 Richard Biener <rguenther@suse.de>
18993 PR tree-optimization/83563
18994 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
18997 2018-01-08 Richard Biener <rguenther@suse.de>
18999 PR middle-end/83713
19000 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
19002 2018-01-08 Richard Biener <rguenther@suse.de>
19004 PR tree-optimization/83685
19005 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
19006 references to abnormals.
19008 2018-01-08 Richard Biener <rguenther@suse.de>
19011 * dwarf2out.c (output_indirect_strings): Handle empty
19012 skeleton_debug_str_hash.
19013 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
19015 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
19017 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
19018 (emit_store_direct): Likewise.
19019 (arc_trampoline_adjust_address): Likewise.
19020 (arc_asm_trampoline_template): New function.
19021 (arc_initialize_trampoline): Use asm_trampoline_template.
19022 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
19023 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
19024 * config/arc/arc.md (flush_icache): Delete pattern.
19026 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
19028 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
19029 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
19032 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
19035 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
19036 by not USED_FOR_TARGET.
19037 (make_pass_resolve_sw_modes): Likewise.
19039 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
19041 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
19044 2018-01-08 Richard Biener <rguenther@suse.de>
19046 PR middle-end/83580
19047 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
19049 2018-01-08 Richard Biener <rguenther@suse.de>
19051 PR middle-end/83517
19052 * match.pd ((t * 2) / 2) -> t): Add missing :c.
19054 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
19056 PR middle-end/81897
19057 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
19058 basic blocks with a small number of successors.
19059 (convert_control_dep_chain_into_preds): Improve handling of
19061 (dump_predicates): Split apart into...
19062 (dump_pred_chain): ...here...
19063 (dump_pred_info): ...and here.
19064 (can_one_predicate_be_invalidated_p): Add debugging printfs.
19065 (can_chain_union_be_invalidated_p): Improve check for invalidation
19067 (uninit_uses_cannot_happen): Avoid unnecessary if
19068 convert_control_dep_chain_into_preds yielded nothing.
19070 2018-01-06 Martin Sebor <msebor@redhat.com>
19072 PR tree-optimization/83640
19073 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
19074 subtracting negative offset from size.
19075 (builtin_access::overlap): Adjust offset bounds of the access to fall
19076 within the size of the object if possible.
19078 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
19080 PR rtl-optimization/83699
19081 * expmed.c (extract_bit_field_1): Restrict the vector usage of
19082 extract_bit_field_as_subreg to cases in which the extracted
19083 value is also a vector.
19085 * lra-constraints.c (process_alt_operands): Test for the equivalence
19086 substitutions when detecting a possible reload cycle.
19088 2018-01-06 Jakub Jelinek <jakub@redhat.com>
19091 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
19092 by default if flag_selective_schedling{,2}. Formatting fixes.
19094 PR rtl-optimization/83682
19095 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
19096 if it has non-VECTOR_MODE element mode.
19097 (vec_duplicate_p): Likewise.
19099 PR middle-end/83694
19100 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
19101 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
19103 2018-01-05 Jakub Jelinek <jakub@redhat.com>
19106 * config/i386/i386-builtin.def
19107 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
19108 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
19109 Require also OPTION_MASK_ISA_AVX512F in addition to
19110 OPTION_MASK_ISA_GFNI.
19111 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
19112 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
19113 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
19114 to OPTION_MASK_ISA_GFNI.
19115 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
19116 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
19117 OPTION_MASK_ISA_AVX512BW.
19118 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
19119 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
19120 addition to OPTION_MASK_ISA_GFNI.
19121 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
19122 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
19123 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
19124 to OPTION_MASK_ISA_GFNI.
19125 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
19126 a requirement for all ISAs rather than any of them with a few
19128 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
19130 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
19131 bitmasks to be enabled with 3 exceptions, instead of requiring any
19132 enabled ISA with lots of exceptions.
19133 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
19134 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
19135 Change avx512bw in isa attribute to avx512f.
19136 * config/i386/sgxintrin.h: Add license boilerplate.
19137 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
19138 to __AVX512F__ and __AVX512VL to __AVX512VL__.
19139 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
19140 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
19142 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
19143 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
19144 temporarily sse2 rather than sse if not enabled already.
19147 * config/i386/sse.md (VI248_VLBW): Rename to ...
19148 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
19149 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
19150 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
19151 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
19152 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
19153 mode iterator instead of VI248_VLBW.
19155 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
19157 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
19158 (record_modified): Skip clobbers; add debug output.
19159 (param_change_prob): Use sreal frequencies.
19161 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
19163 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
19164 punt for user-aligned variables.
19166 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
19168 * tree-chrec.c (chrec_contains_symbols): Return true for
19171 2018-01-05 Sudakshina Das <sudi.das@arm.com>
19174 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
19175 of (x|y) == x for BICS pattern.
19177 2018-01-05 Jakub Jelinek <jakub@redhat.com>
19179 PR tree-optimization/83605
19180 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
19181 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
19184 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
19186 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
19187 * config/epiphany/rtems.h: New file.
19189 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19190 Uros Bizjak <ubizjak@gmail.com>
19193 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
19194 QIreg_operand instead of register_operand predicate.
19195 * config/i386/i386.c (ix86_rop_should_change_byte_p,
19196 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
19197 comments instead of -fmitigate[-_]rop.
19199 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19202 * cgraphunit.c (symbol_table::compile): Switch to text_section
19203 before calling assembly_start debug hook.
19204 * run-rtl-passes.c (run_rtl_passes): Likewise.
19207 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19209 * tree-vrp.c (extract_range_from_binary_expr_1): Check
19210 range_int_cst_p rather than !symbolic_range_p before calling
19211 extract_range_from_multiplicative_op_1.
19213 2018-01-04 Jeff Law <law@redhat.com>
19215 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
19216 redundant test in assertion.
19218 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19220 * doc/rtl.texi: Document machine_mode wrapper classes.
19222 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19224 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
19225 using tree_to_uhwi.
19227 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19229 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
19230 the VEC_PERM_EXPR fold to fail.
19232 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19235 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
19236 to switched_sections.
19238 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19241 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
19242 test for d.testing.
19244 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
19247 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
19248 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
19250 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19253 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
19254 is BLKmode and bitpos not zero or mode change is needed.
19256 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19259 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
19262 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
19265 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
19266 instead of MULT rtx. Update all corresponding splitters.
19267 (*saddl_se): Ditto.
19268 (*ssub<modesuffix>): Ditto.
19269 (*ssubl_se): Ditto.
19270 (*cmp_sadd_di): Update split patterns.
19271 (*cmp_sadd_si): Ditto.
19272 (*cmp_sadd_sidi): Ditto.
19273 (*cmp_ssub_di): Ditto.
19274 (*cmp_ssub_si): Ditto.
19275 (*cmp_ssub_sidi): Ditto.
19276 * config/alpha/predicates.md (const23_operand): New predicate.
19277 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
19278 Look for ASHIFT, not MULT inner operand.
19279 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
19281 2018-01-04 Martin Liska <mliska@suse.cz>
19283 PR gcov-profile/83669
19284 * gcov.c (output_intermediate_file): Add version to intermediate
19286 * doc/gcov.texi: Document new field 'version' in intermediate
19287 file format. Fix location of '-k' option of gcov command.
19289 2018-01-04 Martin Liska <mliska@suse.cz>
19292 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
19294 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19296 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
19298 2018-01-03 Martin Sebor <msebor@redhat.com>
19300 PR tree-optimization/83655
19301 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
19302 checking calls with invalid arguments.
19304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19306 * tree-vect-stmts.c (vect_get_store_rhs): New function.
19307 (vectorizable_mask_load_store): Delete.
19308 (vectorizable_call): Return false for masked loads and stores.
19309 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
19310 instead of gimple_assign_rhs1.
19311 (vectorizable_load): Handle IFN_MASK_LOAD.
19312 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
19314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19316 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
19318 (vectorizable_mask_load_store): ...here.
19319 (vectorizable_load): ...and here.
19321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19323 * tree-vect-stmts.c (vect_build_all_ones_mask)
19324 (vect_build_zero_merge_argument): New functions, split out from...
19325 (vectorizable_load): ...here.
19327 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19329 * tree-vect-stmts.c (vect_check_store_rhs): New function,
19331 (vectorizable_mask_load_store): ...here.
19332 (vectorizable_store): ...and here.
19334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19336 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
19338 (vectorizable_mask_load_store): ...here.
19340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19342 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
19343 (vect_model_store_cost): Take a vec_load_store_type instead of a
19345 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
19346 (vect_model_store_cost): Take a vec_load_store_type instead of a
19348 (vectorizable_mask_load_store): Update accordingly.
19349 (vectorizable_store): Likewise.
19350 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
19352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19354 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
19355 IFN_MASK_LOAD calls here rather than...
19356 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
19358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19359 Alan Hayward <alan.hayward@arm.com>
19360 David Sherwood <david.sherwood@arm.com>
19362 * expmed.c (extract_bit_field_1): For vector extracts,
19363 fall back to extract_bit_field_as_subreg if vec_extract
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 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
19371 they are variable or constant sized.
19372 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
19373 slots for constant-sized data.
19375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19376 Alan Hayward <alan.hayward@arm.com>
19377 David Sherwood <david.sherwood@arm.com>
19379 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
19380 handling COND_EXPRs with boolean comparisons, try to find a better
19381 basis for the mask type than the boolean itself.
19383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19385 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
19386 is calculated and how it can be overridden.
19387 * genmodes.c (max_bitsize_mode_any_mode): New variable.
19388 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
19390 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
19393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19394 Alan Hayward <alan.hayward@arm.com>
19395 David Sherwood <david.sherwood@arm.com>
19397 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
19398 Remove the mode argument.
19399 (aarch64_simd_valid_immediate): Remove the mode and inverse
19401 * config/aarch64/iterators.md (bitsize): New iterator.
19402 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
19403 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
19404 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
19405 aarch64_simd_valid_immediate.
19406 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
19407 (aarch64_reg_or_bic_imm): Likewise.
19408 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
19409 with an insn_type enum and msl with a modifier_type enum.
19410 Replace element_width with a scalar_mode. Change the shift
19411 to unsigned int. Add constructors for scalar_float_mode and
19412 scalar_int_mode elements.
19413 (aarch64_vect_float_const_representable_p): Delete.
19414 (aarch64_can_const_movi_rtx_p)
19415 (aarch64_simd_scalar_immediate_valid_for_move)
19416 (aarch64_simd_make_constant): Update call to
19417 aarch64_simd_valid_immediate.
19418 (aarch64_advsimd_valid_immediate_hs): New function.
19419 (aarch64_advsimd_valid_immediate): Likewise.
19420 (aarch64_simd_valid_immediate): Remove mode and inverse
19421 arguments. Rewrite to use the above. Use const_vec_duplicate_p
19422 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
19423 and aarch64_float_const_representable_p on the result.
19424 (aarch64_output_simd_mov_immediate): Remove mode argument.
19425 Update call to aarch64_simd_valid_immediate and use of
19426 simd_immediate_info.
19427 (aarch64_output_scalar_simd_mov_immediate): Update call
19430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19431 Alan Hayward <alan.hayward@arm.com>
19432 David Sherwood <david.sherwood@arm.com>
19434 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
19435 (mode_nunits): Likewise CONST_MODE_NUNITS.
19436 * machmode.def (ADJUST_NUNITS): Document.
19437 * genmodes.c (mode_data::need_nunits_adj): New field.
19438 (blank_mode): Update accordingly.
19439 (adj_nunits): New variable.
19440 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
19442 (emit_mode_size_inline): Set need_bytesize_adj for all modes
19443 listed in adj_nunits.
19444 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
19445 listed in adj_nunits. Don't emit case statements for such modes.
19446 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
19447 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
19448 nothing if adj_nunits is nonnull.
19449 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
19450 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
19451 (emit_mode_fbit): Update use of print_maybe_const_decl.
19452 (emit_move_size): Likewise. Treat the array as non-const
19454 (emit_mode_adjustments): Handle adj_nunits.
19456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19458 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
19459 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
19460 (VECTOR_MODES): Use it.
19461 (make_vector_modes): Take the prefix as an argument.
19463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19464 Alan Hayward <alan.hayward@arm.com>
19465 David Sherwood <david.sherwood@arm.com>
19467 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
19468 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
19469 for MODE_VECTOR_BOOL.
19470 * machmode.def (VECTOR_BOOL_MODE): Document.
19471 * genmodes.c (VECTOR_BOOL_MODE): New macro.
19472 (make_vector_bool_mode): New function.
19473 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
19475 * lto-streamer-in.c (lto_input_mode_table): Likewise.
19476 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
19478 * stor-layout.c (int_mode_for_mode): Likewise.
19479 * tree.c (build_vector_type_for_mode): Likewise.
19480 * varasm.c (output_constant_pool_2): Likewise.
19481 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
19482 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
19483 for MODE_VECTOR_BOOL.
19484 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
19485 of mode class checks.
19486 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
19487 instead of a list of mode class checks.
19488 (expand_vector_scalar_condition): Likewise.
19489 (type_for_widest_vector_mode): Handle BImode as an inner mode.
19491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19492 Alan Hayward <alan.hayward@arm.com>
19493 David Sherwood <david.sherwood@arm.com>
19495 * machmode.h (mode_size): Change from unsigned short to
19497 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
19498 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
19499 or if measurement_type is not polynomial.
19500 (fixed_size_mode::includes_p): Check for constant-sized modes.
19501 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
19502 return a poly_uint16 rather than an unsigned short.
19503 (emit_mode_size): Change the type of mode_size from unsigned short
19504 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
19505 (emit_mode_adjustments): Cope with polynomial vector sizes.
19506 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19508 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19510 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
19511 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
19512 * caller-save.c (setup_save_areas): Likewise.
19513 (replace_reg_with_saved_mem): Likewise.
19514 * calls.c (emit_library_call_value_1): Likewise.
19515 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
19516 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
19517 (gen_lowpart_for_combine): Likewise.
19518 * convert.c (convert_to_integer_1): Likewise.
19519 * cse.c (equiv_constant, cse_insn): Likewise.
19520 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
19521 (cselib_subst_to_values): Likewise.
19522 * dce.c (word_dce_process_block): Likewise.
19523 * df-problems.c (df_word_lr_mark_ref): Likewise.
19524 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
19525 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
19526 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
19527 (rtl_for_decl_location): Likewise.
19528 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
19529 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
19530 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
19531 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
19532 (expand_expr_real_1): Likewise.
19533 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
19534 (pad_below): Likewise.
19535 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
19536 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
19537 * ira.c (get_subreg_tracking_sizes): Likewise.
19538 * ira-build.c (ira_create_allocno_objects): Likewise.
19539 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
19540 (ira_sort_regnos_for_alter_reg): Likewise.
19541 * ira-costs.c (record_operand_costs): Likewise.
19542 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
19543 (resolve_simple_move): Likewise.
19544 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
19545 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
19546 (lra_constraints): Likewise.
19547 (CONST_POOL_OK_P): Reject variable-sized modes.
19548 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
19549 (add_pseudo_to_slot, lra_spill): Likewise.
19550 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19551 * optabs-query.c (get_best_extraction_insn): Likewise.
19552 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
19553 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
19554 (expand_mult_highpart, valid_multiword_target_p): Likewise.
19555 * recog.c (offsettable_address_addr_space_p): Likewise.
19556 * regcprop.c (maybe_mode_change): Likewise.
19557 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
19558 * regrename.c (build_def_use): Likewise.
19559 * regstat.c (dump_reg_info): Likewise.
19560 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
19561 (find_reloads, find_reloads_subreg_address): Likewise.
19562 * reload1.c (eliminate_regs_1): Likewise.
19563 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
19564 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
19565 (simplify_binary_operation_1, simplify_subreg): Likewise.
19566 * targhooks.c (default_function_arg_padding): Likewise.
19567 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
19568 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
19569 (verify_gimple_assign_ternary): Likewise.
19570 * tree-inline.c (estimate_move_cost): Likewise.
19571 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19572 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
19573 (get_address_cost_ainc): Likewise.
19574 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
19575 (vect_supportable_dr_alignment): Likewise.
19576 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
19577 (vectorizable_reduction): Likewise.
19578 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
19579 (vectorizable_operation, vectorizable_load): Likewise.
19580 * tree.c (build_same_sized_truth_vector_type): Likewise.
19581 * valtrack.c (cleanup_auto_inc_dec): Likewise.
19582 * var-tracking.c (emit_note_insn_var_location): Likewise.
19583 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
19584 (ADDR_VEC_ALIGN): Likewise.
19586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19587 Alan Hayward <alan.hayward@arm.com>
19588 David Sherwood <david.sherwood@arm.com>
19590 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
19592 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
19593 or if measurement_type is polynomial.
19594 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
19595 * combine.c (make_extraction): Likewise.
19596 * dse.c (find_shift_sequence): Likewise.
19597 * dwarf2out.c (mem_loc_descriptor): Likewise.
19598 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
19599 (extract_bit_field, extract_low_bits): Likewise.
19600 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
19601 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
19602 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
19603 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
19604 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
19605 * reload.c (find_reloads): Likewise.
19606 * reload1.c (alter_reg): Likewise.
19607 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
19608 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
19609 * tree-if-conv.c (predicate_mem_writes): Likewise.
19610 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19611 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
19612 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
19613 * valtrack.c (dead_debug_insert_temp): Likewise.
19614 * varasm.c (mergeable_constant_section): Likewise.
19615 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
19617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19618 Alan Hayward <alan.hayward@arm.com>
19619 David Sherwood <david.sherwood@arm.com>
19621 * expr.c (expand_assignment): Cope with polynomial mode sizes
19622 when assigning to a CONCAT.
19624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19625 Alan Hayward <alan.hayward@arm.com>
19626 David Sherwood <david.sherwood@arm.com>
19628 * machmode.h (mode_precision): Change from unsigned short to
19630 (mode_to_precision): Return a poly_uint16 rather than an unsigned
19632 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
19633 or if measurement_type is not polynomial.
19634 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
19635 in which the mode is already known to be a scalar_int_mode.
19636 * genmodes.c (emit_mode_precision): Change the type of mode_precision
19637 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
19639 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19640 for GET_MODE_PRECISION.
19641 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19642 for GET_MODE_PRECISION.
19643 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
19645 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
19646 (expand_field_assignment, make_extraction): Likewise.
19647 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
19648 (get_last_value): Likewise.
19649 * convert.c (convert_to_integer_1): Likewise.
19650 * cse.c (cse_insn): Likewise.
19651 * expr.c (expand_expr_real_1): Likewise.
19652 * lra-constraints.c (simplify_operand_subreg): Likewise.
19653 * optabs-query.c (can_atomic_load_p): Likewise.
19654 * optabs.c (expand_atomic_load): Likewise.
19655 (expand_atomic_store): Likewise.
19656 * ree.c (combine_reaching_defs): Likewise.
19657 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
19658 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
19659 * tree.h (type_has_mode_precision_p): Likewise.
19660 * ubsan.c (instrument_si_overflow): Likewise.
19662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19663 Alan Hayward <alan.hayward@arm.com>
19664 David Sherwood <david.sherwood@arm.com>
19666 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
19667 polynomial numbers of units.
19668 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
19669 (valid_vector_subparts_p): New function.
19670 (build_vector_type): Remove temporary shim and take the number
19671 of units as a poly_uint64 rather than an int.
19672 (build_opaque_vector_type): Take the number of units as a
19673 poly_uint64 rather than an int.
19674 * tree.c (build_vector_from_ctor): Handle polynomial
19675 TYPE_VECTOR_SUBPARTS.
19676 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
19677 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
19678 (build_vector_from_val): If the number of units is variable,
19679 use build_vec_duplicate_cst for constant operands and
19680 VEC_DUPLICATE_EXPR otherwise.
19681 (make_vector_type): Remove temporary is_constant ().
19682 (build_vector_type, build_opaque_vector_type): Take the number of
19683 units as a poly_uint64 rather than an int.
19684 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
19686 * cfgexpand.c (expand_debug_expr): Likewise.
19687 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
19688 (store_constructor, expand_expr_real_1): Likewise.
19689 (const_scalar_mask_from_tree): Likewise.
19690 * fold-const-call.c (fold_const_reduction): Likewise.
19691 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
19692 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
19693 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
19694 (fold_relational_const): Likewise.
19695 (native_interpret_vector): Likewise. Change the size from an
19696 int to an unsigned int.
19697 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
19698 TYPE_VECTOR_SUBPARTS.
19699 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
19700 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
19701 duplicating a non-constant operand into a variable-length vector.
19702 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
19703 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
19704 * ipa-icf.c (sem_variable::equals): Likewise.
19705 * match.pd: Likewise.
19706 * omp-simd-clone.c (simd_clone_subparts): Likewise.
19707 * print-tree.c (print_node): Likewise.
19708 * stor-layout.c (layout_type): Likewise.
19709 * targhooks.c (default_builtin_vectorization_cost): Likewise.
19710 * tree-cfg.c (verify_gimple_comparison): Likewise.
19711 (verify_gimple_assign_binary): Likewise.
19712 (verify_gimple_assign_ternary): Likewise.
19713 (verify_gimple_assign_single): Likewise.
19714 * tree-pretty-print.c (dump_generic_node): Likewise.
19715 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19716 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
19717 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
19718 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
19719 (vect_shift_permute_load_chain): Likewise.
19720 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
19721 (expand_vector_condition, optimize_vector_constructor): Likewise.
19722 (lower_vec_perm, get_compute_type): Likewise.
19723 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
19724 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
19725 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
19726 (vect_recog_mask_conversion_pattern): Likewise.
19727 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
19728 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
19729 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19730 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
19731 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
19732 (vectorizable_shift, vectorizable_operation, vectorizable_store)
19733 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
19734 (supportable_widening_operation): Likewise.
19735 (supportable_narrowing_operation): Likewise.
19736 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
19738 * varasm.c (output_constant): Likewise.
19740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19741 Alan Hayward <alan.hayward@arm.com>
19742 David Sherwood <david.sherwood@arm.com>
19744 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
19745 so that both the length == 3 and length != 3 cases set up their
19746 own permute vectors. Add comments explaining why we know the
19747 number of elements is constant.
19748 (vect_permute_load_chain): Likewise.
19750 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19751 Alan Hayward <alan.hayward@arm.com>
19752 David Sherwood <david.sherwood@arm.com>
19754 * machmode.h (mode_nunits): Change from unsigned char to
19756 (ONLY_FIXED_SIZE_MODES): New macro.
19757 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
19758 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
19759 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
19761 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
19762 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
19763 or if measurement_type is not polynomial.
19764 * genmodes.c (ZERO_COEFFS): New macro.
19765 (emit_mode_nunits_inline): Make mode_nunits_inline return a
19767 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
19768 Use ZERO_COEFFS when emitting initializers.
19769 * data-streamer.h (bp_pack_poly_value): New function.
19770 (bp_unpack_poly_value): Likewise.
19771 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19772 for GET_MODE_NUNITS.
19773 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19774 for GET_MODE_NUNITS.
19775 * tree.c (make_vector_type): Remove temporary shim and make
19776 the real function take the number of units as a poly_uint64
19777 rather than an int.
19778 (build_vector_type_for_mode): Handle polynomial nunits.
19779 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
19780 * emit-rtl.c (const_vec_series_p_1): Likewise.
19781 (gen_rtx_CONST_VECTOR): Likewise.
19782 * fold-const.c (test_vec_duplicate_folding): Likewise.
19783 * genrecog.c (validate_pattern): Likewise.
19784 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
19785 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
19786 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
19787 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
19788 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
19789 * rtlanal.c (subreg_get_info): Likewise.
19790 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19791 (vect_grouped_load_supported): Likewise.
19792 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19793 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
19794 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19795 (simplify_const_unary_operation, simplify_binary_operation_1)
19796 (simplify_const_binary_operation, simplify_ternary_operation)
19797 (test_vector_ops_duplicate, test_vector_ops): Likewise.
19798 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
19799 instead of CONST_VECTOR_NUNITS.
19800 * varasm.c (output_constant_pool_2): Likewise.
19801 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
19802 explicit-encoded elements in the XVEC for variable-length vectors.
19804 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19806 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
19808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19809 Alan Hayward <alan.hayward@arm.com>
19810 David Sherwood <david.sherwood@arm.com>
19812 * coretypes.h (fixed_size_mode): Declare.
19813 (fixed_size_mode_pod): New typedef.
19814 * builtins.h (target_builtins::x_apply_args_mode)
19815 (target_builtins::x_apply_result_mode): Change type to
19816 fixed_size_mode_pod.
19817 * builtins.c (apply_args_size, apply_result_size, result_vector)
19818 (expand_builtin_apply_args_1, expand_builtin_apply)
19819 (expand_builtin_return): Update accordingly.
19821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19823 * cse.c (hash_rtx_cb): Hash only the encoded elements.
19824 * cselib.c (cselib_hash_rtx): Likewise.
19825 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
19826 CONST_VECTOR encoding.
19828 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19829 Jeff Law <law@redhat.com>
19832 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
19833 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
19834 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
19835 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
19838 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
19839 explicitly probe *sp in a noreturn function if there were any callee
19840 register saves or frame pointer is needed.
19842 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19845 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
19846 BLKmode for ternary, binary or unary expressions.
19849 * var-tracking.c (delete_vta_debug_insn): New inline function.
19850 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
19851 insns from get_insns () to NULL instead of each bb separately.
19852 Use delete_vta_debug_insn. No longer static.
19853 (vt_debug_insns_local, variable_tracking_main_1): Adjust
19854 delete_vta_debug_insns callers.
19855 * rtl.h (delete_vta_debug_insns): Declare.
19856 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
19857 instead of variable_tracking_main.
19859 2018-01-03 Martin Sebor <msebor@redhat.com>
19861 PR tree-optimization/83603
19862 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
19863 arguments past the endof the argument list in functions declared
19864 without a prototype.
19865 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
19866 Avoid checking when arguments are null.
19868 2018-01-03 Martin Sebor <msebor@redhat.com>
19871 * doc/extend.texi (attribute const): Fix a typo.
19872 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
19873 issuing -Wsuggest-attribute for void functions.
19875 2018-01-03 Martin Sebor <msebor@redhat.com>
19877 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
19878 offset_int::from instead of wide_int::to_shwi.
19879 (maybe_diag_overlap): Remove assertion.
19880 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
19881 * gimple-ssa-sprintf.c (format_directive): Same.
19882 (parse_directive): Same.
19883 (sprintf_dom_walker::compute_format_length): Same.
19884 (try_substitute_return_value): Same.
19886 2018-01-03 Jeff Law <law@redhat.com>
19888 PR middle-end/83654
19889 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
19890 non-constant residual for zero at runtime and avoid probing in
19891 that case. Reorganize code for trailing problem to mirror handling
19894 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19896 PR tree-optimization/83501
19897 * tree-ssa-strlen.c (get_string_cst): New.
19898 (handle_char_store): Call get_string_cst.
19900 2018-01-03 Martin Liska <mliska@suse.cz>
19902 PR tree-optimization/83593
19903 * tree-ssa-strlen.c: Include tree-cfg.h.
19904 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
19905 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
19906 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
19908 (strlen_dom_walker::before_dom_children): Call
19909 gimple_purge_dead_eh_edges. Dump tranformation with details
19911 (strlen_dom_walker::before_dom_children): Update call by adding
19912 new argument cleanup_eh.
19913 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
19915 2018-01-03 Martin Liska <mliska@suse.cz>
19918 * cif-code.def (VARIADIC_THUNK): New enum value.
19919 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
19922 2018-01-03 Jan Beulich <jbeulich@suse.com>
19924 * sse.md (mov<mode>_internal): Tighten condition for when to use
19925 vmovdqu<ssescalarsize> for TI and OI modes.
19927 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19929 Update copyright years.
19931 2018-01-03 Martin Liska <mliska@suse.cz>
19934 * ipa-visibility.c (function_and_variable_visibility): Skip
19935 functions with noipa attribure.
19937 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19939 * gcc.c (process_command): Update copyright notice dates.
19940 * gcov-dump.c (print_version): Ditto.
19941 * gcov.c (print_version): Ditto.
19942 * gcov-tool.c (print_version): Ditto.
19943 * gengtype.c (create_file): Ditto.
19944 * doc/cpp.texi: Bump @copying's copyright year.
19945 * doc/cppinternals.texi: Ditto.
19946 * doc/gcc.texi: Ditto.
19947 * doc/gccint.texi: Ditto.
19948 * doc/gcov.texi: Ditto.
19949 * doc/install.texi: Ditto.
19950 * doc/invoke.texi: Ditto.
19952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19954 * vector-builder.h (vector_builder::m_full_nelts): Change from
19955 unsigned int to poly_uint64.
19956 (vector_builder::full_nelts): Update prototype accordingly.
19957 (vector_builder::new_vector): Likewise.
19958 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
19959 (vector_builder::operator ==): Likewise.
19960 (vector_builder::finalize): Likewise.
19961 * int-vector-builder.h (int_vector_builder::int_vector_builder):
19962 Take the number of elements as a poly_uint64 rather than an
19964 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
19965 from unsigned int to poly_uint64.
19966 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
19967 (vec_perm_indices::new_vector): Likewise.
19968 (vec_perm_indices::length): Likewise.
19969 (vec_perm_indices::nelts_per_input): Likewise.
19970 (vec_perm_indices::input_nelts): Likewise.
19971 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
19972 number of elements per input as a poly_uint64 rather than an
19973 unsigned int. Use the original encoding for variable-length
19974 vectors, rather than clamping each individual element.
19975 For the second and subsequent elements in each pattern,
19976 clamp the step and base before clamping their sum.
19977 (vec_perm_indices::series_p): Handle polynomial element counts.
19978 (vec_perm_indices::all_in_range_p): Likewise.
19979 (vec_perm_indices_to_tree): Likewise.
19980 (vec_perm_indices_to_rtx): Likewise.
19981 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19982 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
19983 (tree_vector_builder::new_binary_operation): Handle polynomial
19984 element counts. Return false if we need to know the number
19985 of elements at compile time.
19986 * fold-const.c (fold_vec_perm): Punt if the number of elements
19987 isn't known at compile time.
19989 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19991 * vec-perm-indices.h (vec_perm_builder): Change element type
19992 from HOST_WIDE_INT to poly_int64.
19993 (vec_perm_indices::element_type): Update accordingly.
19994 (vec_perm_indices::clamp): Handle polynomial element_types.
19995 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19996 (vec_perm_indices::all_in_range_p): Likewise.
19997 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
19999 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
20000 polynomial vec_perm_indices element types.
20001 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
20002 * fold-const.c (fold_vec_perm): Likewise.
20003 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
20004 * tree-vect-generic.c (lower_vec_perm): Likewise.
20005 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
20006 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
20007 element type to HOST_WIDE_INT.
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 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
20014 rather than an int. Use plus_constant.
20015 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
20016 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
20018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20019 Alan Hayward <alan.hayward@arm.com>
20020 David Sherwood <david.sherwood@arm.com>
20022 * calls.c (emit_call_1, expand_call): Change struct_value_size from
20023 a HOST_WIDE_INT to a poly_int64.
20025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20026 Alan Hayward <alan.hayward@arm.com>
20027 David Sherwood <david.sherwood@arm.com>
20029 * calls.c (load_register_parameters): Cope with polynomial
20030 mode sizes. Require a constant size for BLKmode parameters
20031 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
20032 forces a parameter to be padded at the lsb end in order to
20033 fill a complete number of words, require the parameter size
20034 to be ordered wrt UNITS_PER_WORD.
20036 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20037 Alan Hayward <alan.hayward@arm.com>
20038 David Sherwood <david.sherwood@arm.com>
20040 * reload1.c (spill_stack_slot_width): Change element type
20041 from unsigned int to poly_uint64_pod.
20042 (alter_reg): Treat mode sizes as polynomial.
20044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20045 Alan Hayward <alan.hayward@arm.com>
20046 David Sherwood <david.sherwood@arm.com>
20048 * reload.c (complex_word_subreg_p): New function.
20049 (reload_inner_reg_of_subreg, push_reload): Use it.
20051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20052 Alan Hayward <alan.hayward@arm.com>
20053 David Sherwood <david.sherwood@arm.com>
20055 * lra-constraints.c (process_alt_operands): Reject matched
20056 operands whose sizes aren't ordered.
20057 (match_reload): Refer to this check here.
20059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20060 Alan Hayward <alan.hayward@arm.com>
20061 David Sherwood <david.sherwood@arm.com>
20063 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
20064 that the mode size is in the set {1, 2, 4, 8, 16}.
20066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20067 Alan Hayward <alan.hayward@arm.com>
20068 David Sherwood <david.sherwood@arm.com>
20070 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
20071 Use plus_constant instead of gen_rtx_PLUS.
20073 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20074 Alan Hayward <alan.hayward@arm.com>
20075 David Sherwood <david.sherwood@arm.com>
20077 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
20078 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
20079 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
20080 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
20081 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
20082 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
20083 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
20084 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
20085 * config/i386/i386.c (ix86_push_rounding): ...this new function.
20086 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
20088 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
20089 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
20090 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
20091 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
20092 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
20093 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
20094 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
20095 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
20096 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
20097 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
20099 * expr.c (emit_move_resolve_push): Treat the input and result
20100 of PUSH_ROUNDING as a poly_int64.
20101 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
20102 (emit_push_insn): Likewise.
20103 * lra-eliminations.c (mark_not_eliminable): Likewise.
20104 * recog.c (push_operand): Likewise.
20105 * reload1.c (elimination_effects): Likewise.
20106 * rtlanal.c (nonzero_bits1): Likewise.
20107 * calls.c (store_one_arg): Likewise. Require the padding to be
20108 known at compile time.
20110 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20111 Alan Hayward <alan.hayward@arm.com>
20112 David Sherwood <david.sherwood@arm.com>
20114 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
20115 Use plus_constant instead of gen_rtx_PLUS.
20117 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20118 Alan Hayward <alan.hayward@arm.com>
20119 David Sherwood <david.sherwood@arm.com>
20121 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
20122 rather than an int.
20124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20125 Alan Hayward <alan.hayward@arm.com>
20126 David Sherwood <david.sherwood@arm.com>
20128 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
20129 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
20130 via stack temporaries. Treat the mode size as polynomial too.
20132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20133 Alan Hayward <alan.hayward@arm.com>
20134 David Sherwood <david.sherwood@arm.com>
20136 * expr.c (expand_expr_real_2): When handling conversions involving
20137 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
20138 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
20139 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
20152 REGMODE_NATURAL_SIZE.
20153 * expmed.c (store_bit_field_1): Likewise.
20154 * expr.c (store_constructor): Likewise.
20155 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
20156 and polynomial REGMODE_NATURAL_SIZE.
20157 (gen_lowpart_common): Likewise.
20158 * reginfo.c (record_subregs_of_mode): Likewise.
20159 * rtlanal.c (read_modify_subreg_p): Likewise.
20161 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20162 Alan Hayward <alan.hayward@arm.com>
20163 David Sherwood <david.sherwood@arm.com>
20165 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
20166 numbers of elements.
20168 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20169 Alan Hayward <alan.hayward@arm.com>
20170 David Sherwood <david.sherwood@arm.com>
20172 * match.pd: Cope with polynomial numbers of vector elements.
20174 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20175 Alan Hayward <alan.hayward@arm.com>
20176 David Sherwood <david.sherwood@arm.com>
20178 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
20179 in a POINTER_PLUS_EXPR.
20181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20182 Alan Hayward <alan.hayward@arm.com>
20183 David Sherwood <david.sherwood@arm.com>
20185 * omp-simd-clone.c (simd_clone_subparts): New function.
20186 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
20187 (ipa_simd_modify_function_body): Likewise.
20189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20190 Alan Hayward <alan.hayward@arm.com>
20191 David Sherwood <david.sherwood@arm.com>
20193 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
20194 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
20195 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
20196 (expand_vector_condition, vector_element): Likewise.
20197 (subparts_gt): New function.
20198 (get_compute_type): Use subparts_gt.
20199 (count_type_subparts): Delete.
20200 (expand_vector_operations_1): Use subparts_gt instead of
20201 count_type_subparts.
20203 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20204 Alan Hayward <alan.hayward@arm.com>
20205 David Sherwood <david.sherwood@arm.com>
20207 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
20208 (vect_compile_time_alias): ...this new function. Do the calculation
20209 on poly_ints rather than trees.
20210 (vect_prune_runtime_alias_test_list): Update call accordingly.
20212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20213 Alan Hayward <alan.hayward@arm.com>
20214 David Sherwood <david.sherwood@arm.com>
20216 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
20218 (vect_schedule_slp_instance): Likewise.
20220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20221 Alan Hayward <alan.hayward@arm.com>
20222 David Sherwood <david.sherwood@arm.com>
20224 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
20225 constant and extern definitions for variable-length vectors.
20226 (vect_get_constant_vectors): Note that the number of units
20227 is known to be constant.
20229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20230 Alan Hayward <alan.hayward@arm.com>
20231 David Sherwood <david.sherwood@arm.com>
20233 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
20234 of units as polynomial. Choose between WIDE and NARROW based
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-stmts.c (simd_clone_subparts): New function.
20242 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
20244 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20245 Alan Hayward <alan.hayward@arm.com>
20246 David Sherwood <david.sherwood@arm.com>
20248 * tree-vect-stmts.c (vectorizable_call): Treat the number of
20249 vectors as polynomial. Use build_index_vector for
20250 IFN_GOMP_SIMD_LANE.
20252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20253 Alan Hayward <alan.hayward@arm.com>
20254 David Sherwood <david.sherwood@arm.com>
20256 * tree-vect-stmts.c (get_load_store_type): Treat the number of
20257 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
20258 for variable-length vectors.
20259 (vectorizable_mask_load_store): Treat the number of units as
20260 polynomial, asserting that it is constant if the condition has
20261 already been enforced.
20262 (vectorizable_store, vectorizable_load): Likewise.
20264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20265 Alan Hayward <alan.hayward@arm.com>
20266 David Sherwood <david.sherwood@arm.com>
20268 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
20269 of units as polynomial. Punt if we can't tell at compile time
20270 which vector contains the final result.
20272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20273 Alan Hayward <alan.hayward@arm.com>
20274 David Sherwood <david.sherwood@arm.com>
20276 * tree-vect-loop.c (vectorizable_induction): Treat the number
20277 of units as polynomial. Punt on SLP inductions. Use an integer
20278 VEC_SERIES_EXPR for variable-length integer reductions. Use a
20279 cast of such a series for variable-length floating-point
20282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20283 Alan Hayward <alan.hayward@arm.com>
20284 David Sherwood <david.sherwood@arm.com>
20286 * tree.h (build_index_vector): Declare.
20287 * tree.c (build_index_vector): New function.
20288 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
20289 of units as polynomial, forcibly converting it to a constant if
20290 vectorizable_reduction has already enforced the condition.
20291 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
20292 to create a {1,2,3,...} vector.
20293 (vectorizable_reduction): Treat the number of units as polynomial.
20294 Choose vectype_in based on the largest scalar element size rather
20295 than the smallest number of units. Enforce the restrictions
20298 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20299 Alan Hayward <alan.hayward@arm.com>
20300 David Sherwood <david.sherwood@arm.com>
20302 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
20303 number of units as polynomial.
20305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20306 Alan Hayward <alan.hayward@arm.com>
20307 David Sherwood <david.sherwood@arm.com>
20309 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
20310 * target.def (autovectorize_vector_sizes): Return the vector sizes
20311 by pointer, using vector_sizes rather than a bitmask.
20312 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
20313 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
20314 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
20316 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
20317 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
20318 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
20319 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
20320 * omp-general.c (omp_max_vf): Likewise.
20321 * omp-low.c (omp_clause_aligned_alignment): Likewise.
20322 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
20323 * tree-vect-loop.c (vect_analyze_loop): Likewise.
20324 * tree-vect-slp.c (vect_slp_bb): Likewise.
20325 * doc/tm.texi: Regenerate.
20326 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
20328 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
20329 the vector size as a poly_uint64 rather than an unsigned int.
20330 (current_vector_size): Change from an unsigned int to a poly_uint64.
20331 (get_vectype_for_scalar_type): Update accordingly.
20332 * tree.h (build_truth_vector_type): Take the size and number of
20333 units as a poly_uint64 rather than an unsigned int.
20334 (build_vector_type): Add a temporary overload that takes
20335 the number of units as a poly_uint64 rather than an unsigned int.
20336 * tree.c (make_vector_type): Likewise.
20337 (build_truth_vector_type): Take the number of units as a poly_uint64
20338 rather than an unsigned int.
20340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20341 Alan Hayward <alan.hayward@arm.com>
20342 David Sherwood <david.sherwood@arm.com>
20344 * target.def (get_mask_mode): Take the number of units and length
20345 as poly_uint64s rather than unsigned ints.
20346 * targhooks.h (default_get_mask_mode): Update accordingly.
20347 * targhooks.c (default_get_mask_mode): Likewise.
20348 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
20349 * doc/tm.texi: Regenerate.
20351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20352 Alan Hayward <alan.hayward@arm.com>
20353 David Sherwood <david.sherwood@arm.com>
20355 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
20356 * omp-general.c (omp_max_vf): Likewise.
20357 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
20358 (expand_omp_simd): Handle polynomial safelen.
20359 * omp-low.c (omplow_simd_context): Add a default constructor.
20360 (omplow_simd_context::max_vf): Change from int to poly_uint64.
20361 (lower_rec_simd_input_clauses): Update accordingly.
20362 (lower_rec_input_clauses): Likewise.
20364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20365 Alan Hayward <alan.hayward@arm.com>
20366 David Sherwood <david.sherwood@arm.com>
20368 * tree-vectorizer.h (vect_nunits_for_cost): New function.
20369 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
20370 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
20371 (vect_analyze_slp_cost): Likewise.
20372 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
20373 (vect_model_load_cost): Likewise.
20375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20376 Alan Hayward <alan.hayward@arm.com>
20377 David Sherwood <david.sherwood@arm.com>
20379 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
20380 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
20381 from an unsigned int * to a poly_uint64_pod *.
20382 (calculate_unrolling_factor): New function.
20383 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
20385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20386 Alan Hayward <alan.hayward@arm.com>
20387 David Sherwood <david.sherwood@arm.com>
20389 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
20390 from an unsigned int to a poly_uint64.
20391 (_loop_vec_info::slp_unrolling_factor): Likewise.
20392 (_loop_vec_info::vectorization_factor): Change from an int
20394 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
20395 (vect_get_num_vectors): New function.
20396 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
20397 (vect_get_num_copies): Use vect_get_num_vectors.
20398 (vect_analyze_data_ref_dependences): Change max_vf from an int *
20399 to an unsigned int *.
20400 (vect_analyze_data_refs): Change min_vf from an int * to a
20402 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
20403 than an unsigned HOST_WIDE_INT.
20404 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
20405 (vect_analyze_data_ref_dependence): Change max_vf from an int *
20406 to an unsigned int *.
20407 (vect_analyze_data_ref_dependences): Likewise.
20408 (vect_compute_data_ref_alignment): Handle polynomial vf.
20409 (vect_enhance_data_refs_alignment): Likewise.
20410 (vect_prune_runtime_alias_test_list): Likewise.
20411 (vect_shift_permute_load_chain): Likewise.
20412 (vect_supportable_dr_alignment): Likewise.
20413 (dependence_distance_ge_vf): Take the vectorization factor as a
20414 poly_uint64 rather than an unsigned HOST_WIDE_INT.
20415 (vect_analyze_data_refs): Change min_vf from an int * to a
20417 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
20418 vfm1 as a poly_uint64 rather than an int. Make the same change
20419 for the returned bound_scalar.
20420 (vect_gen_vector_loop_niters): Handle polynomial vf.
20421 (vect_do_peeling): Likewise. Update call to
20422 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
20423 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
20425 * tree-vect-loop.c (vect_determine_vectorization_factor)
20426 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
20427 (vect_get_known_peeling_cost): Likewise.
20428 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
20429 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
20430 (vect_transform_loop): Likewise. Use the lowest possible VF when
20431 updating the upper bounds of the loop.
20432 (vect_min_worthwhile_factor): Make static. Return an unsigned int
20433 rather than an int.
20434 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
20435 polynomial unroll factors.
20436 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
20437 (vect_make_slp_decision): Likewise.
20438 (vect_supported_load_permutation_p): Likewise, and polynomial
20440 (vect_analyze_slp_cost): Handle polynomial vf.
20441 (vect_slp_analyze_node_operations): Likewise.
20442 (vect_slp_analyze_bb_1): Likewise.
20443 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
20444 than an unsigned HOST_WIDE_INT.
20445 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
20446 (vectorizable_load): Handle polynomial vf.
20447 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
20449 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
20451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20452 Alan Hayward <alan.hayward@arm.com>
20453 David Sherwood <david.sherwood@arm.com>
20455 * match.pd: Handle bit operations involving three constants
20456 and try to fold one pair.
20458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20460 * tree-vect-loop-manip.c: Include gimple-fold.h.
20461 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
20462 niters_maybe_zero parameters. Handle other cases besides a step of 1.
20463 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
20464 Add a path that uses a step of VF instead of 1, but disable it
20466 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
20467 and niters_no_overflow parameters. Update calls to
20468 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
20469 Create a new SSA name if the latter choses to use a ste other
20470 than zero, and return it via niters_vector_mult_vf_var.
20471 * tree-vect-loop.c (vect_transform_loop): Update calls to
20472 vect_do_peeling, vect_gen_vector_loop_niters and
20473 slpeel_make_loop_iterate_ntimes.
20474 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
20475 (vect_gen_vector_loop_niters): Update declarations after above changes.
20477 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
20479 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
20480 128-bit round to integer instructions.
20481 (ceil<mode>2): Likewise.
20482 (btrunc<mode>2): Likewise.
20483 (round<mode>2): Likewise.
20485 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20487 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
20488 unaligned VSX load/store on P8/P9.
20489 (expand_block_clear): Allow the use of unaligned VSX
20490 load/store on P8/P9.
20492 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20494 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
20496 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
20497 swap associated with both a load and a store.
20499 2018-01-02 Andrew Waterman <andrew@sifive.com>
20501 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
20502 * config/riscv/riscv.md (clear_cache): Use it.
20504 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
20506 * web.c: Remove out-of-date comment.
20508 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20510 * expr.c (fixup_args_size_notes): Check that any existing
20511 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
20512 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
20513 (emit_single_push_insn): ...here.
20515 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20517 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
20518 (const_vector_encoded_nelts): New function.
20519 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
20520 (const_vector_int_elt, const_vector_elt): Declare.
20521 * emit-rtl.c (const_vector_int_elt_1): New function.
20522 (const_vector_elt): Likewise.
20523 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
20524 of CONST_VECTOR_ELT.
20526 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20528 * expr.c: Include rtx-vector-builder.h.
20529 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
20530 directly on the tree encoding.
20531 (const_vector_from_tree): Likewise.
20532 * optabs.c: Include rtx-vector-builder.h.
20533 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
20534 sequence of "u" values.
20535 * vec-perm-indices.c: Include rtx-vector-builder.h.
20536 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
20537 directly on the vec_perm_indices encoding.
20539 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20541 * doc/rtl.texi (const_vector): Describe new encoding scheme.
20542 * Makefile.in (OBJS): Add rtx-vector-builder.o.
20543 * rtx-vector-builder.h: New file.
20544 * rtx-vector-builder.c: Likewise.
20545 * rtl.h (rtx_def::u2): Add a const_vector field.
20546 (CONST_VECTOR_NPATTERNS): New macro.
20547 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
20548 (CONST_VECTOR_DUPLICATE_P): Likewise.
20549 (CONST_VECTOR_STEPPED_P): Likewise.
20550 (CONST_VECTOR_ENCODED_ELT): Likewise.
20551 (const_vec_duplicate_p): Check for a duplicated vector encoding.
20552 (unwrap_const_vec_duplicate): Likewise.
20553 (const_vec_series_p): Check for a non-duplicated vector encoding.
20554 Say that the function only returns true for integer vectors.
20555 * emit-rtl.c: Include rtx-vector-builder.h.
20556 (gen_const_vec_duplicate_1): Delete.
20557 (gen_const_vector): Call gen_const_vec_duplicate instead of
20558 gen_const_vec_duplicate_1.
20559 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
20560 (gen_const_vec_duplicate): Use rtx_vector_builder.
20561 (gen_const_vec_series): Likewise.
20562 (gen_rtx_CONST_VECTOR): Likewise.
20563 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
20564 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
20565 Build a new vector rather than modifying a CONST_VECTOR in-place.
20566 (handle_special_swappables): Update call accordingly.
20567 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
20568 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
20569 Build a new vector rather than modifying a CONST_VECTOR in-place.
20570 (handle_special_swappables): Update call accordingly.
20572 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20574 * simplify-rtx.c (simplify_const_binary_operation): Use
20575 CONST_VECTOR_ELT instead of XVECEXP.
20577 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20579 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
20580 the selector elements to be different from the data elements
20581 if the selector is a VECTOR_CST.
20582 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
20583 ssizetype for the selector.
20585 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20587 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
20588 before testing each element individually.
20589 * tree-vect-generic.c (lower_vec_perm): Likewise.
20591 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20593 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
20594 * selftest-run-tests.c (selftest::run_tests): Call it.
20595 * vector-builder.h (vector_builder::operator ==): New function.
20596 (vector_builder::operator !=): Likewise.
20597 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
20598 (vec_perm_indices::all_from_input_p): New function.
20599 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
20600 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
20601 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
20602 instead of reading the VECTOR_CST directly. Detect whether both
20603 vector inputs are the same before constructing the vec_perm_indices,
20604 and update the number of inputs argument accordingly. Use the
20605 utility functions added above. Only construct sel2 if we need to.
20607 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20609 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
20610 the broadcast of the low byte.
20611 (expand_mult_highpart): Use an explicit encoding for the permutes.
20612 * optabs-query.c (can_mult_highpart_p): Likewise.
20613 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
20614 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20615 (vectorizable_bswap): Likewise.
20616 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
20617 explicit encoding for the power-of-2 permutes.
20618 (vect_permute_store_chain): Likewise.
20619 (vect_grouped_load_supported): Likewise.
20620 (vect_permute_load_chain): Likewise.
20622 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20624 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
20625 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
20626 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
20627 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
20628 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
20629 (vect_gen_perm_mask_any): Likewise.
20631 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20633 * int-vector-builder.h: New file.
20634 * vec-perm-indices.h: Include int-vector-builder.h.
20635 (vec_perm_indices): Redefine as an int_vector_builder.
20636 (auto_vec_perm_indices): Delete.
20637 (vec_perm_builder): Redefine as a stand-alone class.
20638 (vec_perm_indices::vec_perm_indices): New function.
20639 (vec_perm_indices::clamp): Likewise.
20640 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
20641 (vec_perm_indices::new_vector): New function.
20642 (vec_perm_indices::new_expanded_vector): Update for new
20643 vec_perm_indices class.
20644 (vec_perm_indices::rotate_inputs): New function.
20645 (vec_perm_indices::all_in_range_p): Operate directly on the
20646 encoded form, without computing elided elements.
20647 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
20648 encoding. Update for new vec_perm_indices class.
20649 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
20650 the given vec_perm_builder.
20651 (expand_vec_perm_var): Update vec_perm_builder constructor.
20652 (expand_mult_highpart): Use vec_perm_builder instead of
20653 auto_vec_perm_indices.
20654 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
20655 vec_perm_indices instead of auto_vec_perm_indices. Use a single
20656 or double series encoding as appropriate.
20657 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
20658 vec_perm_indices instead of auto_vec_perm_indices.
20659 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20660 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20661 (vect_permute_store_chain): Likewise.
20662 (vect_grouped_load_supported): Likewise.
20663 (vect_permute_load_chain): Likewise.
20664 (vect_shift_permute_load_chain): Likewise.
20665 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
20666 (vect_transform_slp_perm_load): Likewise.
20667 (vect_schedule_slp_instance): Likewise.
20668 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20669 (vectorizable_mask_load_store): Likewise.
20670 (vectorizable_bswap): Likewise.
20671 (vectorizable_store): Likewise.
20672 (vectorizable_load): Likewise.
20673 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
20674 vec_perm_indices instead of auto_vec_perm_indices. Use
20675 tree_to_vec_perm_builder to read the vector from a tree.
20676 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
20677 vec_perm_builder instead of a vec_perm_indices.
20678 (have_whole_vector_shift): Use vec_perm_builder and
20679 vec_perm_indices instead of auto_vec_perm_indices. Leave the
20680 truncation to calc_vec_perm_mask_for_shift.
20681 (vect_create_epilog_for_reduction): Likewise.
20682 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
20683 from auto_vec_perm_indices to vec_perm_indices.
20684 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
20685 instead of changing individual elements.
20686 (aarch64_vectorize_vec_perm_const): Use new_vector to install
20687 the vector in d.perm.
20688 * config/arm/arm.c (expand_vec_perm_d::perm): Change
20689 from auto_vec_perm_indices to vec_perm_indices.
20690 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
20691 instead of changing individual elements.
20692 (arm_vectorize_vec_perm_const): Use new_vector to install
20693 the vector in d.perm.
20694 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
20695 Update vec_perm_builder constructor.
20696 (rs6000_expand_interleave): Likewise.
20697 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
20698 (rs6000_expand_interleave): Likewise.
20700 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20702 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
20703 to qimode could truncate the indices.
20704 * optabs.c (expand_vec_perm_var): Likewise.
20706 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20708 * Makefile.in (OBJS): Add vec-perm-indices.o.
20709 * vec-perm-indices.h: New file.
20710 * vec-perm-indices.c: Likewise.
20711 * target.h (vec_perm_indices): Replace with a forward class
20713 (auto_vec_perm_indices): Move to vec-perm-indices.h.
20714 * optabs.h: Include vec-perm-indices.h.
20715 (expand_vec_perm): Delete.
20716 (selector_fits_mode_p, expand_vec_perm_var): Declare.
20717 (expand_vec_perm_const): Declare.
20718 * target.def (vec_perm_const_ok): Replace with...
20719 (vec_perm_const): ...this new hook.
20720 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
20721 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
20722 * doc/tm.texi: Regenerate.
20723 * optabs.def (vec_perm_const): Delete.
20724 * doc/md.texi (vec_perm_const): Likewise.
20725 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
20726 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
20727 expand_vec_perm for constant permutation vectors. Assert that
20728 the mode of variable permutation vectors is the integer equivalent
20729 of the mode that is being permuted.
20730 * optabs-query.h (selector_fits_mode_p): Declare.
20731 * optabs-query.c: Include vec-perm-indices.h.
20732 (selector_fits_mode_p): New function.
20733 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
20734 is defined, instead of checking whether the vec_perm_const_optab
20735 exists. Use targetm.vectorize.vec_perm_const instead of
20736 targetm.vectorize.vec_perm_const_ok. Check whether the indices
20737 fit in the vector mode before using a variable permute.
20738 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
20739 vec_perm_indices instead of an rtx.
20740 (expand_vec_perm): Replace with...
20741 (expand_vec_perm_const): ...this new function. Take the selector
20742 as a vec_perm_indices rather than an rtx. Also take the mode of
20743 the selector. Update call to shift_amt_for_vec_perm_mask.
20744 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
20745 Use vec_perm_indices::new_expanded_vector to expand the original
20746 selector into bytes. Check whether the indices fit in the vector
20747 mode before using a variable permute.
20748 (expand_vec_perm_var): Make global.
20749 (expand_mult_highpart): Use expand_vec_perm_const.
20750 * fold-const.c: Includes vec-perm-indices.h.
20751 * tree-ssa-forwprop.c: Likewise.
20752 * tree-vect-data-refs.c: Likewise.
20753 * tree-vect-generic.c: Likewise.
20754 * tree-vect-loop.c: Likewise.
20755 * tree-vect-slp.c: Likewise.
20756 * tree-vect-stmts.c: Likewise.
20757 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
20759 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
20760 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
20761 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
20762 (aarch64_vectorize_vec_perm_const): ...this new function.
20763 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
20764 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20765 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
20766 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
20767 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
20768 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20769 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
20771 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
20772 check for NEON modes.
20773 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
20774 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
20775 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
20776 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
20778 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
20779 the old VEC_PERM_CONST conditions.
20780 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
20781 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
20782 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
20783 (ia64_vectorize_vec_perm_const_ok): Merge into...
20784 (ia64_vectorize_vec_perm_const): ...this new function.
20785 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
20786 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
20787 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
20788 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
20789 * config/mips/mips.c (mips_expand_vec_perm_const)
20790 (mips_vectorize_vec_perm_const_ok): Merge into...
20791 (mips_vectorize_vec_perm_const): ...this new function.
20792 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
20793 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
20794 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
20795 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
20796 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
20797 (rs6000_expand_vec_perm_const): Delete.
20798 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
20800 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20801 (altivec_expand_vec_perm_const_le): Take each operand individually.
20802 Operate on constant selectors rather than rtxes.
20803 (altivec_expand_vec_perm_const): Likewise. Update call to
20804 altivec_expand_vec_perm_const_le.
20805 (rs6000_expand_vec_perm_const): Delete.
20806 (rs6000_vectorize_vec_perm_const_ok): Delete.
20807 (rs6000_vectorize_vec_perm_const): New function.
20808 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
20809 an element count and rtx array.
20810 (rs6000_expand_extract_even): Update call accordingly.
20811 (rs6000_expand_interleave): Likewise.
20812 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
20813 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
20814 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
20815 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
20816 (rs6000_expand_vec_perm_const): Delete.
20817 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
20818 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20819 (altivec_expand_vec_perm_const_le): Take each operand individually.
20820 Operate on constant selectors rather than rtxes.
20821 (altivec_expand_vec_perm_const): Likewise. Update call to
20822 altivec_expand_vec_perm_const_le.
20823 (rs6000_expand_vec_perm_const): Delete.
20824 (rs6000_vectorize_vec_perm_const_ok): Delete.
20825 (rs6000_vectorize_vec_perm_const): New function. Remove stray
20826 reference to the SPE evmerge intructions.
20827 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
20828 an element count and rtx array.
20829 (rs6000_expand_extract_even): Update call accordingly.
20830 (rs6000_expand_interleave): Likewise.
20831 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
20832 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
20834 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20836 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20838 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
20839 vector mode and that that mode matches the mode of the data
20841 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
20842 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
20843 directly using expand_vec_perm_1 when forcing selectors into
20845 (expand_vec_perm_var): New function, split out from expand_vec_perm.
20847 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20849 * optabs-query.h (can_vec_perm_p): Delete.
20850 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
20851 * optabs-query.c (can_vec_perm_p): Split into...
20852 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
20853 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
20854 particular selector is valid.
20855 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20856 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20857 (vect_grouped_load_supported): Likewise.
20858 (vect_shift_permute_load_chain): Likewise.
20859 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
20860 (vect_transform_slp_perm_load): Likewise.
20861 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20862 (vectorizable_bswap): Likewise.
20863 (vect_gen_perm_mask_checked): Likewise.
20864 * fold-const.c (fold_ternary_loc): Likewise. Don't take
20865 implementations of variable permutation vectors into account
20866 when deciding which selector to use.
20867 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
20868 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
20869 with a false third argument.
20870 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
20871 to test whether the constant selector is valid and can_vec_perm_var_p
20872 to test whether a variable selector is valid.
20874 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20876 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
20877 * optabs-query.c (can_vec_perm_p): Likewise.
20878 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
20879 instead of vec_perm_indices.
20880 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
20881 (vect_gen_perm_mask_checked): Likewise,
20882 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
20883 (vect_gen_perm_mask_checked): Likewise,
20885 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20887 * optabs-query.h (qimode_for_vec_perm): Declare.
20888 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
20889 (qimode_for_vec_perm): ...this new function.
20890 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
20892 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20894 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
20895 does not have a conditional at the top.
20897 2018-01-02 Richard Biener <rguenther@suse.de>
20899 * ipa-inline.c (big_speedup_p): Fix expression.
20901 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
20904 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
20907 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
20911 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
20912 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
20913 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
20914 cond_taken_branch_cost 3->4.
20916 2018-01-01 Jakub Jelinek <jakub@redhat.com>
20918 PR tree-optimization/83581
20919 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
20920 TODO_cleanup_cfg if any changes have been made.
20922 PR middle-end/83608
20923 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
20924 convert_modes if target mode has the right side, but different mode
20927 PR middle-end/83609
20928 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
20929 last argument when extracting from CONCAT. If either from_real or
20930 from_imag is NULL, use expansion through memory. If result is not
20931 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
20932 the parts directly to inner mode, if even that fails, use expansion
20935 PR middle-end/83623
20936 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
20937 check for bswap in mode rather than HImode and use that in expand_unop
20940 Copyright (C) 2018 Free Software Foundation, Inc.
20942 Copying and distribution of this file, with or without modification,
20943 are permitted in any medium without royalty provided the copyright
20944 notice and this notice are preserved.