1 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3 * tree-vect-loop.c (vectorizable_reduction): Assert that the
4 function is not called for second and subsequent members of
7 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
10 cases for nested loops from here to ...
11 (vect_create_epilog_for_reduction): ...here. Only call
12 vect_is_simple_use for inner-loop reductions.
14 2018-07-31 Martin Liska <mliska@suse.cz>
19 * profile.c (struct location_triplet): New.
20 (struct location_triplet_hash): Likewise.
21 (output_location): Do not output a BB that
22 is already recorded for a line.
23 (branch_prob): Use streamed_locations.
25 2018-07-31 Martin Liska <mliska@suse.cz>
28 * coverage.c (coverage_begin_function): Do not mark target
29 clones as artificial functions.
31 2018-07-31 Martin Liska <mliska@suse.cz>
37 * profile.c (branch_prob): Do not record GOTO expressions
38 for GIMPLE statements which locations are already streamed.
40 2018-07-31 Olivier Hainque <hainque@adacore.com>
42 * gcc.c (handle_spec_function): Accept a soft_matched_part
43 argument, as do_spec_1. Pass it down to ...
44 (eval_spec_function): Accept a soft_matched_part argument,
45 and pass it down to ...
46 (do_spec_2): Accept a soft_matched_part argument, and pass
48 (do_spec_1): Pass soft_matched_part to handle_spec_function.
49 (handle_braces): Update call to handle_spec_function.
50 (driver::set_up_specs): Update calls to do_spec_2.
51 (compare_debug_dump_opt_spec_function): Likewise.
52 (compare_debug_self_opt_spec_function): Likewise.
54 2018-07-31 Olivier Hainque <hainque@adacore.com>
56 * common.opt (nolibc): New option.
57 * doc/invoke.texi (Link Options): Document it.
58 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
59 * config/alpha/linux.h: Likewise.
60 * config/arc/elf.h: Likewise.
61 * config/arm/uclinux-elf.h: Likewise.
62 * config/arm/unknown-elf.h: Likewise.
63 * config/avr/avrlibc.h: Likewise.
64 * config/bfin/bfin.h: Likewise.
65 * config/bfin/linux.h: Likewise.
66 * config/bfin/uclinux.h: Likewise.
67 * config/darwin.h: Likewise.
68 * config/darwin10.h: Likewise.
69 * config/darwin12.h: Likewise.
70 * config/gnu-user.h: Likewise.
71 * config/lm32/uclinux-elf.h: Likewise.
72 * config/pa/pa-hpux11.h: Likewise.
73 * config/pa/pa64-hpux.h: Likewise.
74 * config/sparc/sparc.h: Likewise.
76 2018-07-31 Olivier Hainque <hainque@adacore.com>
78 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
81 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
84 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
87 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
89 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
90 terminated string literal.
92 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
94 PR rtl-optimization/85160
95 * combine.c (is_just_move): New function.
96 (try_combine): Allow combining two instructions into two if neither of
97 the original instructions was a move.
99 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
102 * doc/extend.texi (Global Register Variables): Discourage use of type
104 (Local Register Variables): Likewise.
106 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
108 PR tree-optimization/86506
109 * hwint.h (ceil_log2): Resync with hwint.c implementation.
111 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
114 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
115 hard_regno, make sure no insn between `from` and `to` clobbers it.
117 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
118 Tom de Vries <tdevries@suse.de>
120 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
121 (PTX_DEFAULT_RUNTIME_DIM): ... this.
122 (nvptx_goacc_validate_dims): Set default worker and gang dims to
123 PTX_DEFAULT_RUNTIME_DIM.
124 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
126 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
128 * config/pa/pa.c (pa_output_addr_vec): Align address table.
129 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
130 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
132 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
134 * config/rs6000/constraints.md (wG constraint): Delete, no longer
136 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
137 predicate to reflect toc fusion has been deleted.
138 (toc_fusion_mem_raw): Delete, no longer used.
139 (toc_fusion_mem_wrapped): Likewise.
140 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
142 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
143 Delete, no longer used.
144 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
145 meant to be used for toc fusion.
146 (rs6000_debug_print_mode): Delete toc fusion debugging.
147 (rs6000_debug_reg_global): Likewise.
148 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
149 fusion and secondary reload support that were never used.
150 (rs6000_option_override_internal): Delete TOC fusion, that was only
151 partially defined, and it did not work unless you also used the
153 (rs6000_legitimate_address_p): Delete TOC fusion support.
154 (rs6000_opt_masks): Likewise.
155 (fusion_wrap_memory_address): Delete function, no longer used.
156 (fusion_split_address); Delete TOC fusion support.
157 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
158 longer used with toc fusion being deleted.
159 (TARGET_TOC_FUSION_FP): Likewise.
160 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
162 (toc fusion spliter): Delete TOC fusion support.
163 (toc_fusionload_<mode>): Likewise.
164 (toc_fusionload_di): Likewise.
165 (fusion_gpr_load_<mode>): Delete generator function, this insn no
166 longer needs to be named. Rename predicate to delete TOC fusion.
167 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
168 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
169 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
170 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
171 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
173 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
175 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
176 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
177 __int128 in built-in function prototypes.
178 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
179 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
181 2018-07-27 Martin Sebor <msebor@redhat.com>
183 PR tree-optimization/86696
184 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
185 types, including enums.
186 (handle_char_store): Be prepared for the above function to fail.
188 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
190 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
191 when optimization level is lower than 2 or optimize for size.
193 2018-07-26 Martin Sebor <msebor@redhat.com>
195 PR tree-optimization/86043
196 PR tree-optimization/86042
197 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
198 (get_string_cst_length): Rename...
199 (get_min_string_length): ...to this. Add argument.
200 (handle_char_store): Extend to handle multi-character stores by
202 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
203 * tree.h (initializer_zerop): Add argument.
205 2018-07-26 Jakub Jelinek <jakub@redhat.com>
208 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
209 declare target to variables if they have always,{to,from,tofrom} map
212 2018-07-26 Martin Liska <mliska@suse.cz>
215 * lto-wrapper.c: Add linker_output as prefix
216 for ltrans_output_file.
218 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
220 PR rtl-optimization/85805
221 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
222 value for hard registers if that was written in the same mode.
224 2018-07-26 Martin Liska <mliska@suse.cz>
226 PR gcov-profile/86536
227 * gcov.c (format_gcov): Use printf format %.*f directly
228 and do not handle special values.
230 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
232 * common/config/arc/arc-common.c (arc_option_optimization_table):
233 Update default optimizations for size.
235 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
237 * config/arc/arc.md (movsf_insn): Add short instruction selection.
238 * config/arc/constraints.md (CfZ): New constraint.
239 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
240 (subsf3_fpu): Likewise.
241 (cmpsf_fpu): Likewise.
242 (cmpsf_fpu_uneq): Likewise.
244 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
246 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
248 (arc_is_uncached_mem_p): uncached applies to both the variable and
251 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
253 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
256 2018-07-25 David Malcolm <dmalcolm@redhat.com>
258 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
259 field "m_scopes" from vec to auto_vec.
261 2018-07-25 Martin Liska <mliska@suse.cz>
263 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
266 2018-07-25 Richard Biener <rguenther@suse.de>
269 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
270 special wrt context_die late.
271 (gen_subprogram_die): Re-use DIEs in local scope.
273 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
275 PR tree-optimization/86644
276 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
278 2018-07-25 Martin Liska <mliska@suse.cz>
281 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
282 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
284 2018-07-25 Martin Liska <mliska@suse.cz>
287 * params.def: Explain ASan abbreviation and provide
288 a documentation link.
290 2018-07-24 Martin Sebor <msebor@redhat.com>
292 PR tree-optimization/86622
293 PR tree-optimization/86532
294 * builtins.h (string_length): Declare.
295 * builtins.c (c_strlen): Correct handling of non-constant offsets.
296 (check_access): Be prepared for non-constant length ranges.
297 (string_length): Make extern.
298 * expr.c (string_constant): Only handle the minor non-constant
299 array index. Use string_constant to compute the length of
300 a generic string constant.
302 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
304 PR tree-optimization/86618
305 * tree-vect-stmts.c (vectorizable_call): Don't take the address
306 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
308 2018-07-24 David Malcolm <dmalcolm@redhat.com>
310 PR tree-optimization/86636
311 * json.cc (json::object::set): Fix comment. Add assertions.
312 (json::array::append): Move here from json.h. Add comment and an
314 (json::string::string): Likewise.
315 * json.h (json::array::append): Move to json.cc.
316 (json::string::string): Likewise.
317 * optinfo-emit-json.cc
318 (optrecord_json_writer::impl_location_to_json): Assert that we
319 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
320 wrapper around it. Expand the location once, rather than three
322 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
323 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
325 (optrecord_json_writer::optinfo_to_json): Likewise, in four
326 places. Fix some overlong lines.
328 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
330 * config/aarch64/aarch64-simd.md
331 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
332 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
333 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
334 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
335 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
336 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
337 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
338 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
339 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
341 2018-07-24 Jakub Jelinek <jakub@redhat.com>
344 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
345 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
346 and abs_d == d, do the power of two handling if profitable.
348 2018-07-24 Richard Biener <rguenther@suse.de>
350 * match.pd: Add BIT_FIELD_REF canonicalizations.
352 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
355 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
357 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
359 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
360 terminated STRING_CST object.
362 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
364 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
366 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
368 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
369 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
370 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
371 the elements into a register.
372 (rs6000_split_v4si_init_di_reg): Delete.
373 (rs6000_split_v4si_init): Delete.
374 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
375 (vsx_init_v4si): Rewrite as a define_expand.
377 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
379 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
380 zero_extend argument from memory): New.
382 2018-07-22 Martin Sebor <msebor@redhat.com>
385 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
386 diagnosing calls with unknown arguments unless -Walloca-larger-than
387 is restricted to less than PTRDIFF_MAX bytes.
389 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
391 * doc/gcov.texi (Invoking Gcov): Editorial changes.
393 2018-07-20 David Malcolm <dmalcolm@redhat.com>
395 * pretty-print.c (text_info::set_location): Remove redundant
396 "line_table" parameter from call to rich_location::set_range.
398 2018-07-20 Martin Sebor <msebor@redhat.com>
401 * builtins.c (expand_builtin_alloca): Adjust.
402 * calls.c (alloc_max_size): Simplify.
403 * cgraphunit.c (cgraph_node::expand): Adjust.
404 * common.opt (larger_than_size, warn_frame_larger_than): Remove
406 (frame_larger_than_size): Same.
407 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
408 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
410 * doc/invoke.texi (GCC Command Options): Document option arguments.
411 Explain byte-size arguments and suffixes.
412 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
413 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
414 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
415 * doc/options.texi (UInteger): Expand.
416 (Host_Wide_Int, ByteSize): Document new properties.
417 * final.c (final_start_function_1): Include sizes in an error message.
418 * function.c (frame_offset_overflow): Same.
419 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
420 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
421 Diagnose unbounded alloca calls only for limits of less than
423 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
424 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
426 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
427 only for limits of less than PTRDIFF_MAX.
428 * langhooks-def.h (lhd_handle_option): Change function argument
430 * langhooks.c (lhd_handle_option): Same.
431 * langhooks.h (handle_option): Same.
432 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
434 (var_type, var_type_struct): Same.
435 (var_set): Handle ByteSize flag.
436 * optc-gen.awk: Add comments to output to ease debugging. Make
437 use of HOST_WIDE_INT where appropriate.
438 * opts-gen-save.awk: Use %lx to format unsigned long.
439 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
440 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
441 arguments. Parse bytes-size suffixes.
442 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
443 (enum_value_to_arg): Same.
444 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
445 (handle_option): Adjust.
446 (generate_option): Change function argument to HOST_WIDE_INT.
447 (cmdline_handle_error): Adjust.
448 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
449 (set_option): Change function argument to HOST_WIDE_INT.
450 (option_enabled): Handle cl_host_wide_int.
451 (get_option_state): Handle CLVC_SIZE.
452 (control_warning_option): Same.
453 * opts.c (common_handle_option): Change function argument to
454 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
455 OPT_Wvla_larger_than_.
456 * opts.h (enum cl_var_type): Add an enumerator.
457 * stor-layout.c (layout_decl): Print a more meaningful warning.
458 * toplev.c (output_stack_usage): Adjust.
460 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
462 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
463 call to inline_expand_builtin_string_cmp.
464 (expand_builtin_strcmp): Likewise.
465 (expand_builtin_strncmp): Likewise.
466 (inline_string_cmp): Delete the last parameter, change char_type_node
467 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
469 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
470 the inlining expansion on target where the type of the call has same or
471 narrower precision than unsigned char.
473 2018-07-20 David Malcolm <dmalcolm@redhat.com>
475 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
476 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
477 * common.opt (fsave-optimization-record): New option.
478 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
479 * doc/invoke.texi (-fsave-optimization-record): New option.
480 * dumpfile.c: Include "optinfo-emit-json.h".
481 (struct kv_pair): Move to coretypes.h.
482 (optgroup_options): Make non-static.
483 (dump_context::end_scope): Call
484 optimization_records_maybe_pop_dump_scope.
485 * dumpfile.h (optgroup_options): New decl.
488 * optinfo-emit-json.cc: New file.
489 * optinfo-emit-json.h: New file.
490 * optinfo.cc: Include "optinfo-emit-json.h".
491 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
492 (optinfo_enabled_p): Check optimization_records_enabled_p.
493 (optinfo_wants_inlining_info_p): Likewise.
494 * optinfo.h: Update comment.
495 * profile-count.c (profile_quality_as_string): New function.
496 * profile-count.h (profile_quality_as_string): New decl.
497 (profile_count::quality): New accessor.
498 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
499 and optinfo_emit_json_cc_tests.
500 * selftest.h (selftest::json_cc_tests): New decl.
501 (selftest::optinfo_emit_json_cc_tests): New decl.
502 * toplev.c: Include "optinfo-emit-json.h".
503 (compile_file): Call optimization_records_finish.
504 (do_compile): Call optimization_records_start.
505 * tree-ssa-live.c: Include optinfo.h.
506 (remove_unused_scope_block_p): Retain inlining information if
507 optinfo_wants_inlining_info_p returns true.
509 2018-07-20 Richard Biener <rguenther@suse.de>
512 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
513 to cover -flto-partition=none.
515 2018-07-20 Martin Liska <mliska@suse.cz>
517 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
518 (get_decl_source_range): Remove unused function.
520 2018-07-20 Richard Biener <rguenther@suse.de>
522 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
523 (struct vn_phi_s): Likewise.
524 (struct vn_reference_s): Likewise.
525 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
526 for searching the slot of an entry known to be in the hash itself.
527 (vn_phi_hasher::equal): Likewise.
528 (vn_reference_hasher::equal): Likewise.
529 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
531 (optimistic_info, current_info): Remove, keeping only valid_info.
532 (vn_reference_lookup_1): Remove fallback lookup.
533 (vn_reference_lookup_2): Likewise.
534 (vn_nary_op_lookup_1): Likewise.
535 (vn_phi_lookup): Likewise.
536 (vn_nary_build_or_lookup_1): Make sure to not chain the built
538 (vn_reference_insert): Adjust, chain the inserted hash element
539 at last_inserted_ref.
540 (vn_reference_insert_pieces): Likewise.
541 (visit_reference_op_call): Likewise.
542 (vn_nary_op_insert_into): Chain the inserted hash element at
544 (vn_nary_op_insert_pieces): Adjust.
545 (vn_nary_op_insert): Likewise.
546 (vn_nary_op_insert_stmt): Likewise.
547 (vn_phi_insert): Adjust, chain the inserted hash element at
549 (process_scc): Remove clearing and copying the optimistic
550 table. Instead remove elements inserted during an optimistic
551 iteration from the single table we maintain.
552 (init_scc_vn): Adjust.
553 (free_scc_vn): Likewise.
554 (sccvn_dom_walker::record_cond): Likewise.
555 (sccvn_dom_walker::after_dom_children): Likewise.
557 2018-07-19 Martin Sebor <msebor@redhat.com>
559 PR tree-optimization/84047
560 PR tree-optimization/83776
561 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
562 (check_array_bounds): Call it.
564 2018-07-19 Martin Sebor <msebor@redhat.com>
566 * align.h (align_flags): Use member initialization.
568 2018-07-19 David Malcolm <dmalcolm@redhat.com>
570 * Makefile.in (OBJS): Add optinfo.o.
571 * coretypes.h (class symtab_node): New forward decl.
572 (struct cgraph_node): New forward decl.
573 (class varpool_node): New forward decl.
574 * dump-context.h: New file.
575 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
577 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
578 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
579 (set_alt_dump_file): Likewise.
580 (dump_context::~dump_context): New dtor.
581 (dump_gimple_stmt): Move implementation to...
582 (dump_context::dump_gimple_stmt): ...this new member function.
583 Add the stmt to any pending optinfo, creating one if need be.
584 (dump_gimple_stmt_loc): Move implementation to...
585 (dump_context::dump_gimple_stmt_loc): ...this new member function.
586 Start a new optinfo and add the stmt to it.
587 (dump_gimple_expr): Move implementation to...
588 (dump_context::dump_gimple_expr): ...this new member function.
589 Add the stmt to any pending optinfo, creating one if need be.
590 (dump_gimple_expr_loc): Move implementation to...
591 (dump_context::dump_gimple_expr_loc): ...this new member function.
592 Start a new optinfo and add the stmt to it.
593 (dump_generic_expr): Move implementation to...
594 (dump_context::dump_generic_expr): ...this new member function.
595 Add the tree to any pending optinfo, creating one if need be.
596 (dump_generic_expr_loc): Move implementation to...
597 (dump_context::dump_generic_expr_loc): ...this new member
598 function. Add the tree to any pending optinfo, creating one if
600 (dump_printf): Move implementation to...
601 (dump_context::dump_printf_va): ...this new member function. Add
602 the text to any pending optinfo, creating one if need be.
603 (dump_printf_loc): Move implementation to...
604 (dump_context::dump_printf_loc_va): ...this new member function.
605 Start a new optinfo and add the stmt to it.
606 (dump_dec): Move implementation to...
607 (dump_context::dump_dec): ...this new member function. Add the
608 value to any pending optinfo, creating one if need be.
609 (dump_context::dump_symtab_node): New member function.
610 (dump_context::get_scope_depth): New member function.
611 (dump_context::begin_scope): New member function.
612 (dump_context::end_scope): New member function.
613 (dump_context::ensure_pending_optinfo): New member function.
614 (dump_context::begin_next_optinfo): New member function.
615 (dump_context::end_any_optinfo): New member function.
616 (dump_context::s_current): New global.
617 (dump_context::s_default): New global.
618 (dump_scope_depth): Delete global.
619 (dumpfile_ensure_any_optinfo_are_flushed): New function.
620 (dump_symtab_node): New function.
621 (get_dump_scope_depth): Reimplement in terms of dump_context.
622 (dump_begin_scope): Likewise.
623 (dump_end_scope): Likewise.
624 (selftest::temp_dump_context::temp_dump_context): New ctor.
625 (selftest::temp_dump_context::~temp_dump_context): New dtor.
626 (selftest::verify_item): New function.
627 (ASSERT_IS_TEXT): New macro.
628 (ASSERT_IS_TREE): New macro.
629 (ASSERT_IS_GIMPLE): New macro.
630 (selftest::test_capture_of_dump_calls): New test.
631 (selftest::dumpfile_c_tests): Call it.
632 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
633 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
634 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
636 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
637 (dump_node, dump_bb): Move these unrelated decls.
638 (class dump_manager): Add leading comment.
639 * optinfo.cc: New file.
640 * optinfo.h: New file.
642 2018-07-19 Michael Collison <michael.collison@arm.com>
643 Richard Henderson <rth@redhat.com>
645 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
646 (subti): Handle op1 zero.
647 (subvti4, usub4ti4): New.
648 (*sub<GPI>3_compare1_imm): New.
649 (sub<GPI>3_carryinCV): New.
650 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
651 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
653 2018-07-19 Michael Collison <michael.collison@arm.com>
654 Richard Henderson <rth@redhat.com>
656 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
657 (addti3): Create simpler code if low part is already known to be 0.
658 (addvti4, uaddvti4): New.
659 (*add<GPI>3_compareC_cconly_imm): New.
660 (*add<GPI>3_compareC_cconly): New.
661 (*add<GPI>3_compareC_imm): New.
662 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
663 handle constants within this pattern..
664 (*add<GPI>3_compareV_cconly_imm): New.
665 (*add<GPI>3_compareV_cconly): New.
666 (*add<GPI>3_compareV_imm): New.
667 (add<GPI>3_compareV): New.
668 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
669 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
670 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
671 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
672 with 'comparison' operator.
673 (*add<GPI>3_compareV_cconly_imm): Ditto.
674 (*add<GPI>3_compareV_cconly): Ditto.
675 (*add<GPI>3_compareV_imm): Ditto.
676 (add<GPI>3_compareV): Ditto.
677 (add<mode>3_carryinC): Ditto.
678 (*add<mode>3_carryinC_zero): Ditto.
679 (*add<mode>3_carryinC): Ditto.
680 (add<mode>3_carryinV): Ditto.
681 (*add<mode>3_carryinV_zero): Ditto.
682 (*add<mode>3_carryinV): Ditto.
684 2018-07-19 Michael Collison <michael.collison@arm.com>
685 Richard Henderson <rth@redhat.com>
687 * config/aarch64/aarch64-modes.def (CC_V): New.
688 * config/aarch64/aarch64-protos.h
689 (aarch64_addti_scratch_regs): Declare
690 (aarch64_subvti_scratch_regs): Declare.
691 (aarch64_expand_subvti): Declare.
692 (aarch64_gen_unlikely_cbranch): Declare
693 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
694 for signed overflow using CC_Vmode.
695 (aarch64_get_condition_code_1): Handle CC_Vmode.
696 (aarch64_gen_unlikely_cbranch): New function.
697 (aarch64_addti_scratch_regs): New function.
698 (aarch64_subvti_scratch_regs): New function.
699 (aarch64_expand_subvti): New function.
701 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
703 * config/aarch64/aarch64-option-extensions.def: New entry for profile
705 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
706 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
709 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
712 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
713 address check not strict.
715 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
717 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
719 (store_pair_lanes<mode>): Likewise.
720 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
721 enum value 'ADDR_QUERY_LDP_STP_N'.
722 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
723 (aarch64_print_address_internal): Add declaration.
724 (aarch64_print_ldpstp_address): Remove.
725 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
726 (aarch64_print_operand): Change printing of 'y'.
727 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
728 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
729 'true' rather than '1'.
730 * config/aarch64/constraints.md (Uml): Likewise.
731 (Uml): Rename to Umn.
734 2018-07-19 Richard Biener <rguenther@suse.de>
736 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
738 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
739 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
740 (vn_reference_hasher): Likewise.
741 (struct vn_tables_s): Remove obstack and alloc-pool members.
742 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
743 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
744 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
745 (vn_reference_insert_pieces): Likewise.
746 (alloc_vn_nary_op_noinit): Adjust.
747 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
749 (shared_lookup_phiargs): Remove.
750 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
751 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
752 (visit_reference_op_call): Likewise.
753 (copy_nary, copy_phi, copy_reference): Remove.
754 (process_scc): Rewind the obstack when iterating. Do not
755 copy the elements to valid_info but just move them from one
756 hashtable to the other.
757 (allocate_vn_table): Adjust.
758 (free_vn_table): Likewise.
759 (init_scc_vn): Likewise.
760 (free_scc_vn): Likewise.
762 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
765 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
766 indirect_return as function type attribute.
767 (ix86_attribute_table): Change indirect_return to function
769 * doc/extend.texi: Update indirect_return attribute.
771 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
773 * wide-int.h (widest2_int): New.
774 * gimple-fold.c (arith_overflowed_p): Use it.
775 * tree.h (widest2_int_cst): New.
776 * tree-vrp.c (wide_int_binop_overflow): Rename from
778 Rewrite to work on trees.
779 (extract_range_from_multiplicative_op_1): Abstract code to...
780 (wide_int_range_min_max): ...here.
781 (wide_int_range_cross_product): ...and here.
782 (extract_range_from_binary_expr_1): Abstract overflow code to...
783 (wide_int_range_mult_wrapping): ...here.
784 * tree-vrp.h (wide_int_range_cross_product): New.
785 (wide_int_range_mult_wrapping): New.
787 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
788 Julia Koval <julia.koval@intel.com>
790 * config/i386/x86-tune-costs.h (skylake_memcpy,
791 skylake_memset): Replace rep_prefix with unrolling for size 512.
793 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
796 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle comparision with EQ_EXPR
799 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
801 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
802 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
803 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
804 previously known as "PowerPC AltiVec Built-in Functions". Move
805 some material to new subsubsections "PowerPC AltiVec Built-in
806 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
808 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
809 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
810 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
811 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
813 2018-07-18 Richard Biener <rguenther@suse.de>
815 PR tree-optimization/86557
816 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
819 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
821 * config/s390/s390.c (s390_function_profiler): Generate CFI.
823 2018-07-17 Jeff Law <law@redhat.com>
825 * config/arm/arm.c (get_label_padding): Update for recent
826 changes to label_to_alignment.
828 PR tree-optimization/86010
829 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
831 * config/mips/mips.c (vr4130_align_insns): Update for recent
832 changes to label_to_alignment.
834 * config/frv/frv.c (frv_label_align): Update for recent changes
835 to label_to_alignment.
837 * config/nios2/nios2.c (nios2_label_align): Update for recent
838 changes which dropped ALIGN_LABELS_LOG.
840 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
842 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
843 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
845 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
847 * config/arc/arc.c (arc_label_align): Use align_labels instead of
848 deprecated align_labels_log.
850 2018-07-17 Richard Biener <rguenther@suse.de>
853 * dwarf2out.c (init_sections_and_labels): Always generate
854 a debug_line_str_section for early LTO debug.
855 (dwarf2out_finish): Reset debug_line_str_hash output early.
856 Bump counter for extra dwarf5 .debug_loc labels to not conflict
858 (dwarf2out_early_finish): Output debug_line_str.
860 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
862 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
863 index register on z196 or later.
865 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
867 * config/s390/s390.c (s390_default_align): Set default function
869 (s390_override_options_after_change): Call s390_default align.
870 (s390_option_override_internal): Call s390_default_align.
871 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
873 2018-07-17 Jakub Jelinek <jakub@redhat.com>
876 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
877 to _looptemp_ clauses, other than the first two.
879 2018-07-17 Martin Liska <mliska@suse.cz>
881 * opts.c: Do not enable OPT_falign_* for -Os.
883 2018-07-17 Martin Liska <mliska@suse.cz>
885 * align.h (MAX_CODE_ALIGN): New.
886 (MAX_CODE_ALIGN_VALUE): New.
887 * common/config/i386/i386-common.c (ix86_handle_option):
888 (MAX_CODE_ALIGN): Moved to align.h.
889 * final.c (MAX_CODE_ALIGN): Likewise.
890 * opts.c (parse_and_check_align_values):
891 (MAX_CODE_ALIGN): Likewise.
892 (MAX_CODE_ALIGN_VALUE): Likewise.
894 2018-07-17 Martin Liska <mliska@suse.cz>
896 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
897 in order to fulfil coding style.
898 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
899 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
900 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
901 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
902 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
903 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
904 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
905 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
906 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
908 2018-07-17 Martin Liska <mliska@suse.cz>
911 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
912 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
913 align_flags of label_to_alignment.
914 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
916 * config/m68k/m68k.c: Do not use removed align_labels_value and
918 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
919 (LOOP_ALIGN): Likewise.
920 (LABEL_ALIGN): Likewise.
921 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
922 Remove not used macro.
923 (rs6000_loop_align): Change return type to align_flags.
924 (rs6000_loop_align_max_skip): Remove.
925 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
926 Change return type to align_flags.
927 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
928 Remove not used macro.
929 (rs6000_loop_align): Change return type to align_flags.
930 (rs6000_loop_align_max_skip): Remove.
931 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
932 * config/rx/rx-protos.h (rx_align_for_label): Make it
934 * config/rx/rx.c (rx_align_for_label): Change return type
936 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
938 into align_flags class.
939 (LABEL_ALIGN): Likewise.
940 (LOOP_ALIGN): Likewise.
941 * config/s390/s390.c (s390_label_align): Use align_flags
943 (s390_asm_output_function_label): Likewise.
944 * config/sh/sh.c (sh_override_options_after_change):
945 Use align_flags class directly without macros.
946 (find_barrier): Likewise.
947 (barrier_align): Likewise.
948 (sh_loop_align): Likewise.
949 * config/spu/spu.c (spu_option_override):
950 Use align_flags_tuple::get_value instead of removed macros.
951 (spu_sched_init): Likewise.
952 * config/spu/spu.h (GTY): Likewise.
953 * config/visium/visium.c (visium_option_override):
954 Set "8" as default secondary alignment.
955 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
956 in order to guarantee secondary alignment of 8.
957 * coretypes.h: Include align.h header file.
958 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
959 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
960 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
961 * doc/tm.texi.in: Likewise.
962 * final.c (struct label_alignment): Remove not used structure.
963 (LABEL_ALIGN): Change type to align_flags.
964 (LOOP_ALIGN): Likewise.
965 (JUMP_ALIGN): Likewise.
966 (default_loop_align_max_skip): Remove.
967 (default_label_align_max_skip): Likewise.
968 (default_jump_align_max_skip): Likewise.
969 (default_label_align_after_barrier_max_skip):
970 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
971 (LABEL_TO_MAX_SKIP): Remove.
972 (label_to_alignment): Return align_flags type instead of integer.
973 (label_to_max_skip): Remove.
974 (align_fuzz): Use align_flags type.
975 (compute_alignments): Use align_flags type and use align_flags::max
976 to combine multiple alignments.
977 (grow_label_align): Grow vec instead of C array.
978 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
979 (shorten_branches): Use align_flags type and use align_flags::max
980 to combine multiple alignments.
981 (final_scan_insn_1): Remove usage of secondary alignment that comes
982 from label alignment, but instead use proper secondary alignment
983 which is computed in grow_label_align.
984 * flags.h (struct align_flags_tuple): Move to align.h.
985 (struct align_flags): Likewise.
986 (state_align_loops): Rename to align_loops.
987 (state_align_jumps): Rename to align_jumps.
988 (state_align_labels): Rename to align_labels.
989 (state_align_functions): Rename to align_functions.
990 (align_loops_log): Remove.
991 (align_jumps_log): Remove.
992 (align_labels_log): Remove.
993 (align_functions_log): Remove.
994 (align_loops_max_skip): Remove.
995 (align_jumps_max_skip): Remove.
996 (align_labels_max_skip): Remove.
997 (align_functions_max_skip): Remove.
998 (align_loops_value): Remove.
999 (align_jumps_value): Remove.
1000 (align_labels_value): Remove.
1001 (align_functions_value): Remove.
1002 * output.h (label_to_alignment): Change return type to align_flags.
1003 (label_to_max_skip): Remove.
1004 * target.def: Remove loop_align_max_skip, label_align_max_skip,
1005 jump_align_max_skip macros.
1006 * targhooks.h (default_loop_align_max_skip): Remove.
1007 (default_label_align_max_skip): Likewise.
1008 (default_jump_align_max_skip): Likewise.
1009 (default_label_align_after_barrier_max_skip): Remove.
1010 * toplev.c (read_log_maxskip): Use ::normalize function.
1011 (parse_N_M): Remove not used argument and also call ::normalize.
1012 (parse_alignment_opts): Do not pass unused arguments.
1013 * varasm.c (assemble_start_function): Use directly align_functions
1014 instead of removed macros.
1015 * system.h: Do not poison removed macros.
1017 2018-07-17 Jakub Jelinek <jakub@redhat.com>
1020 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
1021 and cond temporaries don't have reference type if iterator has
1022 pointer type. For init use &for_pre_body instead of pre_p if
1023 for_pre_body is non-empty.
1025 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1027 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
1028 double-double modes to SFmode directly directly.
1029 (trunc<mode>sf2_fprs): Delete.
1031 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1033 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
1034 for conversions between IFmode and the decimal floating point modes.
1035 (init_float128_ieee): Use the correct names for conversions between
1036 KFmode and the decimal floating point modes.
1038 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1040 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
1041 for the conversions between TDmode and IFmode.
1042 (init_float128_ieee): Use more correct names for the conversions
1043 between TDmode and KFmode.
1045 2018-07-16 Jakub Jelinek <jakub@redhat.com>
1047 PR tree-optimization/86526
1048 * builtins.c (expand_builtin_memcmp): Formatting fixes.
1049 (inline_expand_builtin_string_cmp): Likewise.
1050 (inline_string_cmp): Likewise. Use c_readstr instead of
1051 builtin_memcpy_read_str. Add unit_mode temporary.
1053 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
1056 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
1057 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
1059 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
1061 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
1062 Alphabetize prototypes of built-in functions, separating out
1063 built-in functions that are listed in this section but should be
1064 described elsewhere.
1066 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
1069 * expmed.c (emit_store_flag): Do not emit setcc followed by a
1070 conditional move when trapping comparison was split to a
1071 non-trapping one (and vice versa).
1073 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1075 * config/s390/s390.c (s390_function_profiler): Generate nops
1076 instead of profiler call sequences.
1077 * config/s390/s390.opt: Add the new option.
1079 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1081 * config/s390/s390.c (s390_function_profiler): Generate
1082 __mcount_loc section.
1083 * config/s390/s390.opt: Add the new option.
1085 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1087 * common.opt: Add the new warning.
1088 * config/s390/s390.c (s390_function_profiler): Emit "brasl
1089 %r0,__fentry__" when -mfentry is specified.
1090 (s390_option_override_internal): Disallow -mfentry for 31-bit
1092 * config/s390/s390.opt: Add the new option.
1094 2018-07-16 Richard Biener <rguenther@suse.de>
1097 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
1098 for function-local FUNCTION_DECL and RESULT_DECL immediately.
1100 2018-07-16 Martin Liska <mliska@suse.cz>
1103 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
1106 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
1108 * config/arc/arcHS.md: Update ARCHS scheduling rules.
1110 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
1112 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
1114 * config/arc/arc-cpus.def (hs4x): New cpu.
1116 * config/arc/arc-tables.opt: Regenerate.
1117 * config/arc/arc.c (arc_sched_issue_rate): New function.
1118 (TARGET_SCHED_ISSUE_RATE): Define.
1119 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
1120 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
1122 (attr tune): Add ARCHS4x tune values.
1123 (attr tune_dspmpy): Define.
1124 (*tst): Correct instruction type.
1125 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
1126 * config/arc/arcHS4x.md: New file.
1127 * config/arc/fpu.md: Update instruction type attributes.
1128 * config/arc/t-multilib: Regenerate.
1130 2018-07-16 Tom de Vries <tdevries@suse.de>
1133 * var-tracking.c (vt_initialize): Fix pre_dec handling.
1135 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1137 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
1140 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
1142 PR tree-optimization/86514
1143 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
1144 conversion to a boolean type from a type with greater precision.
1146 2018-07-16 Tom de Vries <tdevries@suse.de>
1148 * var-tracking.c (vt_initialize): Print adjusted insn slim if
1149 dump_flags request TDF_SLIM.
1151 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
1153 * fold-const.c (int_const_binop_1): Abstract...
1154 (wide_int_binop): ...wide int code here.
1155 (poly_int_binop): ...poly int code here.
1156 Abstract the rest of int_const_binop_1 into int_const_binop.
1157 * fold-const.h (wide_int_binop): New.
1158 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
1159 Remove useless PLUS/MINUS_EXPR case.
1160 (zero_nonzero_bits_from_vr): Move wide int code...
1161 (zero_nonzero_bits_from_bounds): ...here.
1162 (extract_range_from_binary_expr_1): Move mask optimization code...
1163 (range_easy_mask_min_max): ...here.
1164 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
1165 (range_easy_mask_min_max): New.
1167 2018-07-15 Jeff Law <law@redhat.com>
1170 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
1173 2018-07-14 Jim Wilson <jimw@sifive.com>
1175 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
1177 2018-07-14 Paul Koning <ni1d@arrl.net>
1179 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
1181 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
1183 * lto-streamer-out.c (copy_function_or_variable): Dump info about
1186 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
1187 Steve Munroe <munroesj52@gmail.com>
1189 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
1190 (_mm_andnot_si128): Likewise.
1191 (_mm_or_si128): Likewise.
1192 (_mm_xor_si128): Likewise.
1194 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
1197 * builtins.c (expand_builtin_memcmp): Inline the calls first
1198 when result_eq is false.
1199 (expand_builtin_strcmp): Inline the calls first.
1200 (expand_builtin_strncmp): Likewise.
1201 (inline_string_cmp): New routine. Expand a string compare
1202 call by using a sequence of char comparison.
1203 (inline_expand_builtin_string_cmp): New routine. Inline expansion
1204 a call to str(n)cmp/memcmp.
1205 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
1207 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
1209 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
1211 * config/arm/driver-arm.c: Include arm-native.h.
1212 (host_detect_local_cpu): Use auto-generated data tables.
1213 (vendors, arm_cpu_table): Delete. Move part information to ...
1214 * config/arm/arm-cpus.in: ... here.
1215 * config/arm/parsecpu.awk (gen_native): New function.
1216 (vendor, part): New CPU fields.
1217 (END): Add support for building the native CPU detection tables.
1218 * config/arm/t-arm (arm-native.h): Add build rule.
1219 (driver-arm.o): Add dependency on arm-native.h.
1221 2018-07-13 Richard Biener <rguenther@suse.de>
1224 * match.pd (addr1 - addr2): Allow either of the operand to
1227 2018-07-13 Tom de Vries <tdevries@suse.de>
1229 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
1232 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
1234 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
1235 arrays instead of numbered variables.
1237 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
1239 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
1240 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
1242 2018-07-13 Richard Biener <rguenther@suse.de>
1245 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
1246 instead of get_context_die.
1248 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
1249 Richard Biener <rguenther@suse.de>
1252 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
1253 that the loop latch destination where phi is defined.
1255 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
1257 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
1258 (riscv_expand_epilogue): Add assertion to check interrupt mode.
1259 (riscv_set_current_function): Extract getting interrupt type to new
1261 (riscv_get_interrupt_type): New function.
1262 (riscv_merge_decl_attributes): New function, checking interrupt type is
1264 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
1266 2018-07-12 Paul Koning <ni1d@arrl.net>
1268 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
1271 2018-07-12 Paul Koning <ni1d@arrl.net>
1273 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
1274 zero reference, add doloop_end instead.
1275 * doc/md.texi (decrement_and_branch_until_zero): Remove.
1276 (Looping patterns): Remove decrement_and_branch_until_zero. Add
1277 detail for doloop_end.
1279 2018-07-12 Martin Sebor <msebor@redhat.com>
1282 * attribs.c (decl_attributes): Reject conflicting attributes before
1283 calling attribute handlers.
1285 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
1287 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
1289 (gcc::dump_manager::get_dump_file_name): likewise.
1290 (dump_begin): Likewise.
1291 * dumpfile.h (dump_begin): Update prototype.
1292 (gcc::dump_manager::get_dump_file_name,
1293 gcc::dump_manager::get_dump_file_name): Update prototype.
1295 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1297 * internal-fn.h (vectorizable_internal_fn_p): New function.
1298 * tree-vect-slp.c (compatible_calls_p): Likewise.
1299 (vect_build_slp_tree_1): Remove nops argument. Handle calls
1300 to internal functions.
1301 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
1303 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1305 * fold-const.h (inverse_conditions_p): Declare.
1306 * fold-const.c (inverse_conditions_p): New function.
1307 * match.pd: Use inverse_conditions_p. Add folds of view_converts
1308 that test the inverse condition of a conditional internal function.
1309 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
1310 * internal-fn.c (internal_fn_mask_index): Handle conditional
1312 (vectorized_internal_fn_supported_p): New function.
1313 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
1314 (any_pred_load_store): Replace with...
1315 (need_to_predicate): ...this new variable.
1316 (redundant_ssa_names): New variable.
1317 (ifcvt_can_use_mask_load_store): Move initial checks to...
1318 (ifcvt_can_predicate): ...this new function. Handle tree codes
1319 for which a conditional internal function exists.
1320 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
1321 instead of ifcvt_can_use_mask_load_store. Update after variable
1323 (predicate_load_or_store): New function, split out from
1324 predicate_mem_writes.
1325 (check_redundant_cond_expr): New function.
1326 (value_available_p): Likewise.
1327 (predicate_rhs_code): Likewise.
1328 (predicate_mem_writes): Rename to...
1329 (predicate_statements): ...this. Use predicate_load_or_store
1330 and predicate_rhs_code.
1331 (combine_blocks, tree_if_conversion): Update after above name changes.
1332 (ifcvt_local_dce): Handle redundant_ssa_names.
1333 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
1334 general conditional functions.
1335 * tree-vect-stmts.c (vectorizable_call): Likewise.
1337 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1338 Alan Hayward <alan.hayward@arm.com>
1339 David Sherwood <david.sherwood@arm.com>
1341 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
1342 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
1343 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
1344 plus and minus and convert them into IFN_COND_FMA-based sequences.
1345 (convert_mult_to_fma): Handle conditional plus and minus.
1347 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1349 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
1350 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
1351 (cond_fnms_optab): New optabs.
1352 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
1354 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
1355 * internal-fn.h (get_conditional_internal_fn): Declare.
1356 (get_unconditional_internal_fn): Likewise.
1357 * internal-fn.c (cond_ternary_direct): New macro.
1358 (expand_cond_ternary_optab_fn): Likewise.
1359 (direct_cond_ternary_optab_supported_p): Likewise.
1360 (FOR_EACH_COND_FN_PAIR): Likewise.
1361 (get_conditional_internal_fn): New function.
1362 (get_unconditional_internal_fn): Likewise.
1363 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
1364 (gimple_match_op::gimple_match_op): Add a new overload for 5
1366 (gimple_match_op::set_op): Likewise.
1367 (gimple_resimplify5): Declare.
1368 * genmatch.c (decision_tree::gen): Generate simplifications for
1370 * gimple-match-head.c (gimple_simplify): Define an overload for
1371 5 operands. Handle calls with 5 arguments in the top-level overload.
1372 (convert_conditional_op): Handle conversions from unconditional
1373 internal functions to conditional ones.
1374 (gimple_resimplify5): New function.
1375 (build_call_internal): Pass a fifth operand.
1376 (maybe_push_res_to_seq): Likewise.
1377 (try_conditional_simplification): Try converting conditional
1378 internal functions to unconditional internal functions.
1379 Handle 3-operand unconditional forms.
1380 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
1381 Define ternary equivalents of the current rules for binary conditional
1383 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
1385 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
1386 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
1387 (optab): Handle them.
1388 (SVE_COND_FP_TERNARY): New int iterator.
1389 (sve_fmla_op, sve_fmad_op): New int attributes.
1390 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
1391 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
1392 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
1394 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1396 * target.def (preferred_else_value): New target hook.
1397 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
1398 * doc/tm.texi: Regenerate.
1399 * targhooks.h (default_preferred_else_value): Declare.
1400 * targhooks.c (default_preferred_else_value): New function.
1401 * internal-fn.h (conditional_internal_fn_code): Declare.
1402 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
1403 (get_conditional_internal_fn): Use it.
1404 (conditional_internal_fn_code): New function.
1405 * gimple-match.h (gimple_match_cond): New struct.
1406 (gimple_match_op): Add a cond member function.
1407 (gimple_match_op::gimple_match_op): Update all forms to take a
1409 * genmatch.c (expr::gen_transform): Use the same condition as res_op
1410 for the suboperation, but don't specify a particular else_value.
1411 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
1412 (visit_nary_op, visit_reference_op_load): Pass
1413 gimple_match_cond::UNCOND to the gimple_match_op constructor.
1414 * gimple-match-head.c: Include tree-eh.h
1415 (convert_conditional_op): New function.
1416 (maybe_resimplify_conditional_op): Likewise.
1417 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
1418 (gimple_resimplify2): Likewise.
1419 (gimple_resimplify3): Likewise.
1420 (gimple_resimplify4): Likewise.
1421 (maybe_push_res_to_seq): Return null for conditional operations.
1422 (try_conditional_simplification): New function.
1423 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
1425 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
1427 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
1429 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
1431 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
1433 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1435 (hash_tree): Do not hash DECL_FCONTEXT
1436 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
1437 Do not stream DECL_FCONTEXT.
1438 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
1439 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
1441 2018-07-12 Richard Biener <rguenther@suse.de>
1444 * dwarf2out.c (gen_block_die): Only output blocks when they have
1445 at least one !DECL_IGNORED_P variable.
1447 2018-07-12 Richard Biener <rguenther@suse.de>
1450 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
1451 Remove -mieee-fp handling.
1453 2018-07-12 Richard Biener <rguenther@suse.de>
1455 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
1456 left-over from last patch.
1458 2018-07-12 Jakub Jelinek <jakub@redhat.com>
1460 PR tree-optimization/86492
1461 * gimple-ssa-store-merging.c
1462 (imm_store_chain_info::coalesce_immediate_stores): Call
1463 check_no_overlap even for the merge_overlapping case. Formatting fix.
1465 2018-07-12 Richard Biener <rguenther@suse.de>
1468 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
1469 move possibly trapping operations into the conditional.
1471 2018-07-12 Richard Biener <rguenther@suse.de>
1473 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
1474 (vn_lookup_simplify_result): Remove recursion limit applied
1476 (vn_nary_build_or_lookup_1): Adjust.
1477 (try_to_simplify): Likewise.
1478 * gimple-match-head.c (gimple_resimplify1): Instead apply one
1480 (gimple_resimplify2): Likewise.
1481 (gimple_resimplify3): Likewise.
1482 (gimple_resimplify4): Likewise.
1484 2018-07-11 Jakub Jelinek <jakub@redhat.com>
1486 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
1487 Use __mmask64 type instead of __mmask8 for __M argument.
1488 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
1489 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
1491 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
1492 __mmask16 for __M argument.
1493 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
1494 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
1495 to __mmask16 instead of __mmask8.
1496 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
1497 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
1498 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
1499 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
1500 instead of __mmask16 for __U argument.
1501 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
1502 __mmask16 instead of __mmask8 for __U argument.
1503 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
1505 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
1507 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
1509 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
1511 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
1513 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
1515 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
1517 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
1519 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
1521 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
1523 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
1524 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
1525 return type as well as __M argument type and all casts from __mmask8
1527 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
1528 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
1529 return type as well as __M argument type and all casts from __mmask8
1531 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
1532 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
1533 return type as well as __M argument type and all casts from __mmask8
1535 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
1536 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
1537 return type as well as __M argument type and all casts from __mmask8
1539 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
1540 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
1543 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
1545 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
1546 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
1549 2018-07-11 Paul Koning <ni1d@arrl.net>
1551 * doc/md.texi (define_subst): Document how multiple occurrences of
1552 the same argument in the replacement pattern are handled.
1554 2018-07-11 Paul Koning <ni1d@arrl.net>
1556 * doc/extend.texi (Common Variable Attributes): Move "mode" into
1558 (Common Type Attributes): Add "mode" attribute.
1560 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
1562 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
1563 stream DECL_ORIGINAL_TYPE.
1564 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
1565 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
1566 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
1567 Do not walk original type.
1568 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
1570 (write_ts_decl_non_common_tree_pointers): Do not stream
1572 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
1573 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
1575 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
1577 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
1578 thread twice from the same starting edge.
1580 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
1582 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
1583 * gimple.c (gimple_call_nonnull_result_p): ...here...
1584 (gimple_call_nonnull_arg): ...and here.
1585 * gimple.h (gimple_call_nonnull_result_p): New.
1586 (gimple_call_nonnull_arg): New.
1588 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
1590 * config/arm/arm-cpus.in: Move information from fpu field of each
1591 cpu definition to the isa field.
1592 * config/arm/parsecpu.awk (fpu): Delete match rule.
1593 (gen_comm_data): Don't add bits from the CPU's FPU entry.
1595 2018-07-11 Richard Biener <rguenther@suse.de>
1598 * dwarf2out.c (init_sections_and_labels): Use
1599 output_asm_line_debug_info consistently.
1600 (dwarf2out_early_finish): Likewise.
1601 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
1604 2018-07-11 Richard Biener <rguenther@suse.de>
1606 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
1607 Rework father_bb setting in a way to avoid propagating constants
1608 multiple times on a loop body.
1610 2018-07-10 Mark Wielaard <mark@klomp.org>
1613 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
1615 2018-07-10 Richard Biener <rguenther@suse.de>
1617 * hash-map.h (hash_map::iterator::operator*): Return
1618 references to key and value.
1620 2018-07-10 Jakub Jelinek <jakub@redhat.com>
1623 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
1624 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
1625 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
1626 (gimplify_omp_for): For composite loops, move outer
1627 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
1628 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
1629 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
1630 TREE_LIST for both the original class iterator and the "last" helper
1631 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
1632 loop, remember has_decl_expr from outer composite loops for the
1633 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
1635 2018-07-09 Martin Sebor <msebor@redhat.com>
1639 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
1640 accessing implicitly initialized array elements.
1641 * expr.c (string_constant): Handle string initializers of
1642 character arrays within aggregates.
1643 * gimple-fold.c (fold_array_ctor_reference): Add argument.
1644 Store element offset. As a special case, handle zero size.
1645 (fold_nonarray_ctor_reference): Same.
1646 (fold_ctor_reference): Add argument. Store subobject offset.
1647 * gimple-fold.h (fold_ctor_reference): Add argument.
1649 2018-07-09 Paul Koning <ni1d@arrl.net>
1651 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
1652 (pdp11_insn_cost): New function.
1653 (pdp11_md_asm_adjust): New function.
1654 (TARGET_INVALID_WITHIN_DOLOOP): Define.
1655 (pdp11_rtx_costs): Update to match machine better.
1656 (output_addr_const_pdp11): Correct format mismatch warnings.
1657 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
1658 * config/pdp11/pdp11.md: General change to add base_cost and/or
1659 length attributes for use by new pdp11_insn_cost function.
1660 (MIN_BRANCH): Correct definition.
1662 (doloop_end): Use standard pattern name for looping pattern.
1663 (doloop_end_nocc): New.
1664 (movsf): Add another constraint alternative.
1665 (zero_extendqihi2): Add constraint alternatives for not in place
1667 (zero_extendhisi2): Remove.
1668 (shift patterns): Add CC handling variants.
1672 (define_peephole2): New peephole to recognize mov that sets CC for
1675 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1677 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
1678 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
1681 2018-07-09 Jakub Jelinek <jakub@redhat.com>
1684 * real.c (real_nextafter): Return true if result is denormal.
1686 2018-07-09 Martin Liska <mliska@suse.cz>
1688 * common.opt: Add back wrongly removed attribute.
1690 2018-07-09 Richard Biener <rguenther@suse.de>
1693 * dwarf2out.c (gen_block_die): For an early generated DIE
1694 always output high/low PC attributes.
1696 2018-07-09 Tom de Vries <tdevries@suse.de>
1698 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
1699 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
1700 onto VAR_DECL with abstract origin.
1702 2018-07-07 Jim Wilson <jimw@sifive.com>
1704 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
1706 2018-07-07 Tom de Vries <tdevries@suse.de>
1708 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
1710 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
1712 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
1714 (combine_bound): Use wide-int overflow calculation instead of
1716 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
1718 * fold-const.c (int_const_binop_2): Same.
1719 (extract_muldiv_1): Same.
1720 (fold_div_compare): Same.
1721 (fold_abs_const): Same.
1723 * poly-int.h (add): Same.
1727 * predict.c (predict_iv_comparison): Same.
1728 * profile-count.c (slow_safe_scale_64bit): Same.
1729 * simplify-rtx.c (simplify_const_binary_operation): Same.
1730 * tree-chrec.c (tree_fold_binomial): Same.
1731 * tree-data-ref.c (split_constant_offset_1): Same.
1732 * tree-if-conv.c (idx_within_array_bound): Same.
1733 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
1734 * tree-ssa-phiopt.c (minmax_replacement): Same.
1735 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
1736 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
1737 * vr-values.c (vr_values::adjust_range_with_scev): Same.
1738 * wide-int.cc (wi::add_large): Same.
1739 (wi::mul_internal): Same.
1740 (wi::sub_large): Same.
1741 (wi::divmod_internal): Same.
1742 * wide-int.h: Change overflow type to overflow_type for neg, add,
1743 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
1744 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
1745 mul_internal, divmod_internal.
1746 (overflow_type): New enum.
1747 (accumulate_overflow): New.
1749 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
1751 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
1752 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
1754 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
1756 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
1757 argument is checked for zero before entering loop, avoid checking again.
1759 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
1761 * gimplify.h (generic_expr_could_trap_p): Set as global function.
1762 * gimplify.h (generic_expr_could_trap_p): Likwise.
1763 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
1765 2018-07-06 Jakub Jelinek <jakub@redhat.com>
1767 PR tree-optimization/86401
1768 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
1769 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
1770 (fold_bit_and_mask): ... here. New helper function for match.pd.
1771 * fold-const.h (fold_bit_and_mask): Declare.
1772 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
1774 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
1777 * target.def (translate_mode_attribute): New hook.
1778 * targhooks.h (default_translate_mode_attribute): Declare.
1779 * targhooks.c (default_translate_mode_attribute): New function.
1780 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
1781 * doc/tm.texi: Regenerate.
1782 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
1783 (rs6000_translate_mode_attribute): New function.
1785 2018-07-06 Paul Koning <ni1d@arrl.net>
1787 * doc/md.texi (define_split): Document DONE and FAIL.
1788 (define_peephole2): Ditto.
1790 2018-07-05 Jeff Law <law@redhat.com>
1792 PR tree-optimization/86010
1793 * tree-ssa-dse.c (compute_trims): More aggressively trim at
1794 both the head and tail of mem* and str* calls.
1796 2018-07-05 Jim Wilson <jimw@sifive.com>
1798 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
1800 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
1802 * config/aarch64/aarch64-simd.md: correct flags text for
1803 MIN_EXPR replacement.
1805 2018-07-05 James Clarke <jrtc27@jrtc27.com>
1807 * configure: Regenerated.
1809 2018-07-05 Carl Love <cel@us.ibm.com>
1811 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
1812 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
1813 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
1814 VSX_BUILTIN_DOUBLEL_V4SF.
1816 2018-07-05 Martin Sebor <msebor@redhat.com>
1819 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
1820 than its domain to compute its the upper bound of a char array.
1822 2018-07-05 Nathan Sidwell <nathan@acm.org>
1824 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
1825 * doc/cpp.texi: Update comment.
1826 * doc/tm.texi: Rebuilt.
1827 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
1828 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
1829 * doc/extend.texi (Backwards Compatibility): Clarify it is system
1830 headers affected by extern "C".
1831 * system.h: Poison NO_IMPLICIT_EXTERN_C.
1832 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
1833 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
1834 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
1835 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
1836 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
1837 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
1838 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
1839 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
1840 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
1841 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
1842 config/sparc/sp64-elf.h, config/spu/spu.h,
1843 config/stormy16/stormy16.h, config/v850/v850.h,
1844 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
1845 define NO_IMPLICIT_EXTERN_C.
1846 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
1848 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1851 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1852 instead of GET_MODE_SIZE when comparing Units.
1854 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1857 * rtlanal.c (set_noop_p): Constrain on mode change,
1858 include hard-reg-set.h
1860 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1862 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
1864 2018-07-05 Jakub Jelinek <jakub@redhat.com>
1867 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
1870 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
1872 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
1874 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
1877 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
1879 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
1881 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
1884 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
1886 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
1888 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
1889 neg_*_op* variables.
1891 2018-07-04 Martin Liska <mliska@suse.cz>
1893 * tree-switch-conversion.c: Define
1894 max_ratio_for_speed and max_ratio_for_size constants.
1896 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
1897 Martin Liska <mliska@suse.cz>
1902 * common.opt: Rename align options with 'str_' prefix.
1903 * common/config/i386/i386-common.c (set_malign_value): New
1905 (ix86_handle_option): Use it to set -falign-* options/
1906 * config/aarch64/aarch64-protos.h (struct tune_params): Change
1907 type from int to string.
1908 * config/aarch64/aarch64.c: Update default values from int
1910 * config/alpha/alpha.c (alpha_override_options_after_change):
1912 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
1913 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1914 max skip conditionally.
1915 * config/i386/freebsd.h (SUBALIGN_LOG): New.
1916 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1917 max skip conditionally.
1918 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1919 max skip conditionally.
1920 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
1921 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1922 max skip conditionally.
1923 * config/i386/i386.c (struct ptt): Change type from int to
1925 (ix86_default_align): Set default values.
1926 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
1927 max skip conditionally.
1928 * config/i386/iamcu.h (SUBALIGN_LOG): New.
1929 (ASM_OUTPUT_MAX_SKIP_ALIGN):
1930 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
1931 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1932 max skip conditionally.
1933 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
1934 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
1935 * config/i386/x86-64.h (SUBALIGN_LOG): New.
1936 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1937 max skip conditionally.
1938 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1939 * config/ia64/ia64.c (ia64_option_override): Set default values
1940 for alignment options.
1941 * config/m68k/m68k.c: Handle new str_align_* options.
1942 * config/mips/mips.c (mips_set_compression_mode): Change
1944 (mips_option_override): Set default values for options.
1945 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
1947 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1949 * config/rx/rx.c (rx_option_override): Likewise.
1950 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
1951 (LABEL_ALIGN): Use align_labels_log.
1952 (LOOP_ALIGN): Use align_loops_align.
1953 * config/s390/s390.c (s390_asm_output_function_label): Use new
1955 * config/sh/sh.c (sh_override_options_after_change):
1956 Change type of constants.
1957 * config/spu/spu.c (spu_sched_init): Likewise.
1958 * config/sparc/sparc.c (sparc_option_override): Set default
1960 * config/visium/visium.c (visium_option_override): Likewise.
1961 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
1962 emit p2align format with last argument if it's not needed.
1963 * doc/invoke.texi: Document extended format of -falign-*.
1964 * final.c: Use align_labels alignment.
1965 * flags.h (struct target_flag_state): Change type to use
1967 (struct align_flags_tuple): New.
1968 (struct align_flags): Likewise.
1969 (align_loops_log): Redefine macro to use new types.
1970 (align_loops_max_skip): Redefine macro to use new types.
1971 (align_jumps_log): Redefine macro to use new types.
1972 (align_jumps_max_skip): Redefine macro to use new types.
1973 (align_labels_log): Redefine macro to use new types.
1974 (align_labels_max_skip): Redefine macro to use new types.
1975 (align_functions_log): Redefine macro to use new types.
1976 (align_loops): Redefine macro to use new types.
1977 (align_jumps): Redefine macro to use new types.
1978 (align_labels): Redefine macro to use new types.
1979 (align_functions): Redefine macro to use new types.
1980 (align_functions_max_skip): Redefine macro to use new types.
1981 (align_loops_value): New macro.
1982 (align_jumps_value): New macro.
1983 (align_labels_value): New macro.
1984 (align_functions_value): New macro.
1985 * function.c (invoke_set_current_function_hook): Propagate
1986 alignment values from flags to global variables default in
1988 * ipa-icf.c (sem_function::equals_wpa): Use
1989 cl_optimization_option_eq instead of memcmp.
1990 * lto-streamer.h (cl_optimization_stream_out): Support streaming
1992 (cl_optimization_stream_in): Likewise.
1993 * optc-save-gen.awk: Support strings in cl_optimization.
1994 * opth-gen.awk: Likewise.
1995 * opts.c (finish_options): Remove error checking of invalid
1997 (MAX_CODE_ALIGN): Remove.
1998 (MAX_CODE_ALIGN_VALUE): Likewise.
1999 (parse_and_check_align_values): New function.
2000 (check_alignment_argument): Likewise.
2001 (common_handle_option): Use check_alignment_argument.
2002 * opts.h (parse_and_check_align_values): Declare.
2003 * toplev.c (init_alignments): Remove.
2004 (read_log_maxskip): New.
2005 (parse_N_M): Likewise.
2006 (parse_alignment_opts): Likewise.
2007 (backend_init_target): Remove usage of init_alignments.
2008 * toplev.h (parse_alignment_opts): Declare.
2009 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
2011 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
2012 * tree.c (cl_option_hasher::equal): New.
2013 * varasm.c: Use new global macros.
2015 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
2017 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
2018 Use a simpler align directive also if MAXSKIP = ALIGN-1.
2019 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2020 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2021 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2022 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
2023 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
2024 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
2025 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
2026 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2027 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2028 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2029 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2031 2018-07-04 Martin Liska <mliska@suse.cz>
2032 Jonathan Wakely <jwakely@redhat.com>
2034 * coverage.c: Use correct type.
2035 * doc/invoke.texi: Language correction.
2037 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
2040 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
2041 ENDBRANCH for non-tail call which may return via indirect branch.
2042 * doc/extend.texi: Document indirect_return attribute.
2044 2018-07-03 Martin Sebor <msebor@redhat.com>
2046 PR tree-optimization/86274
2047 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
2049 (format_floating): Correct handling of infinities and NaNs.
2051 2018-07-03 Martin Sebor <msebor@redhat.com>
2053 * print-tree.c (print_real_cst): New function.
2054 (print_node_brief): Call it.
2055 (print_node): Ditto.
2057 2018-07-03 Jeff Law <law@redhat.com>
2059 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
2060 into a single pattern.
2062 * config/h8300/h8300.md (ors code_iterator): New.
2063 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
2064 a single pattern and single splitter.
2065 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
2066 (iorqi3_1, xorqi3_1): Likewise.
2067 (iorqi3, xorqi3 expanders): Similarly.
2069 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
2070 (movmd_internal) into a single pattern using the P mode iterator.
2071 (movmd splitters): Similarly.
2072 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
2073 (movsd splitters): Similarly.
2075 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
2076 ADDB, ADDW and ADDL into a single ADD attribute which selects the
2077 right table based on the size of the operand.
2078 * config/h8300/h8300.md (length_table): Corresponding changes. All
2079 references to "addb", "addw" and "addl" changed to "add".
2080 (btst patterns): Merge two variants into a single pattern.
2081 (tstqi, tsthi): Likewise.
2082 (addhi3_incdec, addsi3_incdec): Likewise.
2083 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
2084 (mulhi3, mulsi3): Likewise.
2085 (udivhi3, udivsi3): Likewise.
2086 (divhi3, divsi3): Likewise.
2087 (andorqi3, andorhi3, andorsi3): Likewise.
2089 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
2092 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
2093 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
2095 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2097 PR tree-optimization/85694
2098 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
2099 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
2101 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
2102 (<u>avg<mode>3_ceil): New patterns.
2104 2018-07-03 David Malcolm <dmalcolm@redhat.com>
2106 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
2107 scan-tree-dump directive.
2108 * gcc.dg/vect/slp-perm-2.c: Likewise.
2109 * gcc.dg/vect/slp-perm-3.c: Likewise.
2110 * gcc.dg/vect/slp-perm-5.c: Likewise.
2111 * gcc.dg/vect/slp-perm-6.c: Likewise.
2112 * gcc.dg/vect/slp-perm-7.c: Likewise.
2113 * gcc.dg/vect/slp-perm-8.c: Likewise.
2115 2018-07-03 Marek Polacek <polacek@redhat.com>
2118 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
2120 2018-07-03 Richard Biener <rguenther@suse.de>
2123 * tree-ssa-structalias.c (find_func_clobbers): Properly
2124 handle indirect calls.
2126 2018-07-03 Jeff Law <law@redhat.com>
2128 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
2129 (shifts): New code iterator.
2130 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
2131 expander. Fix HImode handling on H8/SX.
2132 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
2133 (subqi3, subhi3, subsi3 expanders): Likewise.
2134 (andqi3, andhi3, andsi3 expanders): Likewise.
2135 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
2136 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
2137 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
2138 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
2139 (zero_extendqihi2, zero_extendqisi2): Likewise.
2140 (extendqihi2, extendqisi2): Likewise.
2141 (rotlqi3, rotlhi3, rotlsi3): Likewise.
2142 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
2143 (rotlqi3_1, rotlhi3_1): Likewise.
2144 (logicalhi3_sn, logicalsi3_sn): Likewise.
2145 (logicalhi3, logicalsi3): Likewise.
2147 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2149 * tree-vect-patterns.c (vect_recog_rotate_pattern)
2150 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
2151 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
2152 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
2153 type to append_pattern_def_seq instead of creating a stmt_vec_info
2155 (build_mask_conversion): Likewise. Remove vinfo argument.
2156 (vect_add_conversion_to_patterm): Likewise, renaming to...
2157 (vect_add_conversion_to_pattern): ...this.
2158 (vect_recog_mask_conversion_pattern): Update call to
2159 build_mask_conversion. Pass the vector type to
2160 append_pattern_def_seq here too.
2161 (vect_recog_gather_scatter_pattern): Update call to
2162 vect_add_conversion_to_pattern.
2164 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2166 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
2167 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2168 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
2169 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
2170 STMT_VINFO_PATTERN_DEF_SEQ to null here.
2171 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
2172 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
2173 append_pattern_def_seq instead of new_pattern_def_seq.
2174 (vect_recog_divmod_pattern): Do both of the above.
2175 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
2178 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2180 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
2181 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
2182 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
2183 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
2184 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
2185 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
2186 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
2187 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2188 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2189 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
2190 parameter with a single stmt_vec_info.
2191 (vect_recog_func_ptr): Likewise.
2192 (vect_recog_gather_scatter_pattern): Likewise, folding in...
2193 (vect_try_gather_scatter_pattern): ...this.
2194 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
2195 the stmt_vec_info of the statement to be matched. Don't clear
2196 STMT_VINFO_RELATED_STMT.
2197 (vect_pattern_recog): Update call accordingly.
2199 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2201 PR tree-optimization/85694
2202 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
2203 (uavgM3_ceil): Document new optabs.
2204 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
2205 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
2207 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
2208 (savg_ceil_optab): New optabs.
2209 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
2210 (vect_vect_recog_func_ptrs): Add it.
2211 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
2212 constant directly from the associated lhs.
2214 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2216 * tree-vect-patterns.c (vect_split_statement): New function.
2217 (vect_convert_input): Use it to try to split an existing cast.
2219 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2221 * poly-int.h (print_hex): New function.
2222 * dumpfile.h (dump_dec, dump_hex): Declare.
2223 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
2224 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
2225 min_input_precision, operation_precision and operation_sign.
2226 * tree-vect-patterns.c (vect_get_range_info): New function.
2227 (vect_same_loop_or_bb_p, vect_single_imm_use)
2228 (vect_operation_fits_smaller_type): Delete.
2229 (vect_look_through_possible_promotion): Add an optional
2230 single_use_p parameter.
2231 (vect_recog_over_widening_pattern): Rewrite to use new
2232 stmt_vec_info infomration. Handle one operation at a time.
2233 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
2234 (vect_truncatable_operation_p, vect_set_operation_type)
2235 (vect_set_min_input_precision): New functions.
2236 (vect_determine_min_output_precision_1): Likewise.
2237 (vect_determine_min_output_precision): Likewise.
2238 (vect_determine_precisions_from_range): Likewise.
2239 (vect_determine_precisions_from_users): Likewise.
2240 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
2241 (vect_vect_recog_func_ptrs): Put over_widening first.
2243 (vect_pattern_recog): Call vect_determine_precisions.
2245 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2247 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
2248 statements that have been replaced by further pattern statements.
2249 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
2251 2018-07-03 Richard Biener <rguenther@suse.de>
2253 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
2254 always set *dt. Dump vectype in vectype overload.
2255 * dumpfile.h (dump_gimple_expr): New function.
2256 (dump_gimple_expr_loc): Likewise.
2257 * dumpfile.c (dump_gimple_expr): New function.
2258 (dump_gimple_expr_loc): Likewise.
2260 2018-07-02 Jeff Law <law@redhat.com>
2262 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
2263 the H8/300, H8/300H and H8/S variants into a single pattern.
2264 (movhi_h8300, movqi_h8300hs): Similarly.
2265 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
2266 (QHI mode iterator): New.
2268 * config/h8300/h8300.md: Remove trailing whitespace.
2270 2018-07-02 Jim Wilson <jimw@sifive.com>
2272 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
2273 instead of emit_insn for interrupt returns.
2274 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
2275 (riscv_sret, riscv_uret): Likewise.
2277 2018-07-02 David Malcolm <dmalcolm@redhat.com>
2279 * pretty-print.c (selftest::test_pp_format): Move save and restore
2280 of quotes to class auto_fix_quotes, and add an instance.
2281 * selftest.c: Include "intl.h".
2282 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
2283 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
2284 * selftest.h (selftest::auto_fix_quotes): New class.
2286 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
2288 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
2289 (aarch64_sve_prepare_conditional_op): Remove.
2290 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
2291 Allow aarch64_simd_reg_or_zero as select operand; remove
2292 the aarch64_sve_prepare_conditional_op call.
2293 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
2294 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
2295 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
2296 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
2297 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
2298 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
2299 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
2300 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
2301 and a splitters to match all of the *_any patterns.
2302 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
2304 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
2305 (SVE_COND_FP_BINARY_REV): Remove.
2306 (sve_int_op_rev, sve_fp_op_rev): New.
2307 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
2308 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
2309 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
2310 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
2311 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
2312 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
2313 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
2314 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
2315 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
2317 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
2318 Remove match_dup 1 from the inner unspec.
2319 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
2321 * config/aarch64/aarch64.md (movprfx): New attr.
2322 (length): Default movprfx to 8.
2323 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
2324 (*madd<SVE_I>, *msub<SVE_I): Likewise.
2325 (*<su>mul<SVE_I>3_highpart): Likewise.
2326 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
2327 (*v<ASHIFT><SVE_I>3): Likewise.
2328 (*<su><MAXMIN><SVE_I>3): Likewise.
2329 (*<su><MAXMIN><SVE_F>3): Likewise.
2330 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
2331 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
2332 (*div<SVE_F>4): Likewise.
2334 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
2336 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
2339 2018-07-02 Richard Biener <rguenther@suse.de>
2341 PR tree-optimization/86363
2342 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
2343 memset argument refers to a non-variable address.
2345 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
2347 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
2348 {PLUS,MINUS}_EXPR code to...
2349 (adjust_symbolic_bound): ...here,
2350 (combine_bound): ...here,
2351 (set_value_range_with_overflow): ...and here.
2353 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
2355 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
2357 (extract_range_from_abs_expr): ...here.
2359 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
2361 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
2362 -fno-omit-frame-pointer when not optimizing.
2364 2018-07-02 Martin Liska <mliska@suse.cz>
2367 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
2368 (propagate_nothrow): Likewise.
2370 2018-07-02 Martin Liska <mliska@suse.cz>
2373 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
2375 2018-07-02 David Malcolm <dmalcolm@redhat.com>
2377 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
2378 function in r262149, changing "loc" param from source_location to
2379 const dump_location_t &.
2380 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
2381 declaration, as above.
2383 2018-07-01 Paul Koning <ni1d@arrl.net>
2385 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
2386 -munit-asm, -mgnu-asm, -mdec-asm.
2387 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
2388 (pdp11_output_labelref): New.
2389 (pdp11_output_def): New.
2390 (pdp11_output_addr_vec_elt): New.
2391 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
2392 %# and %@ format codes.
2393 (pdp11_option_override): New.
2394 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
2395 (pdp11_output_ident): New.
2396 (pdp11_asm_named_section): New.
2397 (pdp11_asm_init_sections): New.
2398 (pdp11_file_start): New.
2399 (pdp11_file_end): New.
2400 (output_ascii): Use .ascii/.asciz for -mdec-asm.
2401 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
2402 %o, like %c but octal.
2403 (pdp11_option_override): New.
2404 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
2406 (DATA_SECTION_ASM_OP): Ditto.
2407 (READONLY_DATA_SECTION_ASM_OP): New.
2408 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
2409 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
2410 (ASM_OUTPUT_LABELREF): Ditto.
2411 (ASM_OUTPUT_DEF): Ditto.
2412 (ASM_OUTPUT_EXTERNAL): New.
2413 (ASM_OUTPUT_SOURCE_FILENAME): New.
2414 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
2415 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
2416 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
2417 %# and %@ format codes.
2418 * config/pdp11/pdp11.opt (mgnu-asm): New.
2419 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
2420 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
2421 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
2423 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
2425 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
2426 dereferencing path[] beyond its length.
2428 (debug_all_paths): New.
2429 (rewire_first_differing_edge): New.
2430 (adjust_paths_after_duplication): New.
2431 (duplicate_thread_path): Call adjust_paths_after_duplication.
2433 (thread_through_all_blocks): Add new argument to
2434 duplicate_thread_path.
2436 2018-06-30 Jim Wilson <jimw@sifive.com>
2438 * config/riscv/predicates.md (p2m1_shift_operand): New.
2439 (high_mask_shift_operand): New.
2440 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
2441 pattern using p2m1_shift_operand.
2442 (lshsi3_zero_extend_3+2): New combiner pattern using
2443 high_mask_shift_operand.
2445 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2447 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
2449 (vect_recog_rotate_pattern): ...here.
2450 (vect_convert_input): Try to insert casts of invariants in the
2452 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
2453 preheader to be empty.
2455 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2457 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
2458 vector type. If given, install it in the new statement's
2460 (vect_element_precision): New function.
2461 (vect_unpromoted_value): New struct.
2462 (vect_unpromoted_value::vect_unpromoted_value): New function.
2463 (vect_unpromoted_value::set_op): Likewise.
2464 (vect_look_through_possible_promotion): Likewise.
2465 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
2466 (vect_widened_op_tree, vect_convert_input): Likewise.
2467 (vect_convert_inputs, vect_convert_output): Likewise.
2468 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
2469 to handle the optional cast of the multiplication result and
2470 vect_widened_op_tree to detect the widened multiplication itself.
2471 Do not require the input and output of promotion casts to have
2472 the same sign, but base the signedness of the operation on the
2473 input rather than the result. If the pattern includes two
2474 promotions, check that those promotions have the same sign.
2475 Do not restrict the MULT_EXPR handling to a double-width result;
2476 handle quadruple-width results and wider. Use vect_convert_inputs
2477 to convert the inputs to the common type.
2478 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
2479 to handle the optional cast of the ABS result. Also allow a sign
2480 change or a sign extension between the ABS and MINUS.
2481 Use vect_widened_op_tree to detect the widened subtraction and use
2482 vect_convert_inputs to convert the inputs to the common type.
2483 (vect_handle_widen_op_by_const): Delete.
2484 (vect_recog_widen_op_pattern): New function.
2485 (vect_recog_widen_mult_pattern): Use it.
2486 (vect_recog_widen_shift_pattern): Likewise.
2487 (vect_recog_widen_sum_pattern): Use
2488 vect_look_through_possible_promotion to handle the promoted
2491 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2493 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
2494 the containing gimple_seq *.
2495 * gimple-iterator.h (gsi_for_stmt): Declare it.
2496 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
2497 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
2498 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
2499 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
2500 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
2502 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
2504 (vect_mark_pattern_stmts): ...here. Handle cases in which the
2505 statement being replaced is part of an existing pattern
2506 definition sequence, inserting the new pattern statements before
2508 (vect_pattern_recog_1): Don't return a bool. If the statement
2509 is already part of a pattern, instead apply pattern matching
2510 to the pattern definition statements. Don't clear the
2511 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
2512 (vect_pattern_recog): Don't break after the first match;
2513 continue processing the pattern definition statements instead.
2514 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
2516 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2518 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2519 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2520 (vect_recog_widen_sum_pattern): Use it.
2522 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2524 * tree-vect-loop.c (vectorizable_reduction): Assert that the
2525 phi is not a pattern statement and has not been replaced by
2526 a pattern statement.
2527 * tree-vect-patterns.c (type_conversion_p): Don't check
2528 STMT_VINFO_IN_PATTERN_P.
2529 (vect_recog_vector_vector_shift_pattern): Likewise.
2530 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
2531 the pattern statement rather than the original statement; check
2532 directly for a WIDEN_MULT_EXPR here.
2533 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
2534 vect_is_simple_use to return the pattern statement rather
2535 than the original statement; use is_pattern_stmt_p to check
2536 for such a pattern statement.
2537 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
2538 to return the pattern statement rather than the original statement;
2539 don't do the same transformation here.
2540 (vect_is_simple_use): If the defining statement has been replaced
2541 by a pattern statement, return the pattern statement instead.
2542 Remove the corresponding (local) transformation from the vectype
2545 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
2547 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
2548 end and default to null.
2549 * tree-vect-loop.c (vect_create_epilog_for_reduction)
2550 (vectorizable_reduction): Update calls accordingly, dropping the
2551 gimple ** argument if the passed-back statement isn't needed.
2552 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
2553 (vect_recog_rotate_pattern): Likewise.
2554 (vect_recog_mask_conversion_pattern): Likewise.
2555 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2556 (vect_mask_constant_operand_p): Likewise.
2557 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
2558 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
2559 (get_group_load_store_type, get_load_store_type): Likewise.
2560 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
2561 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
2562 (vectorizable_conversion, vectorizable_assignment): Likewise.
2563 (vectorizable_shift, vectorizable_operation): Likewise.
2564 (vectorizable_store, vect_is_simple_cond): Likewise.
2565 (vectorizable_condition, vectorizable_comparison): Likewise.
2566 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
2567 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
2568 and move it to the end. Cope with null def_stmt_outs.
2570 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2572 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
2574 2018-06-29 Jeff Law <law@redhat.com>
2576 * config/v850/v850.c (v850_legitimate_address_p): Handle large
2577 displacements for TARGET_V850E2V3 and newer.
2578 (TARGET_LRA_P): Remove. Defaults to LRA now.
2579 * config/v850/v850.md (sign23byte_load): Remove.
2580 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
2581 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
2583 2018-06-29 Martin Liska <mliska@suse.cz>
2586 * coverage.c (coverage_init): Mangle full path name.
2587 * doc/invoke.texi: Document the change.
2588 * gcov-io.c (mangle_path): New.
2589 * gcov-io.h (mangle_path): Likewise.
2590 * gcov.c (mangle_name): Use mangle_path for path mangling.
2592 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2594 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
2595 if starting source register is not even.
2597 2018-06-29 Martin Liska <mliska@suse.cz>
2599 PR tree-optimization/86263
2600 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
2601 Make edge redirection.
2603 2018-06-29 David Malcolm <dmalcolm@redhat.com>
2605 * dumpfile.c (dump_loc): Add indentation based on scope depth.
2606 (dump_scope_depth): New variable.
2607 (get_dump_scope_depth): New function.
2608 (dump_begin_scope): New function.
2609 (dump_end_scope): New function.
2610 * dumpfile.h (get_dump_scope_depth): New declaration.
2611 (dump_begin_scope): New declaration.
2612 (dump_end_scope): New declaration.
2613 (class auto_dump_scope): New class.
2614 (AUTO_DUMP_SCOPE): New macro.
2615 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
2618 2018-06-29 Richard Biener <rguenther@suse.de>
2620 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
2621 compute_all_dependences succeeds.
2622 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
2623 exceed --param loop-max-datarefs-for-datadeps.
2625 2018-06-29 Jakub Jelinek <jakub@redhat.com>
2627 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
2629 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
2632 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
2633 alternative 0 in preferred_for_speed attribute.
2635 2018-06-28 Paul Koning <ni1d@arrl.net>
2637 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
2638 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
2639 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
2640 * config/pdp11/pdp11.md: Correct "length" attribute calculation
2641 for shift insn patterns.
2643 2018-06-28 David Malcolm <dmalcolm@redhat.com>
2645 * cgraph.c (cgraph_node::get_body): Replace assignments to
2646 "dump_file" with calls to set_dump_file.
2647 * dumpfile.c (alt_dump_file): Make static, and group with...
2648 (alt_flags): ...this definition.
2649 (dumps_are_enabled): New variable.
2650 (refresh_dumps_are_enabled): New function.
2651 (set_dump_file): New function.
2652 (set_alt_dump_file): New function.
2653 (gcc::dump_manager::dump_start): Replace assignments to
2654 "dump_file" and "alt_dump_file" with calls to set_dump_file and
2656 (gcc::dump_manager::dump_finish): Likewise.
2657 * dumpfile.h (alt_dump_file): Delete decl.
2658 (dumps_are_enabled): New variable decl.
2659 (set_dump_file): New function decl.
2660 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
2662 * tree-nested.c (lower_nested_functions): Replace assignments to
2663 "dump_file" with calls to set_dump_file.
2665 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
2667 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
2668 goto_locus of each outgoing edge of each basic block.
2670 2018-06-28 Richard Biener <rguenther@suse.de>
2672 * dwarf2out.c (decl_scope_table): Remove.
2673 (push_decl_scope): Likewise.
2674 (pop_decl_scope): Likewise.
2675 (gen_type_die_for_member): Do not call push/pop_decl_scope.
2676 (gen_struct_or_union_type_die): Likewise.
2677 (gen_tagged_type_die): Likewise.
2678 (dwarf2out_init): Do not initialize decl_scope_table.
2679 (dwarf2out_c_finalize): Do not free it.
2681 2018-06-28 Richard Biener <rguenther@suse.de>
2683 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
2684 deciding whether to not re-use a DIE.
2686 2018-06-28 Richard Biener <rguenther@suse.de>
2688 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
2689 DW_AT_abstract_origin attribute.
2691 2018-06-28 Martin Liska <mliska@suse.cz>
2693 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2694 Use newly introduced constants.
2695 * tree-switch-conversion.h (struct jump_table_cluster):
2696 Define max_ratio_for_size and max_ratio_for_speed.
2698 2018-06-28 Martin Liska <mliska@suse.cz>
2700 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2701 Add new checking assert to catch invalid state.
2702 (jump_table_cluster::can_be_handled): Handle single case
2704 (jump_table_cluster::is_beneficial): Bail out for such case.
2705 (bit_test_cluster::find_bit_tests):
2706 Add new checking assert to catch invalid state.
2707 (bit_test_cluster::can_be_handled): Handle single case
2709 (bit_test_cluster::is_beneficial): Bail out for such case.
2710 (switch_decision_tree::analyze_switch_statement):
2713 2018-06-28 Martin Liska <mliska@suse.cz>
2715 * common.opt: Introduce -completion option.
2716 * gcc.c (driver_handle_option): Handle it.
2717 (driver::main): Print completions if completion
2719 * opt-suggestions.c (option_proposer::get_completions):
2721 (option_proposer::suggest_completion): Likewise.
2722 (option_proposer::find_param_completions): Likewise.
2723 (verify_autocompletions): Likewise.
2724 (test_completion_valid_options): Likewise.
2725 (test_completion_valid_params): Likewise.
2726 (in_completion_p): Likewise.
2727 (empty_completion_p): Likewise.
2728 (test_completion_partial_match): Likewise.
2729 (test_completion_garbage): Likewise.
2730 (opt_proposer_c_tests): Likewise.
2731 * opt-suggestions.h: Declare new functions.
2732 * opts.c (common_handle_option): Handle OPT__completion_.
2733 * selftest-run-tests.c (selftest::run_tests): Add
2734 opt_proposer_c_tests.
2735 * selftest.c (assert_str_startswith): New.
2736 * selftest.h (assert_str_startswith): Likewise.
2737 (opt_proposer_c_tests): New.
2738 (ASSERT_STR_STARTSWITH): Likewise.
2740 2018-06-28 Martin Liska <mliska@suse.cz>
2742 * Makefile.in: Add opt-suggestions.o.
2743 * gcc-main.c: Include opt-suggestions.h.
2744 * gcc.c (driver::driver): Likewise.
2745 (driver::~driver): Remove m_option_suggestions.
2746 (driver::build_option_suggestions): Moved to option_proposer.
2747 (driver::suggest_option): Likewise.
2748 (driver::handle_unrecognized_options): Use option_proposer.
2749 * gcc.h (class driver): Add new memver m_option_proposer.
2750 * opt-suggestions.c: New file.
2751 * opt-suggestions.h: New file.
2753 2018-06-28 Martin Liska <mliska@suse.cz>
2755 * vec.h (class auto_string_vec): New (moved from auto_argvec).
2756 (auto_string_vec::~auto_string_vec): Likewise.
2758 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
2760 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
2761 prevent_decl_creation_for_types fields up and add reset_location field.
2762 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
2763 statement if id->reset_location is true.
2764 (copy_edges_for_bb): Do not set goto_locus on the new edges if
2765 id->reset_location is true.
2766 (copy_phis_for_bb): Force input_location on the arguments if
2767 id->reset_location is true.
2768 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
2769 is set on the function to be inlined.
2771 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
2773 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
2775 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
2777 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
2778 registers for Pmode.
2779 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
2780 hard registers for the clobbered pseudo.
2782 2018-06-27 Paul Koning <ni1d@arrl.net>
2784 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
2785 mutually exclusive options.
2786 * config/pdp11/constraints.md (h): New constraint.
2787 (O): Update definition to match shift code generation.
2788 (D): New constraint.
2789 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
2791 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
2793 (output_jump): Change arguments.
2794 (pdp11_fixed_cc_regs): New function.
2795 (pdp11_cc_mode): Ditto.
2796 (pdp11_expand_shift): Ditto.
2797 (pdp11_assemble_shift): Ditto.
2798 (pdp11_small_shift): Ditto.
2799 (pdp11_branch_cost): Remove.
2800 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
2802 (pdp11_register_move_cost): Update for CC registers.
2803 (pdp11_rtx_costs): Add case for LSHIFTRT.
2804 (pdp11_output_jump): Add CCNZ mode conditional branches.
2805 (notice_update_cc_on_set): Remove.
2806 (pdp11_cc_mode): New function.
2807 (simple_memory_operand): Correct pre/post decrement case.
2808 (no_side_effect_operand): New function.
2809 (pdp11_regno_reg_class): Add CC_REGS class.
2810 (pdp11_fixed_cc_regs): New function.
2811 (pdp11_small_shift): New function.
2812 (pdp11_expand_shift): New function to expand shift insns.
2813 (pdp11_assemble_shift): New function to output shifts.
2814 (pdp11_branch_cost): Remove.
2815 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
2816 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
2817 (WCHAR_TYPE): Ditto.
2818 (PTRDIFF_TYPE): Ditto.
2819 (ADJUST_INSN_LENGTH): New macro.
2820 (FIXED_REGISTERS): Add CC registers.
2821 (CALL_USED_REGISTERS): Ditto.
2823 (REG_CLASS_NAMES): Ditto.
2824 (REG_CLASS_CONTENTS): Ditto.
2825 (SELECT_CC_MODE): Use new function.
2826 (TARGET_FLAGS_REGNUM): New macro.
2827 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
2828 (cc0_reg_rtx): Remove.
2829 (CC_STATUS_MDEP): Remove.
2830 (CC_STATUS_MDEFP_INIT): Remove.
2831 (CC_IN_FPU): Remove.
2832 (NOTICE_UPDATE_CC): Remove.
2833 (REGISTER_NAMES): Add CC registers.
2834 (BRANCH_COST): Change to constant 1.
2835 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
2837 * config/pdp11/pdp11.opt (mbcopy): Remove.
2838 (mbcopy-builtin): Remove.
2839 (mbranch-cheap): Remove.
2840 (mbranch-expensive): Remove.
2841 * config/pdp11/predicates.md (expand_shift_operand): Update to
2842 match shift code generation.
2843 (ccnz_operator): New predicate.
2844 * doc/invoke.texi (PDP-11 Options): Remove deleted options
2845 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
2846 Remove non-existent option -mabshi, -mno-abshi. Document mutually
2848 * doc/md.texi (PDP-11): Document new D and h constraints. Update
2849 description of O constraint.
2851 2018-06-27 Jeff Law <law@redhat.com>
2852 Austin Law <austinklaw@gmail.com>
2854 * config/v850/v850.md (addsi3_set_flags): New pattern.
2855 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
2856 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
2857 (zero_extendhisi2_v850_set_flags): Likewise.
2858 (zero_extendqisi2_v850_set_flags): Likewise.
2859 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
2860 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
2861 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
2863 * config/v850/v850-protos.h (notice_update_cc): Remove.
2864 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
2865 (v850_print_operand): Handle 'D' and "d".
2866 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
2867 Add handling of arithmetic/logical operations compared against zero.
2868 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
2869 Do not look at v850_compare_op, instead get mode from last argument.
2870 (v850_gen_compare): Remove
2871 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
2872 after reload for prologue insns.
2873 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
2875 (construct_save_jarl): Likewise.
2876 (TARGET_FLAGS_REGNUM): Define.
2877 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
2878 (NOTICE_UPDATE_CC): Remove.
2879 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
2880 than cc0. Conditionalize on reload_completed.
2881 (cmpsi_insn, setfcc_insn): Likewise.
2882 (tst1 splitter): Turn into define_and_split which sets the flags
2884 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
2885 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
2886 (cstoresf4, cstoredf4): Clobber the flags.
2887 (cmpsi, cmpsf, cmpdf): Remove expanders.
2888 (setf_insn): Remove pattern.
2889 (addsi3): Turn into define_and_split which clobbers the flags after
2890 reload and a suitable pattern (addsi3_clobber_flags) for use after
2892 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
2893 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
2894 (ashrsi3, ashrsi3_v850e2): Likewise.
2895 (bins): Clobber the flags.
2896 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
2897 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
2898 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
2899 (call_value_internal_short, call_value_internal_long): Likewise.
2900 (callt_save_interrupt, callt_return_interrupt): Likewise.
2901 (save_interrupt, return_interrupt): Likewise.
2902 (callt_save_all_interrupt, save_all_interrupt): Likewise.
2903 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
2904 (restore_all_interrupt, _restore_all_interrupt): Likewise.
2905 (All FP comparisons): Only allow after reload has completed.
2907 (divh, divhu): Tweak output template.
2908 (branch_z_normal, branch_z_invert): Remove
2909 (branch_nz_normal, branch_nz_invert): Likewise.
2910 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
2912 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
2913 * config/v850/v850.c (notice_update_cc): Remove.
2914 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
2915 (CC_NO_CARRY): Likewise.
2916 (NOTICE_UPDATE_CC): Define to nothing.
2917 * config/v850/v850.md: Remove block comment on cc0 handling
2918 Remove "cc" attribute from all patterns. Remove cc_status handling
2919 from all patterns. Minor formatting fixes.
2921 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2923 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
2924 (cortex-a76.cortex-a55): Likewise.
2925 * config/aarch64/aarch64-tune.md: Regenerate.
2926 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
2927 cortex-a76.cortex-a55.
2929 2018-06-27 Jeff Law <law@redhat.com>
2931 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
2932 (MULTILIB_DIRNAMES): Similarly.
2934 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
2936 * gimple.h (gimple_return_retbnd): Delete.
2937 (gimple_return_set_retbnd): Likewise.
2938 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
2939 gimple_return_set_retbnd.
2940 * gimple-pretty-print.c (dump_gimple_return): Remove call to
2941 gimple_return_retbnd and adjust.
2942 * tree-inline.h (struct copy_body_data): Remove retbnd field.
2943 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
2944 Explicitly return NULL in a couple more cases. Move assertion
2945 on debug statements and remove unreachable code.
2946 (reset_debug_binding): Do not test id->retbnd.
2947 (expand_call_inline): Do not set it.
2949 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2951 * configure.ac: Add --disable-gcov option.
2952 * configure: Regenerate.
2953 * Makefile.in: Honour @enable_gcov@.
2954 * doc/install.texi: Document --disable-gcov.
2956 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2958 * config/arm/arm-cpus.in (cortex-a76): New entry.
2959 (cortex-a76.cortex-a55): Likewise.
2960 * config/arm/arm-tables.opt: Regenerate.
2961 * config/arm/arm-tune.md: Likewise.
2962 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
2963 * doc/invoke.texi (ARM Options): Document cortex-a76 and
2964 cortex-a76.cortex-a55.
2966 2018-06-27 Tamar Christina <tamar.christina@arm.com>
2969 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
2971 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
2973 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
2975 (EPILOGUE_USES): Likewise.
2977 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
2979 * tree-inline.c (remap_location): New function extracted from...
2980 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
2981 (copy_phis_for_bb): ...here. Call remap_location.
2982 (copy_cfg_body): Adjust call to copy_edges_for_bb.
2984 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
2986 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
2987 unaligned vsx for 16B memset.
2989 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
2992 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
2993 ieee128_float_type_node to long_double_type_node unless
2994 TARGET_LONG_DOUBLE_128 is set.
2996 2018-06-26 David Malcolm <dmalcolm@redhat.com>
2998 * cfgloop.c (get_loop_location): Convert return type from
2999 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
3000 by implicit construction from rtx_insn *, and using
3001 dump_user_location_t::from_function_decl for the fallback case.
3002 * cfgloop.h (get_loop_location): Convert return type from
3003 location_t to dump_user_location_t.
3004 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
3005 dump_printf_loc to pass in a dump_location_t rather than a
3006 location_t, via the gimple stmt.
3007 * coverage.c (get_coverage_counts): Update calls to
3008 dump_printf_loc to pass in dump_location_t rather than a
3010 * doc/optinfo.texi (Dump types): Convert example of
3011 dump_printf_loc from taking "locus" to taking "insn". Update
3012 description of the "_loc" calls to cover dump_location_t.
3013 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
3015 (dump_user_location_t::dump_user_location_t): New constructors,
3016 from gimple *stmt and rtx_insn *.
3017 (dump_user_location_t::from_function_decl): New function.
3018 (dump_loc): Make static.
3019 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
3020 const dump_location_t &.
3021 (dump_generic_expr_loc): Delete.
3022 (dump_printf_loc): Convert param "loc" from location_t to
3023 const dump_location_t &.
3024 (selftest::test_impl_location): New function.
3025 (selftest::dumpfile_c_tests): New function.
3026 * dumpfile.h: Include "profile-count.h".
3027 (class dump_user_location_t): New class.
3028 (struct dump_impl_location_t): New struct.
3029 (class dump_location_t): New class.
3030 (dump_printf_loc): Convert 2nd param from source_location to
3031 const dump_location_t &.
3032 (dump_generic_expr_loc): Delete.
3033 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
3034 const dump_location_t &.
3035 * gimple-fold.c (fold_gimple_assign): Update call to
3036 dump_printf_loc to pass in a dump_location_t rather than a
3037 location_t, via the gimple stmt.
3038 (gimple_fold_call): Likewise.
3039 * gimple-loop-interchange.cc
3040 (loop_cand::analyze_iloop_reduction_var): Update for change to
3041 check_reduction_path.
3042 (tree_loop_interchange::interchange): Update for change to
3044 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
3045 change in return-type of find_loop_location.
3046 (graphite_regenerate_ast_isl): Likewise.
3047 * graphite-optimize-isl.c (optimize_isl): Likewise.
3048 * graphite.c (graphite_transform_loops): Likewise.
3049 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
3050 pass in a dump_location_t rather than a location_t, via the
3052 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3053 * ipa.c (walk_polymorphic_call_targets): Likewise.
3054 * loop-unroll.c (report_unroll): Convert "locus" param from
3055 location_t to dump_location_t.
3056 (decide_unrolling): Update for change to get_loop_location's
3058 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
3059 location_t to dump_user_location_t.
3060 (grid_find_single_omp_among_assignments_1): Updates calls to
3061 dump_printf_loc to pass in a dump_location_t rather than a
3062 location_t, via the gimple stmt.
3063 (grid_parallel_clauses_gridifiable): Convert "tloc" from
3064 location_t to dump_location_t. Updates calls to dump_printf_loc
3065 to pass in a dump_location_t rather than a location_t, via the
3067 (grid_inner_loop_gridifiable_p): Likewise.
3068 (grid_dist_follows_simple_pattern): Likewise.
3069 (grid_gfor_follows_tiling_pattern): Likewise.
3070 (grid_target_follows_gridifiable_pattern): Likewise.
3071 (grid_attempt_target_gridification): Convert initialization
3072 of local "grid" from memset to zero-initialization; FIXME: does
3073 this require C++11? Update call to dump_printf_loc to pass in a
3074 optinfo_location rather than a location_t, via the gimple stmt.
3075 * profile.c (read_profile_edge_counts): Updates call to
3076 dump_printf_loc to pass in a dump_location_t rather than a
3078 (compute_branch_probabilities): Likewise.
3079 * selftest-run-tests.c (selftest::run_tests): Call
3081 * selftest.h (dumpfile_c_tests): New decl.
3082 * tree-loop-distribution.c (pass_loop_distribution::execute):
3083 Update for change in return type of find_loop_location.
3084 * tree-parloops.c (parallelize_loops): Likewise.
3085 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
3086 "locus" from location_t to dump_user_location_t.
3087 (canonicalize_loop_induction_variables): Likewise.
3088 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
3089 for change in return type of find_loop_location.
3090 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
3091 to dump_printf_loc to pass in a dump_location_t rather than a
3092 location_t, via the stmt.
3093 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
3095 * tree-vect-loop-manip.c (find_loop_location): Convert return
3096 type from source_location to dump_user_location_t.
3097 (vect_do_peeling): Update for above change.
3098 (vect_loop_versioning): Update for change in type of
3100 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
3101 from location_t to dump_user_location_t.
3102 (vect_estimate_min_profitable_iters): Update for change in type
3104 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
3105 location_t to dump_location_t.
3106 (vect_slp_bb): Update for change in type of vect_location.
3107 * tree-vectorizer.c (vect_location): Convert from source_location
3108 to dump_user_location_t.
3109 (try_vectorize_loop_1): Update for change in vect_location's type.
3110 (vectorize_loops): Likewise.
3111 (increase_alignment): Likewise.
3112 * tree-vectorizer.h (vect_location): Convert from source_location
3113 to dump_user_location_t.
3114 (find_loop_location): Convert return type from source_location to
3115 dump_user_location_t.
3116 (check_reduction_path): Convert 1st param from location_t to
3117 dump_user_location_t.
3118 * value-prof.c (check_counter): Update call to dump_printf_loc to
3119 pass in a dump_user_location_t rather than a location_t; update
3120 call to error_at for change in type of "locus".
3121 (check_ic_target): Update call to dump_printf_loc to
3122 pass in a dump_user_location_t rather than a location_t, via the
3125 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
3127 * config/s390/s390.h (enum processor_flags): Do not use
3128 default tune parameter when -march was specified.
3130 2018-06-26 Jakub Jelinek <jakub@redhat.com>
3133 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
3134 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
3137 2018-06-26 Richard Biener <rguenther@suse.de>
3139 PR tree-optimization/86287
3141 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
3142 (vect_analyze_loop): Initialize n_stmts.
3144 2018-06-26 Richard Biener <rguenther@suse.de>
3147 * fold-const.c (fold_convertible_p): Pointer extension
3150 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
3153 * dwarf2out.c (loc_list_has_views): Adjust comments.
3154 (dw_loc_list): Split single cross-partition range with
3157 2018-06-25 Jeff Law <law@redhat.com>
3159 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
3160 on -mbig-switch by default.
3162 * config/v850/predicates.md (const_float_1_operand): Fix match_code
3164 (const_float_0_operand): Remove unused predicate.
3165 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
3166 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
3167 (recipsf2): New expander. Original pattern now called
3169 (recipdf2, recipdf2_insn): Similarly.
3170 (rsqrtsf2, rsqrtsf2_insn): Similarly
3171 (rsqrtdf2, rsqrtdf2_insn): Similarly
3173 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
3175 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
3176 Simplify logic for FreeBSD (twice).
3178 2018-06-25 Martin Sebor <msebor@redhat.com>
3180 PR tree-optimization/86204
3181 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
3182 a strnlen result if it's less than the length of the string.
3184 2018-06-25 Martin Sebor <msebor@redhat.com>
3186 PR tree-optimization/85700
3187 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
3188 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
3189 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
3191 2018-06-25 Martin Sebor <msebor@redhat.com>
3193 * doc/extend.texi (Zero-length arrays): Update and clarify.
3195 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
3197 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
3198 added IEEE/IBM long double multilib support on PowerPC little
3199 endian Linux systems.
3200 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
3201 (MULTILIB_DEFAULTS): Likewise.
3202 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3204 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
3205 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
3206 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
3208 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
3211 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
3212 (REORDER_45): Likewise.
3214 2018-06-25 Jeff Law <law@redhat.com>
3216 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
3217 dividend to 32 bits. Adjust length.
3218 (udivmodhi4): Cleanup output template. Fix length.
3220 2018-06-25 Carl Love <cel@us.ibm.com>
3222 * config/rs6000/vsx.md: Change word selector to prefered location.
3224 2018-06-25 Richard Biener <rguenther@suse.de>
3226 PR tree-optimization/86304
3227 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
3228 epilogue-if-converted loops as well.
3230 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
3232 * lto-section-out.c (lto_begin_section): Do not print section
3233 name for noaddr and unnumbered dumps.
3235 2018-06-25 Richard Biener <rguenther@suse.de>
3237 * tree-vectorizer.h (struct vec_info_shared): New structure
3238 with parts split out from struct vec_info and loop_nest from
3239 struct _loop_vec_info.
3240 (struct vec_info): Adjust accordingly.
3241 (struct _loop_vec_info): Likewise.
3242 (LOOP_VINFO_LOOP_NEST): Adjust.
3243 (LOOP_VINFO_DATAREFS): Likewise.
3244 (LOOP_VINFO_DDRS): Likewise.
3245 (struct _bb_vec_info): Likewise.
3246 (BB_VINFO_DATAREFS): Likewise.
3247 (BB_VINFO_DDRS): Likewise.
3248 (struct _stmt_vec_info): Add dr_aux member.
3249 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
3250 (DR_MISALIGNMENT_UNINITIALIZED): New.
3251 (set_dr_misalignment): Adjust.
3252 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
3253 (vect_analyze_loop): Adjust prototype.
3254 (vect_analyze_loop_form): Likewise.
3255 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
3256 Compute dependences lazily.
3257 (vect_record_base_alignments): Use shared datarefs/ddrs.
3258 (vect_verify_datarefs_alignment): Likewise.
3259 (vect_analyze_data_refs_alignment): Likewise.
3260 (vect_analyze_data_ref_accesses): Likewise.
3261 (vect_analyze_data_refs): Likewise.
3262 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
3263 constructor parameter for shared part.
3264 (vect_analyze_loop_form): Pass in shared part and adjust.
3265 (vect_analyze_loop_2): Pass in storage for the number of
3266 stmts. Move loop nest finding to the caller. Compute
3268 (vect_analyze_loop): Pass in shared part.
3269 (vect_transform_loop): Verify shared datarefs are unchanged.
3270 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
3271 constructor parameter for shared part.
3272 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
3273 (vect_slp_bb): Verify shared datarefs are unchanged before
3275 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
3277 (new_stmt_vec_info): Initialize DR_AUX misalignment to
3278 DR_MISALIGNMENT_UNINITIALIZED.
3279 * tree-vectorizer.c (vec_info::vec_info): Add constructor
3280 parameter for shared part.
3281 (vec_info::~vec_info): Adjust.
3282 (vec_info_shared::vec_info_shared): New.
3283 (vec_info_shared::~vec_info_shared): Likewise.
3284 (vec_info_shared::save_datarefs): Likewise.
3285 (vec_info_shared::check_datarefs): Likewise.
3286 (try_vectorize_loop_1): Construct shared part live for analyses
3287 of a single loop for multiple vector sizes.
3288 * tree-parloops.c (gather_scalar_reductions): Adjust.
3290 2018-06-25 Richard Biener <rguenther@suse.de>
3292 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
3293 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
3294 (vect_analyze_data_refs): Remove similar code from here and
3295 simplify accordingly.
3297 2018-06-25 Richard Biener <rguenther@suse.de>
3299 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
3300 for reverse storage order accesses rather than asserting
3301 they cannot happen here.
3303 2018-06-25 Tom de Vries <tdevries@suse.de>
3306 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
3307 Use data16 instead of .byte for insn prefix.
3309 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
3312 * parser.c (make_char_string_pack): Pass this literal chars
3313 through cpp_interpret_string.
3314 (cp_parser_userdef_numeric_literal): Check the result of
3315 make_char_string_pack.
3317 2018-06-24 Maya Rashish <coypu@sdf.org>
3319 * ginclude/stddef.h: Simplify conditions around avoiding
3320 re-definition of __size_t.
3322 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3324 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
3325 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
3327 2018-06-22 Maya Rashish <coypu@sdf.org>
3329 * doc/invoke.texi (mno-fancy-math-387): Update for changes
3330 made to OpenBSD and NetBSD through the years.
3332 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3334 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3335 behavior of vec_pack (vector double, vector double) to match
3336 behavior of vec_float2 (vector double, vector double).
3338 2018-06-22 Olivier Hainque <hainque@adacore.com>
3340 * gimplify.c (gimplify_function_tree): Prevent creation
3341 of a trampoline for the address of the current function
3342 passed to entry/exit instrumentation hooks.
3344 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
3347 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
3350 2018-06-22 Martin Liska <mliska@suse.cz>
3352 PR tree-optimization/86263
3353 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3354 Bail out if is_enabled is false.
3355 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
3357 (jump_table_cluster::is_enabled): New function.
3359 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3361 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
3362 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
3363 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
3364 (lto_input_ts_binfo_tree_pointers): Likewise.
3365 * tree-streamer-out.c (streamer_write_tree_bitfields,
3366 write_ts_binfo_tree_pointers): Likewise.
3367 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
3369 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3371 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
3373 2018-06-22 Martin Liska <mliska@suse.cz>
3375 * symbol-summary.h (get): Make it pure and inline move
3376 functionality from ::get function.
3377 (get): Remove and inline into ::get and ::get_create.
3378 (get_create): Move code from ::get function.
3380 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3383 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
3384 -x assembler-with-cpp.
3386 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3388 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
3389 _FILE_OFFSET_BITS=64 for C++.
3391 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
3393 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
3394 conversion insn that shows up when pr85657-3.c is compiled using
3395 IEEE 128-bit long double.
3396 (neg<mode>2_internal): Use the correct mode to check whether the
3397 mode is IBM extended.
3398 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
3399 multiply and divide external functions from being created more
3402 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3404 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
3406 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
3407 the edge can be forwarded.
3408 (cfg_layout_merge_blocks): Likewise.
3410 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3412 * except.c (finish_eh_generation): Commit edge insertions only after
3413 the EH edges have been redirected from post-landing to landing pads.
3415 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3417 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
3418 create_tmp_var_for to create the FRAME decl.
3419 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
3421 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3423 * tree-inline.c (copy_edges_for_bb): Minor tweak.
3424 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
3425 debug statement when resetting its value.
3426 (expand_call_inline): Copy the locus of the call onto the assignment
3427 of the return value, if any. Use local variable in more cases.
3429 2018-06-21 Martin Liska <mliska@suse.cz>
3431 * ipa-pure-const.c (propagate_nothrow): Use
3432 funct_state_summaries->get.
3433 (dump_malloc_lattice): Likewise.
3434 (propagate_malloc): Likewise.
3436 2018-06-21 Richard Biener <rguenther@suse.de>
3438 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
3439 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
3440 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
3442 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
3443 BLOCK_ABSTRACT_ORIGIN unconditionally.
3445 2018-06-21 David Malcolm <dmalcolm@redhat.com>
3447 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
3449 * ipa-reference.c (ipa_reference_c_finalize): Delete
3450 ipa_ref_opt_sum_summaries and set it to NULL.
3452 2018-06-21 Tom de Vries <tdevries@suse.de>
3454 PR tree-optimization/85859
3455 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
3456 test with comment from bb_no_side_effects_p.
3458 2018-06-21 Richard Biener <rguenther@suse.de>
3460 PR tree-optimization/86232
3461 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
3462 max for constant niter.
3464 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
3466 * config/aarch64/aarch64-simd.md
3467 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
3469 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
3471 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
3472 Make opernads of the unspec commutative.
3474 2018-06-21 Richard Biener <rguenther@suse.de>
3476 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
3477 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
3478 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
3479 (vect_analyze_data_ref_dependence): Re-order checks to deal with
3481 (vect_record_base_alignments): Do not record base alignment
3482 for gathers or scatters.
3483 (vect_compute_data_ref_alignment): Drop return value that is always
3484 true. Bail out early for gathers or scatters.
3485 (vect_enhance_data_refs_alignment): Bail out early for gathers
3487 (vect_find_same_alignment_drs): Likewise.
3488 (vect_analyze_data_refs_alignment): Remove dead code.
3489 (vect_slp_analyze_and_verify_node_alignment): Likewise.
3490 (vect_analyze_data_refs): For possible gathers or scatters do
3491 not create an alternate DR, just check their possible validity
3492 and mark them. Adjust DECL_NONALIASED handling to not rely
3494 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
3495 update inits of gathers or scatters.
3496 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
3497 Also copy gather/scatter flag to pattern vinfo.
3499 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
3501 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3502 behavior of vec_packsu (vector unsigned long long, vector unsigned
3503 long long) to match behavior of vec_packs with same signature.
3505 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
3506 Thomas Schwinge <thomas@codesourcery.com>
3507 Cesar Philippidis <cesar@codesourcery.com>
3509 * gimplify.c (gimplify_scan_omp_clauses): Add support for
3510 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
3511 (gimplify_adjust_omp_clauses): Likewise.
3512 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
3513 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
3514 (gimplify_omp_target_update): Update handling of acc update and
3516 * omp-low.c (install_var_field): Remove unused parameter
3517 base_pointers_restrict.
3518 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
3519 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
3521 (omp_target_base_pointers_restrict_p): Delete.
3522 (scan_omp_target): Update call to scan_sharing_clauses.
3523 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
3525 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
3526 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
3527 (convert_local_omp_clauses): Likewise.
3528 * tree-pretty-print.c (dump_omp_clause): Likewise.
3529 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
3531 (omp_clause_code_name): Likewise.
3533 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3536 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
3539 PR tree-optimization/86231
3540 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
3541 anti-range don't overwrite *vr0min before using it to compute *vr0max.
3543 2018-06-20 Tom de Vries <tdevries@suse.de>
3545 PR tree-optimization/86097
3546 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
3547 iv type if signedness of iv type is not the same as that of *nit.
3549 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3551 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
3552 EDGE_EH edges, verify they are all EDGE_EH.
3554 2018-06-20 Maya Rashish <coypu@sdf.org>
3556 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
3558 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3560 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
3561 * config/aarch64/aarch64.c (xgene1_tunings): Add
3562 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
3563 (aarch64_mode_valid_for_sched_fusion_p):
3564 Allow 16-byte modes.
3565 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
3566 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
3568 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
3570 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
3571 * config/aarch64/iterators.md (VQ2): New mode iterator.
3573 2018-06-20 Martin Liska <mliska@suse.cz>
3575 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
3576 Change default ratio from 10 to 8.
3578 2018-06-20 Martin Liska <mliska@suse.cz>
3580 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3582 (bit_test_cluster::find_bit_tests): Likewise.
3583 (switch_decision_tree::analyze_switch_statement): Find clusters.
3584 * tree-switch-conversion.h (struct jump_table_cluster): Document
3587 2018-06-20 Martin Liska <mliska@suse.cz>
3589 * tree-switch-conversion.c (switch_conversion::collect):
3590 Record m_uniq property.
3591 (switch_conversion::expand): Bail out for special conditions.
3592 (group_cluster::~group_cluster): New.
3593 (group_cluster::group_cluster): Likewise.
3594 (group_cluster::dump): Likewise.
3595 (jump_table_cluster::emit): New.
3596 (switch_decision_tree::fix_phi_operands_for_edges): New.
3597 (struct case_node): Remove struct.
3598 (jump_table_cluster::can_be_handled): New.
3599 (case_values_threshold): Moved to header.
3600 (reset_out_edges_aux): Likewise.
3601 (jump_table_cluster::is_beneficial): New.
3602 (bit_test_cluster::can_be_handled): Likewise.
3603 (add_case_node): Remove.
3604 (bit_test_cluster::is_beneficial): New.
3605 (case_bit_test::cmp): New.
3606 (bit_test_cluster::emit): New.
3607 (expand_switch_as_decision_tree_p): Remove.
3608 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
3609 (fix_phi_operands_for_edge): Likewise.
3610 (switch_decision_tree::analyze_switch_statement): New.
3611 (compute_cases_per_edge): Move ...
3612 (switch_decision_tree::compute_cases_per_edge): ... here.
3613 (try_switch_expansion): Likewise.
3614 (switch_decision_tree::try_switch_expansion): Likewise.
3615 (record_phi_operand_mapping): Likewise.
3616 (switch_decision_tree::record_phi_operand_mapping): Likewise.
3617 (emit_case_decision_tree): Likewise.
3618 (switch_decision_tree::emit): Likewise.
3619 (balance_case_nodes): Likewise.
3620 (switch_decision_tree::balance_case_nodes): Likewise.
3621 (dump_case_nodes): Likewise.
3622 (switch_decision_tree::dump_case_nodes): Likewise.
3623 (emit_jump): Likewise.
3624 (switch_decision_tree::emit_jump): Likewise.
3625 (emit_cmp_and_jump_insns): Likewise.
3626 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
3627 (emit_case_nodes): Likewise.
3628 (switch_decision_tree::emit_case_nodes): Likewise.
3629 (conditional_probability): Remove.
3630 * tree-switch-conversion.h (enum cluster_type): New.
3632 (struct cluster): Likewise.
3633 (cluster::cluster): Likewise.
3634 (struct simple_cluster): Likewise.
3635 (simple_cluster::simple_cluster): Likewise.
3636 (struct group_cluster): Likewise.
3637 (struct jump_table_cluster): Likewise.
3638 (struct bit_test_cluster): Likewise.
3639 (struct min_cluster_item): Likewise.
3640 (struct case_tree_node): Likewise.
3641 (case_tree_node::case_tree_node): Likewise.
3642 (jump_table_cluster::case_values_threshold): Likewise.
3643 (struct case_bit_test): Likewise.
3644 (struct switch_decision_tree): Likewise.
3645 (struct switch_conversion): Likewise.
3646 (switch_decision_tree::reset_out_edges_aux): Likewise.
3648 2018-06-20 Martin Liska <mliska@suse.cz>
3650 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
3651 (hoist_edge_and_branch_if_true): Likewise.
3652 (expand_switch_using_bit_tests_p): Likewise.
3653 (struct case_bit_test): Likewise.
3654 (case_bit_test_cmp): Likewise.
3655 (emit_case_bit_tests): Likewise.
3656 (switch_conversion::switch_conversion): New class.
3657 (struct switch_conv_info): Remove old struct.
3658 (collect_switch_conv_info): More to ...
3659 (switch_conversion::collect): ... this.
3660 (check_range): Likewise.
3661 (switch_conversion::check_range): Likewise.
3662 (check_all_empty_except_final): Likewise.
3663 (switch_conversion::check_all_empty_except_final): Likewise.
3664 (check_final_bb): Likewise.
3665 (switch_conversion::check_final_bb): Likewise.
3666 (create_temp_arrays): Likewise.
3667 (switch_conversion::create_temp_arrays): Likewise.
3668 (free_temp_arrays): Likewise.
3669 (gather_default_values): Likewise.
3670 (switch_conversion::gather_default_values): Likewise.
3671 (build_constructors): Likewise.
3672 (switch_conversion::build_constructors): Likewise.
3673 (constructor_contains_same_values_p): Likewise.
3674 (switch_conversion::contains_same_values_p): Likewise.
3675 (array_value_type): Likewise.
3676 (switch_conversion::array_value_type): Likewise.
3677 (build_one_array): Likewise.
3678 (switch_conversion::build_one_array): Likewise.
3679 (build_arrays): Likewise.
3680 (switch_conversion::build_arrays): Likewise.
3681 (gen_def_assigns): Likewise.
3682 (switch_conversion::gen_def_assigns): Likewise.
3683 (prune_bbs): Likewise.
3684 (switch_conversion::prune_bbs): Likewise.
3685 (fix_phi_nodes): Likewise.
3686 (switch_conversion::fix_phi_nodes): Likewise.
3687 (gen_inbound_check): Likewise.
3688 (switch_conversion::gen_inbound_check): Likewise.
3689 (process_switch): Use the newly created class.
3690 (switch_conversion::expand): New.
3691 (switch_conversion::~switch_conversion): New.
3692 * tree-switch-conversion.h: New file.
3694 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3696 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
3697 tree-vect-patterns.c.
3698 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
3699 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
3700 (vect_recog_sad_pattern): Likewise.
3701 (vect_recog_widen_sum_pattern): Likewise.
3702 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
3703 (vect_recog_widen_shift_pattern): Remove the type_in argument.
3704 (vect_recog_rotate_pattern): Likewise.
3705 (vect_recog_mult_pattern): Likewise.
3706 (vect_recog_vector_vector_shift_pattern): Likewise.
3707 (vect_recog_divmod_pattern): Likewise.
3708 (vect_recog_mixed_size_cond_pattern): Likewise.
3709 (vect_recog_bool_pattern): Likewise.
3710 (vect_recog_mask_conversion_pattern): Likewise.
3711 (vect_try_gather_scatter_pattern): Likewise.
3712 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
3713 (vect_recog_over_widening_pattern): Likewise.
3714 (vect_recog_gather_scatter_pattern): Likewise.
3715 (vect_recog_func_ptr): Move from tree-vectorizer.h
3716 (vect_vect_recog_func_ptrs): Move further down the file.
3717 (vect_recog_func): Likewise. Remove the third argument.
3718 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
3719 (vect_pattern_recog_1): Expect the pattern function to do any
3720 necessary target tests. Also expect it to provide a vector type.
3721 Remove the type_in handling.
3723 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3725 * tree-vect-patterns.c (vect_pattern_detected): New function.
3726 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
3727 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
3728 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
3729 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
3730 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
3731 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
3732 (vect_recog_mask_conversion_pattern)
3733 (vect_try_gather_scatter_pattern): Likewise.
3735 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3737 * tree-vect-patterns.c (vect_get_internal_def): New function.
3738 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
3739 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
3740 (search_type_for_mask_1): Use it.
3742 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3744 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
3745 redundant WIDEN_SUM_EXPR handling.
3746 (vect_recog_sad_pattern): Likewise.
3748 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3750 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
3751 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
3752 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
3753 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
3754 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
3756 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3758 * tree-vect-stmts.c (vectorizable_call): Make sure that we
3759 use the stmt_vec_info of the original bb statement for the
3760 new zero assignment, even if the call is part of a pattern.
3762 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3764 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
3765 that the sequence is attached to the original statement rather
3766 than the pattern statement.
3767 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
3768 PATTERN_DEF_SEQ from the original statement rather than
3769 the main pattern statement.
3770 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
3771 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
3772 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
3774 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
3776 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
3777 definition statements before the early exit for statements that aren't
3779 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
3781 (vect_transform_loop): ...here. Process pattern definition
3782 statements without first checking whether the main pattern
3783 statement is live or relevant.
3785 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
3787 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
3788 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
3790 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
3792 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
3793 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
3794 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
3795 (expand_block_compare): Change select_block_compare_mode call.
3796 (expand_strncmp_align_check): Use new functions, fix comment.
3797 (emit_final_str_compare_gpr): New function.
3798 (expand_strn_compare): Refactor and clean up code.
3799 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
3801 2018-06-19 Tony Reix <tony.reix@atos.com>
3802 Damien Bergamini <damien.bergamini@atos.com>
3803 David Edelsohn <dje.gcc@gmail.com>
3805 * collect2.c (static_obj): New variable.
3806 (static_libs): New variable.
3807 (is_in_list): Uncomment declaration.
3808 (main): Track AIX libraries linked statically.
3809 (is_in_list): Uncomment definition.
3810 (scan_prog_file): Don't add AIX shared libraries initializer
3811 to constructor list if linking statically.
3813 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
3815 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
3817 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
3819 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
3821 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
3824 2018-06-19 Martin Liska <mliska@suse.cz>
3826 * config/i386/i386.c (ix86_can_inline_p): Do not use
3827 ipa_fn_summaries::get_create.
3828 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
3830 (devirtualization_time_bonus): Likewise.
3831 (ipcp_propagate_stage): Likewise.
3832 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3833 (edge_set_predicate): Likewise.
3834 (evaluate_conditions_for_known_args): Likewise.
3835 (evaluate_properties_for_edge): Likewise.
3836 (ipa_call_summary::reset): Tranform to ...
3837 (ipa_call_summary::~ipa_call_summary): ... this.
3838 (ipa_fn_summary::reset): Transform to ...
3839 (ipa_fn_summary::~ipa_fn_summary): ... this.
3840 (ipa_fn_summary_t::remove): Rename to ...
3841 (ipa_fn_summary_t::remove_callees): ... this.
3842 (ipa_fn_summary_t::duplicate): Use placement new
3843 instead of memory copy.
3844 (ipa_call_summary_t::duplicate): Likewise.
3845 (ipa_call_summary_t::remove): Remove.
3846 (dump_ipa_call_summary): Change get_create to get.
3847 (ipa_dump_fn_summary): Dump only when summary exists.
3848 (analyze_function_body): Use symbol_summary::get instead
3850 (compute_fn_summary): Likewise.
3851 (estimate_edge_devirt_benefit): Likewise.
3852 (estimate_edge_size_and_time): Likewise.
3853 (inline_update_callee_summaries): Likewise.
3854 (remap_edge_change_prob): Likewise.
3855 (remap_edge_summaries): Likewise.
3856 (ipa_merge_fn_summary_after_inlining): Likewise.
3857 (write_ipa_call_summary): Likewise.
3858 (ipa_fn_summary_write): Likewise.
3859 (ipa_free_fn_summary): Likewise.
3860 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
3861 (struct ipa_call_summary): Likewise.
3862 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
3864 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
3865 (estimate_size_after_inlining): Likewise.
3866 (estimate_growth): Likewise.
3867 (growth_likely_positive): Likewise.
3868 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3869 (inline_call): Likewise.
3870 * ipa-inline.c (caller_growth_limits): Likewise.
3871 (can_inline_edge_p): Likewise.
3872 (can_inline_edge_by_limits_p): Likewise.
3873 (compute_uninlined_call_time): Likewise.
3874 (compute_inlined_call_time): Likewise.
3875 (want_inline_small_function_p): Likewise.
3876 (edge_badness): Likewise.
3877 (update_caller_keys): Likewise.
3878 (update_callee_keys): Likewise.
3879 (inline_small_functions): Likewise.
3880 (inline_to_all_callers_1): Likewise.
3881 (dump_overall_stats): Likewise.
3882 (early_inline_small_functions): Likewise.
3883 (early_inliner): Likewise.
3884 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3885 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3886 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3887 * ipa-split.c (execute_split_functions): Likewise.
3888 * symbol-summary.h: Likewise.
3889 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3891 2018-06-19 Richard Biener <rguenther@suse.de>
3893 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
3894 (vectorize_loops): ... here. Fix dbgcnt handling.
3895 (try_vectorize_loop): Wrap try_vectorize_loop_1.
3897 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
3900 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
3901 ieee128 argument takes up only one (vector) register, not two (floating
3904 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
3906 * gimplify.c (gimplify_init_constructor): Really never clear for an
3907 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
3909 2018-06-19 Richard Biener <rguenther@suse.de>
3911 PR tree-optimization/86179
3912 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
3913 after failed recognition.
3915 2018-06-18 Martin Sebor <msebor@redhat.com>
3918 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
3919 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
3920 Handle integer subtraction.
3921 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
3922 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
3924 2018-06-18 David Malcolm <dmalcolm@redhat.com>
3926 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
3927 param from rtx to rtx_insn *.
3928 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
3930 (frv_ifcvt_modify_insn): Likwise.
3931 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
3932 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
3933 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
3934 as_a <rtx_insn *> cast to local "unprotected_region" once
3935 it's been established that it's not NULL or pc_rtx.
3936 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
3937 param "sethi" from rtx to rtx_insn *.
3938 (nds32_group_float_insns): Likewise for param "insn".
3939 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
3941 (vax_output_int_subtract): Likewise.
3942 * config/vax/vax.c (vax_output_int_add): Likewise for param
3944 (vax_output_int_subtract): Likewise.
3945 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
3946 (emit_pattern_after): Likewise for param "after".
3947 (emit_insn_after): Likewise.
3948 (emit_jump_insn_after): Likewise.
3949 (emit_call_insn_after): Likewise.
3950 (emit_debug_insn_after): Likewise.
3951 (emit_pattern_before): Likewise for param "before".
3952 (emit_insn_before): Likewise.
3953 (emit_jump_insn_before): Likewise.
3954 * final.c (get_insn_template): Likewise for param "insn", removing
3956 * output.h (get_insn_template): Likewise for 2nd param.
3957 * rtl.h (emit_insn_before): Likewise.
3958 (emit_jump_insn_before): Likewise.
3959 (emit_debug_insn_before_noloc): Likewise.
3960 (emit_insn_after): Likewise.
3961 (emit_jump_insn_after): Likewise.
3962 (emit_call_insn_after): Likewise.
3963 (emit_debug_insn_after): Likewise.
3964 (set_insn_deleted): Likewise for param.
3966 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
3969 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
3970 floating point modes, so that IFmode is numerically greater than
3971 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
3972 to declare the ordering. This prevents IFmode from being
3973 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
3974 machine. Include rs6000-modes.h to share the fractional values
3975 between genmodes* and the rest of the compiler.
3979 * config/rs6000/rs6000-modes.h: New file.
3980 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
3981 meaning of rs6000_long_double_size so that 126..128 selects an
3982 appropriate 128-bit floating point type.
3983 (rs6000_option_override_internal): Likewise.
3984 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
3985 (TARGET_LONG_DOUBLE_128): Change the meaning of
3986 rs6000_long_double_size so that 126..128 selects an appropriate
3987 128-bit floating point type.
3988 (LONG_DOUBLE_TYPE_SIZE): Update comment.
3989 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
3990 source and destination to match the standard usage.
3991 (truncifkf2): Likewise.
3992 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
3993 ISA 2.07 to use an explicit clobber, instead of passing in a
3995 (copysign<mode>3_soft): Likewise.
3997 2018-06-18 David Malcolm <dmalcolm@redhat.com>
3999 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
4000 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
4001 (vect_slp_analyze_instance_dependence): Likewise.
4002 (vect_enhance_data_refs_alignment): Likewise.
4003 (vect_analyze_data_refs_alignment): Likewise.
4004 (vect_slp_analyze_and_verify_instance_alignment
4005 (vect_analyze_data_ref_accesses): Likewise.
4006 (vect_prune_runtime_alias_test_list): Likewise.
4007 (vect_analyze_data_refs): Likewise.
4008 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
4009 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
4010 (vect_analyze_scalar_cycles_1): Likewise.
4011 (vect_get_loop_niters): Likewise.
4012 (vect_analyze_loop_form_1): Likewise.
4013 (vect_update_vf_for_slp): Likewise.
4014 (vect_analyze_loop_operations): Likewise.
4015 (vect_analyze_loop): Likewise.
4016 (vectorizable_induction): Likewise.
4017 (vect_transform_loop): Likewise.
4018 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
4019 * tree-vect-slp.c (vect_analyze_slp): Likewise.
4020 (vect_make_slp_decision): Likewise.
4021 (vect_detect_hybrid_slp): Likewise.
4022 (vect_slp_analyze_operations): Likewise.
4023 (vect_slp_bb): Likewise.
4024 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
4025 (vectorizable_bswap): Likewise.
4026 (vectorizable_call): Likewise.
4027 (vectorizable_simd_clone_call): Likewise.
4028 (vectorizable_conversion): Likewise.
4029 (vectorizable_assignment): Likewise.
4030 (vectorizable_shift): Likewise.
4031 (vectorizable_operation): Likewise.
4032 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
4034 2018-06-18 Martin Sebor <msebor@redhat.com>
4036 PR tree-optimization/81384
4037 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
4038 * builtins.c (expand_builtin_strnlen): New function.
4039 (expand_builtin): Call it.
4040 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
4041 * builtins.def (BUILT_IN_STRNLEN): New.
4042 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
4043 Warn for bounds in excess of maximum object size.
4044 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
4045 single-value ranges. Handle strnlen.
4046 (handle_builtin_strlen): Handle strnlen.
4047 (strlen_check_and_optimize_stmt): Same.
4048 * doc/extend.texi (Other Builtins): Document strnlen.
4050 2018-06-18 Maya Rashish <coypu@sdf.org>
4052 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
4053 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
4054 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
4056 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
4058 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
4060 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4062 * tree.c (escaped_string::escape): Replace cast to char * by
4063 const_cast<char *> (unescaped).
4065 2018-06-18 Nick Clifton <nickc@redhat.com>
4068 * tree.c (escaped_string): New class. Converts an unescaped
4069 string into its escaped equivalent.
4070 (warn_deprecated_use): Use the new class to convert the
4071 deprecation message, if present.
4072 (test_escaped_strings): New self test.
4073 (test_c_tests): Add test_escaped_strings.
4074 * doc/extend.texi (deprecated): Add a note that the
4075 deprecation message is affected by the -fmessage-length
4076 option, and that control characters will be escaped.
4077 (#pragma GCC error): Document this pragma.
4078 (#pragma GCC warning): Likewise.
4079 * doc/invoke.texi (-fmessage-length): Document this option's
4080 effect on the #warning and #error preprocessor directives and
4081 the deprecated attribute.
4083 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
4085 * tree.c (decl_value_expr_lookup): Revert latest change.
4086 (decl_value_expr_insert): Likewise.
4088 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
4090 * gimplify.c (nonlocal_vlas): Delete.
4091 (nonlocal_vla_vars): Likewise.
4092 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
4094 (gimplify_body): Do not create and destroy nonlocal_vlas.
4095 * tree-nested.c: Include diagnostic.h.
4096 (use_pointer_in_frame): Tweak.
4097 (lookup_field_for_decl): Add assertion and declare the transformation.
4098 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
4099 internal error when the reference is in a wrong context. Do not
4100 create a debug decl by default.
4101 (note_nonlocal_block_vlas): Delete.
4102 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
4103 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
4104 create a debug decl by default.
4105 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
4106 call is in a wrong context.
4107 (fixup_vla_decls): New function.
4108 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
4109 debug variables were created.
4110 * tree.c (decl_value_expr_lookup): Add checking assertion.
4111 (decl_value_expr_insert): Likewise.
4113 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
4116 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
4117 * tree-scalar-evolution.c (interpret_expr): Likewise.
4118 (expression_expensive_p): Likewise.
4119 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
4120 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
4121 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
4122 (ssa_defined_by_minus_one_stmt_p): New.
4124 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
4127 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
4128 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
4129 * dojump.c (do_jump): Likewise.
4130 * expr.c (expand_expr_real_2): Check operand type's sign.
4131 * fold-const.c (const_unop): Handle ABSU_EXPR.
4132 (fold_abs_const): Likewise.
4133 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
4134 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
4135 (strip_sign_op_1): Likesise.
4136 * match.pd: Add new pattern to generate ABSU_EXPR.
4137 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
4138 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4139 * tree-eh.c (operation_could_trap_helper_p): Likewise.
4140 * tree-inline.c (estimate_operator_cost): Likewise.
4141 * tree-pretty-print.c (dump_generic_node): Likewise.
4142 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
4143 * tree.def (ABSU_EXPR): New.
4145 2018-06-16 Jakub Jelinek <jakub@redhat.com>
4148 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
4149 documented as preserved for backward compatibility only.
4150 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
4152 PR rtl-optimization/86108
4153 * bb-reorder.c (create_forwarder_block): Renamed to ...
4154 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
4155 jump from new landing pad to the second part.
4156 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
4159 2018-06-15 Jakub Jelinek <jakub@redhat.com>
4162 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
4163 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
4164 Only call store_expr for halves if the mode is the same.
4167 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
4168 Fix up comment formatting.
4170 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
4172 * typed-splay-tree.h (typed_splay_tree::remove): New function.
4173 (typed_splay_tree::closure,
4174 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
4175 (typed_splay_tree::typed_splay_tree,
4176 typed_splay_tree::operator =): Declared private.
4177 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
4178 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
4179 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
4180 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
4181 typed_splay_tree::splay_tree_splay,
4182 typed_splay_tree::splay_tree_foreach_helper,
4183 typed_splay_tree::splay_tree_insert,
4184 typed_splay_tree::splay_tree_remove,
4185 typed_splay_tree::splay_tree_lookup,
4186 typed_splay_tree::splay_tree_predecessor,
4187 typed_splay_tree::splay_tree_successor,
4188 typed_splay_tree::splay_tree_min,
4189 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
4190 (typed_splay_tree::root, typed_splay_tree::comp,
4191 typed_splay_tree::delete_key,
4192 typed_splay_tree::delete_value): New data members.
4193 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
4194 typed_splay_tree::remove.
4196 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
4198 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
4199 -mginv and -mno-ginv to the assembler.
4200 * config/mips/mips.opt (-mcrc): New option.
4202 * doc/invoke.text (-mcrc): Document.
4205 2018-06-15 Nick Clifton <nickc@redhat.com>
4208 * tree.c (escaped_string): New class. Converts an unescaped
4209 string into its escaped equivalent.
4210 (warn_deprecated_use): Use the new class to convert the
4211 deprecation message, if present.
4212 (test_escaped_strings): New self test.
4213 (test_c_tests): Add test_escaped_strings.
4214 * doc/extend.texi (deprecated): Add a note that the
4215 deprecation message is affected by the -fmessage-length
4216 option, and that control characters will be escaped.
4217 (#pragma GCC error): Document this pragma.
4218 (#pragma GCC warning): Likewise.
4219 * doc/invoke.texi (-fmessage-length): Document this option's
4220 effect on the #warning and #error preprocessor directives and
4221 the deprecated attribute.
4223 2018-06-15 Richard Biener <rguenther@suse.de>
4225 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
4226 here, also noting vector size used.
4227 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
4228 size used in MSG_OPTIMIZED_LOCATIONS dump.
4229 (pass_slp_vectorize::execute): Adjust.
4231 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
4234 * config/arc/arc.c (arc_return_address_register): Fix
4237 2018-06-15 Richard Biener <rguenther@suse.de>
4240 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
4241 leave useless conversion stripping to force_gimple_operand_gsi.
4242 (gimplify_build2): Likewise.
4243 (gimplify_build1): Likewise.
4245 2018-06-15 Richard Biener <rguenther@suse.de>
4248 * tree-cfg.c (move_stmt_op): unshare invariant addresses
4249 before adjusting their block.
4251 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
4253 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
4254 multilibs for *-*-rtems*.
4255 * config/riscv/t-rtems: New file.
4257 2018-06-14 Jakub Jelinek <jakub@redhat.com>
4260 * match.pd ((A +- CST1) +- CST2): Punt if last resort
4261 unsigned_type_for returns NULL.
4264 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
4265 subregs of multi-word pseudos unless the float mode has word size.
4267 2018-06-14 Richard Biener <rguenther@suse.de>
4270 * tree-vect-generic.c (build_word_mode_vector_type): Remove
4271 duplicate and harmful type_hash_canon.
4272 * tree.c (type_hash_canon): Assert we didn't find ourselves.
4274 2018-06-14 Richard Biener <rguenther@suse.de>
4277 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
4280 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
4282 * config/rtems.h (STDINT_LONG32): Define.
4284 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
4285 Prachi Godbole <prachi.godbole@imgtec.com>
4287 * config/mips/mips-cpus.def: Define P6600.
4288 * config/mips/mips-tables.opt: Regenerate.
4289 * config/mips/mips.c (mips_ucbranch_type): New enum.
4290 (mips_rtx_cost_data): Add support for P6600.
4291 (mips_issue_rate): Likewise.
4292 (mips_multipass_dfa_lookahead): Likewise.
4293 (mips_avoid_hazard): Likewise.
4294 (mips_reorg_process_insns): Likewise.
4295 (mips_classify_branch_p6600): New function.
4296 * config/mips/mips.h (TUNE_P6600): New define.
4297 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
4298 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
4299 * config/mips/mips.md: Include p6600.md.
4300 (processor): Add p6600.
4301 * config/mips/p6600.md: New file.
4302 * doc/invoke.texi: Add p6600 to supported architectures.
4304 2018-06-13 Martin Sebor <msebor@redhat.com>
4306 PR tree-optimization/86114
4307 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
4309 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
4311 2018-06-13 Richard Biener <rguenther@suse.de>
4313 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
4314 Properly set vector type of the intermediate stmt.
4315 * tree-vect-stmts.c (vectorizable_operation): The destination
4316 var always has vectype_out type.
4318 2018-06-13 Jeff Law <law@redhat.com>
4320 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
4321 integer 0 for argument to print_rtl_with_bb.
4322 (rl78_reorg): Likewise.
4324 2018-06-13 David Malcolm <dmalcolm@redhat.com>
4326 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
4327 from rtx to rtx_insn *.
4328 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
4330 (add_sched_insns_for_speculation): Likewise for local "target",
4331 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
4332 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
4333 from rtx_insn ** to rtx_code_label **.
4334 (reorg_emit_nops): Likewise.
4335 (c6x_reorg): Likewise for local "call_labels".
4336 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
4338 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
4339 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
4340 the loops over LABEL_REFS.
4341 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
4343 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
4344 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
4345 (split_branches): Strengthen local "olabel" from rtx to
4346 rtx_insn *, adding a safe_as_a cast.
4347 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
4349 (add_insn_after): Likewise for first two params.
4350 (add_insn_before): Likewise.
4351 (remove_insn): Likewise for param.
4352 (emit_pattern_before_noloc): Likewise for second and third params.
4353 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
4354 (emit_call_insn_before_noloc): Likewise.
4355 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
4357 (emit_barrier_before): Likewise.
4358 (emit_label_before): Strengthen "label" param from "rtx" to
4359 "rtx_code_label *". Strengthen "before" param from "rtx" to
4361 (emit_insn_after_1): Strengthen "after" param from "rtx" to
4363 (emit_pattern_after_noloc): Likewise.
4364 (emit_insn_after_noloc): Likewise.
4365 (emit_jump_insn_after_noloc): Likewise.
4366 (emit_call_insn_after_noloc): Likewise.
4367 (emit_debug_insn_after_noloc): Likewise.
4368 (emit_barrier_after): Likewise.
4369 (emit_label_after): Likewise for both params.
4370 (emit_pattern_after_setloc): Likewise for "after" param. Convert
4371 "loc" param from "int" to "location_t".
4372 (emit_insn_after_setloc): Likewise.
4373 (emit_jump_insn_after_setloc): Likewise.
4374 (emit_call_insn_after_setloc): Likewise.
4375 (emit_debug_insn_after_setloc): Likewise.
4376 (emit_pattern_before_setloc): Likewise for "before" param. Convert
4377 "loc" param from "int" to "location_t".
4378 (emit_pattern_before): Convert NULL_RTX to NULL.
4379 (emit_insn_before_setloc): Convert "loc" param from "int" to
4381 (emit_jump_insn_before_setloc): Likewise.
4382 (emit_call_insn_before_setloc): Likewise.
4383 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
4384 rtx_insn *. Convert "loc" param from "int" to "location_t".
4385 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
4386 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
4387 Convert 3rd param from "int" to "location_t".
4388 (emit_barrier_before, emit_barrier_after, next_real_insn):
4389 Strengthen param from rtx to rtx_insn *.
4390 (emit_label_before): Strengthen 1st param from "rtx" to
4391 "rtx_code_label *". Strengthen 2nd param from "rtx" to
4393 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
4394 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
4395 Strengthen 2nd param from "rtx" to "rtx_insn *".
4396 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
4397 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
4398 Likewise. Convert 3rd param from "int" to "location_t".
4399 (emit_label_after): Strengthen 1st param from "rtx" to
4401 (next_real_insn, remove_insn): Strengthen param from "rtx" to
4403 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
4404 from "rtx" to "rtx_insn *".
4406 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
4408 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
4409 bodies streamed in with -Q.
4410 * dumpfile.c (dump_files): Add lto-stream-out dump file.
4411 * dumpfile.h (tree_dump_index): Add lto_stream_out.
4412 * gimple-streamer-out.c: Include gimple-pretty-print.h
4413 (output_bb): Dump stmts streamed.
4414 * lto-section-out.c: Include print-tree.h
4415 (lto_begin_section): Dump sections created.
4416 (lto_output_decl_index): Dump decl encoded.
4417 * lto-streamer-out.c: Include print-tree.h
4418 (create_output_block): Dump output block created.
4419 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
4420 (output_function): Dump function output.
4421 (output_constructor): Dump constructor streamed.
4422 (write_global_stream): Output indexes encoded.
4423 (produce_asm_for_decls): Dump streams encoded.
4424 * lto-streamer.c (streamer_dump_file): New global var.
4425 * lto-streamer.h (streamer_dump_file): Declare.
4426 * passes.c (ipa_write_summaries): Initialize streamer dump.
4427 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
4430 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
4433 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
4434 offsets for register save directives. Emit a second batch of save
4435 directives, if need be, when the function accesses prior frames.
4437 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4439 * config/arc/fpu.md (fmasf4): Force operand to register.
4440 (fnmasf4): Likewise.
4442 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4444 * config/arc/arc-protos.h (arc_pad_return): Remove.
4445 * config/arc/arc.c (machine_function): Remove force_short_suffix
4447 (arc_print_operand): Adjust printing of '&'.
4448 (arc_verify_short): Remove conditional printing of short suffix.
4449 (arc_final_prescan_insn): Remove reference to size_reason.
4450 (pad_return): New function.
4451 (arc_reorg): Call pad_return.
4452 (arc_pad_return): Remove.
4453 (arc_init_machine_status): Remove reference to force_short_suffix.
4454 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
4455 (attr length): When attribute iscompact is true force to 2
4456 regardless; in the case of maybe check if we want to force the
4457 instruction to have 4 bytes length.
4458 (nopv): Change it to generate 4 byte long nop as well.
4459 (blockage): New pattern.
4460 (simple_return): Remove call to arc_pad_return.
4461 (p_return_i): Likewise.
4463 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4465 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
4467 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4469 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
4472 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4474 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
4475 for ARC700 and ARCv2.
4477 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
4480 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
4481 operands[2] instead of operands[1].
4484 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
4486 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
4487 case, check whether the outer register overlaps an unallocatable
4488 register, not just whether it fits the required class.
4490 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
4492 * poly-int.h (can_div_trunc_p): Add new overload in which all values
4494 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
4495 (memrefs_conflict_p): Likewise.
4496 (init_alias_analysis): Likewise.
4497 * cfgexpand.c (expand_debug_expr): Likewise.
4498 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
4499 * cse.c (fold_rtx): Likewise.
4500 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
4501 * expr.c (emit_block_move_hints): Likewise.
4502 (clear_storage_hints, push_block, emit_push_insn): Likewise.
4503 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
4504 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
4505 (emit_group_store): Likewise.
4506 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
4507 to read the PRE/POST_MODIFY increment.
4508 * calls.c (store_one_arg): Use strip_offset.
4509 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
4511 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
4513 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
4514 (simplify_binary_operation_1): Extend CONST_INT handling to
4516 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
4517 than a HOST_WIDE_INT.
4518 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
4520 (adjust_mems, add_stores): Update accodingly.
4521 (vt_canonicalize_addr): Track polynomial offsets.
4522 (emit_note_insn_var_location): Likewise.
4523 (vt_add_function_parameter): Likewise.
4524 (vt_initialize): Likewise.
4526 2018-06-12 Jeff Law <law@redhat.com>
4528 * config.gcc (alpha*-*-freebsd*): Remove.
4529 * config/alpha/freebsd.h: Remove.
4531 2018-06-12 David Malcolm <dmalcolm@redhat.com>
4534 * spellcheck-tree.c (levenshtein_distance): Rename to...
4535 (get_edit_distance): ...this, and update for underlying renaming.
4536 * spellcheck-tree.h (levenshtein_distance): Rename to...
4537 (get_edit_distance): ...this.
4538 * spellcheck.c (levenshtein_distance): Rename to...
4539 (get_edit_distance): ...this. Convert from Levenshtein distance
4540 to Damerau-Levenshtein distance by supporting transpositions of
4541 adjacent characters. Rename "v1" to "v_next" and "v0" to
4543 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
4544 (selftest::test_edit_distance_unit_test_oneway): ...this, and
4545 update for underlying renaming.
4546 (selftest::levenshtein_distance_unit_test): Rename to...
4547 (selftest::test_get_edit_distance_unit): ...this, and update for
4548 underlying renaming.
4549 (selftest::test_find_closest_string): Add example from PR 69968
4550 where transposition helps
4551 (selftest::test_metric_conditions): Update for renaming.
4552 (selftest::test_metric_conditions): Likewise.
4553 (selftest::spellcheck_c_tests): Likewise.
4554 * spellcheck.h (levenshtein_distance): Rename both overloads to...
4555 (get_edit_distance): ...this.
4556 (best_match::consider): Update for renaming.
4558 2018-06-12 Martin Sebor <msebor@redhat.com>
4560 PR tree-optimization/85259
4561 * builtins.c (compute_objsize): Handle constant offsets.
4562 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
4563 true iff a warning has been issued.
4564 * gimple.h (gimple_nonartificial_location): New function.
4565 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
4566 gimple_nonartificial_location and handle -Wno-system-headers.
4567 (handle_builtin_stxncpy): Same.
4569 2018-06-12 Martin Sebor <msebor@redhat.com>
4572 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
4574 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
4576 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
4577 BUILTIN_VEC_XST entries for pointer to double and long long.
4579 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
4582 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
4584 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
4587 2018-06-12 Martin Liska <mliska@suse.cz>
4589 * doc/options.texi: Document IntegerRange.
4591 2018-06-12 Martin Liska <mliska@suse.cz>
4593 * config/i386/i386.opt: Make MPX-related options as Deprecated.
4594 * opt-functions.awk: Handle Deprecated flag.
4595 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
4597 (read_cmdline_option): Report warning for a deprecated option.
4598 * opts.h (struct cl_option): Add new field cl_deprecated.
4599 (CL_ERR_DEPRECATED): New.
4601 2018-06-12 Martin Liska <mliska@suse.cz>
4603 * doc/options.texi: Document Deprecated option flag.
4605 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4607 * config/arc/arc-arch.h (arc_extras): New enum.
4608 (arc_cpu_t):Add field extra.
4609 (arc_cpu_types): Consider the extras.
4610 * config/arc/arc-cpus.def: Add extras info.
4611 * config/arc/arc-opts.h (processor_type): Consider extra field.
4612 * config/arc/arc.c (arc_override_options): Handle extra field.
4614 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4616 * config/arc/arc-arch.h: Update ARC_OPTX macro.
4617 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
4619 * config/arc/arc.c (arc_init): Update pic warning.
4620 (irq_range): Update irq range parsing warnings.
4621 (arc_override_options): Update various warning messages.
4622 (arc_handle_aux_attribute): Likewise.
4624 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
4626 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
4628 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4630 * doc/sourcebuild.texi: Document usage of line number 0 in verify
4631 compiler messages directives.
4633 2018-06-12 Matthew Fortune <mfortune@gmail.com>
4635 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
4636 * config/mips/mips-tables.opt: Regenerate.
4637 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
4639 * doc/invoke.texi: Document -march=i6500.
4641 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
4643 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
4644 (i6400_gpmul): Add cpu_unit.
4645 (i6400_gpdiv): Likewise.
4646 (i6400_msa_add_d): Update reservations.
4647 (i6400_msa_int_add) Likewise.
4648 (i6400_msa_short_logic3) Likewise.
4649 (i6400_msa_short_logic2) Likewise.
4650 (i6400_msa_short_logic) Likewise.
4651 (i6400_msa_move) Likewise.
4652 (i6400_msa_cmp) Likewise.
4653 (i6400_msa_short_float2) Likewise.
4654 (i6400_msa_div_d) Likewise.
4655 (i6400_msa_long_logic1) Likewise.
4656 (i6400_msa_long_logic2) Likewise.
4657 (i6400_msa_mult) Likewise.
4658 (i6400_msa_long_float2) Likewise.
4659 (i6400_msa_long_float4) Likewise.
4660 (i6400_msa_long_float5) Likewise.
4661 (i6400_msa_long_float8) Likewise.
4662 (i6400_fpu_fadd): Include frint type.
4663 (i6400_fpu_store): New define_insn_reservation.
4664 (i6400_fpu_load): Likewise.
4665 (i6400_fpu_move): Likewise.
4666 (i6400_fpu_fcmp): Likewise.
4667 (i6400_fpu_fmadd): Likewise.
4668 (i6400_int_mult): Include imul3nc type and update reservation.
4669 (i6400_int_div): Include idiv3 type and update reservation.
4670 (i6400_int_load): Update to check type not move_type.
4671 (i6400_int_store): Likewise.
4672 (i6400_int_prefetch): Set zero latency.
4674 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
4676 * gcc.c: Document new %@{...} sequence.
4677 (LINK_COMMAND_SPEC): Use it for the -L switches.
4678 (cpp_unique_options): Use it for the -I switches.
4679 (at_file_argbuf): New global variable.
4680 (in_at_file): Likewise.
4681 (alloc_args): Create at_file_argbuf.
4682 (clear_args): Truncate at_file_argbuf.
4683 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
4684 (open_at_file): New function.
4685 (close_at_file): Likewise.
4686 (create_at_file): Delete.
4687 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
4690 (validate_switches_from_spec): Deal with %@{...} sequence.
4691 (validate_switches): Likewise.
4692 (driver::finalize): Call clear_args.
4694 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
4696 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
4698 2018-06-11 Martin Sebor <msebor@redhat.com>
4700 * doc/invoke.texi (-Wall): List -Wc++17-compat.
4701 (Wno-class-memaccess): Add @opindex.
4702 (Wno-templates, Wno-multiple-inheritance): Same.
4703 (Wno-virtual-inheritance, Wno-namespaces): Same.
4704 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
4705 (Wno-format-overflow, Wno-format-truncation): Same.
4706 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
4707 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
4708 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
4709 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
4710 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
4711 (Wno-misspelled-isr): Same.
4713 2018-06-11 Martin Sebor <msebor@redhat.com>
4715 * PR tree-optimization/86083
4716 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
4718 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
4720 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
4722 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
4725 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
4726 on the correct operand.
4727 (*movdi_internal64): Ditto.
4729 2018-06-11 Martin Liska <mliska@suse.cz>
4731 PR tree-optimization/86089
4732 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
4734 2018-06-11 Julia Koval <julia.koval@intel.com>
4736 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
4737 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
4738 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
4740 2018-06-11 Olivier Hainque <hainque@adacore.com>
4742 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
4743 for Ada with strict dwarf2.
4745 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
4748 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
4751 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
4753 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
4755 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
4757 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
4760 2018-06-08 Martin Liska <mliska@suse.cz>
4762 * tree-cfg.h (debug_function): Fix argument type to match
4765 2018-06-08 Martin Liska <mliska@suse.cz>
4767 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
4768 Remove usage of MPX-related (and removed) fields.
4769 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
4771 2018-06-08 David Malcolm <dmalcolm@redhat.com>
4773 * cfg.c (debug): Use TDF_NONE rather than 0.
4774 * cfghooks.c (debug): Likewise.
4775 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
4776 (struct dump_option_value_info): Convert to...
4777 (struct kv_pair): ...this template type.
4778 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
4780 (optinfo_verbosity_options): Likewise.
4781 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
4783 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
4784 than int for "optgroup_flags" param.
4785 (dump_generic_expr_loc): Use dump_flags_t rather than int for
4787 (dump_dec): Likewise.
4788 (dump_finish): Use TDF_NONE rather than 0.
4789 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
4790 rather than int for "optgroup_flags" param. Use TDF_NONE rather
4791 than 0. Update for change to option_ptr.
4792 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
4793 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
4794 0. Update for changes to optinfo_verbosity_options and
4796 (opt_info_switch_p): Convert optgroup_flags from int to
4798 (dump_basic_block): Use dump_flags_t rather than int
4799 for "dump_kind" param.
4800 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
4801 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
4802 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
4803 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
4804 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
4805 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
4806 TDF_NONE): Convert from macros to...
4807 (enum dump_flag): ...this new enum.
4808 (dump_flags_t): Update to use enum.
4809 (operator|, operator&, operator~, operator|=, operator&=):
4810 Implement for dump_flags_t.
4811 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
4812 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
4813 Convert from macros to...
4814 (enum optgroup_flag): ...this new enum.
4815 (optgroup_flags_t): New typedef.
4816 (operator|, operator|=): Implement for optgroup_flags_t.
4817 (struct dump_file_info): Convert field "alt_flags" to
4818 dump_flags_t. Convert field "optgroup_flags" to
4820 (dump_basic_block): Use dump_flags_t rather than int for param.
4821 (dump_generic_expr_loc): Likewise.
4822 (dump_dec): Likewise.
4823 (dump_register): Convert param "optgroup_flags" to
4825 (opt_info_enable_passes): Likewise.
4826 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
4828 * gimple-pretty-print.c (debug): Likewise.
4829 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
4830 (merged_store_group::apply_stores): Likewise.
4831 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
4832 * gimple.c (verify_gimple_pp): Likewise.
4833 * graphite-poly.c (print_pbb_body): Likewise.
4834 * passes.c (pass_manager::register_one_dump_file): Convert
4835 local "optgroup_flags" to optgroup_flags_t.
4836 * print-tree.c (print_node): Use TDF_NONE rather than 0.
4838 (debug_body): Likewise.
4839 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
4840 to optgroup_flags_t.
4841 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
4843 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
4844 (convert_mult_to_fma): Likewise.
4845 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
4846 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
4847 * tree-vect-data-refs.c (dump_lower_bound): Convert param
4848 "dump_kind" to dump_flags_t.
4850 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
4852 * config/rs6000/rs6000.c (min, max): Delete.
4854 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
4856 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
4857 -mabi=spe and -mabi=no-spe.
4859 2018-06-08 Martin Liska <mliska@suse.cz>
4861 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
4862 where we expect an existing summary.
4864 2018-06-08 Martin Liska <mliska@suse.cz>
4866 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
4867 * ipa-inline.h (estimate_edge_growth): Likewise.
4869 2018-06-08 Martin Liska <mliska@suse.cz>
4871 * cgraph.c (function_version_hasher::hash): Use
4872 cgraph_node::get_uid ().
4873 (function_version_hasher::equal):
4874 * cgraph.h (cgraph_node::get_uid): New method.
4875 * ipa-inline.c (update_caller_keys): Use
4876 cgraph_node::get_uid ().
4877 (update_callee_keys): Likewise.
4878 * ipa-utils.c (searchc): Likewise.
4879 (ipa_reduced_postorder): Likewise.
4880 * lto-cgraph.c (input_node): Likewise.
4881 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
4882 * symbol-summary.h (symtab_insertion): Likewise.
4883 (symtab_removal): Likewise.
4884 (symtab_duplication): Likewise.
4885 * tree-pretty-print.c (dump_function_header): Likewise.
4886 * tree-sra.c (convert_callers_for_node): Likewise.
4888 2018-06-08 Martin Liska <mliska@suse.cz>
4890 * cgraph.c (symbol_table::create_edge): Always assign a new
4892 (symbol_table::free_edge): Do not recycle numbers.
4893 * cgraph.h (cgraph_edge::get): New method.
4894 * symbol-summary.h (symtab_removal): Use it.
4895 (symtab_duplication): Likewise.
4896 (call_summary::hashable_uid): Remove.
4898 2018-06-08 Martin Liska <mliska@suse.cz>
4900 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
4901 (initialize_growth_caches): Remove.
4902 (free_growth_caches): Likewise.
4903 (do_estimate_edge_time): Use edge_growth_cache.
4904 (do_estimate_edge_size): Likewise.
4905 (do_estimate_edge_hints): Likewise.
4906 * ipa-inline.c (reset_edge_caches): Likewise.
4907 (recursive_inlining): Likewise.
4908 (inline_small_functions): Likewise.
4909 * ipa-inline.h (initialize_growth_caches): Remove.
4910 (estimate_edge_size): Likewise.
4911 (estimate_edge_time): Likewise.
4912 (estimate_edge_hints): Likewise.
4913 (reset_edge_growth_cache): Likewise.
4914 * symbol-summary.h (call_summary::remove): New method.
4916 2018-06-08 Martin Liska <mliska@suse.cz>
4918 * ipa-cp.c (class edge_clone_summary): New summary.
4919 (grow_edge_clone_vectors): Remove.
4920 (ipcp_edge_duplication_hook): Remove.
4921 (class edge_clone_summary_t): New call_summary class.
4922 (ipcp_edge_removal_hook): Remove.
4923 (edge_clone_summary_t::duplicate): New function.
4924 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
4925 (create_specialized_node): Likewise.
4926 (ipcp_driver): Initialize edge_clone_summaries and do not
4929 2018-06-08 Martin Liska <mliska@suse.cz>
4931 * symbol-summary.h (get): New function.
4932 (call_summary::m_initialize_when_cloning): New class member.
4934 2018-06-08 Martin Liska <mliska@suse.cz>
4936 * cgraph.c (cgraph_node::remove): Do not recycle uid.
4937 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
4938 (symbol_table::allocate_cgraph_symbol): Do not set uid.
4939 * passes.c (uid_hash_t): Record removed_nodes by their uids.
4940 (remove_cgraph_node_from_order): Use the removed_nodes set.
4941 (do_per_function_toporder): Likwise.
4942 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
4943 instead of summary_uid.
4944 (symtab_removal): Likewise.
4945 (symtab_duplication): Likewise.
4947 2018-06-08 Martin Liska <mliska@suse.cz>
4949 * ipa-cp.c (ipcp_store_bits_results): Use
4950 ipcp_transformation_sum.
4951 (ipcp_store_vr_results): Likewise.
4952 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
4954 (ipcp_transformation_initialize): ... this.
4955 (ipa_set_node_agg_value_chain):
4956 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
4957 (write_ipcp_transformation_info): Likewise.
4958 (read_ipcp_transformation_info): Likewise.
4959 (ipcp_update_bits): Likewise.
4960 (ipcp_update_vr): Likewise.
4961 (ipcp_transform_function): Likewise.
4962 * ipa-prop.h: Rename ipcp_transformation_summary to
4963 ipcp_transformation.
4964 (class ipcp_transformation_t): New function summary.
4965 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
4966 (ipa_get_agg_replacements_for_node): Likewise.
4968 2018-06-08 Martin Liska <mliska@suse.cz>
4970 * ipa-pure-const.c (struct funct_state_d): Do it class instead
4972 (class funct_state_summary_t): New function_summary class.
4973 (has_function_state): Remove.
4974 (get_function_state): Likewise.
4975 (set_function_state): Likewise.
4976 (add_new_function): Likewise.
4977 (funct_state_summary_t::insert): New function.
4978 (duplicate_node_data): Remove.
4979 (remove_node_data): Remove.
4980 (funct_state_summary_t::duplicate): New function.
4981 (register_hooks): Create new funct_state_summaries.
4982 (pure_const_generate_summary): Use it.
4983 (pure_const_write_summary): Likewise.
4984 (pure_const_read_summary): Likewise.
4985 (propagate_pure_const): Likewise.
4986 (propagate_nothrow): Likewise.
4987 (dump_malloc_lattice): Likewise.
4988 (propagate_malloc): Likewise.
4989 (execute): Do not register hooks, just remove summary
4991 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
4994 2018-06-08 Martin Liska <mliska@suse.cz>
4996 * ipa-reference.c (remove_node_data): Remove.
4997 (duplicate_node_data): Likewise.
4998 (class ipa_ref_var_info_summary_t): New class.
4999 (class ipa_ref_opt_summary_t): Likewise.
5000 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
5001 (get_reference_optimization_summary): Use
5002 ipa_ref_opt_sum_summaries.
5003 (set_reference_vars_info): Remove.
5004 (set_reference_optimization_summary): Likewise.
5005 (ipa_init): Create summaries.
5006 (init_function_info): Use function summary.
5007 (ipa_ref_opt_summary_t::duplicate): New function.
5008 (ipa_ref_opt_summary_t::remove): New function.
5009 (get_read_write_all_from_node): Fix GNU coding style.
5010 (propagate): Use function summary.
5011 (write_node_summary_p): Fix GNU coding style.
5012 (stream_out_bitmap): Likewise.
5013 (ipa_reference_read_optimization_summary): Use function summary.
5014 (ipa_reference_c_finalize): Do not release hooks.
5016 2018-06-08 Martin Liska <mliska@suse.cz>
5018 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
5019 (analyze_function_body): Extract multiple calls of get_create.
5020 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
5021 * ipa-inline.c (recursive_inlining): Use ::get method.
5022 * ipa-inline.h (estimate_edge_growth): Likewise.
5024 2018-06-08 Martin Liska <mliska@suse.cz>
5026 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
5028 (hsa_function_summary::hsa_function_summary): Use the new enum
5030 (hsa_gpu_implementation_p): Use hsa_summaries::get.
5031 * hsa-gen.c (hsa_get_host_function): Likewise.
5032 (get_brig_function_name): Likewise.
5033 * ipa-hsa.c (process_hsa_functions): Likewise.
5034 (ipa_hsa_write_summary): Likewise.
5035 * symbol-summary.h (symtab_duplication): Use ::get function/
5036 (get): New function.
5038 2018-06-08 Martin Liska <mliska@suse.cz>
5040 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
5042 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
5043 (hsa_register_kernel): Likewise.
5044 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
5045 * hsa-gen.c (hsa_get_host_function): Likewise.
5046 (get_brig_function_name): Likewise.
5047 (generate_hsa): Likewise.
5048 (pass_gen_hsail::execute): Likewise.
5049 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
5050 (devirtualization_time_bonus): Likewise.
5051 (ipcp_propagate_stage): Likewise.
5052 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
5053 (edge_set_predicate): Likewise.
5054 (evaluate_conditions_for_known_args): Likewise.
5055 (evaluate_properties_for_edge): Likewise.
5056 (ipa_fn_summary::reset): Likewise.
5057 (ipa_fn_summary_t::duplicate): Likewise.
5058 (dump_ipa_call_summary): Likewise.
5059 (ipa_dump_fn_summary): Likewise.
5060 (analyze_function_body): Likewise.
5061 (compute_fn_summary): Likewise.
5062 (estimate_edge_devirt_benefit): Likewise.
5063 (estimate_edge_size_and_time): Likewise.
5064 (estimate_calls_size_and_time): Likewise.
5065 (estimate_node_size_and_time): Likewise.
5066 (inline_update_callee_summaries): Likewise.
5067 (remap_edge_change_prob): Likewise.
5068 (remap_edge_summaries): Likewise.
5069 (ipa_merge_fn_summary_after_inlining): Likewise.
5070 (ipa_update_overall_fn_summary): Likewise.
5071 (read_ipa_call_summary): Likewise.
5072 (inline_read_section): Likewise.
5073 (write_ipa_call_summary): Likewise.
5074 (ipa_fn_summary_write): Likewise.
5075 (ipa_free_fn_summary): Likewise.
5076 * ipa-hsa.c (process_hsa_functions): Likewise.
5077 (ipa_hsa_write_summary): Likewise.
5078 (ipa_hsa_read_section): Likewise.
5079 * ipa-icf.c (sem_function::merge): Likewise.
5080 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
5081 (do_estimate_edge_time): Likewise.
5082 (estimate_size_after_inlining): Likewise.
5083 (estimate_growth): Likewise.
5084 (growth_likely_positive): Likewise.
5085 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
5086 (inline_call): Likewise.
5087 * ipa-inline.c (caller_growth_limits): Likewise.
5088 (can_inline_edge_p): Likewise.
5089 (can_inline_edge_by_limits_p): Likewise.
5090 (compute_uninlined_call_time): Likewise.
5091 (compute_inlined_call_time): Likewise.
5092 (want_inline_small_function_p): Likewise.
5093 (edge_badness): Likewise.
5094 (update_caller_keys): Likewise.
5095 (update_callee_keys): Likewise.
5096 (recursive_inlining): Likewise.
5097 (inline_small_functions): Likewise.
5098 (inline_to_all_callers_1): Likewise.
5099 (dump_overall_stats): Likewise.
5100 (early_inline_small_functions): Likewise.
5101 (early_inliner): Likewise.
5102 * ipa-inline.h (estimate_edge_growth): Likewise.
5103 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
5104 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5105 * ipa-prop.h (IPA_NODE_REF): Likewise.
5106 (IPA_EDGE_REF): Likewise.
5107 * ipa-pure-const.c (malloc_candidate_p): Likewise.
5108 (propagate_malloc): Likewise.
5109 * ipa-split.c (execute_split_functions): Likewise.
5110 * symbol-summary.h: Rename get to get_create.
5112 (get_create): Likewise.
5113 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
5115 2018-06-08 Martin Liska <mliska@suse.cz>
5117 * symbol-summary.h (release): Move definition out of class
5119 (symtab_removal): Likewise.
5120 (symtab_duplication): Likewise.
5122 2018-06-08 Martin Liska <mliska@suse.cz>
5124 * symbol-summary.h (function_summary): Move constructor
5125 implementation out of class declaration.
5126 (release): Likewise.
5127 (symtab_insertion): Likewise.
5128 (symtab_removal): Likewise.
5129 (symtab_duplication): Likewise.
5132 2018-06-08 Martin Liska <mliska@suse.cz>
5134 * Makefile.in: Remove support for MPX (macros, related functions,
5135 fields in cgraph_node, ...).
5136 * builtin-types.def (BT_BND): Likewise.
5137 (BT_FN_BND_CONST_PTR): Likewise.
5138 (BT_FN_CONST_PTR_BND): Likewise.
5139 (BT_FN_VOID_PTR_BND): Likewise.
5140 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
5141 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
5142 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
5143 (expand_builtin_mempcpy_with_bounds): Likewise.
5144 (expand_builtin_memset_with_bounds): Likewise.
5145 (expand_builtin_memset_args): Likewise.
5146 (std_expand_builtin_va_start): Likewise.
5147 (expand_builtin): Likewise.
5148 (expand_builtin_with_bounds): Likewise.
5149 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
5150 (DEF_LIB_BUILTIN_CHKP): Likewise.
5151 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
5152 (DEF_CHKP_BUILTIN): Likewise.
5153 (BUILT_IN_MEMCPY): Likewise.
5154 (BUILT_IN_MEMMOVE): Likewise.
5155 (BUILT_IN_MEMPCPY): Likewise.
5156 (BUILT_IN_MEMSET): Likewise.
5157 (BUILT_IN_STPCPY): Likewise.
5158 (BUILT_IN_STRCAT): Likewise.
5159 (BUILT_IN_STRCHR): Likewise.
5160 (BUILT_IN_STRCPY): Likewise.
5161 (BUILT_IN_STRLEN): Likewise.
5162 (BUILT_IN_MEMCPY_CHK): Likewise.
5163 (BUILT_IN_MEMMOVE_CHK): Likewise.
5164 (BUILT_IN_MEMPCPY_CHK): Likewise.
5165 (BUILT_IN_MEMSET_CHK): Likewise.
5166 (BUILT_IN_STPCPY_CHK): Likewise.
5167 (BUILT_IN_STRCAT_CHK): Likewise.
5168 (BUILT_IN_STRCPY_CHK): Likewise.
5169 * calls.c (store_bounds): Likewise.
5170 (emit_call_1): Likewise.
5171 (special_function_p): Likewise.
5172 (maybe_warn_nonstring_arg): Likewise.
5173 (initialize_argument_information): Likewise.
5174 (finalize_must_preallocate): Likewise.
5175 (compute_argument_addresses): Likewise.
5176 (expand_call): Likewise.
5177 * cfgexpand.c (expand_call_stmt): Likewise.
5178 (expand_return): Likewise.
5179 (expand_gimple_stmt_1): Likewise.
5180 (pass_expand::execute): Likewise.
5181 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
5182 (cgraph_node::remove): Likewise.
5183 (cgraph_node::dump): Likewise.
5184 (cgraph_node::verify_node): Likewise.
5185 * cgraph.h (chkp_function_instrumented_p): Likewise.
5186 (symtab_node::get_alias_target): Likewise.
5187 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
5188 (cgraph_local_p): Likewise.
5189 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
5190 (cgraph_edge::rebuild_references): Likewise.
5191 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
5192 (walk_polymorphic_call_targets): Likewise.
5193 (cgraph_node::expand_thunk): Likewise.
5194 (symbol_table::output_weakrefs): Likewise.
5195 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
5196 (ix86_handle_option): Likewise.
5197 * config/i386/constraints.md: Likewise.
5198 * config/i386/i386-builtin-types.def (BND): Likewise.
5202 * config/i386/i386-builtin.def (BDESC_END): Likewise.
5203 (BDESC_FIRST): Likewise.
5205 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
5206 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
5207 * config/i386/i386.c (enum reg_class): Likewise.
5208 (ix86_target_string): Likewise.
5209 (ix86_option_override_internal): Likewise.
5210 (ix86_conditional_register_usage): Likewise.
5211 (ix86_valid_target_attribute_inner_p): Likewise.
5212 (ix86_set_indirect_branch_type): Likewise.
5213 (ix86_set_current_function): Likewise.
5214 (ix86_function_arg_regno_p): Likewise.
5215 (init_cumulative_args): Likewise.
5216 (ix86_function_arg_advance): Likewise.
5217 (ix86_function_arg): Likewise.
5218 (ix86_pass_by_reference): Likewise.
5219 (ix86_function_value_regno_p): Likewise.
5220 (ix86_function_value_1): Likewise.
5221 (ix86_function_value_bounds): Likewise.
5222 (ix86_return_in_memory): Likewise.
5223 (ix86_setup_incoming_vararg_bounds): Likewise.
5224 (ix86_va_start): Likewise.
5225 (indirect_thunk_need_prefix): Likewise.
5226 (print_reg): Likewise.
5227 (ix86_print_operand): Likewise.
5228 (ix86_expand_call): Likewise.
5229 (ix86_output_function_return): Likewise.
5230 (reg_encoded_number): Likewise.
5231 (BDESC_VERIFYS): Likewise.
5232 (ix86_init_mpx_builtins): Likewise.
5233 (ix86_init_builtins): Likewise.
5234 (ix86_emit_cmove): Likewise.
5235 (ix86_emit_move_max): Likewise.
5236 (ix86_expand_builtin): Likewise.
5237 (ix86_builtin_mpx_function): Likewise.
5238 (ix86_get_arg_address_for_bt): Likewise.
5239 (ix86_load_bounds): Likewise.
5240 (ix86_store_bounds): Likewise.
5241 (ix86_load_returned_bounds): Likewise.
5242 (ix86_store_returned_bounds): Likewise.
5243 (ix86_class_likely_spilled_p): Likewise.
5244 (ix86_hard_regno_mode_ok): Likewise.
5245 (x86_order_regs_for_local_alloc): Likewise.
5246 (ix86_mitigate_rop): Likewise.
5247 (ix86_bnd_prefixed_insn_p): Likewise.
5248 (ix86_mpx_bound_mode): Likewise.
5249 (ix86_make_bounds_constant): Likewise.
5250 (ix86_initialize_bounds): Likewise.
5251 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
5252 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
5253 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
5254 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
5255 (TARGET_CHKP_BOUND_MODE): Likewise.
5256 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
5257 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
5258 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
5259 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
5260 * config/i386/i386.h (TARGET_MPX): Likewise.
5261 (TARGET_MPX_P): Likewise.
5262 (VALID_BND_REG_MODE): Likewise.
5263 (FIRST_BND_REG): Likewise.
5264 (LAST_BND_REG): Likewise.
5265 (enum reg_class): Likewise.
5266 (BND_REG_P): Likewise.
5267 (BND_REGNO_P): Likewise.
5268 (BNDmode): Likewise.
5269 (ADJUST_INSN_LENGTH): Likewise.
5270 * config/i386/i386.md: Likewise.
5271 * config/i386/i386.opt: Likewise.
5272 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
5273 (defined): Likewise.
5274 (LINK_MPX): Likewise.
5275 (MPX_SPEC): Likewise.
5276 (LIBMPX_SPEC): Likewise.
5277 (LIBMPXWRAPPERS_SPEC): Likewise.
5278 (CHKP_SPEC): Likewise.
5279 * config/i386/predicates.md: Likewise.
5280 * dbxout.c (dbxout_type): Likewise.
5281 * doc/extend.texi: Likewise.
5282 * doc/invoke.texi: Likewise.
5283 * doc/md.texi: Likewise.
5284 * doc/tm.texi: Likewise.
5285 * doc/tm.texi.in: Likewise.
5286 * dwarf2out.c (is_base_type): Likewise.
5287 (gen_formal_types_die): Likewise.
5288 (gen_subprogram_die): Likewise.
5289 (gen_type_die_with_usage): Likewise.
5290 (gen_decl_die): Likewise.
5291 (dwarf2out_late_global_decl): Likewise.
5292 * expr.c (expand_assignment): Likewise.
5293 (emit_storent_insn): Likewise.
5294 (store_expr_with_bounds): Likewise.
5295 (store_expr): Likewise.
5296 (expand_expr_real_1): Likewise.
5297 * expr.h (store_expr_with_bounds): Likewise.
5298 * function.c (use_register_for_decl): Likewise.
5299 (struct bounds_parm_data): Likewise.
5300 (assign_parms_augmented_arg_list): Likewise.
5301 (assign_parm_find_entry_rtl): Likewise.
5302 (assign_parm_is_stack_parm): Likewise.
5303 (assign_parm_load_bounds): Likewise.
5304 (assign_bounds): Likewise.
5305 (assign_parms): Likewise.
5306 (expand_function_start): Likewise.
5307 * gcc.c (CHKP_SPEC): Likewise.
5308 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
5309 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
5310 (wrestrict_dom_walker::check_call): Likewise.
5311 * gimple.c (gimple_build_call_from_tree): Likewise.
5312 * gimple.h (enum gf_mask): Likewise.
5313 (gimple_call_with_bounds_p): Likewise.
5314 (gimple_call_set_with_bounds): Likewise.
5315 * gimplify.c (gimplify_init_constructor): Likewise.
5316 * ipa-cp.c (initialize_node_lattices): Likewise.
5317 (propagate_constants_across_call): Likewise.
5318 (find_more_scalar_values_for_callers_subset): Likewise.
5319 * ipa-hsa.c (process_hsa_functions): Likewise.
5320 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
5321 * ipa-icf.c (sem_function::merge): Likewise.
5322 * ipa-inline.c (early_inliner): Likewise.
5323 * ipa-pure-const.c (warn_function_noreturn): Likewise.
5324 (warn_function_cold): Likewise.
5325 (propagate_pure_const): Likewise.
5326 * ipa-ref.h (enum GTY): Likewise.
5327 * ipa-split.c (find_retbnd): Likewise.
5328 (consider_split): Likewise.
5329 (split_function): Likewise.
5330 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
5331 * ipa.c (walk_polymorphic_call_targets): Likewise.
5332 (symbol_table::remove_unreachable_nodes): Likewise.
5333 (process_references): Likewise.
5334 (cgraph_build_static_cdtor_1): Likewise.
5335 * lto-cgraph.c (lto_output_node): Likewise.
5336 (output_refs): Likewise.
5337 (compute_ltrans_boundary): Likewise.
5338 (input_overwrite_node): Likewise.
5339 (input_node): Likewise.
5340 (input_cgraph_1): Likewise.
5341 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
5342 * passes.c (pass_manager::execute_early_local_passes): Likewise.
5343 (class pass_chkp_instrumentation_passes): Likewise.
5344 (make_pass_chkp_instrumentation_passes): Likewise.
5345 * passes.def: Likewise.
5346 * rtl.h (struct GTY): Likewise.
5347 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
5348 * stor-layout.c (layout_type): Likewise.
5349 * symtab.c: Likewise.
5350 * target.def: Likewise.
5351 * targhooks.c (default_chkp_bound_type): Likewise.
5352 (default_chkp_bound_mode): Likewise.
5353 (default_builtin_chkp_function): Likewise.
5354 (default_chkp_function_value_bounds): Likewise.
5355 (default_chkp_make_bounds_constant): Likewise.
5356 (default_chkp_initialize_bounds): Likewise.
5357 * targhooks.h (default_chkp_bound_type): Likewise.
5358 (default_chkp_bound_mode): Likewise.
5359 (default_builtin_chkp_function): Likewise.
5360 (default_chkp_function_value_bounds): Likewise.
5361 (default_chkp_make_bounds_constant): Likewise.
5362 (default_chkp_initialize_bounds): Likewise.
5363 * toplev.c (compile_file): Likewise.
5364 (process_options): Likewise.
5365 * tree-core.h (DEF_BUILTIN): Likewise.
5366 (DEF_BUILTIN_CHKP): Likewise.
5367 * tree-inline.c (declare_return_variable): Likewise.
5368 (remap_gimple_stmt): Likewise.
5369 (copy_bb): Likewise.
5370 (initialize_inlined_parameters): Likewise.
5371 (expand_call_inline): Likewise.
5372 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
5373 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
5374 (make_pass_ipa_chkp_produce_thunks): Likewise.
5375 (make_pass_chkp): Likewise.
5376 (make_pass_chkp_opt): Likewise.
5377 (make_pass_chkp_instrumentation_passes): Likewise.
5378 * tree-pretty-print.c (dump_generic_node): Likewise.
5379 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
5380 * tree-ssa-dce.c (propagate_necessity): Likewise.
5381 (eliminate_unnecessary_stmts): Likewise.
5382 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
5383 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
5384 * tree-ssa-sccvn.h: Likewise.
5385 * tree-ssa-strlen.c (get_string_length): Likewise.
5386 (valid_builtin_call): Likewise.
5387 (adjust_last_stmt): Likewise.
5388 (handle_builtin_strchr): Likewise.
5389 (handle_builtin_strcpy): Likewise.
5390 (handle_builtin_stxncpy): Likewise.
5391 (handle_builtin_memcpy): Likewise.
5392 (handle_builtin_strcat): Likewise.
5393 (strlen_check_and_optimize_stmt): Likewise.
5394 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
5395 * tree-streamer-in.c: Likewise.
5396 * tree-streamer.c (record_common_node): Likewise.
5397 * tree.c (tree_code_size): Likewise.
5398 (wide_int_to_tree_1): Likewise.
5399 (type_contains_placeholder_1): Likewise.
5400 (build_common_tree_nodes): Likewise.
5401 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
5402 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
5403 (POINTER_BOUNDS_P): Likewise.
5404 (BOUNDED_TYPE_P): Likewise.
5405 (BOUNDED_P): Likewise.
5406 (CALL_WITH_BOUNDS_P): Likewise.
5407 (pointer_bounds_type_node): Likewise.
5408 * value-prof.c (gimple_ic): Likewise.
5409 * var-tracking.c (vt_add_function_parameters): Likewise.
5410 * varasm.c (make_decl_rtl): Likewise.
5411 (assemble_start_function): Likewise.
5412 (output_constant): Likewise.
5413 (maybe_assemble_visibility): Likewise.
5414 * varpool.c (ctor_for_folding): Likewise.
5415 * chkp-builtins.def: Remove.
5416 * ipa-chkp.c: Remove.
5417 * ipa-chkp.h: Remove.
5418 * rtl-chkp.c: Remove.
5419 * rtl-chkp.h: Remove.
5420 * tree-chkp-opt.c: Remove.
5421 * tree-chkp.c: Remove.
5422 * tree-chkp.h: Remove.
5424 2018-06-07 Carl Love <cel@us.ibm.com>
5426 * config/rs6000/vsx.md (vextract_fp_from_shorth,
5427 vextract_fp_from_shortl): Add BE support.
5429 2018-06-07 Paul Koning <ni1d@arrl.net>
5431 * compare-elim.c (try_merge_compare): Don't merge compare if
5432 address contains a side effect.
5433 (try_eliminate_compare): Likewise.
5435 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
5437 * config.gcc: Support "tremont".
5438 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
5439 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5441 * config/i386/i386.c (m_TREMONT): Define.
5442 (processor_target_table): Add "tremont".
5443 (PTA_TREMONT): Define.
5444 (ix86_lea_outperforms): Add TARGET_TREMONT.
5445 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
5446 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
5447 and M_INTEL_GOLDMONT_PLUS.
5448 (fold_builtin_cpu): Add "tremont".
5449 (ix86_add_stmt_cost): Add TARGET_TREMONT.
5450 (ix86_option_override_internal): Add "tremont".
5451 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
5452 (processor_type): Add PROCESSOR_TREMONT.
5453 * config/i386/x86-tune.def: Add m_TREMONT.
5454 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
5456 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5458 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
5459 symbol defined for msp430i* devices to be lower case.
5461 2018-06-07 Richard Biener <rguenther@suse.de>
5463 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
5464 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
5465 Properly wrap signed arithmetic if overflow wraps.
5467 2018-06-07 Jakub Jelinek <jakub@redhat.com>
5469 PR tree-optimization/69615
5470 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
5471 of a cast from a same precision integral SSA_NAME in a bb dominated
5472 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
5473 cast to utype if rhs2 has already a compatible type.
5475 2018-06-07 Richard Biener <rguenther@suse.de>
5477 PR tree-optimization/85935
5478 * graphite-scop-detection.c (find_params_in_bb): Analyze
5479 condition operands with respect to the correct loop. Assert
5480 the analysis doesn't fail.
5482 2018-06-04 Carl Love <cel@us.ibm.com>
5484 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
5485 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
5486 as it is slightly cheaper.
5487 (first_match_or_eos_index_<mode>):
5488 Calculate index using natural element order.
5489 (first_match_index_<mode>):
5490 Calculate index using natural element order.
5491 (first_match_or_eos_index_<mode>):
5492 Calculate index using natural order.
5493 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
5494 for BE and LE modes.
5495 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
5496 P9V_BUILTIN_VCLZLSBB_V16QI.
5497 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
5500 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
5502 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
5503 indentation and line wrap for many prototypes. Add missing
5504 @smallexample directives around block of prototypes for vec_xl and
5507 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
5509 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
5510 track if we pass or return IEEE 128-bit floating point.
5511 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
5512 C++ mangling that is compatible with GCC 8.1.
5513 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
5514 (init_cumulative_args): Note if we pass or return IEEE 128-bit
5515 floating point types.
5516 (rs6000_function_arg_advance_1): Likewise.
5517 (rs6000_mangle_type): Optionally generate mangled names that match
5518 what GCC 8.1 generated for IEEE 128-bit floating point types.
5519 (rs6000_globalize_decl_name): If we have an external function that
5520 passes or returns IEEE 128-bit types, generate a weak reference
5521 from the mangled name used in GCC 8.1 to the current mangled
5523 (rs6000_init_builtins): Make __ibm128 use the long double type if
5524 long double is IBM extended double. Make __float128 use the long
5525 double type if long double is IEEE 128-bit.
5528 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
5529 macro for __ibm128 built-in functions.
5530 (PACK_IF): Add __ibm128 pack/unpack functions.
5531 (UNPACK_IF): Likewise.
5532 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
5533 enable long double built-in functions if long double is IEEE
5534 128-bit floating point.
5535 (rs6000_invalid_builtin): Update long double built-in function
5537 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
5538 functions, adjust the built-in function to use the long double
5539 built-in function if __ibm128 and long double are the same type.
5540 * doc/extend.texi (PowerPC builtins): Update documention for
5541 __builtin_{,un}pack_longdouble. Add documentation for
5542 __builtin_{,un}pack_ibm128.
5544 2018-06-06 Jim Wilson <jimw@sifive.com>
5546 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
5547 (struct machine_function): New field interrupt_mode.
5548 (riscv_handle_type_attribute): New function. Add forward declaration.
5549 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
5550 (riscv_expand_epilogue): Check interrupt_mode field.
5551 (riscv_set_current_function): Check interrupt attribute args and
5552 set interrupt_mode field.
5553 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
5554 (riscv_sret, riscv_uret): New.
5555 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
5556 new arguments to interrupt attribute.
5558 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
5561 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
5562 Don't handle -mcpu=power8 if -mpower9-vector is also used.
5564 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
5566 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
5567 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
5568 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5569 several redundant entries.
5571 2018-06-06 David Malcolm <dmalcolm@redhat.com>
5573 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
5574 type from "rtx" to "rtx_insn *".
5575 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
5576 for local "call_insn", removing cast.
5577 (ix86_expand_call): Likewise, introducing a "call_insn" local.
5579 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
5581 PR tree-optimization/86066
5582 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
5583 for BIT_INSERT_EXPR stores.
5585 2018-06-06 Richard Biener <rguenther@suse.de>
5587 PR tree-optimization/86062
5588 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5589 component refs ontop
5590 of to be offsetted base.
5592 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5594 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
5595 to be static and remove check on interrupt attribute name.
5597 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
5599 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5600 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
5602 2018-06-05 Steve Ellcey <sellcey@cavium.com>
5605 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
5607 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
5608 Remove second argument, change how error is called.
5609 (aarch64_layout_arg): Remove second argument from
5610 aarch64_err_no_fpadvsimd call.
5611 (aarch64_init_cumulative_args): Ditto.
5612 (aarch64_gimplify_va_arg_expr): Ditto.
5613 * config/aarch64/aarch64.md (mov<mode>): Ditto.
5615 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
5617 * config/i386/i386.md (simple_return_indirect_internal): New expander.
5618 (*simple_return_indirect_internal<mode>): Rename from
5619 simple_return_indirect_internal. Use W mode iterator.
5620 (rstorssp): New expander.
5621 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
5622 (clrssbsy): New expander.
5623 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
5625 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
5627 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
5629 (cmse_check_pointed_object): Likewise.
5631 2018-06-05 Martin Liska <mliska@suse.cz>
5633 PR gcov-profile/47618
5634 * doc/invoke.texi: Document how -fprofile-dir format
5637 2018-06-05 Richard Biener <rguenther@suse.de>
5639 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
5640 removal pretend DOM info isn't available so we do not update
5641 it and only remove edges, not dominated blocks. Actually free
5642 DOM info in case we removed something. Remove unreachable blocks.
5643 (mfb_keep_latches): Work with either DOM info or marked backedges.
5644 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
5645 first. Mark backedges if DOM info isn't available.
5646 (Re-)compute DOM info after cleanup_control_flow_pre.
5648 2018-06-05 Richard Biener <rguenther@suse.de>
5650 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
5651 (locus_discrim_hasher::hash): Adjust.
5652 (locus_discrim_hasher::equal): Likewise.
5653 (next_discriminator_for_locus): Work on line directly.
5654 (same_line_p): Pass in expanded locus1 as well.
5655 (assign_discriminators): Avoid redundant location expansions.
5657 2018-06-05 Richard Biener <rguenther@suse.de>
5659 PR tree-optimization/86046
5660 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
5661 if required after clearing TREE_ADDRESSABLE.
5663 2018-06-05 Richard Biener <rguenther@suse.de>
5665 PR tree-optimization/86047
5666 * tree-ssa-loop.c (for_each_index): Glob handling of all
5667 decls and constants and really handle all of them.
5669 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5672 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
5673 qualifier_void_pointer and qualifier_const_void_pointer.
5674 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
5675 (arm_init_builtins): Handle the above.
5676 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
5677 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
5678 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
5681 2018-06-05 Martin Liska <mliska@suse.cz>
5683 * auto-profile.c (read_autofdo_file): Do not use
5684 gcov_ctr_summary struct.
5685 (afdo_callsite_hot_enough_for_early_inline): Likewise.
5686 * coverage.c (struct counts_entry): Likewise.
5687 (read_counts_file): Read just single summary entry.
5688 (get_coverage_counts): Use gcov_summary struct.
5689 * coverage.h (get_coverage_counts): Likewise.
5690 * gcov-dump.c (dump_working_sets): Likewise.
5691 (tag_summary): Dump just single summary.
5692 * gcov-io.c (gcov_write_summary): Write just histogram
5694 (gcov_read_summary): Read just single summary.
5695 (compute_working_sets): Use gcov_summary struct.
5696 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
5697 of GCOV_COUNTERS_SUMMABLE.
5698 (GCOV_COUNTERS_SUMMABLE): Remove.
5699 (GCOV_FIRST_VALUE_COUNTER): Replace with
5700 GCOV_COUNTER_V_INTERVAL.
5701 (struct gcov_ctr_summary): Remove.
5702 (struct gcov_summary): Directly use fields of former
5704 (compute_working_sets): Use gcov_summary struct.
5705 * gcov.c (read_count_file): Do not use ctrs fields.
5706 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
5708 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
5710 * profile.c: Likewise.
5711 * profile.h: Likewise.
5713 2018-06-05 Martin Liska <mliska@suse.cz>
5715 PR gcov-profile/84846
5716 * gcov.c (output_lines): Print working directory only
5717 in intermediate format.
5719 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
5721 * config/s390/s390-builtin-types.def: Add void function type.
5722 * config/s390/s390-builtins.def: Use the function type for the
5725 2018-06-04 Jim Wilson <jimw@sifive.com>
5727 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
5729 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
5730 and maybe_eh_return. Change regno to unsigned int. Use new args to
5731 handle EH_RETURN_DATA_REGNO registers properly.
5732 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
5733 (riscv_expand_epilogue): Update comment. Change argument name and
5734 type. Update code to use new name and type. Pass new args to
5735 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
5737 * config/riscv/riscv.md (NORMAL_RETURN): New.
5738 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
5739 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
5740 (eh_return): Call gen_eh_return_internal and emit barrier.
5741 (eh_return_internal): Call riscv_expand_epilogue.
5743 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
5745 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
5746 bit_insertion field and declare can_be_merged_into method.
5747 (merged_store_group::can_be_merged_into): New method.
5748 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
5749 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
5750 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
5752 2018-06-04 Richard Biener <rguenther@suse.de>
5754 PR tree-optimization/85955
5755 * builtins.c (fold_builtin_sincos): Convert pointers to
5756 destination to appropriate type before dereferencing.
5758 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
5760 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
5762 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
5764 * expr.c (expand_expr_real_1): Force the operand into memory if
5765 its TYPE_MODE is BLKmode and if there is no integer mode for
5766 the number of bits being extracted.
5768 2018-06-04 Jakub Jelinek <jakub@redhat.com>
5772 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5773 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
5775 2018-06-04 Richard Biener <rguenther@suse.de>
5777 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
5778 (cleanup_tree_cfg_noloop): ... single caller. Do
5779 start_recording_case_labels later.
5781 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
5783 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
5784 to _IMMINTRIN_H_INCLUDED.
5785 * config/i386/pconfigintrin.h: Ditto.
5786 * config/i386/waitpkgintrin.h: Ditto.
5787 * config/i386/immintrin.h: Add includes for sgxintrin.h,
5788 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
5789 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
5790 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
5791 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
5792 waitpkgintrin.h and cldemoteintrin.h.
5794 2018-06-04 Richard Biener <rguenther@suse.de>
5796 PR tree-optimization/86038
5797 * tracer.c (find_best_successor): Check probability for
5798 being initialized, bail out if not.
5800 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
5803 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
5804 of bits to ignore when comparing architectures.
5806 2018-06-04 Jakub Jelinek <jakub@redhat.com>
5808 PR tree-optimization/69615
5809 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
5810 maximum or minimum of the type, try to merge it also as if
5811 range1 is + [-, x - 1] or + [x + 1, -].
5814 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
5816 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
5818 PR tree-optimization/86034
5819 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
5820 the unsigned bitfield type in a bit insertion sequence if it does not
5821 have a larger precision than the bitfield size.
5822 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
5824 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
5826 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
5828 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
5830 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
5831 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
5832 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
5833 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
5835 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
5837 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
5838 Disable -fdelete-null-pointer-checks for ELF toolchain.
5840 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
5841 Kito Cheng <kito.cheng@gmail.com>
5843 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
5844 (nds32le-*-*, nds32be-*-*): Integrate checking process.
5845 (nds32*-*-*): Add glibc and uclibc conditions.
5846 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
5847 (TARGET_EXCEPT_UNWIND_INFO): Define.
5848 * config/nds32/elf.h: New file.
5849 * config/nds32/linux.h: New file.
5850 * config/nds32/nds32-elf.opt: New file.
5851 * config/nds32/nds32-linux.opt: New file.
5852 * config/nds32/nds32-fp-as-gp.c
5853 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
5854 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
5856 (nds32_asm_file_end): Ditto.
5857 (nds32_print_operand): Ditto.
5858 (nds32_insert_attributes): Ditto.
5859 (nds32_init_libfuncs): New function.
5860 (TARGET_HAVE_TLS): Define.
5861 (TARGET_INIT_LIBFUNCS): Define.
5862 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
5864 (TARGET_ELF): Apply different mcmodel setting.
5865 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
5866 been migrated into elf.h and linux.h files.
5867 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
5868 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
5869 (mcmodel): The content has been migrated into nds32-elf.opt and
5870 nds32-linux.opt files.
5871 * config/nds32/t-elf: New file.
5872 * config/nds32/t-linux: New file.
5874 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
5875 Shiva Chen <shiva0217@gmail.com>
5877 * config/nds32/constants.md (unspec_volatile_element): Add
5878 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
5879 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
5881 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
5882 (make_pass_nds32_fp_as_gp): Declare.
5883 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
5885 (nds32_asm_function_end_prologue): Remove unused asm output.
5886 (nds32_asm_function_begin_epilogue): Remove unused asm output.
5887 (nds32_asm_file_start): Output necessary fp_as_gp information.
5888 (nds32_option_override): Adjust register usage.
5889 (nds32_expand_prologue): Consider fp_as_gp situation.
5890 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
5891 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
5894 (simple_return): Ditto.
5895 (omit_fp_begin): Output special directive for fp_as_gp.
5896 (omit_fp_end): Output special directive for fp_as_gp.
5897 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
5898 mforbid-fp-as-gp): New options.
5900 2018-06-01 Mark Wielaard <mark@klomp.org>
5902 * dwarf2out.c (dwarf2out_finish): Remove generation of
5903 DW_AT_loclists_base.
5905 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
5907 * gimple-ssa-store-merging.c: Include gimple-fold.h.
5908 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
5909 (struct merged_store_group): Add bit_insertion field.
5910 (dump_char_array): Use standard hexadecimal format.
5911 (merged_store_group::merged_store_group): Set bit_insertion to false.
5912 (merged_store_group::apply_stores): Use optimal buffer size. Deal
5913 with BIT_INSERT_EXPR stores. Move up code updating the mask and
5914 also print the mask in the dump file.
5915 (pass_store_merging::gate): Minor tweak.
5916 (imm_store_chain_info::coalesce_immediate): Fix wrong association
5917 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
5918 stores with INTEGER_CST stores.
5919 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
5920 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
5921 and use it throughout. Generate bit insertion sequences if need be.
5922 (pass_store_merging::process_store): Remove redundant condition.
5923 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
5925 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
5927 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
5928 the 128-bit floating point types. Fix function comment.
5930 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5932 * config/aarch64/aarch64-simd.md
5933 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
5935 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
5938 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
5940 PR tree-optimization/85989
5941 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
5943 (backprop::intersect_uses): Check it when deciding whether this
5944 is a backedge reference.
5945 (backprop::process_block): Add each phi to m_visited_phis
5946 after visiting it, then clear it at the end.
5948 2018-06-01 Richard Biener <rguenther@suse.de>
5950 * tree-vectorizer.h (vect_dr_stmt): New function.
5951 (vect_get_load_cost): Adjust.
5952 (vect_get_store_cost): Likewise.
5953 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
5954 Use vect_dr_stmt instead of DR_SMTT.
5955 (vect_record_base_alignments): Likewise.
5956 (vect_calculate_target_alignment): Likewise.
5957 (vect_compute_data_ref_alignment): Likewise and make static.
5958 (vect_update_misalignment_for_peel): Likewise.
5959 (vect_verify_datarefs_alignment): Likewise.
5960 (vector_alignment_reachable_p): Likewise.
5961 (vect_get_data_access_cost): Likewise. Pass down
5962 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
5963 (vect_get_peeling_costs_all_drs): Likewise.
5964 (vect_peeling_hash_get_lowest_cost): Likewise.
5965 (vect_enhance_data_refs_alignment): Likewise.
5966 (vect_find_same_alignment_drs): Likewise.
5967 (vect_analyze_data_refs_alignment): Likewise.
5968 (vect_analyze_group_access_1): Likewise.
5969 (vect_analyze_group_access): Likewise.
5970 (vect_analyze_data_ref_access): Likewise.
5971 (vect_analyze_data_ref_accesses): Likewise.
5972 (vect_vfa_segment_size): Likewise.
5973 (vect_small_gap_p): Likewise.
5974 (vectorizable_with_step_bound_p): Likewise.
5975 (vect_prune_runtime_alias_test_list): Likewise.
5976 (vect_analyze_data_refs): Likewise.
5977 (vect_supportable_dr_alignment): Likewise.
5978 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
5979 (vect_gen_prolog_loop_niters): Likewise.
5980 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5981 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
5983 (vect_recog_mask_conversion_pattern): Likewise.
5984 (vect_try_gather_scatter_pattern): Likewise.
5985 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
5986 to vect_get_store_cost.
5987 (vect_get_store_cost): Get stmt_info instead of DR.
5988 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
5989 (vect_get_load_cost): Get stmt_info instead of DR.
5991 2018-06-01 Richard Biener <rguenther@suse.de>
5994 * gimple-fold.c (var_decl_component_p): Also allow offsetted
5995 vars wrapped in MEM_REFs.
5997 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
5999 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
6000 Fix subreg tests so that we only return a choice between
6001 GENERAL_REGS and FP_REGS if the original classes included both.
6003 2018-06-01 Richard Biener <rguenther@suse.de>
6006 * tree-ssa-structalias.c (get_function_part_constraint):
6007 Handle NULL fi->decl.
6008 (find_func_aliases_for_call): Properly handle indirect
6009 fi from direct call.
6010 (find_func_clobbers): Likewise.
6011 (ipa_pta_execute): Likewise.
6012 (create_variable_info_for): For functions that are ifunc_resolver
6013 resolve to a varinfo that contains the result of the resolver call.
6014 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
6017 2018-05-31 Michael Collison <michael.collison@arm.com>
6019 * config/aarch64/aarch64.md:
6020 (*fix_to_zero_extenddfdi2): New pattern.
6021 * gcc.target/aarch64/fix_extend1.c: New testcase.
6023 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
6027 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
6028 and BUILT_IN_STRNCMP_EQ.
6029 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
6030 BUILT_IN_STRNCMP_EQ.
6031 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
6032 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
6033 (gimple_fold_builtin): Likewise.
6034 * tree-ssa-strlen.c (compute_string_length): New function.
6035 (determine_min_obsize): New function.
6036 (handle_builtin_string_cmp): New function to handle calls to
6037 string compare functions.
6038 (strlen_optimize_stmt): Add handling to builtin string compare
6040 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
6041 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
6042 * tree.c (build_common_builtin_nodes): Add new defines of
6043 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
6045 2018-05-31 Jakub Jelinek <jakub@redhat.com>
6048 * bb-reorder.c (pass_partition_blocks::gate): Return false for
6049 functions with naked attribute.
6051 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
6053 * config/i386/sse.md (avx_vec_concat<mode>):
6054 Substitute concat_tg_mode mode attribute with xtg_mode.
6055 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
6056 (concat_tg_mode): Remove mode attribute.
6058 2018-05-31 Martin Sebor <msebor@redhat.com>
6061 * calls.c (alloc_max_size): Correct a logic error/typo.
6062 Treat excessive arguments as infinite. Warn for invalid arguments.
6063 * doc/invoke.texi (-Walloc-size-larger-than): Update.
6065 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
6068 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
6069 and movx for Haswell.
6071 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
6072 Cesar Philippidis <cesar@codesourcery.com>
6075 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
6076 when emitting error on private/firstprivate reductions.
6077 * omp-low.c (lower_omp_target): Avoid reference-type processing
6078 on pointers for firstprivate clause.
6080 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
6082 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
6085 * config/aarch64/aarch64-simd.md
6086 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
6087 (aarch64_ld1_x3_<mode>): Likewise
6088 (aarch64_st1x2<VALLDIF:mode>): Likewise
6089 (aarch64_st1_x2_<mode>): Likewise
6090 (aarch64_st1x3<VALLDIF:mode>): Likewise
6091 (aarch64_st1_x3_<mode>): Likewise
6092 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
6093 (vld1_s8_x3): Likewise.
6094 (vld1_u16_x3): Likewise.
6095 (vld1_s16_x3): Likewise.
6096 (vld1_u32_x3): Likewise.
6097 (vld1_s32_x3): Likewise.
6098 (vld1_u64_x3): Likewise.
6099 (vld1_s64_x3): Likewise.
6100 (vld1_f16_x3): Likewise.
6101 (vld1_f32_x3): Likewise.
6102 (vld1_f64_x3): Likewise.
6103 (vld1_p8_x3): Likewise.
6104 (vld1_p16_x3): Likewise.
6105 (vld1_p64_x3): Likewise.
6106 (vld1q_u8_x3): Likewise.
6107 (vld1q_s8_x3): Likewise.
6108 (vld1q_u16_x3): Likewise.
6109 (vld1q_s16_x3): Likewise.
6110 (vld1q_u32_x3): Likewise.
6111 (vld1q_s32_x3): Likewise.
6112 (vld1q_u64_x3): Likewise.
6113 (vld1q_s64_x3): Likewise.
6114 (vld1q_f16_x3): Likewise.
6115 (vld1q_f32_x3): Likewise.
6116 (vld1q_f64_x3): Likewise.
6117 (vld1q_p8_x3): Likewise.
6118 (vld1q_p16_x3): Likewise.
6119 (vld1q_p64_x3): Likewise.
6120 (vst1_s64_x2): Likewise.
6121 (vst1_u64_x2): Likewise.
6122 (vst1_f64_x2): Likewise.
6123 (vst1_s8_x2): Likewise.
6124 (vst1_p8_x2): Likewise.
6125 (vst1_s16_x2): Likewise.
6126 (vst1_p16_x2): Likewise.
6127 (vst1_s32_x2): Likewise.
6128 (vst1_u8_x2): Likewise.
6129 (vst1_u16_x2): Likewise.
6130 (vst1_u32_x2): Likewise.
6131 (vst1_f16_x2): Likewise.
6132 (vst1_f32_x2): Likewise.
6133 (vst1_p64_x2): Likewise.
6134 (vst1q_s8_x2): Likewise.
6135 (vst1q_p8_x2): Likewise.
6136 (vst1q_s16_x2): Likewise.
6137 (vst1q_p16_x2): Likewise.
6138 (vst1q_s32_x2): Likewise.
6139 (vst1q_s64_x2): Likewise.
6140 (vst1q_u8_x2): Likewise.
6141 (vst1q_u16_x2): Likewise.
6142 (vst1q_u32_x2): Likewise.
6143 (vst1q_u64_x2): Likewise.
6144 (vst1q_f16_x2): Likewise.
6145 (vst1q_f32_x2): Likewise.
6146 (vst1q_f64_x2): Likewise.
6147 (vst1q_p64_x2): Likewise.
6148 (vst1_s64_x3): Likewise.
6149 (vst1_u64_x3): Likewise.
6150 (vst1_f64_x3): Likewise.
6151 (vst1_s8_x3): Likewise.
6152 (vst1_p8_x3): Likewise.
6153 (vst1_s16_x3): Likewise.
6154 (vst1_p16_x3): Likewise.
6155 (vst1_s32_x3): Likewise.
6156 (vst1_u8_x3): Likewise.
6157 (vst1_u16_x3): Likewise.
6158 (vst1_u32_x3): Likewise.
6159 (vst1_f16_x3): Likewise.
6160 (vst1_f32_x3): Likewise.
6161 (vst1_p64_x3): Likewise.
6162 (vst1q_s8_x3): Likewise.
6163 (vst1q_p8_x3): Likewise.
6164 (vst1q_s16_x3): Likewise.
6165 (vst1q_p16_x3): Likewise.
6166 (vst1q_s32_x3): Likewise.
6167 (vst1q_s64_x3): Likewise.
6168 (vst1q_u8_x3): Likewise.
6169 (vst1q_u16_x3): Likewise.
6170 (vst1q_u32_x3): Likewise.
6171 (vst1q_u64_x3): Likewise.
6172 (vst1q_f16_x3): Likewise.
6173 (vst1q_f32_x3): Likewise.
6174 (vst1q_f64_x3): Likewise.
6175 (vst1q_p64_x3): Likewise.
6177 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6179 * config/msp430/msp430.c (msp430_output_labelref): Prepend
6180 user_label_prefix to name.
6182 * tree-core.h: Update comment about the format of NAME string
6183 passed to handler in attribute_spec.
6185 * config/msp430/msp430.md: Remove erroneous subreg expression from
6186 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
6187 zero_extend{q,h}isi2.
6189 2018-05-30 Borislav Petkov <bp@suse.de>
6191 * doc/extend.texi: Document some architecture specific
6192 constraints and sort entries.
6194 2018-05-30 Martin Sebor <msebor@redhat.com>
6197 * builtins.c (expand_builtin_stpcpy_1): New function.
6198 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
6199 only if the former succeeds.
6201 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
6203 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
6206 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6208 * doc/invoke.texi (-flinker-output): Document
6210 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6212 * passes.c (ipa_write_summaries): Only modify statements if body
6214 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
6215 incrementally linking.
6216 (ipa_passes): Likewise.
6217 * lto-cgraph.c (lto_output_node): When incrementally linking do not
6218 pass down resolution info.
6219 * common.opt (flag_incremental_link): Update info.
6220 * gcc.c (plugin specs): Turn flinker-output=* to
6221 -plugin-opt=-linker-output-known
6222 * toplev.c (compile_file): Also cut compilation when doing incremental
6224 * flag-types. (enum lto_partition_model): Add
6225 LTO_LINKER_OUTPUT_NOLTOREL.
6226 (invoke.texi): Add -flinker-output docs.
6227 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
6228 link same way as WPA; do not stream in dead initializers.
6230 * dwarf2out.c (dwarf2out_die_ref_for_decl,
6231 darf2out_register_external_decl): Support incremental link.
6233 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6235 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
6237 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6239 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
6240 it down to simple_object_copy_lto_debug_sections.
6241 (run_gcc): Determine incremental LTO link time and configure
6242 lto1 into non-wpa mode, disable renaming of debug sections.
6244 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
6246 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
6247 descriptions of various incorrectly documented functions.
6249 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
6252 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6253 address check not strict.
6255 2018-05-30 Richard Biener <rguenther@suse.de>
6257 PR tree-optimization/85964
6258 * tracer.c (better_p): Drop initialized count check, we only
6259 call the function with initialized counts now.
6260 (find_best_successor): Do find a best edge if one
6261 has uninitialized count.
6262 (find_best_predecessor): Likewise. Do BB frequency check only
6263 if count is initialized.
6265 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
6267 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
6268 (aarch64_ldrstr_offset_compare): New.
6269 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
6270 load/store orderings.
6271 (aarch64_gen_adjusted_ldpstp): Likewise.
6273 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
6275 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
6276 Check for subset of GENERAL_REGS and FP_REGS.
6277 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
6280 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
6282 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
6283 and wi::to_poly_offset. Add the current offset and then check
6284 whether the sum fits, rather than using an unchecked addition of
6285 a checked term. Check for a shwi rather than a uhwi.
6286 * expr.c (get_bit_range): Use tree_to_poly_uint64.
6287 (store_constructor): Use poly_int_tree_p.
6288 (expand_expr_real_1): Likewise.
6289 * function.c (assign_temp): Likewise.
6290 * fold-const.c (const_binop): Use poly_int_tree_p and
6292 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
6294 * ipa-icf-gimple.c (func_checker::compare_operand): Use
6295 to_poly_offset for MEM offsets.
6296 * ipa-icf.c (sem_variable::equals): Likewise.
6297 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
6298 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
6299 wi::to_poly_offset for BIT_FIELD_REF offsets.
6300 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
6302 * var-tracking.c (emit_note_insn_var_location): Use
6303 tree_to_poly_uint64.
6305 2018-05-29 Jim Wilson <jimw@sifive.com>
6307 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
6309 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
6312 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
6313 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
6315 (sse4_1_round<mode>2): Use nonimmediate_operand
6316 for operand 1 predicate.
6318 2018-05-29 Martin Sebor <msebor@redhat.com>
6319 Richard Biener <rguenther@suse.de>
6322 * calls.c (get_size_range): Call determine_value_range instead
6323 of get_value_range..
6324 * tree-vrp.h (determine_value_range): Declared new function.
6325 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
6327 2018-05-29 Richard Biener <rguenther@suse.de>
6329 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
6330 sure to use non-pattern stmts for get_earlier_stmt arguments.
6331 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
6332 called on pattern stmts.
6333 (get_later_stmt): Likewise.
6335 2018-05-29 Martin Liska <mliska@suse.cz>
6337 PR gcov-profile/85759
6338 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
6341 2018-05-29 Jakub Jelinek <jakub@redhat.com>
6343 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
6345 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
6349 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
6350 VEC_PACK_FLOAT_EXPR): New tree codes.
6351 * tree-pretty-print.c (op_code_prio): Handle
6352 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
6353 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
6354 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
6355 * tree-inline.c (estimate_operator_cost): Likewise.
6356 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
6357 * fold-const.c (const_binop): Likewise.
6358 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
6359 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
6360 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6361 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
6362 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
6363 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
6364 * expr.c (expand_expr_real_2): Likewise.
6365 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
6366 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
6367 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
6369 * optabs.c (expand_widen_pattern_expr): For
6370 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
6371 sign from result type rather than operand's type.
6372 (expand_binop_directly): For vec_packu_float_optab and
6373 vec_packs_float_optab allow result type to be different from operand's
6375 * optabs-tree.c (optab_for_tree_code): Handle
6376 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
6377 VEC_PACK_FLOAT_EXPR. Formatting fixes.
6378 * tree-vect-generic.c (expand_vector_operations_1): Handle
6379 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
6380 VEC_PACK_FLOAT_EXPR.
6381 * tree-vect-stmts.c (supportable_widening_operation): Handle
6383 (supportable_narrowing_operation): Handle FLOAT_EXPR.
6384 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
6385 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
6386 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
6387 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
6389 (vec_pack<floatprefix>_float_<mode>): New expander.
6390 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
6392 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
6393 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
6394 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
6395 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
6396 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
6398 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
6399 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
6400 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
6401 VEC_PACK_FLOAT_EXPR): Document.
6403 2018-05-29 Richard Biener <rguenther@suse.de>
6405 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
6407 (stmt_vec_info_vec): Make pointer.
6408 (init_stmt_vec_info_vec): Remove.
6409 (free_stmt_vec_info_vec): Likewise.
6410 (set_stmt_vec_info_vec): New function.
6411 (free_stmt_vec_infos): Likewise.
6412 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
6413 (set_vinfo_for_stmt): Likewise.
6414 (get_earlier_stmt): Likewise.
6415 (get_later_stmt): Likewise.
6416 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
6417 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
6418 (vec_info::~vec_info): Free stmt_vec_infos.
6419 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
6420 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
6421 (pass_slp_vectorize::execute): Likewise.
6422 * tree-vect-stmts.c (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 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
6427 the global stmt_vec_info_vec.
6428 * tree-parloops.c (gather_scalar_reductions): Use
6429 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
6432 2018-05-29 Richard Biener <rguenther@suse.de>
6434 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
6436 2018-05-29 Martin Liska <mliska@suse.cz>
6437 David Malcolm <dmalcolm@redhat.com>
6439 * vec.c (test_reverse): New.
6440 (vec_c_tests): Add new test.
6441 * vec.h (vl_ptr>::reverse): New function.
6443 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
6445 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
6447 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
6450 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
6452 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
6454 2018-05-28 Richard Biener <rguenther@suse.de>
6456 PR tree-optimization/85933
6457 * tree-vect-data-refs.c (vect_record_base_alignments): Only
6458 look at stmts marked as vectorizable.
6460 2018-05-28 Richard Biener <rguenther@suse.de>
6462 PR tree-optimization/85934
6463 * tree-vect-generic.c (expand_vector_operations_1): Hoist
6464 vector boolean check before scalar optimization.
6466 2018-05-28 Jakub Jelinek <jakub@redhat.com>
6468 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
6471 2018-05-28 Mark Wielaard <mark@klomp.org>
6473 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
6474 if it is an expression containing a minus sign.
6476 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
6478 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
6480 2018-05-27 Paul Koning <ni1d@arrl.net>
6482 * config/pdp11/pdp11.md (truncsihi2): Remove.
6484 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
6485 Chung-Ju Wu <jasonwucj@gmail.com>
6487 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
6489 (unaligned_store_dw): Ditto.
6490 * config/nds32/nds32-memory-manipulation.c
6491 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
6492 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
6493 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
6494 (emit_setmem_word_loop): Rename to ...
6495 (emit_setmem_doubleword_loop): ... this.
6496 (nds32_gen_dup_4_byte_to_word_value): New function.
6497 (nds32_gen_dup_8_byte_to_double_word_value): New function.
6498 (nds32_expand_setmem_loop): Refine implementation.
6499 (nds32_expand_setmem_loop_v3m): Ditto.
6500 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
6503 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
6505 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
6507 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
6509 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
6510 (nds32_init_machine_status): Initialize machine->attr_naked_p and
6511 machine->attr_no_prologue_p.
6512 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
6513 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
6514 (nds32_expand_epilogue): Consider attr_naked_p.
6515 (nds32_expand_epilogue_v3pop): Likewise.
6516 (nds32_can_use_return_insn): Likewise.
6517 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
6518 attr_no_prologue_p fields.
6519 * config/nds32/nds32.opt (mret-in-naked-func): New option.
6521 2018-05-27 Jakub Jelinek <jakub@redhat.com>
6524 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
6526 * config/i386/sse.md
6527 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
6529 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
6531 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
6533 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
6535 (*<floatsuffix>floatv2div2sf2): Rename to ...
6536 (*float<floatunssuffix>v2div2sf2): ... this.
6537 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
6538 (float<floatunssuffix>v2div2sf2_mask): ... this.
6539 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
6540 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
6541 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
6543 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
6545 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
6547 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
6549 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
6551 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
6553 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
6554 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
6555 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
6556 gen_ufix_truncv8dfv8si2.
6557 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
6558 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
6559 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
6560 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
6561 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
6562 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
6563 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
6564 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
6566 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
6570 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
6572 2018-05-25 Jim Wilson <jimw@sifive.com>
6574 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
6575 * config/riscv/riscv.c (struct machine_function): Add
6576 interrupt_handler_p and attribute_checked_p fields.
6577 (riscv_attribute_table): Add interrupt.
6578 (riscv_interrupt_type_p): New.
6579 (riscv_save_reg_p): Save extra regs for interrupt handler.
6580 (riscv_use_save_libcall): Return false for interrupt handler.
6581 (riscv_first_stack_step): Add forward declaration.
6582 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
6583 for interrupt handler with large frame. Use it for saved reg list.
6584 (riscv_expand_prologue): Move flag_stack_usage_info support to
6585 eliminate duplication.
6586 (riscv_expand_epilogue): Generate mret for interrupt handler.
6587 (riscv_epilogue_uses): New.
6588 (riscv_can_use_return_insn): Return false for interrupt handler.
6589 (riscv_function_ok_for_sibcall): Likewise.
6590 (riscv_set_current_function): Add interrupt handler support.
6591 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
6592 * config/riscv/riscv.md (UNSPECV_MRET): New.
6594 (riscv_frflags, riscv_fsflags): Use tab after opcode.
6596 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
6598 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
6600 PR tree-optimization/85712
6601 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
6602 this candidate has already been replaced in-situ by a copy.
6604 2018-05-25 Jason Merrill <jason@redhat.com>
6606 PR c++/80485 - inline function non-zero address.
6607 * symtab.c (nonzero_address): Check DECL_COMDAT.
6609 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
6612 * config/alpha/alpha.md (ashlsi3): New insn pattern.
6613 (*ashlsi_se): Rename from *ashldi_se. Define as sign
6614 extension of SImode operation. Use const123_operand predicate.
6615 (*saddsi_1): Remove.
6616 (*saddl_se_1): Ditto.
6618 (*ssubl_se_1): Ditto.
6619 * config/alpha/predicates.md (const123_operand): New predicate.
6620 * config/alpha/constraints.md (P): Use IN_RANGE.
6622 2018-05-25 Richard Biener <rguenther@suse.de>
6624 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
6626 (ref_maybe_used_by_stmt_p): Likewise.
6627 (stmt_may_clobber_ref_p): Likewise.
6628 (stmt_may_clobber_ref_p_1): Likewise.
6629 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
6631 (ref_maybe_used_by_stmt_p): Likewise.
6632 (stmt_may_clobber_ref_p): Likewise.
6633 (stmt_may_clobber_ref_p_1): Likewise.
6634 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
6635 the alias oracle to disambiguate DRs with stmts DR analysis
6637 (vect_analyze_data_refs): Do not give up on not analyzable
6638 DRs for BB vectorization. Remove code truncating the dataref
6641 2018-05-25 Jakub Jelinek <jakub@redhat.com>
6644 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
6645 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
6646 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
6648 2018-05-25 Richard Biener <rguenther@suse.de>
6650 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
6651 function, combining stmt data ref gathering and fatal analysis
6653 (vect_analyze_data_refs): Remove now redudnant code and simplify.
6654 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
6655 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
6656 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
6657 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
6659 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6661 PR tree-optimization/85720
6662 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
6663 SCC if all partitions are builtins.
6664 (version_loop_by_alias_check): New parameter. Generate cancelable
6665 runtime alias check if all partitions are builtins.
6666 (distribute_loop): Update call to above function.
6668 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6670 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
6671 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
6672 (parm_default_def_partition_arg): Ditto.
6673 (set_parm_default_def_partition): Ditto.
6674 (get_parm_default_def_partitions): Ditto and make it static.
6675 (get_undefined_value_partitions): Ditto and make it static.
6676 (remove_ssa_form): Refactor call to init_var_map here.
6677 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
6678 computation for loop region.
6679 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
6680 (register_default_def): Delete.
6681 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
6682 (parm_default_def_partition_arg): Ditto.
6683 (set_parm_default_def_partition): Ditto.
6684 (get_parm_default_def_partitions): Ditto and make it static.
6685 (get_undefined_value_partitions): Ditto and make it static.
6686 (coalesce_with_default, coalesce_with_default): Update comment.
6687 (create_coalesce_list_for_region): New func factored out from
6688 create_outofssa_var_map.
6689 (populate_coalesce_list_for_outofssa): New func factored out from
6690 create_outofssa_var_map and coalesce_ssa_name.
6691 (create_outofssa_var_map): Delete.
6692 (coalesce_ssa_name): Refactor to support live range computation.
6693 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
6694 (get_parm_default_def_partitions): Delete.
6695 (get_undefined_value_partitions): Ditto.
6696 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
6697 computation for loop region.
6698 (new_tree_live_info, loe_visit_block): Ditto.
6699 (live_worklist, set_var_live_on_entry): Ditto.
6700 (calculate_live_on_exit, verify_live_on_entry): Ditto.
6701 * tree-ssa-live.h (struct _var_map): New fields.
6702 (init_var_map): Change decl.
6703 (region_contains_p): New.
6705 2018-05-25 Bin Cheng <bin.cheng@arm.com>
6707 * tree-ssa-live.h (live_merge_and_clear): Delete.
6709 2018-05-25 Richard Biener <rguenther@suse.de>
6712 * tree-dump.c (dequeue_and_dump): Remove access to removed
6713 operand 2 of a SWITCH_EXPR.
6715 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6717 * doc/sourcebuild.texi (vect_double_cond_arith): Include
6718 multiplication and division.
6719 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
6720 (cond_udiv@var{m}, cond_umod@var{m}): Document.
6721 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
6722 (cond_udiv_optab, cond_umod_optab): New optabs.
6723 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
6724 (IFN_COND_RDIV): New internal functions.
6725 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
6726 TRUNC_MOD_EXPR and RDIV_EXPR.
6727 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
6728 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
6730 (SVE_INT_BINARY): Include mult.
6731 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
6732 (optab, sve_int_op): Handle mult.
6733 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
6735 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
6736 for SVE_INT_BINARY_SD.
6738 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6740 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
6741 (optab, sve_int_op): Handle div and udiv.
6742 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
6743 for SVE_INT_BINARY_SD.
6744 (*<optab><mode>3): New insn for the same.
6746 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6748 * tree-vect-patterns.c: Include predict.h.
6749 (vect_recog_divmod_pattern): Restrict check for division support
6750 to when optimizing for size.
6752 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6754 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
6755 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
6756 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
6757 (gimple_match_op::set_op): Likewise.
6758 (gimple_resimplify4): Declare.
6759 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
6760 (expr::gen_transform): Likewise.
6761 (decision_tree::gen): Generate a simplification routine for 4 operands.
6762 * gimple-match-head.c (gimple_simplify): Add an overload for
6763 4 operands. In the top-level function, handle up to 4 call
6764 arguments and call gimple_resimplify4.
6765 (gimple_resimplify4): New function.
6766 (build_call_internal): Pass a fourth operand.
6767 (maybe_push_to_seq): Likewise.
6768 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
6769 Fold VEC_COND_EXPRs of an operation and a default value into
6770 an IFN_COND_* function if possible.
6771 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
6773 (SVE_COND_FP_BINARY): Include them.
6774 (optab, sve_fp_op): Handle them.
6775 (SVE_INT_BINARY_REV): New code iterator.
6776 (SVE_COND_FP_BINARY_REV): New int iterator.
6777 (commutative): New int attribute.
6778 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
6780 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
6782 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
6783 (*cond_<optab><mode>): New patterns for reversed operands.
6785 2018-05-25 Richard Biener <rguenther@suse.de>
6787 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
6788 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
6789 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
6790 (STMT_VINFO_GROUPED_ACCESS): Adjust.
6791 * tree-vect-data-refs.c (everywhere): Adjust users.
6792 * tree-vect-loop.c (everywhere): Likewise.
6793 * tree-vect-slp.c (everywhere): Likewise.
6794 * tree-vect-stmts.c (everywhere): Likewise.
6795 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
6797 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6799 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
6801 (gcc_cv_as_section_exclude): ... this.
6802 Try Solaris as #exclude syntax.
6803 * configure: Regenerate.
6804 * config.in: Regenerate.
6805 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
6807 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
6808 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
6810 * varasm.c (default_elf_asm_named_section): Don't check if
6811 HAVE_GAS_SECTION_EXCLUDE is defined.
6813 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6815 * doc/md.texi: Update the documentation of the cond_* optabs
6816 to mention the new final operand. Fix GET_MODE_NUNITS call.
6817 Describe the scalar case too.
6818 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
6819 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
6821 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
6822 (get_conditional_internal_fn): Update comment.
6823 * tree-vect-loop.c (vectorizable_reduction): Pass the original
6824 accumulator value as a final argument to conditional functions.
6825 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
6826 a define_expand and add an "else" operand. Assert for now that
6827 the else operand is equal to operand 2. Use SVE_INT_BINARY and
6828 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
6829 (*cond_<optab><mode>): New patterns.
6830 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
6831 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6832 (UNSPEC_COND_EOR): Delete.
6833 (optab): Remove associated mappings.
6834 (SVE_INT_BINARY): New code iterator.
6835 (sve_int_op): Remove int attribute and add "minus" to the code
6837 (SVE_COND_INT_OP): Delete.
6838 (SVE_COND_FP_OP): Rename to...
6839 (SVE_COND_FP_BINARY): ...this.
6841 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6843 * optabs.c (can_reuse_operands_p): New function.
6844 (maybe_legitimize_operands): Try to reuse the results for
6847 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
6849 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
6850 Add {q} suffix to insn mnemonic.
6852 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6854 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
6855 (msp430_warn_func_return): New.
6857 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
6859 * fold-const.c (tree_nonzero_bits): New function.
6860 * fold-const.h (tree_nonzero_bits): Likewise.
6861 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
6862 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
6864 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
6868 * varasm.c (assemble_alias): Check ifunc_resolver only on
6871 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
6874 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
6875 when memory input operand is handled.
6877 2018-05-24 Luis Machado <luis.machado@linaro.org>
6879 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
6881 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
6883 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
6885 * match.pd: Delay FMA folds until after vectorization.
6887 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
6890 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6891 address check not strict.
6893 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
6895 * gimple-match.h (gimple_match_op): New class.
6896 (mprts_hook): Replace parameters with a gimple_match_op *.
6897 (maybe_build_generic_op): Likewise.
6898 (gimple_simplified_result_is_gimple_val): Replace parameters with
6899 a const gimple_match_op *.
6900 (gimple_simplify): Replace code_helper * and tree * parameters with
6901 a gimple_match_op * parameter.
6902 (gimple_resimplify1): Replace code_helper *, tree and tree *
6903 parameters with a gimple_match_op * parameter.
6904 (gimple_resimplify2): Likewise.
6905 (gimple_resimplify3): Likewise.
6906 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
6907 parameters with a gimple_match_op * parameter.
6908 * gimple-match-head.c (gimple_simplify): Change prototypes of
6909 auto-generated functions to take a gimple_match_op * instead of
6910 separate code_helper * and tree * parameters. Make the same
6911 change in the top-level overload and update calls to the
6912 gimple_resimplify routines. Update calls to the auto-generated
6913 functions and to maybe_push_res_to_seq in the publicly-facing
6914 operation-specific gimple_simplify overloads.
6915 (gimple_match_op::MAX_NUM_OPS): Define.
6916 (gimple_resimplify1): Replace rcode and ops with a single res_op
6917 parameter. Update call to gimple_simplify.
6918 (gimple_resimplify2): Likewise.
6919 (gimple_resimplify3): Likewise.
6920 (mprts_hook): Replace parameters with a gimple_match_op *.
6921 (maybe_build_generic_op): Likewise.
6922 (build_call_internal): Replace type, nargs and ops with
6923 a gimple_match_op *.
6924 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
6925 with a single gimple_match_op *. Update calls to mprts_hook,
6926 build_call_internal and gimple_simplified_result_is_gimple_val.
6927 Factor out code that is common to the tree_code and combined_fn cases.
6928 * genmatch.c (expr::gen_transform): Replace tem_code and
6929 tem_ops with a gimple_match_op called tem_op. Update calls
6930 to the gimple_resimplify functions and maybe_push_res_to_seq.
6931 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
6932 res_ops. Update call to the gimple_resimplify functions.
6933 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
6934 (decision_tree::gen): Make the functions take a gimple_match_op *
6935 called res_op instead of separate res_code and res_ops parameters.
6936 Update call accordingly.
6937 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
6938 and ops with a single res_op parameter. Update calls to
6939 maybe_build_generic_op and maybe_push_res_to_seq.
6940 (fold_stmt_1): Update calls to gimple_simplify and
6941 replace_stmt_with_simplification.
6942 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
6943 and gimple_simplified_result_is_gimple_val.
6944 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
6946 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
6947 with a gimple_match_op *.
6948 (vn_nary_build_or_lookup): Likewise. Update call to
6949 vn_nary_build_or_lookup_1.
6950 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
6951 gimple_match_op *. Update calls to the gimple_resimplify routines
6952 and to gimple_simplified_result_is_gimple_val.
6953 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
6954 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
6955 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
6956 (visit_nary_op): Likewise.
6957 (visit_reference_op_load): Likewise.
6959 2018-05-23 Luis Machado <luis.machado@linaro.org>
6961 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
6962 modifier for printing the step amount.
6964 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
6967 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
6970 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
6972 * doc/sourcebuild.texi (Endianness): New subsubsection.
6974 2018-05-23 Luis Machado <luis.machado@linaro.org>
6976 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6977 <prefetch_dynamic_strides>: New const bool field.
6978 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6979 prefetch_dynamic_strides.
6980 (exynosm1_prefetch_tune): Likewise.
6981 (thunderxt88_prefetch_tune): Likewise.
6982 (thunderx_prefetch_tune): Likewise.
6983 (thunderx2t99_prefetch_tune): Likewise.
6984 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
6986 (aarch64_override_options_internal): Update to set
6987 PARAM_PREFETCH_DYNAMIC_STRIDES.
6988 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6989 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6990 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6991 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6992 prefetch-dynamic-strides setting.
6994 2018-05-23 Luis Machado <luis.machado@linaro.org>
6996 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6997 <minimum_stride>: New const int field.
6998 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6999 minimum_stride field defaulting to -1.
7000 (exynosm1_prefetch_tune): Likewise.
7001 (thunderxt88_prefetch_tune): Likewise.
7002 (thunderx_prefetch_tune): Likewise.
7003 (thunderx2t99_prefetch_tune): Likewise.
7004 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
7005 <default_opt_level>: Set to 3.
7006 (aarch64_override_options_internal): Update to set
7007 PARAM_PREFETCH_MINIMUM_STRIDE.
7008 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7009 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7010 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7011 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7012 stride is constant and is below the minimum stride threshold.
7014 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7016 * config/arm/arm-cpus.in (mode26): Delete.
7017 (armv4): Delete mode26 reference.
7018 * config/arm/arm.c (arm_configure_build_target): Delete use of
7021 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
7023 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
7025 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
7026 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
7028 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
7029 (floatunsdidf2): Ditto.
7031 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
7033 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
7034 (fixuns_trunc<mode>si2_avx512f): Ditto.
7035 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
7036 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
7037 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
7039 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
7041 PR rtl-optimization/79985
7042 * df-scan.c (df_insn_refs_collect): Remove special case for
7043 global registers and asm statements.
7045 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
7047 * extend.texi (Global Register Variables): Rewrite the bullet list.
7048 Note that the register is available for allocation. Note that access
7049 via inline asm must use constraints. Add note about async-signal
7050 handlers. Remove paragraph about automagic register selection.
7052 2018-05-23 Richard Biener <rguenther@suse.de>
7054 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
7055 of fixed offset from memset VN.
7057 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
7059 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
7061 (alloc_cand_and_find_basis): Initialize first_interp field.
7062 (slsr_process_mul): Modify first_interp field.
7063 (slsr_process_add): Likewise.
7064 (slsr_process_cast): Modify first_interp field for each new
7066 (slsr_process_copy): Likewise.
7067 (dump_candidate): Dump first_interp field.
7068 (replace_mult_candidate): Process all interpretations, not just
7070 (replace_rhs_if_not_dup): Likewise.
7071 (replace_one_candidate): Likewise.
7073 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
7075 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
7077 (aarch64_needs_frame_chain): New function.
7078 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
7080 2018-05-23 Sudakshina Das <sudi.das@arm.com>
7083 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
7084 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
7085 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
7086 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
7087 as true for strict-align.
7088 (aarch64_can_inline_p): Perform checks even when callee has no
7089 attributes to check for strict alignment.
7090 * doc/extend.texi (AArch64 Function Attributes): Document
7092 * doc/invoke.texi: (AArch64 Options): Likewise.
7094 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
7096 PR tree-optimization/85853
7097 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
7098 the handling of the root of the node to...
7099 (vect_slp_analyze_node_operations_1): ...this new function,
7100 and run the whole thing with the child nodes' def types
7101 set according to their SLP node's def type.
7103 2018-05-23 Richard Biener <rguenther@suse.de>
7106 * tree-data-ref.c (create_runtime_alias_checks): Defer
7107 and ignore overflow warnings.
7109 2018-05-23 Yury Gribov <tetra2005@gmail.com>
7111 PR tree-optimization/85822
7112 * tree-vrp.c (is_masked_range_test): Fix handling of negative
7115 2018-05-23 Richard Biener <rguenther@suse.de>
7117 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
7118 memset constants via native_interpret_expr.
7120 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
7123 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
7125 (cgraph_node::create_alias): Likewise.
7126 (cgraph_node::get_availability): Check ifunc_resolver instead
7127 of looking up ifunc attribute.
7128 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
7129 * varasm.c (do_assemble_alias): Likewise.
7130 (assemble_alias): Likewise.
7131 (default_binds_local_p_3): Likewise.
7132 * cgraph.h (cgraph_node): Add ifunc_resolver.
7133 (cgraph_node::only_called_directly_or_aliased_p): Return false
7135 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
7137 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
7138 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
7139 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
7140 instead of looking up ifunc attribute.
7142 2018-05-22 Luis Machado <luis.machado@linaro.org>
7144 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
7146 2018-05-22 Martin Sebor <msebor@redhat.com>
7149 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
7150 only when expasion succeeds.
7151 (expand_builtin_strcmp): Same.
7152 (expand_builtin_strncmp): Same.
7154 2018-05-22 Martin Sebor <msebor@redhat.com>
7156 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
7158 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
7159 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7161 * config/aarch64/aarch64-ldpstp.md: Replace uses of
7162 aarch64_mem_pair_operand with memory_operand and delete operand swapping
7164 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
7165 Add check for legitimate_address.
7166 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
7167 (aarch64_swap_ldrstr_operands): New.
7168 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
7171 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
7172 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7174 * config/aarch64/aarch64.md: New patterns to generate stp
7176 (store_pair_sw, store_pair_dw): New patterns to generate stp for
7177 single words and double words.
7178 (load_pair_sw, load_pair_dw): Likewise.
7179 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
7181 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
7183 * config/aarch64/aarch64-ldpstp.md: Modify peephole
7184 for different mode ldpstp and add peephole for merged zero stores.
7186 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
7188 (aarch64_gen_store_pair): Rename calls to match new patterns.
7189 (aarch64_gen_load_pair): Rename calls to match new patterns.
7190 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
7191 (load_pair<DREG:mode><DREG2:mode>): ... This.
7192 (store_pair<mode>): Rename to...
7193 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
7194 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
7196 (V_INT_EQUIV): Handle SImode.
7197 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
7200 2018-05-22 Martin Sebor <msebor@redhat.com>
7203 * calls.c (maybe_warn_nonstring_arg): Use string length to set
7204 or ajust the presumed bound on an operation to avoid unnecessary
7207 2018-05-22 Martin Sebor <msebor@redhat.com>
7209 PR tree-optimization/85826
7210 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
7211 assuming that a DECL necesarily has a constant size.
7213 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
7216 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
7218 2018-05-22 Richard Biener <rguenther@suse.de>
7220 PR tree-optimization/85834
7221 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
7222 non-constant and non-zero memset arguments.
7224 2018-05-22 Martin Liska <mliska@suse.cz>
7227 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
7229 2018-05-22 Richard Biener <rguenther@suse.de>
7231 PR tree-optimization/85863
7232 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
7233 comparisons when vectype is specified.
7234 (vectorizable_condition): Do not specify vectype for
7235 vect_is_simple_cond when SLP vectorizing.
7237 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
7240 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
7241 define __ibm128 as long double.
7242 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
7243 as a distinct type when IEEE 128-bit support is enabled.
7244 (init_float128_ieee): Fix up conversions between IFmode and IEEE
7245 128-bit types to use the correct functions.
7246 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
7247 convert between 128-bit floating point types that have different
7248 modes but the same representation, instead of using gen_lowpart to
7250 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
7252 (IFKF_reg): New attributes to give the register constraints for
7254 (extend<mode>tf2_internal): New insns to mark an explicit
7255 conversion between 128-bit floating point types that have a
7256 different mode but share the same representation.
7258 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
7260 PR tree-optimization/85814
7261 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
7262 a null return from get_strinfo when unsharing the next
7263 strinfo in the chain.
7265 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7268 * varasm.c (weak_finish): Clean up weak_decls.
7270 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7272 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
7273 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
7274 UNSPEC_UADALP values.
7275 * config/aarch64/iterators.md (ABAL): New int iterator.
7278 (sur): Add mappings for the above.
7279 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
7281 (aarch64_<sur>abal<mode>_4): Likewise.
7282 (aarch64_<sur>adalp<mode>_3): Likewise.
7283 (<sur>sadv16qi): New define_expand.
7285 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
7287 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
7288 (*movdf_internal): Ditto.
7289 (*rcpsf2_sse): Ditto.
7290 (*rsqrtsf2_sse): Ditto.
7291 (*sqrt<mode>2_sse): Ditto.
7293 2018-05-21 Tamar Christina <tamar.christina@arm.com>
7295 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
7297 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
7298 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
7299 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
7300 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
7302 * config/aarch64/arm_neon.h: Likewise.
7303 * config/aarch64/iterators.md (VQ_I): New.
7305 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
7307 * config.gcc: Add arc/t-multilib-linux to tmake_file for
7309 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
7312 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
7314 * config/nds32/constraints.md (S): New constraint.
7315 * config/nds32/nds32.md (call_internal): Use constraint S.
7316 (call_value_internal): Likewise.
7317 (sibcall_internal): Likewise.
7318 (sibcall_value_internal): Likewise.
7320 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
7321 Chung-Ju Wu <jasonwucj@gmail.com>
7323 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
7326 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
7327 Chung-Ju Wu <jasonwucj@gmail.com>
7329 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
7330 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
7331 (nds32_rtx_costs_impl): Simplify.
7332 (nds32_address_cost_impl): Simplify.
7333 (nds32_init_rtx_costs): New function.
7334 (nds32_rtx_costs_speed_prefer): Likewise.
7335 (nds32_rtx_costs_size_prefer): Likewise.
7336 (nds32_address_cost_speed_prefer): Likewise.
7337 (nds32_address_cost_speed_fwprop): Likewise.
7338 (nds32_address_cost_size_prefer): Likewise.
7339 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
7340 * config/nds32/nds32.c (nds32_option_override): Use
7341 nds32_init_rtx_costs function.
7343 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
7345 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
7346 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
7347 (TARGET_PIPELINE_N8): Likewise.
7348 (TARGET_PIPELINE_N10): Likewise.
7349 (TARGET_PIPELINE_N13): Likewise.
7350 (TARGET_PIPELINE_GRAYWOLF): Likewise.
7352 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
7354 * config/nds32/nds32-fpu.md: Update copyright year.
7356 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7358 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
7360 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7362 * config/nds32/nds32.c
7363 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
7364 * config/nds32/nds32.opt (minline-asm-r15): New option.
7366 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7368 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
7370 * config/nds32/nds32.md (abssi2): New pattern.
7372 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
7374 * config/i386/i386.md (rex64namesuffix): New mode attribute.
7375 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
7376 Merge insn pattern from sse_cvtsi2ss<round_name> and
7377 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
7378 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
7379 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
7380 using SWI48 mode iterator.
7381 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
7382 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
7383 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
7384 pattern from sse_cvttss2si<round_saeonly_name>
7385 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
7386 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
7387 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
7388 using SWI48 mode iterator.
7389 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
7390 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
7391 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
7392 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
7393 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
7394 using SWI48 mode iterator.
7395 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
7396 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
7397 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
7398 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
7399 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
7400 SWI48 mode iterator.
7401 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
7402 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
7403 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
7404 pattern from sse_cvttsd2si<round_saeonly_name>
7405 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
7407 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7409 * config/nds32/nds32-md-auxiliary.c
7410 (nds32_valid_smw_lwm_base_p): Refine.
7411 (nds32_output_smw_single_word): Refine.
7412 (nds32_output_smw_double_word): New.
7413 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
7415 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7417 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
7418 (nds32_output_stack_pop): Refine.
7419 (nds32_expand_unaligned_load): Refine.
7420 (nds32_expand_unaligned_store): Refine.
7422 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7423 Chung-Ju Wu <jasonwucj@gmail.com>
7425 * config/nds32/constants.md: Add TP_REGNUM constant.
7426 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
7427 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
7429 * config/nds32/nds32-doubleword.md: Consider flag_pic.
7430 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
7431 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
7432 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
7433 and PIC code generation.
7434 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
7436 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
7438 * config/nds32/nds32.md: Support TLS and PIC.
7439 * config/nds32/nds32.c: Support TLS and PIC.
7440 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
7441 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
7444 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7446 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
7447 mode with E_ prefix.
7449 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7450 Chung-Ju Wu <jasonwucj@gmail.com>
7452 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
7453 * config/nds32/nds32-md-auxiliary.c
7454 (symbolic_reference_mentioned_p): New.
7455 (nds32_legitimize_ict_address): New.
7456 (nds32_expand_ict_move): New.
7457 (nds32_indirect_call_referenced_p): New.
7458 (nds32_symbol_binds_local_p): Delete.
7459 (nds32_long_call_p): Modify.
7460 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
7461 * config/nds32/nds32-protos.h
7462 (symbolic_reference_mentioned_p): Declare.
7463 (nds32_legitimize_ict_address): Declare.
7464 (nds32_expand_ict_move): Declare.
7465 (nds32_indirect_call_referenced_p): Declare.
7466 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
7467 (nds32_relax_group): Use nds32_ict_const_p as condition.
7468 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
7469 (nds32_asm_file_start): Output ict_model directive in asm code.
7470 (nds32_legitimate_address_p): Consider indirect call.
7471 (nds32_print_operand): Consider indirect call.
7472 (nds32_print_operand_address): Consider indirect call.
7473 (nds32_insert_attributes): Handle "indirect_call" attribute.
7474 (TARGET_LEGITIMATE_ADDRESS_P): Define.
7475 (TARGET_LEGITIMATE_CONSTANT_P): Define.
7476 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
7477 (TARGET_DELEGITIMIZE_ADDRESS): Define.
7478 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
7479 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
7480 (TARGET_ICT_MODEL_SMALL): Define.
7481 (TARGET_ICT_MODEL_LARGE): Define.
7482 * config/nds32/nds32.md (movsi): Consider ict model.
7483 (call, call_value): Consider ict model.
7484 (sibcall, sibcall_value): Consider ict model.
7485 * config/nds32/nds32.opt (mict-model): New option.
7486 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
7489 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
7490 Monk Chiang <sh.chiang04@gmail.com>
7491 Jim Wilson <jimw@sifive.com>
7493 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
7494 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
7495 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
7496 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
7497 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
7498 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
7499 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
7500 compute save_libcall_adjustment properly.
7501 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
7502 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
7503 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
7504 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
7505 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
7506 (ABI_SPEC): Handle mabi=ilp32e.
7507 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
7508 (RVE): Add RVE mask.
7509 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
7510 <-march>: Add rv32e as an example.
7512 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
7515 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
7516 (intra_create_variable_infos): Handle C++ constructors.
7518 2018-05-18 Martin Liska <mliska@suse.cz>
7520 * passes.def: Remove a redundant pass.
7522 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
7525 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
7527 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7529 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
7531 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
7533 (armv2, armv2a, armv3, armv3m): Delete architectures.
7534 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
7535 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
7536 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
7538 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
7539 (*mulsidi3adddi): Likewise.
7540 (mulsidi3): Likewise.
7541 (*mulsidi3_nov6): Likewise.
7542 (umulsidi3): Likewise.
7543 (umulsidi3_nov6): Likewise.
7544 (umaddsidi4): Likewise.
7545 (*umulsidi3adddi): Likewise.
7546 (smulsi3_highpart): Likewise.
7547 (*smulsi3_highpart_nov6): Likewise.
7548 (umulsi3_highpart): Likewise.
7549 (*umulsi3_highpart_nov6): Likewise.
7550 * config/arm/arm.h (arm_arch3m): Delete.
7551 * config/arm/arm.c (arm_arch3m): Delete.
7552 (arm_option_override_internal): Update armv3-related comment.
7553 (arm_configure_build_target): Delete use of isa_bit_mode32.
7554 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
7555 (arm_rtx_costs_internal): Delete check of arm_arch3m.
7556 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
7558 (mulusa3): Likewise.
7559 * config/arm/arm-protos.h (arm_arch3m): Delete.
7560 * config/arm/arm-tables.opt: Regenerate.
7561 * config/arm/arm-tune.md: Likewise.
7562 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
7563 deleted architectures.
7565 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7567 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
7568 (armv5t, armv5te): New features.
7569 (ARMv5, ARMv5e): Delete fgroups.
7570 (ARMv5t, ARMv5te): Adjust for above changes.
7572 (armv5, armv5e): Delete arches.
7573 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
7575 (*call_reg_arm): Likewise.
7576 (*call_value_reg_armv5): Likewise.
7577 (*call_value_reg_arm): Likewise.
7578 (*call_symbol): Likewise.
7579 (*call_value_symbol): Likewise.
7580 (*sibcall_insn): Likewise.
7581 (*sibcall_value_insn): Likewise.
7583 (prefetch): Likewise.
7584 (define_split and define_peephole2 dependent on arm_arch5):
7586 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
7588 (TARGET_ARM_QBIT): Likewise.
7589 (TARGET_DSP_MULTIPLY): Likewise.
7590 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
7591 (arm_arch5, arm_arch5e): Delete.
7592 (arm_arch5t, arm_arch5te): Declare.
7593 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
7594 (arm_arch5t): Declare.
7595 (arm_option_reconfigure_globals): Update for the above.
7596 (arm_options_perform_arch_sanity_checks): Update comment, replace
7597 use of arm_arch5 with arm_arch5t.
7598 (use_return_insn): Likewise.
7599 (arm_emit_call_insn): Likewise.
7600 (output_return_instruction): Likewise.
7601 (arm_final_prescan_insn): Likewise.
7602 (arm_coproc_builtin_available): Likewise.
7603 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
7604 arm_arch5e with arm_arch5t and arm_arch5te.
7605 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
7606 (arm_arch5t, arm_arch5te): Declare.
7607 * config/arm/arm-tables.opt: Regenerate.
7608 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
7609 * config/arm/t-multilib: Likewise.
7610 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
7611 instead of arm_arch5.
7612 (*call_reg_thumb1): Likewise.
7613 (*call_value_reg_thumb1_v5): Likewise.
7614 (*call_value_reg_thumb1): Likewise.
7615 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
7617 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
7619 2018-05-18 Martin Liska <mliska@suse.cz>
7621 PR gcov-profile/84846
7622 * doc/gcov.texi: Document -t option of gcov tool.
7624 2018-05-18 Martin Liska <mliska@suse.cz>
7626 PR gcov-profile/84846
7627 * gcov.c (print_usage): Add new -t option.
7628 (process_args): Handle the option.
7629 (generate_results): Use stdout as output when requested by
7632 2018-05-18 Martin Liska <mliska@suse.cz>
7634 PR gcov-profile/84846
7635 * coverage.c (coverage_init): Write PWD to .gcno file.
7636 * doc/gcov.texi: Document how working directory is printed.
7637 * gcov-dump.c (dump_gcov_file): Print PWD.
7638 * gcov.c (output_intermediate_file): Likewise.
7639 (read_graph_file): Read PWD string.
7640 (output_lines): Print PWD.
7642 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7645 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
7646 for retval and return false if all args to phi are zero.
7648 2018-05-18 Richard Biener <rguenther@suse.de>
7650 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
7652 (evrp_dom_walker::before_dom_children): Call it.
7654 2018-05-18 Richard Biener <rguenther@suse.de>
7656 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
7657 results when processing array refs with variable index.
7659 2018-05-18 Toon Moene <toon@moene.org>
7661 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
7662 directly after that of -floop-interchange. Indicate that both
7663 options are enabled by default when specifying -O3.
7665 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7667 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
7668 iterator. Delete separate integer-mode vec_set<mode> expander.
7669 (aarch64_simd_vec_setv2di): Delete.
7670 (vec_setv2di): Delete.
7671 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
7672 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
7673 the "w, r" alternative.
7675 2018-05-18 Martin Liska <mliska@suse.cz>
7677 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
7678 * tree-pass.h (make_pass_lower_switch_O0): New function.
7679 * tree-switch-conversion.c (node_has_low_bound): Remove.
7680 (node_has_high_bound): Likewise.
7681 (node_is_bounded): Likewise.
7682 (class pass_lower_switch): Make it a template type and create
7684 (pass_lower_switch::execute): Add template argument.
7685 (make_pass_lower_switch): New function.
7686 (make_pass_lower_switch_O0): New function.
7687 (do_jump_if_equal): Remove.
7688 (emit_case_nodes): Simplify to just handle all 3 cases and leave
7689 all the hard work to tree optimization passes.
7691 2018-05-18 Martin Liska <mliska@suse.cz>
7693 * dbgcnt.c (limit_low): Renamed from limit.
7694 (limit_high): New variable.
7695 (dbg_cnt_is_enabled): Check for upper limit.
7696 (dbg_cnt): Adjust dumping.
7697 (dbg_cnt_set_limit_by_index): Add new argument for high
7699 (dbg_cnt_set_limit_by_name): Likewise.
7700 (dbg_cnt_process_single_pair): Parse new format.
7701 (dbg_cnt_process_opt): Use strtok.
7702 (dbg_cnt_list_all_counters): Remove 'value' and add
7704 * doc/invoke.texi: Document changes.
7706 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
7708 * doc/sourcebuild.texi (scalar_all_fma): Document.
7709 * tree.def (FMA_EXPR): Delete.
7710 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
7711 * internal-fn.c (ternary_direct): New macro.
7712 (expand_ternary_optab_fn): Likewise.
7713 (direct_ternary_optab_supported_p): Likewise.
7714 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
7715 * builtins.c (fold_builtin_fma): Delete.
7716 (fold_builtin_3): Don't call it.
7717 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
7718 * expr.c (expand_expr_real_2): Likewise.
7719 * fold-const.c (operand_equal_p): Likewise.
7720 (fold_ternary_loc): Likewise.
7721 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
7722 * gimple.c (DEFTREECODE): Likewise.
7723 * gimplify.c (gimplify_expr): Likewise.
7724 * optabs-tree.c (optab_for_tree_code): Likewise.
7725 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
7726 * tree-eh.c (operation_could_trap_p): Likewise.
7727 (stmt_could_throw_1_p): Likewise.
7728 * tree-inline.c (estimate_operator_cost): Likewise.
7729 * tree-pretty-print.c (dump_generic_node): Likewise.
7730 (op_code_prio): Likewise.
7731 * tree-ssa-loop-im.c (stmt_cost): Likewise.
7732 * tree-ssa-operands.c (get_expr_operands): Likewise.
7733 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
7734 * fold-const-call.h (fold_fma): Delete.
7735 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
7736 CFN_FNMA and CFN_FNMS.
7738 * genmatch.c (combined_fn): New enum.
7739 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
7740 (commutative_op): New function.
7741 (commutate): Use it. Handle more than 2 operands.
7742 (dt_operand::gen_gimple_expr): Use commutative_op.
7743 (parser::parse_expr): Allow :c to be used with non-binary
7744 operators if the commutative operand is known.
7745 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
7746 CFN_FMS, CFN_FNMA and CFN_FNMS.
7747 (backprop::process_assign_use): Remove FMA_EXPR handling.
7748 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
7749 (gen_hsa_fma): New function.
7750 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
7751 IFN_FNMA and IFN_FNMS.
7752 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
7753 * gimple-fold.h (follow_all_ssa_edges): Declare.
7754 * gimple-fold.c (follow_all_ssa_edges): New function.
7755 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
7756 gimple_build interface and use follow_all_ssa_edges to fold the result.
7757 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
7758 instead of checking for optabs directly.
7759 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
7760 rather than FMA_EXPRs.
7761 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
7762 call to IFN_FMA instead of an FMA_EXPR.
7764 2018-05-17 Jim Wilson <jimw@sifive.com>
7766 * expr.c (do_tablejump): When converting index to Pmode, if we have a
7767 sign extended promoted subreg, and the range does not have the sign bit
7768 set, then do a sign extend.
7770 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
7771 test, check for sign extended subreg and/or constant operands, and
7772 do a sign extend in that case.
7774 2018-05-17 Steve Ellcey <sellcey@cavium.com>
7776 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
7777 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
7779 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
7780 Change logics_shift_reg to logics_shift_imm.
7781 (thunderx2t99_fp_loadpair_basic): Delete.
7782 (thunderx2t99_fp_storepair_basic): Delete.
7783 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
7784 (thunderx2t99_asimd_polynomial): Delete.
7785 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
7786 and neon_fp_mul_d_scalar_q.
7787 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
7788 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
7789 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
7790 (thunderx2t99_asimd_lut): Add missing tbl types.
7791 (thunderx2t99_asimd_ext): Delete.
7792 (thunderx2t99_asimd_load1_1_mult): Delete.
7793 (thunderx2t99_asimd_load1_2_mult): Delete.
7794 (thunderx2t99_asimd_load1_ldp): New.
7795 (thunderx2t99_asimd_load1): New.
7796 (thunderx2t99_asimd_load2): Add missing *load2* types.
7797 (thunderx2t99_asimd_load3): New.
7798 (thunderx2t99_asimd_load4): New.
7799 (thunderx2t99_asimd_store1_1_mult): Delete.
7800 (thunderx2t99_asimd_store1_2_mult): Delete.
7801 (thunderx2t99_asimd_store2_mult): Delete.
7802 (thunderx2t99_asimd_store2_onelane): Delete.
7803 (thunderx2t99_asimd_store_stp): New.
7804 (thunderx2t99_asimd_store1): New.
7805 (thunderx2t99_asimd_store2): New.
7806 (thunderx2t99_asimd_store3): New.
7807 (thunderx2t99_asimd_store4): New.
7809 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
7811 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
7812 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
7814 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
7815 Segher Boessenkool <segher@kernel.crashing.org>
7818 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
7821 2018-05-17 Richard Biener <rguenther@suse.de>
7823 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
7824 for pruning loop and prune defs feeding only already visited PHIs.
7826 2018-05-17 Richard Biener <rguenther@suse.de>
7828 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
7830 2018-05-17 Bin Cheng <bin.cheng@arm.com>
7831 Richard Biener <rguenther@suse.de>
7833 PR tree-optimization/85793
7834 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
7835 for VMAT_ELEMENTWISE.
7837 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
7839 * internal-fn.h (lookup_internal_fn): Declare
7840 * internal-fn.c (lookup_internal_fn): New function.
7841 * gimple.c (gimple_build_call_from_tree): Handle calls to
7843 * gimple-pretty-print.c (dump_gimple_call): Print "." before
7844 internal function names.
7845 * tree-pretty-print.c (dump_generic_node): Likewise.
7846 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
7848 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
7850 * gimple-fold.h (gimple_build): Make the function forms take
7851 combined_fn rather than built_in_function.
7852 (gimple_simplify): Likewise.
7853 * gimple-match-head.c (gimple_simplify): Likewise.
7854 * gimple-fold.c (gimple_build): Likewise.
7855 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
7856 rather than gimple_build_call_internal.
7857 (get_initial_defs_for_reduction): Likewise.
7858 (vect_create_epilog_for_reduction): Likewise.
7859 (vectorizable_live_operation): Likewise.
7861 2018-05-17 Martin Liska <mliska@suse.cz>
7863 * gimple-ssa-sprintf.c (format_directive): Do not use
7864 space in between 'G_' and '('.
7866 2018-05-17 Jakub Jelinek <jakub@redhat.com>
7869 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
7870 even if the mask is not all ones.
7873 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
7875 (ix86_gimple_fold_builtin): Likewise.
7878 * config/i386/i386.c: Include tree-vector-builder.h.
7879 (ix86_vector_shift_count): New function.
7880 (ix86_fold_builtin): Fold shift builtins by scalar count.
7881 (ix86_gimple_fold_builtin): Likewise.
7883 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
7884 _mm512_setzero): New intrinsics.
7886 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
7887 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7889 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
7890 code generation for cases where splatting a value is not useful.
7891 * simplify-rtx.c (simplify_ternary_operation): Simplify
7892 vec_merge across a vec_duplicate and a paradoxical subreg forming
7893 a vector mode to a vec_concat.
7895 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
7897 * config.gcc: Support "goldmont-plus".
7898 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
7900 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7901 PROCESSOR_GOLDMONT_PLUS.
7902 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
7903 (processor_target_table): Add "goldmont-plus".
7904 (PTA_GOLDMONT_PLUS): Define.
7905 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
7906 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
7907 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
7908 (fold_builtin_cpu): Add "goldmont-plus".
7909 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
7910 (ix86_option_override_internal): Add "goldmont-plus".
7911 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
7912 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
7913 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
7914 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
7916 2018-05-17 Richard Biener <rguenther@suse.de>
7918 PR tree-optimization/85757
7919 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
7920 remove defs that only feed that PHI from further processing.
7922 2018-05-16 Jim Wilson <jimw@sifive.com>
7924 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
7926 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
7928 2018-05-16 Mark Wielaard <mark@klomp.org>
7930 * dwarf2out.c (count_index_strings): New function.
7931 (output_indirect_strings): Call count_index_strings and generate
7932 header for dwarf_version >= 5.
7934 2018-05-16 Mark Wielaard <mark@klomp.org>
7936 * dwarf2out.c (dwarf_FORM): New function.
7937 (set_indirect_string): Use dwarf_FORM.
7938 (reset_indirect_string): Likewise.
7939 (size_of_die): Likewise.
7940 (value_format): Likewise.
7941 (output_die): Likewise.
7942 (add_skeleton_AT_string): Likewise.
7943 (output_macinfo_op): Likewise.
7944 (index_string): Likewise.
7945 (output_index_string_offset): Likewise.
7946 (output_index_string): Likewise.
7947 (count_index_strings): Likewise.
7949 2018-05-16 Carl Love <cel@us.ibm.com>
7951 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7952 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
7954 2018-05-16 Martin Jambor <mjambor@suse.cz>
7956 * ipa-prop.c (ipa_free_all_edge_args): Remove.
7957 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
7959 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
7961 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
7962 (fnma<mode>4): Likewise.
7963 (fms<mode>4): Likewise.
7964 (fnms<mode>4): Likewise.
7965 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
7966 (aarch64_fnma<mode>4): Likewise.
7967 (aarch64_fms<mode>4): Likewise.
7968 (aarch64_fnms<mode>4): Likewise.
7969 (aarch64_fnmadd<mode>4): Likewise.
7971 2018-05-16 Jason Merrill <jason@redhat.com>
7973 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
7975 2018-05-16 Richard Biener <rguenther@suse.de>
7977 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
7978 (dump_stmt_cost): Declare.
7979 (add_stmt_cost): Dump cost we add.
7980 (add_stmt_costs): New function.
7981 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
7983 (vect_analyze_stmt): Adjust prototype.
7984 (vectorizable_condition): Likewise.
7985 (vectorizable_live_operation): Likewise.
7986 (vectorizable_reduction): Likewise.
7987 (vectorizable_induction): Likewise.
7988 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
7989 cost vector to pass to vectorizable_ and record afterwards.
7990 (vect_model_reduction_cost): Take cost vector argument and adjust.
7991 (vect_model_induction_cost): Likewise.
7992 (vectorizable_reduction): Likewise.
7993 (vectorizable_induction): Likewise.
7994 (vectorizable_live_operation): Likewise.
7995 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
7996 SLP_TREE_NUMBER_OF_VEC_STMTS.
7997 (vect_analyze_slp_cost_1): Remove.
7998 (vect_analyze_slp_cost): Likewise.
7999 (vect_slp_analyze_node_operations): Take visited args and
8000 a target cost vector. Avoid processing already visited stmt sets.
8001 (vect_slp_analyze_operations): Use a local cost vector to gather
8002 costs and register those of non-discarded instances.
8003 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
8004 (vect_schedule_slp_instance): Remove copying of
8005 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
8007 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
8008 adding cost. Record cost entry location.
8009 (vect_prologue_cost_for_slp_op): Function to compute cost of
8010 a constant or invariant generated for SLP vect in the prologue,
8011 split out from vect_analyze_slp_cost_1.
8012 (vect_model_simple_cost): Make static. Adjust for SLP costing.
8013 (vect_model_promotion_demotion_cost): Likewise.
8014 (vect_model_store_cost): Likewise, make static.
8015 (vect_model_load_cost): Likewise.
8016 (vectorizable_bswap): Add cost vector arg and adjust.
8017 (vectorizable_call): Likewise.
8018 (vectorizable_simd_clone_call): Likewise.
8019 (vectorizable_conversion): Likewise.
8020 (vectorizable_assignment): Likewise.
8021 (vectorizable_shift): Likewise.
8022 (vectorizable_operation): Likewise.
8023 (vectorizable_store): Likewise.
8024 (vectorizable_load): Likewise.
8025 (vectorizable_condition): Likewise.
8026 (vectorizable_comparison): Likewise.
8027 (can_vectorize_live_stmts): Likewise.
8028 (vect_analyze_stmt): Likewise.
8029 (vect_transform_stmt): Adjust calls to vectorizable_*.
8030 * tree-vectorizer.c: Include gimple-pretty-print.h.
8031 (dump_stmt_cost): New function.
8033 2018-05-16 Richard Biener <rguenther@suse.de>
8035 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
8036 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
8037 * tree-ssa-dse.c: Include tree-ssa-loop.h.
8038 (check_name): New callback.
8039 (dse_classify_store): Track cycles via a visited bitmap of PHI
8040 defs and simplify handling of in-loop and across loop dead stores
8041 and properly fail for loop-variant refs. Handle byte-tracking with
8042 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
8045 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
8047 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
8048 (vect_get_mask_type_for_stmt): Likewise.
8049 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
8051 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
8052 to determine the statement's vector type and the vector type that
8053 should be used for calculating nunits. Deal with cases in which
8054 the type has to be deferred.
8055 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
8056 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
8057 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
8058 (vect_determine_vf_for_stmt): New functions, split out from...
8059 (vect_determine_vectorization_factor): ...here.
8060 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
8061 (vect_get_mask_type_for_stmt): New functions, split out from
8062 vect_determine_vectorization_factor.
8064 2018-05-16 Richard Biener <rguenther@suse.de>
8066 * tree-cfg.c (verify_gimple_assign_ternary): Properly
8067 verify the [VEC_]COND_EXPR embedded comparison.
8069 2018-05-15 Martin Sebor <msebor@redhat.com>
8071 PR tree-optimization/85753
8072 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
8073 RECORD_TYPE in addition to ARRAY_TYPE.
8075 2018-05-15 Martin Sebor <msebor@redhat.com>
8078 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
8080 2018-05-15 Richard Biener <rguenther@suse.de>
8082 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
8083 add by_clobber_p one. Change algorithm to collect all defs
8084 representing uses we need to walk and try reducing them to
8085 a single one before failing.
8086 (dse_dom_walker::dse_optimize_stmt): Adjust.
8088 2018-05-13 Mark Wielaard <mark@klomp.org>
8090 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
8091 (size_of_loc_descr): Likewise.
8092 (output_loc_operands): Likewise.
8093 (output_loc_operands_raw): Likewise.
8094 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
8095 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
8096 (hash_loc_operands): Likewise.
8097 (compare_loc_operands): Likewise.
8099 2018-05-14 Mark Wielaard <mark@klomp.org>
8101 * dwarf2out.c (count_index_addrs): New function.
8102 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
8104 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8106 PR tree-optimization/83648
8107 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
8108 return value as malloc candidate.
8110 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8113 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
8114 param as true in call to suggest_attribute.
8116 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
8118 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
8119 -mreadonly-in-sdata.
8121 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8123 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
8125 (aarch64_crypto_aesd_fused): Likewise.
8127 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
8129 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
8130 (movsi_aarch64): Likewise.
8131 (load_pairsi): Likewise.
8132 (load_pairdi): Likewise.
8133 (store_pairsi): Likewise.
8134 (store_pairdi): Likewise.
8135 (load_pairsf): Likewise.
8136 (load_pairdf): Likewise.
8137 (store_pairsf): Likewise.
8138 (store_pairdf): Likewise.
8139 (zero_extend): Likewise.
8140 (trunc): Swap alternatives.
8141 (fcvt_target): Add '?' to prefer w over r.
8143 2018-05-14 Jakub Jelinek <jakub@redhat.com>
8146 * config/i386/i386.md: Disallow non-commutative arithmetics in
8147 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
8148 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
8149 in the peephole2 before it.
8151 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
8153 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
8154 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
8155 (ix86_handle_option): Handle -mcldemote.
8156 * config.gcc: New header.
8157 * config/i386/cldemoteintrin.h: New file.
8158 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
8159 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
8161 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8162 OPTION_MASK_ISA_CLDEMOTE.
8163 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
8164 (ix86_valid_target_attribute_inner_p): Ditto.
8165 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
8166 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
8167 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
8168 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
8169 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
8171 * config/i386/i386.opt: Add -mcldemote.
8172 * config/i386/x86intrin.h: New header.
8173 * doc/invoke.texi: Add -mcldemote.
8175 2018-05-14 Richard Biener <rguenther@suse.de>
8177 * doc/match-and-simplify.texi: Adjust :s documentation.
8179 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
8181 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
8182 intended memcpy size.
8183 (REORDER_45): Likewise.
8185 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
8187 * sort.cc: New file.
8188 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
8189 * vec.c (qsort_chk): Use gcc_qsort.
8190 * Makefile.in (OBJS-libcommon): Add sort.o.
8191 (build/sort.o): New target. Use it...
8192 (BUILD_RTL): ... here, and...
8193 (build/gencfn-macros): ... here, and...
8194 (build/genmatch): ... here.
8196 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8197 Chung-Ju Wu <jasonwucj@gmail.com>
8199 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
8200 * config/nds32/nds32-graywolf.md: New file.
8201 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
8202 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
8204 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
8205 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
8206 * config/nds32/nds32.md (pipeline_model): Add graywolf.
8207 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
8208 * config/nds32/pipelines.md: Include n15 settings.
8210 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8211 Chung-Ju Wu <jasonwucj@gmail.com>
8213 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
8214 * config/nds32/nds32-n13.md: New file.
8215 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
8216 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
8218 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
8219 * config/nds32/nds32.md (pipeline_model): Add n13.
8220 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
8221 * config/nds32/pipelines.md: Include n13 settings.
8223 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8224 Chung-Ju Wu <jasonwucj@gmail.com>
8226 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
8227 * config/nds32/nds32-n10.md: New file.
8228 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
8229 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
8231 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
8232 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
8233 * config/nds32/nds32.md (pipeline_model): Add n10.
8234 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
8235 * config/nds32/pipelines.md: Include n10 settings.
8237 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
8238 Kito Cheng <kito.cheng@gmail.com>
8239 Chung-Ju Wu <jasonwucj@gmail.com>
8241 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
8242 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8243 Add enum values for DSP extension instructions.
8244 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
8246 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
8247 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
8249 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
8250 * config/nds32/nds32-dspext.md: New file for DSP implementation.
8251 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
8252 * config/nds32/nds32-intrinsic.md: Likewise.
8253 * config/nds32/nds32_intrinsic.h: Likewise.
8254 * config/nds32/nds32-md-auxiliary.c: Likewise.
8255 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
8256 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
8257 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
8258 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
8259 * config/nds32/nds32-protos.h: New declarations for DSP extension.
8260 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
8261 TYPE_DMAC in switch statement.
8262 * config/nds32/nds32.c: New checking and implementation for DSP
8263 extension instructions.
8264 * config/nds32/nds32.h: Likewise.
8265 * config/nds32/nds32.md: Likewise.
8266 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
8267 * config/nds32/predicates.md: Implement new predicates for DSP
8270 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8272 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
8273 Reformat alternatives and attributes so it is easier to identify
8274 which constraints/attributes go with which instruction.
8275 (mov<mode>_hardfloat32, FMOVE64): Likewise.
8276 (mov<mode>_softfloat32, FMOVE64): Likewise.
8277 (mov<mode>_hardfloat64, FMOVE64): Likewise.
8278 (mov<mode>_softfloat64, FMOVE64): Likewise.
8280 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
8282 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8284 (Basic PowerPC Built-in Functions): The new name of the
8285 subsection previously known as "PowerPC Built-in Functions".
8286 (Basic PowerPC Built-in Functions Available on all Configurations):
8288 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
8289 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
8290 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
8291 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
8293 2018-05-11 Martin Jambor <mjambor@suse.cz>
8296 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
8299 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
8302 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
8304 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
8306 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
8307 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
8308 (ix86_handle_option): Handle -mwaitpkg.
8309 * config.gcc: New header.
8310 * config/i386/cpuid.h (bit_WAITPKG): New bit.
8311 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
8312 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
8314 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8315 OPTION_MASK_ISA_WAITPKG.
8316 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
8317 (ix86_option_override_internal): Add PTA_WAITPKG.
8318 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
8319 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
8320 IX86_BUILTIN_TPAUSE.
8321 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
8322 __builtin_ia32_umwait and __builtin_ia32_tpause.
8323 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
8324 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
8325 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
8326 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
8327 UNSPECV_TPAUSE): New.
8328 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
8329 * config/i386/i386.opt: Add -mwaitpkg.
8330 * config/i386/waitpkgintrin.h: New file.
8331 * config/i386/x86intrin.h: New header.
8332 * doc/invoke.texi: Add -mwaitpkg.
8334 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
8337 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
8339 (cortex-m0): Use armv6s-m isa.
8340 (cortex-m0plus): Likewise.
8341 (cortex-m1): Likewise.
8342 (cortex-m0.small-multiply): Likewise.
8343 (cortex-m0plus.small-multiply): Likewise.
8344 (cortex-m1.small-multiply): Likewise.
8346 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
8347 Jakub Jelinek <jakub@redhat.com>
8349 PR tree-optimization/85692
8350 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
8351 source permute as well.
8353 2018-05-11 Martin Liska <mliska@suse.cz>
8356 * doc/extend.texi: Document LLVM style format for no_sanitize
8359 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8361 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
8362 mode_supports_vsx_dform_quad to mode_supports_dq_form.
8363 (mode_supports_vsx_dform_quad): Likewise.
8364 (mode_supports_vmx_dform): Move these functions to be next to the
8365 other mode_supports functions.
8366 (mode_supports_dq_form): Likewise.
8367 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
8368 mode_supports_dq_form.
8369 (reg_offset_addressing_ok_p): Likewise.
8370 (offsettable_ok_by_alignment): Likewise.
8371 (rs6000_legitimate_offset_address_p): Likewise.
8372 (legitimate_lo_sum_address_p): Likewise.
8373 (rs6000_legitimize_address): Likewise.
8374 (rs6000_legitimize_reload_address): Likewise.
8375 (rs6000_secondary_reload_inner): Likewise.
8376 (rs6000_preferred_reload_class): Likewise.
8377 (rs6000_output_move_128bit): Likewise.
8379 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
8381 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
8382 Generate SImode target register for null target.
8383 <case IX86_BUILTIN_XGETBV>: Ditto.
8384 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
8385 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
8387 2018-05-10 Carl Love <cel@us.ibm.com>
8389 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
8390 dcbtt and dcbtstt if operands[2] is 0.
8392 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
8395 * config/i386/sse.md (usadv64qi): New expander.
8397 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
8399 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
8400 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
8401 -maltivec=be support.
8402 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
8403 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
8404 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
8405 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
8406 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
8407 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
8408 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
8409 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
8410 altivec_vsumsws): Adjust.
8411 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
8412 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
8413 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
8415 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
8416 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
8417 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
8418 (altivec_lve<VI_char>x): Delete expand.
8419 (*altivec_lve<VI_char>x_internal): Rename to...
8420 (altivec_lve<VI_char>x): ... this.
8421 (altivec_lvxl_<mode>): Delete expand.
8422 (*altivec_lvxl_<mode>_internal): Rename to ...
8423 (altivec_lvxl_<mode>): ... this.
8424 (altivec_stvxl_<mode>): Delete expand.
8425 (*altivec_stvxl_<mode>_internal): Rename to ...
8426 (altivec_stvxl_<mode>): ... this.
8427 (altivec_stve<VI_char>x): Delete expand.
8428 (*altivec_stve<VI_char>x_internal): Rename to ...
8429 (altivec_stve<VI_char>x): ... this.
8430 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
8431 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
8432 reduc_plus_scal_<mode>): Adjust.
8433 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
8435 (rs6000_cpu_cpp_builtins): Adjust.
8436 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
8437 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
8438 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
8439 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
8440 -maltivec=be support.
8441 (rs6000_split_vec_extract_var): Adjust.
8442 (rs6000_split_v4si_init): Adjust.
8443 (swap_selector_for_mode): Delete.
8444 (altivec_expand_lvx_be, altivec_expand_stvx_be,
8445 altivec_expand_stvex_be): Delete.
8446 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
8447 -maltivec=be support.
8448 (rs6000_gimple_fold_builtin): Ditto.
8449 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
8451 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
8452 (TARGET_DIRECT_MOVE_64BIT): Adjust.
8453 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
8454 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
8455 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
8456 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
8457 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
8458 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
8459 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
8460 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
8461 anonymous split): Adjust.
8462 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
8463 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
8465 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
8467 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
8468 when --with-gxx-include-dir is also specified.
8469 * configure: Regenerate.
8471 2018-05-09 Jim Wilson <jimw@sifive.com>
8474 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
8475 * config/riscv/t-withmultilib: New.
8476 * config/riscv/withmultilib.h: New.
8477 * doc/install.texi: Document RISC-V --with-multilib-list support.
8479 2018-05-09 Richard Biener <rguenther@suse.de>
8481 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
8483 (vect_bb_vectorization_profitable_p): Adjust. Compute
8484 actual scalar cost using the cost vector and the add_stmt_cost
8487 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8489 PR rtl-optimization/85645
8490 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
8491 in the REG_CFA_REGISTER note for LR, don't leave it empty.
8493 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8495 PR rtl-optimization/85645
8496 * shrink-wrap.c (spread_components): Return a boolean saying if
8497 anything was changed.
8498 (try_shrink_wrapping_separate): Iterate spread_components until
8499 nothing changes anymore.
8501 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8503 PR rtl-optimization/85645
8504 * regrename.c (build_def_use): Also kill the chains that include the
8505 destination of a REG_CFA_REGISTER note.
8507 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
8509 PR rtl-optimization/85645
8510 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
8511 insn that has a REG_CFA_REGISTER note.
8513 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
8515 * cfgexpand.c (expand_clobber): New function.
8516 (expand_gimple_stmt_1): Use it.
8517 * tree-vect-stmts.c (vect_clobber_variable): New function,
8519 (vectorizable_simd_clone_call): ...here.
8520 (vectorizable_store): Emit a clobber either side of an
8521 IFN_STORE_LANES sequence.
8522 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
8524 2018-05-09 Tom de Vries <tom@codesourcery.com>
8527 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
8528 (define_insn "trap_if_false"): Add exit after trap.
8530 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
8532 PR rtl-optimization/85638
8533 * bb-reorder.c: Include common/common-target.h.
8534 (create_forwarder_block): New function extracted from...
8535 (fix_up_crossing_landing_pad): ...here. Rename into...
8536 (dw2_fix_up_crossing_landing_pad): ...this.
8537 (sjlj_fix_up_crossing_landing_pad): New function.
8538 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
8539 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
8540 from both partitions and exit the loop after one iteration.
8542 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
8545 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8547 (Basic PowerPC Built-in Functions): The new name of the
8548 subsection previously known as "PowerPC Built-in Functions".
8549 (Basic PowerPC Built-in Functions Available on all Configurations):
8551 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
8553 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
8555 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
8557 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
8560 2018-05-08 Jim Wilson <jimw@sifive.com>
8562 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
8563 (LD_EMUL_SUFFIX): New.
8564 (LINK_SPEC): Use it.
8566 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
8568 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8570 (Basic PowerPC Built-in Functions): The new name of the
8571 subsection previously known as "PowerPC Built-in Functions".
8572 (Basic PowerPC Built-in Functions Available on all Configurations):
8574 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
8576 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
8578 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
8580 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
8583 2018-05-08 Jakub Jelinek <jakub@redhat.com>
8586 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
8587 after cmpelim optimization.
8589 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
8591 * config.gcc: Support "goldmont".
8592 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
8593 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8595 * config/i386/i386.c (m_GOLDMONT): Define.
8596 (processor_target_table): Add "goldmont".
8597 (PTA_GOLDMONT): Define.
8598 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
8599 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
8600 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
8601 (fold_builtin_cpu): Add "goldmont".
8602 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
8603 (ix86_option_override_internal): Add "goldmont".
8604 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
8605 (processor_type): Add PROCESSOR_GOLDMONT.
8606 * config/i386/i386.md: Add CPU "glm".
8607 * config/i386/glm.md: New file.
8608 * config/i386/x86-tune.def: Add m_GOLDMONT.
8609 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
8611 2018-05-08 Jakub Jelinek <jakub@redhat.com>
8614 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
8616 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
8617 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
8618 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
8621 * config/i386/i386.c (ix86_fold_builtin): Handle
8622 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
8625 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
8626 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
8628 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
8631 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
8632 (check_arch): Likewise.
8633 (check_fpu): Return the result rather than printing it.
8634 (end arch): Fix operator precedence.
8635 (end cpu): Likewise.
8636 (END): Print the result from check_fpu.
8638 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
8639 Alan Hayward <alan.hayward@arm.com>
8640 David Sherwood <david.sherwood@arm.com>
8642 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
8643 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
8644 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
8645 (*fcmuo<mode>_and): New patterns.
8647 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
8649 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
8650 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
8651 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
8652 (cmp_op, sve_imm_con): New code attributes.
8653 (SVE_COND_INT_CMP, imm_con): Delete.
8654 (cmp_op): Remove above unspecs from int attribute.
8655 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
8657 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
8658 comparison-specific unspecs.
8659 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
8660 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
8661 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
8662 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
8663 (*vec_fcm<cmp_op><mode>): Rename to...
8664 (*fcm<cmp_op><mode>): ...this and adjust likewise.
8665 (*vec_fcmuo<mode>): Rename to...
8666 (*fcmuo<mode>): ...this and adjust likewise.
8667 (*pred_fcm<cmp_op><mode>): New pattern.
8668 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
8669 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
8671 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
8673 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
8674 (aarch64_emit_sve_predicated_cond): New function.
8675 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
8676 (aarch64_emit_unspec_cond_or): Replace with...
8677 (aarch64_emit_sve_or_conds): ...this new function. Use
8678 aarch64_emit_sve_ptrue_op for the individual comparisons and
8679 aarch64_emit_binop to OR them together.
8680 (aarch64_emit_inverted_unspec_cond): Replace with...
8681 (aarch64_emit_sve_inverted_cond): ...this new function. Use
8682 aarch64_emit_sve_ptrue_op for the comparison and
8683 aarch64_emit_unop to invert the result.
8684 (aarch64_expand_sve_vec_cmp_float): Update after the above
8685 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
8687 2018-05-07 Nathan Sidwell <nathan@acm.org>
8689 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
8690 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
8691 (Backwards Compatibility): Likewise.
8693 2018-05-07 Luis Machado <luis.machado@linaro.org>
8697 2018-05-07 Luis Machado <luis.machado@linaro.org>
8699 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8700 <prefetch_dynamic_strides>: New const bool field.
8701 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8702 prefetch_dynamic_strides.
8703 (exynosm1_prefetch_tune): Likewise.
8704 (thunderxt88_prefetch_tune): Likewise.
8705 (thunderx_prefetch_tune): Likewise.
8706 (thunderx2t99_prefetch_tune): Likewise.
8707 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
8709 (aarch64_override_options_internal): Update to set
8710 PARAM_PREFETCH_DYNAMIC_STRIDES.
8711 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
8712 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
8713 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
8714 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
8715 prefetch-dynamic-strides setting.
8717 2018-05-07 Luis Machado <luis.machado@linaro.org>
8719 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8720 <minimum_stride>: New const int field.
8721 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8722 minimum_stride field.
8723 (exynosm1_prefetch_tune): Likewise.
8724 (thunderxt88_prefetch_tune): Likewise.
8725 (thunderx_prefetch_tune): Likewise.
8726 (thunderx2t99_prefetch_tune): Likewise.
8727 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
8728 (aarch64_override_options_internal): Update to set
8729 PARAM_PREFETCH_MINIMUM_STRIDE.
8730 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
8731 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
8732 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
8733 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
8734 stride is constant and is below the minimum stride threshold.
8736 2018-05-07 Luis Machado <luis.machado@linaro.org>
8738 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
8741 2018-05-07 Luis Machado <luis.machado@linaro.org>
8743 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8744 <prefetch_dynamic_strides>: New const bool field.
8745 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8746 prefetch_dynamic_strides.
8747 (exynosm1_prefetch_tune): Likewise.
8748 (thunderxt88_prefetch_tune): Likewise.
8749 (thunderx_prefetch_tune): Likewise.
8750 (thunderx2t99_prefetch_tune): Likewise.
8751 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
8753 (aarch64_override_options_internal): Update to set
8754 PARAM_PREFETCH_DYNAMIC_STRIDES.
8755 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
8756 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
8757 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
8758 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
8759 prefetch-dynamic-strides setting.
8761 2018-05-07 Luis Machado <luis.machado@linaro.org>
8763 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8764 <minimum_stride>: New const int field.
8765 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8766 minimum_stride field.
8767 (exynosm1_prefetch_tune): Likewise.
8768 (thunderxt88_prefetch_tune): Likewise.
8769 (thunderx_prefetch_tune): Likewise.
8770 (thunderx2t99_prefetch_tune): Likewise.
8771 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
8772 (aarch64_override_options_internal): Update to set
8773 PARAM_PREFETCH_MINIMUM_STRIDE.
8774 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
8775 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
8776 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
8777 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
8778 stride is constant and is below the minimum stride threshold.
8780 2018-05-06 Jakub Jelinek <jakub@redhat.com>
8783 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
8784 the type is addressable. Don't force op into register if it has
8787 2018-05-05 Roland McGrath <mcgrathr@google.com>
8790 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
8791 any section for which we don't know a specific type it should have,
8792 regardless of name. Previously this was done only for the exact
8793 names ".init_array", ".fini_array", and ".preinit_array".
8794 (default_elf_asm_named_section): Add comment about
8795 relationship with default_section_type_flags and SECTION_NOTYPE.
8796 (get_section): Don't consider it a type conflict if one side has
8797 SECTION_NOTYPE and the other doesn't, as long as neither has the
8798 SECTION_BSS et al used in the default_section_type_flags logic.
8800 2018-05-05 Tom de Vries <tom@codesourcery.com>
8803 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
8804 (workaround_barsyncs): New function.
8805 (nvptx_reorg): Use workaround_barsyncs.
8806 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
8807 (define_expand "nvptx_membar_cta"): New define_expand.
8808 (define_insn "*nvptx_membar_cta"): New insn.
8810 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
8812 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
8813 To improve optimization opportunities.
8814 * builtin-types.def: The new needed builtin types for the above.
8816 2018-05-04 Richard Biener <rguenther@suse.de>
8818 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
8819 * gimple-ssa-store-merging.c
8820 (imm_store_chain_info::output_merged_store): Remove redundant create,
8821 release split_store vector contents on failure.
8822 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
8823 scalar stmt vector on cache hit.
8825 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
8827 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
8829 * config.gcc (powerpc-xilinx-eabi*): Remove.
8830 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
8832 (fusion_addis_mem_combo_load): Ditto.
8833 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
8835 (rs6000_cpu_cpp_builtins): Ditto.
8836 * config/rs6000/rs6000-linux.c
8837 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
8838 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
8839 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
8841 (rs6000_setup_reg_addr_masks): Ditto.
8842 (rs6000_init_hard_regno_mode_ok): Ditto.
8843 (rs6000_option_override_internal): Ditto.
8844 (legitimate_lo_sum_address_p): Ditto.
8845 (rs6000_legitimize_address): Ditto.
8846 (rs6000_legitimize_reload_address): Ditto.
8847 (rs6000_legitimate_address_p): Ditto.
8848 (abi_v4_pass_in_fpr): Ditto.
8849 (setup_incoming_varargs): Ditto.
8850 (rs6000_gimplify_va_arg): Ditto.
8851 (rs6000_split_multireg_move): Ditto.
8852 (rs6000_savres_strategy): Ditto.
8853 (rs6000_emit_prologue_components): Ditto.
8854 (rs6000_emit_epilogue_components): Ditto.
8855 (rs6000_emit_prologue): Ditto.
8856 (rs6000_emit_epilogue): Ditto.
8857 (rs6000_elf_file_end): Ditto.
8858 (rs6000_function_value): Ditto.
8859 (rs6000_libcall_value): Ditto.
8860 * config/rs6000/rs6000.h: Ditto.
8861 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
8862 (TARGET_MINMAX): ... this. New.
8863 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
8864 * config/rs6000/rs6000.md: Remove Xilinx FP support.
8865 (*movsi_internal1_single): Delete.
8866 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
8867 mfpu=, mxilinx-fpu): Delete.
8868 * config/rs6000/singlefp.h: Delete.
8869 * config/rs6000/sysv4.h: Remove Xilinx FP support.
8870 * config/rs6000/t-rs6000: Ditto.
8871 * config/rs6000/t-xilinx: Delete.
8872 * config/rs6000/titan.md: Adjust for fp_type removal.
8873 * config/rs6000/vsx.md: Remove Xilinx FP support.
8874 (VStype_simple): Delete.
8875 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
8876 * config/rs6000/xfpu.h: Delete.
8877 * config/rs6000/xfpu.md: Delete.
8878 * config/rs6000/xilinx.h: Delete.
8879 * config/rs6000/xilinx.opt: Delete.
8880 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
8881 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
8883 2018-05-04 Tom de Vries <tom@codesourcery.com>
8886 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
8889 2018-05-04 Richard Biener <rguenther@suse.de>
8892 * tree-complex.c (update_complex_assignment): We are always in SSA form.
8893 (expand_complex_div_wide): Likewise.
8894 (expand_complex_operations_1): Likewise.
8895 (expand_complex_libcall): Preserve EH info of the original stmt.
8896 (tree_lower_complex): Handle removed blocks.
8897 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
8898 on complex multiplication and division libcall builtins.
8900 2018-05-04 Richard Biener <rguenther@suse.de>
8903 * fold-const.c (negate_expr_p): Restrict negation of operand
8904 zero of a division to when we know that can happen without
8906 (fold_negate_expr_1): Likewise.
8908 2018-05-04 Jakub Jelinek <jakub@redhat.com>
8911 * real.h (real_nextafter): Declare.
8912 * real.c (real_nextafter): New function.
8913 * fold-const-call.c (fold_const_nextafter): New function.
8914 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
8915 CASE_CFN_NEXTTOWARD.
8916 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
8917 even when arg1_mode is different from arg0_mode.
8919 2018-05-03 Nathan Sidwell <nathan@acm.org>
8921 * doc/extend.texi (Deprecated Features): Remove
8923 (Backwards Compatibility): Likewise.
8924 * doc/invoke.texi (C++ Language Options): Likewise.
8925 (C++ Dialect Options): Likewise.
8927 2018-05-03 Jakub Jelinek <jakub@redhat.com>
8930 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
8931 _mm512_mask_mullox_epi64): New intrinsics.
8933 2018-05-03 Tom de Vries <tom@codesourcery.com>
8936 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8937 dump files): Add offload-tree.
8939 2018-05-03 Richard Biener <rguenther@suse.de>
8941 PR tree-optimization/85615
8942 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
8943 to loops not nested in BBs loop father to avoid creating multi-entry
8946 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8948 PR tree-optimization/70291
8949 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
8950 arguments. Change return type to tree. Emit libcall as a new
8951 statement rather than replacing existing one when inplace_p is true.
8952 (expand_complex_multiplication_components): New function.
8953 (expand_complex_multiplication): Expand floating-point complex
8954 multiplication using the above.
8955 (expand_complex_division): Rename inner_type parameter to type.
8956 Update expand_complex_libcall call-site.
8957 (expand_complex_operations_1): Update expand_complex_multiplication
8958 and expand_complex_division call-sites.
8960 2018-05-02 Jakub Jelinek <jakub@redhat.com>
8963 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
8964 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
8965 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
8966 the highest significant bit of the shift count mask is clear. In
8967 check whether and[sq]i3 is needed verify that all significant bits
8968 of the shift count other than the highest are set.
8970 2018-05-02 Tom de Vries <tom@codesourcery.com>
8973 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
8974 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
8975 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
8976 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
8977 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
8978 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
8979 __builtin_goacc_parlevel_size.
8981 2018-05-02 Richard Biener <rguenther@suse.de>
8983 PR tree-optimization/85597
8984 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
8985 do not use split vect_get_vec_defs call but call vect_get_slp_defs
8988 2018-05-02 Tom de Vries <tom@codesourcery.com>
8991 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8992 dump files): Add ltrans-tree.
8994 2018-05-02 Tom de Vries <tom@codesourcery.com>
8997 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8998 dump files): Add wpa-ipa.
9000 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
9002 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
9003 powerpc*-*-linux*paired* target.
9004 * config/rs6000/750cl.h: Delete.
9005 * config/rs6000/paired.h: Delete.
9006 * config/rs6000/paired.md: Delete.
9007 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
9009 * config/rs6000/rs6000-builtin.def: Remove paired float support.
9010 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
9011 comment. Remove paired float support.
9012 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
9013 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
9015 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
9016 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
9018 * config/rs6000/rs6000.c: Remove paired float support.
9019 (paired_expand_vector_init, paired_expand_vector_move,
9020 paired_emit_vector_compare, paired_emit_vector_cond_expr,
9021 (paired_expand_lv_builtin, paired_expand_stv_builtin,
9022 paired_expand_builtin, paired_expand_predicate_builtin,
9023 paired_init_builtins): Delete.
9024 * config/rs6000/rs6000.h: Remove paired float support.
9025 * config/rs6000/rs6000.md: Remove paired float support.
9026 (move_from_CR_ov_bit): Delete.
9027 * config/rs6000/rs6000.opt (mpaired): Delete.
9028 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
9029 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
9031 2018-05-02 Richard Biener <rguenther@suse.de>
9034 * gimplify.c (gimplify_save_expr): When in SSA form allow
9035 SAVE_EXPRs to compute to SSA vars.
9037 2018-05-02 Jakub Jelinek <jakub@redhat.com>
9040 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
9041 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
9042 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
9043 clobber operands[2], instead use a new pseudo. Formatting fixes.
9045 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
9047 PR tree-optimization/85586
9048 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
9049 exit early for statements in the same group if the accesses are
9052 2018-05-02 Tom de Vries <tom@codesourcery.com>
9055 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
9058 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
9060 PR tree-optimization/85143
9061 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
9063 2018-05-01 Tom de Vries <tom@codesourcery.com>
9066 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
9067 not found" error message.
9069 2018-05-01 Tom de Vries <tom@codesourcery.com>
9072 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
9073 * vec.c (test_ordered_remove_if): New function.
9074 (vec_c_tests): Call test_ordered_remove_if.
9075 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
9076 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
9077 * tree-vect-patterns.c (vect_pattern_recog_1): Use
9078 VEC_ORDERED_REMOVE_IF.
9080 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9082 PR tree-optimization/82665
9083 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
9084 pointer subtraction where arguments come from a memchr call.
9086 2018-05-01 Jakub Jelinek <jakub@redhat.com>
9088 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
9089 --push-state --as-needed and --pop-state instead of --as-needed and
9090 --no-as-needed if ld supports it.
9091 * configure: Regenerated.
9094 * doc/install.texi2html: Replace _002d with - and _002a with * in
9095 generated html files using sed.
9097 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9100 * gcc-rich-location.c (blank_line_before_p): New function.
9101 (use_new_line): New function.
9102 (gcc_rich_location::add_fixit_insert_formatted): New function.
9103 * gcc-rich-location.h
9104 (gcc_rich_location::add_fixit_insert_formatted): New function.
9106 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9108 * selftest.c (assert_streq): Rename "expected" and "actual" to
9109 "val1" and "val2". Extend NULL-handling to cover both inputs
9110 symmetrically, while still requiring both to be non-NULL for a pass.
9111 * selftest.h (assert_streq): Rename "expected" and "actual" to
9113 (ASSERT_EQ): Likewise.
9114 (ASSERT_EQ_AT): Likewise.
9115 (ASSERT_KNOWN_EQ): Likewise.
9116 (ASSERT_KNOWN_EQ_AT): Likewise.
9117 (ASSERT_NE): Likewise.
9118 (ASSERT_MAYBE_NE): Likewise.
9119 (ASSERT_MAYBE_NE_AT): Likewise.
9120 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
9121 the assertion to pass.
9122 (ASSERT_STREQ_AT): Likewise.
9124 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
9126 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
9127 interaction with -pie.
9129 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9131 * selftest.h: Fix alphabetization of per-source-file selftest
9134 2018-04-30 Jason Merrill <jason@redhat.com>
9136 PR c++/61982 - dead stores to destroyed objects.
9137 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
9140 2018-04-30 Jason Merrill <jason@redhat.com>
9142 * tree.c (build_clobber): New.
9143 * tree.h: Declare it.
9144 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
9146 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9148 * diagnostic-show-locus.c (layout::layout): Update for
9149 location_get_source_line returning a char_span.
9150 (struct char_span): Move to input.h.
9151 (struct correction): Update for fields in char_span becoming
9153 (struct source_line): Update for location_get_source_line
9154 returning a char_span.
9155 (layout::print_line): Likewise.
9156 * edit-context.c (edited_file::print_content): Likewise.
9157 (edited_file::print_diff_hunk): Likewise.
9158 (edited_file::print_run_of_changed_lines): Likewise.
9159 (edited_file::get_num_lines): Likewise.
9160 (edited_line::edited_line): Likewise.
9161 * final.c (asm_show_source): Likewise.
9162 * input.c (location_get_source_line): Convert return type
9163 from const char * to char_span, losing the final "line_len"
9165 (dump_location_info): Update for the above.
9166 (get_substring_ranges_for_loc): Likewise. Use a char_span
9167 when handling the literal within the line.
9168 (test_reading_source_line): Update for location_get_source_line
9169 returning a char_span.
9170 * input.h (class char_span): Move here from
9171 diagnostic-show-locus.c, converting from a struct to a class.
9172 Make data members private.
9173 (char_span::operator bool): New.
9174 (char_span::length): New.
9175 (char_span::get_buffer): New.
9176 (char_span::operator[]): New.
9177 (char_span::subspan): Make const.
9178 (char_span::xstrdup): New.
9179 (location_get_source_line): Convert return type from const char *
9180 to char_span, losing the final "line_size" param.
9182 2018-04-30 Jan Hubicka <jh@suse.cz>
9184 * lto-wrapper.c (ltrans_priorities): New static var.
9185 (cmp_priority): New.
9186 (run_gcc): Read priorities and if doing parallel build order
9187 the Makefile by them.
9189 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9191 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
9193 2018-04-30 Richard Biener <rguenther@suse.de>
9195 * tree-cfg.c (verify_address): Remove base argument, add
9196 flag whether to check TREE_ADDRESSABLE and do that.
9197 (verify_expr): Remove.
9198 (verify_types_in_gimple_reference): Add pieces from verify_expr.
9199 (verify_gimple_assign_single): Likewise.
9200 (verify_gimple_switch): Likewise.
9201 (verify_expr_location_1): Dereference tp once. Add (disabled)
9202 piece from verify_expr.
9203 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
9205 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9207 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
9209 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9211 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
9212 (small_data_pattern): Likewise.
9213 (arc_rewrite_small_data): Likewise.
9214 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
9215 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
9216 (get_symbol_alignment): New function.
9217 (legitimate_small_data_address_p): Likewise.
9218 (legitimate_scaled_address): Update, call
9219 legitimate_small_data_address_p.
9220 (output_sdata): New static variable.
9221 (arc_print_operand): Update how we handle small data operands.
9222 (arc_print_operand_address): Likewise.
9223 (arc_legitimate_address_p): Update, use
9224 legitimate_small_data_address_p.
9225 (arc_rewrite_small_data_p): Remove.
9226 (arc_rewrite_small_data_1): Likewise.
9227 (arc_rewrite_small_data): Likewise.
9228 (small_data_pattern): Likewise.
9229 (compact_sda_memory_operand): Update to use
9230 legitimate_small_data_address_p and get_symbol_alignment.
9231 (prepare_move_operands): Don't rewite sdata pattern.
9232 (prepare_extend_operands): Remove.
9233 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
9235 (zero_extendqisi2): Likewise.
9236 (zero_extendhisi2): Likewise.
9237 (extendqihi2): Likewise.
9238 (extendqisi2): Likewise.
9239 (extendhisi2): Likewise.
9243 * config/arc/constraints.md (Usd): Change it to memory constraint.
9245 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9247 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
9248 as source of std instructions.
9249 * config/arc/arc.md (movsi_insn): Update pattern predicate to
9250 allow 6-bit constants as source for store instructions.
9251 (movdi_insn): Update instruction pattern to allow 6-bit constants
9252 as source for store instructions.
9254 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
9256 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
9258 2018-04-30 Nathan Sidwell <nathan@acm.org>
9259 Sandra Loosemore <sandra@codesourcery.com>
9261 * dumpfile.c (dump_open): Allow '-' for stdout.
9262 * doc/invoke.texi (Developer Options): Document dump filename
9263 determination early. Document stdin/stdout selection.
9265 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
9267 Microblaze Target: PIC data text relative
9269 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
9270 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
9272 * config/microblaze/microblaze.h (microblaze_constant_address_p):
9273 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
9274 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
9275 New addressing mode for data-text relative position indepenedent code.
9276 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
9277 'ADDRESS_SYMBOLIC_TXT_REL'.
9278 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
9279 (microblaze_legitimate_pic_operand): Exclude function calls from
9280 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
9281 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
9283 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
9284 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
9285 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
9286 for 'address + offset'.
9287 (microblaze_expand_prologue): Add new function prologue call for
9289 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
9290 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
9291 table in case of TARGET_PIC_DATA_TEXT_REL.
9292 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
9293 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
9294 Add new macros 'UNSPEC_TEXT',
9295 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
9296 + exclude function calls from 'UNSPEC_PLT' in case of data text
9298 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
9299 new target hook for generating address diff vector tables in case of
9301 * doc/tm.texi : Regenerate.
9302 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
9303 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
9304 of addr diff vector generation.
9305 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
9306 target hook definition.
9307 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
9308 Add default function for generate_pic_addr_diff_vec -> flag_pic.
9309 * doc/invoke.texi (Add new pic option): Add new microblaze pic
9310 option for data text relative.
9312 2018-04-30 Richard Biener <rguenther@suse.de>
9314 * tree-chrec.h (evolution_function_is_constant_p): Remove
9316 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
9318 2018-04-30 Richard Biener <rguenther@suse.de>
9321 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
9323 2018-04-30 Richard Biener <rguenther@suse.de>
9325 PR tree-optimization/28364
9326 PR tree-optimization/85275
9327 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
9328 copying first exit test.
9330 2018-04-28 Mark Wielaard <mark@klomp.org>
9332 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
9334 (dwarf_AT): Handle DW_AT_addr_base.
9335 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
9337 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
9340 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
9341 (*ashl<dwi>3_doubleword_mask_1): Ditto.
9342 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
9343 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
9345 2018-04-28 Richard Biener <rguenther@suse.de>
9347 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
9348 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
9349 to reflect use. Only add interesting stmts.
9351 2018-04-27 Martin Jambor <mjambor@suse.cz>
9354 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
9355 the jump function allows for passing through aggregate values.
9357 2018-04-27 David Malcolm <dmalcolm@redhat.com>
9359 * input.h (in_system_header_at): Convert from macro to inline
9361 (from_macro_expansion_at): Likewise.
9362 (from_macro_definition_at): Likewise.
9364 2018-04-27 Jeff Law <law@redhat.com>
9366 * config.gcc: Mark tile* targets as deprecated/obsolete.
9368 2018-04-27 Richard Biener <rguenther@suse.de>
9370 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
9373 2018-04-27 Richard Biener <rguenther@suse.de>
9375 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
9377 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
9379 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
9380 with Yd constraint. Set "preferred_for_speed" attribute from
9381 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
9383 (*movdi_internal): Ditto.
9384 (movti_interunit splitters): Remove
9385 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
9386 (movdi_interunit splitters): Ditto.
9387 * config/i386/constraints.md (Ye): Remove.
9388 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
9390 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9393 * config/aarch64/constraints.md (Usg): Limit to 31.
9396 2018-04-27 Jakub Jelinek <jakub@redhat.com>
9398 PR tree-optimization/85529
9399 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
9400 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
9401 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
9402 zero extension or masking of the MSB bit.
9403 (optimize_range_tests): Add FIRST_BB argument, pass it through
9404 to optimize_range_tests_var_bound.
9405 (maybe_optimize_range_tests, reassociate_bb): Adjust
9406 optimize_range_tests callers.
9408 2018-04-26 Richard Biener <rguenther@suse.de>
9409 Jakub Jelinek <jakub@redhat.com>
9411 * cgraph.h (symbol_table): Just declare debug method here.
9412 * symtab.c (symbol_table::debug): Define.
9414 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
9416 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
9418 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
9420 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
9421 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
9422 (*movdi_internal): Substitute Yi and Yj constraint with x
9423 and Ym and Yn constraint with y constraint. Update "isa"
9424 attribute and set "preferred_for_speed" attribute from
9425 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
9426 (*movsi_internal): Ditto.
9427 (*movdf_internal): Ditto.
9428 (*movsf_internal): Ditto.
9429 (*zero_extendsidi2): Ditto.
9430 * config/i386/sse.md (vec_set<mode>_0): Ditto.
9431 (sse2_loadld): Ditto.
9432 (*vec_extract<ssevecmodelower>_0): Ditto.
9433 (*vec_extractv4si_0_zext_sse4): Ditto.
9434 (vec_concatv2di): Ditto.
9435 (*vec_dup<mode>): Ditto.
9436 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
9437 * config/i386/constraints.md (Yi): Remove.
9442 2018-04-26 Nathan Sidwell <nathan@acm.org>
9444 * dumpfile.c (dump_open): New.
9445 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
9446 (dump_finish): Detect stdio/stderr by value not name.
9448 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
9450 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
9452 2018-04-26 Tom de Vries <tom@codesourcery.com>
9455 * config/nvptx/nvptx.c (verify_neutering_jumps)
9456 (verify_neutering_labels): New function
9457 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
9459 2018-04-26 Tom de Vries <tom@codesourcery.com>
9462 * config/nvptx/nvptx.c (needs_neutering_p): New function.
9463 (nvptx_single): Use needs_neutering_p to skip over insns that do not
9466 2018-04-26 Richard Biener <rguenther@suse.de>
9467 Tom de Vries <tom@codesourcery.com>
9470 * lto-streamer-out.c (output_function): Fixup loops if required to match
9471 discovery done in the reader.
9473 2018-04-26 Richard Biener <rguenther@suse.de>
9475 PR tree-optimization/85116
9476 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
9477 have a loop exit from the single latch predecessor. Remove
9478 case of header with just condition.
9479 (ch_base::copy_headers): Exclude infinite loops from any
9481 (pass_ch::execute): Record exits.
9483 2018-04-26 Richard Biener <rguenther@suse.de>
9485 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
9486 prologue cost vector and pass it to vect_get_load_cost.
9487 (vect_get_peeling_costs_all_drs): Likewise.
9488 (vect_peeling_hash_get_lowest_cost): Likewise.
9489 (vect_enhance_data_refs_alignment): Likewise.
9491 2018-04-26 Richard Biener <rguenther@suse.de>
9494 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
9495 checking of integer<->pointer conversions.
9496 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
9497 sign-/zero-extending pointer types.
9498 (expand_omp_for_static_chunk): Likewise.
9500 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
9501 Jean Lee <xiaoyur347@gmail.com>
9503 * config/mips/mips.c (mips_asan_shadow_offset): New function.
9504 (TARGET_ASAN_SHADOW_OFFSET): Define.
9505 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
9506 true for -fsanitize=address.
9508 2018-04-25 Mark Wielaard <mark@klomp.org>
9510 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
9513 2018-04-25 Jakub Jelinek <jakub@redhat.com>
9515 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
9516 than "alu", remove explicit "memory" and "imm_disp" attributes.
9517 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
9520 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
9521 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
9522 gen_lowpart_no_emit.
9524 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
9527 * config/i386/i386.c (ix86_expand_builtin): Change memory
9528 operand to XI, extend p0 to Pmode.
9529 * config/i386/i386.md: Change unspec volatile and operand
9530 1 mode to XI, change operand 0 mode to P.
9532 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9534 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
9535 GET_MODE_MASK before any checking.
9536 (nds32_can_use_bset_p): Likewise.
9537 (nds32_can_use_btgl_p): Likewise.
9539 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9541 * config/nds32/nds32-doubleword.md: New define_split pattern for
9542 illegal register number.
9544 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9546 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
9548 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
9550 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
9552 2018-04-25 Richard Biener <rguenther@suse.de>
9554 * lto-streamer.h (LTO_major_version): Bump to 8.
9556 2018-04-25 Jakub Jelinek <jakub@redhat.com>
9558 * BASE-VER: Set to 9.0.0.
9560 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
9562 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
9563 in __abskf2 and __powikf2.
9565 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9568 * config/aarch64/constraints.md (Usg, Usj): New constraints.
9569 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
9570 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
9571 Use the above on operand 2. Reindent.
9572 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
9574 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
9577 * common/config/i386/i386-common.c (ix86_handle_option): Don't
9579 * config/i386/i386.opt (mcet): Removed.
9580 * doc/install.texi: Remove -mcet documentation.
9581 * doc/invoke.texi: Likewise.
9583 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
9586 * doc/install.texi: Remove -mcet from bootstrap-cet.
9588 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9591 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
9592 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
9596 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
9597 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
9598 containing a CONST_VECTOR.
9600 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
9602 * doc/install.texi: Update newlib dependency for nvptx.
9604 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9607 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
9608 instead of INTVAL when shifting x left.
9610 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
9612 PR tree-optimization/85478
9613 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
9614 vect_grouped_store_supported for single element vectors.
9616 2018-04-24 Richard Biener <rguenther@suse.de>
9619 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
9620 load cost increase to the case of non-constant step.
9622 2018-04-24 Jakub Jelinek <jakub@redhat.com>
9625 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
9626 destination if any_malformed_asm.
9628 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
9631 * expr.c (store_field): In the bitfield case, if the value comes from
9632 a function call and is returned in registers by means of a PARALLEL,
9633 do not change the mode of the temporary unless BLKmode and VOIDmode.
9635 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
9637 PR rtl-optimization/85423
9638 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
9639 dependencies to debug insns when the previous insn is non-debug.
9641 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
9643 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
9644 enums into a single definition.
9645 (fls): Fix predicates and printing.
9648 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
9650 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
9651 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
9652 and short u6 immediate.
9653 (check_if_valid_sleep_operand): Remove.
9654 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
9656 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9658 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
9659 flag_always_save_lp condition.
9660 * config/nds32/nds32.opt (malways-save-lp): New option.
9662 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9664 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
9665 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
9666 * config/nds32/nds32.h
9667 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
9668 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
9670 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9672 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
9673 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
9675 2018-04-22 Shiva Chen <shiva0217@gmail.com>
9676 Chung-Ju Wu <jasonwucj@gmail.com>
9678 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
9680 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
9681 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
9683 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9685 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
9687 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9689 * config/nds32/nds32-protos.h (nds32_data_alignment,
9690 nds32_local_alignment): Declare.
9691 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
9692 nds32_local_alignment): New functions.
9693 (TARGET_CONSTANT_ALIGNMENT): Define.
9694 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
9696 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9698 * config/nds32/nds32.c
9699 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
9700 (TARGET_MODES_TIEABLE_P): Likewise.
9702 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
9704 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
9707 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
9708 Chung-Ju Wu <jasonwucj@gmail.com>
9710 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9711 for unaligned access.
9712 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9714 * config/nds32/nds32-intrinsic.md: Likewise.
9715 * config/nds32/nds32_intrinsic.h: Likewise.
9716 * config/nds32/nds32.h (nds32_builtins): Likewise.
9717 * config/nds32/nds32.opt (munaligned-access): New option.
9718 * config/nds32/nds32.c (nds32_asm_file_start): Display
9719 flag_unaligned_access status.
9721 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
9723 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
9724 -mno-relax is present.
9725 * config/riscv/linux.h (LINK_SPEC): Ditto.
9727 2018-04-20 Martin Sebor <msebor@redhat.com>
9730 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
9732 (gimple_fold_builtin_stxcpy_chk): Same.
9733 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
9735 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
9738 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
9739 __powikf2 when long double is IEEE 128-bit.
9741 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
9743 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
9744 step to make sure stack always aligned.
9746 2018-04-20 Carl Love <cel@us.ibm.com>
9749 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
9750 size check for arg0.
9752 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
9753 Tom de Vries <tom@codesourcery.com>
9756 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
9757 Emit insns for calls too.
9758 (nvptx_find_par): Always look for worker-level predecessor insn.
9759 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
9761 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
9762 (nvptx_process_pars): Propagate frames for calls.
9764 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
9767 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
9769 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
9770 (ix86_handle_option): Don't handle OPT_mibt.
9771 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
9773 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
9775 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
9776 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
9777 (ix86_target_macros): Define __CET__ with flag_cf_protection
9778 for -fcf-protection.
9779 * config/i386/i386.c (isa2_opts): Remove -mibt.
9780 * config/i386/i386.h (TARGET_IBT): Removed.
9781 (TARGET_IBT_P): Likewise.
9782 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
9783 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
9784 * config/i386/i386.opt (mcet): Update help message.
9787 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
9788 -mcet as an alias for -mshstk.
9790 2018-04-20 Richard Biener <rguenther@suse.de>
9793 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
9794 complexity by forcing a single use of the multiply operand.
9796 2018-04-20 Martin Jambor <mjambor@suse.cz>
9799 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
9800 recursion dependency to only apply to non-clones.
9802 2018-04-20 Martin Jambor <mjambor@suse.cz>
9805 * ipa-cp.c (create_specialized_node): Check that clones of
9806 self-recursive edges exist during IPA-CP.
9808 2018-04-19 Toon Moene <toon@moene.org>
9810 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
9813 2018-04-19 Jakub Jelinek <jakub@redhat.com>
9815 PR tree-optimization/85467
9816 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
9817 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
9818 VECTOR_CST element to type.
9820 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9823 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
9824 * config/i386/i386.md (builtin_setjmp_setup): Removed.
9825 (builtin_longjmp): Likewise.
9826 (save_stack_nonlocal): New pattern.
9827 (restore_stack_nonlocal): Likewise.
9829 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9832 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
9833 Replace ASM_OUTPUT_LABEL with fprintf.
9835 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9838 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
9839 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
9840 * config/i386/i386-c.c (ix86_target_macros_internal): Also
9841 define __IBT__ and __SHSTK__ for -fcf-protection.
9842 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
9844 (ix86_trampoline_init): Likewise.
9845 (x86_output_mi_thunk): Likewise.
9846 (ix86_notrack_prefixed_insn_p): Likewise.
9847 (ix86_option_override_internal): Don't disallow -fcf-protection.
9848 * config/i386/i386.md (rdssp<mode>): Also enable for
9850 (incssp<mode>): Likewise.
9851 (nop_endbr): Likewise.
9852 * config/i386/i386.opt (mcet): Change help message to built-in
9856 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
9857 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
9858 enable CET built-in functions.
9860 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
9862 * common/config/i386/i386-common.c
9863 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
9864 OPTION_MASK_ISA_MOVDIRI_UNSET,
9865 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
9866 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
9867 * config.gcc (movdirintrin.h): New header.
9868 * config/i386/cpuid.h (bit_MOVDIRI,
9869 bit_MOVDIR64B): New bits.
9870 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
9872 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
9873 (VOID, PVOID, PCVOID)): New function types.
9874 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
9875 __builtin_ia32_directstoreu_u64,
9876 __builtin_ia32_movdir64b): New builtins.
9877 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
9878 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
9880 (ix86_valid_target_attribute_inner_p): Ditto.
9881 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
9882 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
9883 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
9884 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
9885 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
9886 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
9887 (movdiri<mode>, movdir64b_<mode>): New.
9888 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
9889 * config/i386/immintrin.h: Include movdirintrin.h.
9890 * config/i386/movdirintrin.h: New file.
9891 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
9893 2018-04-19 Richard Biener <rguenther@suse.de>
9896 * cfg.c (clear_bb_flags): When loop state says we have
9897 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
9899 2018-04-19 Richard Biener <rguenther@suse.de>
9901 PR tree-optimization/84737
9902 * tree-vect-data-refs.c (vect_copy_ref_info): New function
9903 copying restrict info.
9904 (vect_setup_realignment): Use it.
9905 * tree-vectorizer.h (vect_copy_ref_info): Declare.
9906 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
9907 the first DR to all generated stores.
9908 (vectorizable_load): Likewise for loads.
9910 2018-04-19 Jakub Jelinek <jakub@redhat.com>
9912 PR tree-optimization/85446
9913 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
9914 the integral and pointer types to have the same precision.
9916 * doc/install.texi: Document --disable-cet being the default and
9919 2018-04-18 Martin Liska <mliska@suse.cz>
9921 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
9924 2018-04-18 Martin Liska <mliska@suse.cz>
9927 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9930 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9931 arguments if they are comparable.
9933 2018-04-18 Martin Liska <mliska@suse.cz>
9936 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9939 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9942 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
9945 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
9946 ENDBR after calling __morestack.
9948 2018-04-18 David Malcolm <dmalcolm@redhat.com>
9951 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
9952 by using gcc_base_ver to generate a gcc_driver_version, and use
9953 it when generating GCC_DRIVER_NAME.
9954 * configure: Regenerate.
9956 2018-04-18 Jakub Jelinek <jakub@redhat.com>
9959 * config.gcc: Obsolete powerpc*-*-*spe*.
9961 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9964 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
9965 (stabstr_D): Change type of unum from unsigned int to
9966 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
9969 2018-04-17 Jim Wilson <jimw@sifive.com>
9972 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
9973 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
9974 Set arg_pointer_offset after using pretend_args_size.
9976 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9978 PR rtl-optimization/85431
9979 * dse.c (record_store): Ignore zero width stores.
9982 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
9983 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
9984 __builtin_stack_restore rather than after it.
9985 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
9986 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
9987 argument instead of virtual_dynamic_stack_rtx.
9989 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
9991 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
9993 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
9994 Add note to error message to explain internal mapping of overloaded
9995 built-in function name to non-overloaded built-in function name.
9996 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
9999 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10002 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
10003 where the inputs overlap with the output.
10005 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10007 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
10008 (=v, v) alternative and explicit "memory" attribute.
10009 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
10010 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10012 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
10013 "sselog1" type instead of "sselog".
10014 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
10015 "sselog". Remove explicit "memory" attribute.
10016 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
10017 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10019 (vec_extract_hi_v32hi): Merge all alternatives into one, use
10020 "sselog1" type instead of "sselog". Remove explicit "memory"
10022 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
10023 use "sselog1" type instead of "sselog". Remove explicit "memory"
10025 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
10026 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10028 (vec_extract_hi_v64qi): Merge all alternatives into one, use
10029 "sselog1" type instead of "sselog". Remove explicit "memory"
10031 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
10032 use "sselog1" type instead of "sselog". Remove explicit "memory"
10036 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
10038 PR middle-end/85414
10039 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
10042 2018-04-17 Martin Jambor <mjambor@suse.cz>
10045 * ipa-cp.c (create_specialized_node): Call
10046 expand_all_artificial_thunks if necessary.
10048 2018-04-17 Martin Liska <mliska@suse.cz>
10051 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
10052 in message, remote space in between '_G' and '('.
10054 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10057 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
10058 avx512f_vmcmp<mode>3<round_saeonly_name>,
10059 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
10060 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
10061 avx512f_rndscale<mode><round_saeonly_name>,
10062 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
10063 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
10064 Use %<iptr>2 instead of %2 for -masm=intel.
10065 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
10066 avx512f_vcvttss2usi<round_saeonly_name>,
10067 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
10069 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
10070 avx512f_vcvttsd2usi<round_saeonly_name>,
10071 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
10072 Use %q1 instead of %1 for -masm=intel.
10073 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
10074 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
10075 of %3 for -masm=intel.
10076 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
10078 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
10080 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
10082 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
10083 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
10085 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
10087 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
10088 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
10089 %g1 and one with %0 and %1.
10090 (avx512er_vmrcp28<mode><round_saeonly_name>,
10091 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
10092 %1 for -masm=intel.
10093 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
10094 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
10095 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
10096 of %0 and %{%4%} for -masm=intel.
10097 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
10098 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
10099 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
10100 order of %0 and %{%5%}%{z%} for -masm=intel.
10102 2018-04-17 Jan Hubicka <jh@suse.cz>
10105 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
10107 2018-04-17 Martin Liska <mliska@suse.cz>
10110 * multiple_target.c (create_dispatcher_calls): Set apostrophes
10111 for target_clone error message. Make default implementation
10112 clone to be a local declaration.
10113 (separate_attrs): Add new argument and check for an empty
10115 (expand_target_clones): Handle it.
10116 (ipa_target_clone): Make redirection just for target_clones
10119 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
10120 Tom de Vries <tom@codesourcery.com>
10122 PR middle-end/84955
10123 * omp-expand.c (expand_oacc_for): Add dummy false branch for
10124 tiled basic blocks without omp continue statements.
10126 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
10129 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
10130 vec_extract expression as having side effects to make sure it gets
10133 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
10136 * config/i386/i386.c (get_builtin_code_for_version): Check
10139 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
10142 * config.gcc: Support "skylake".
10143 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10145 * config/i386/i386.c (m_SKYLAKE): Define.
10146 (processor_target_table): Add "skylake".
10147 (ix86_option_override_internal): Add "skylake".
10148 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
10149 PROCESSOR_CANNONLAKE.
10150 (get_builtin_code_for_version): Fix priority for
10151 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
10152 PROCESSOR_SKYLAKE-AVX512.
10153 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
10154 (processor_type): Add PROCESSOR_SKYLAKE.
10156 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
10157 Jason Merrill <jason@redhat.com>
10160 * convert.c (convert_to_integer_1): Use direct recursion for
10161 enumeral types and types with a precision less than the number
10162 of bits in their mode.
10164 2018-04-16 Julia Koval <julia.koval@intel.com>
10167 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
10168 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
10170 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
10173 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
10174 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
10175 and -mno-direct-move.
10177 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
10180 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
10181 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
10182 Ensure negative shifts result in {0}.
10184 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
10186 PR rtl-optimization/79916
10187 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
10188 regs (if any) to define how to gnerate SD moves when LRA is in
10191 2018-04-13 Jakub Jelinek <jakub@redhat.com>
10193 PR rtl-optimization/85393
10194 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
10195 * except.c (expand_dw2_landing_pad_for_region): Make static.
10196 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
10197 a label and unconditional jump to old_bb, rather than
10198 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
10201 PR rtl-optimization/85376
10202 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
10203 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
10204 instead of a specific value.
10206 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
10207 Bin Cheng <bin.cheng@arm.com>
10209 PR tree-optimization/82965
10210 PR tree-optimization/83991
10211 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
10212 by_profile_only parameter.
10213 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
10214 information if the loop was predicted to iterate too many times.
10215 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
10217 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
10220 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
10223 2018-04-13 Martin Liska <mliska@suse.cz>
10224 Jakub Jelinek <jakub@redhat.com>
10226 PR middle-end/81657
10227 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
10228 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
10229 * builtins.c (expand_builtin_memory_copy_args): Use
10230 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
10231 handle dest_addr == pc_rtx.
10233 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
10236 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
10237 asked to not generate direct moves.
10238 (fix_trunc<mode>si2_stfiwx): Similar.
10239 (fix_trunc<mode>si2_internal): Similar.
10241 2018-04-12 Jakub Jelinek <jakub@redhat.com>
10244 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
10245 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
10246 lookup if dest in some wider mode is known to be const0_rtx and
10247 if so, record permanent equivalence for it to be ZERO_EXTEND of
10248 the narrower mode destination.
10250 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
10252 * lto-streamer-out.c (output_function): Revert 259346.
10253 * omp-expand.c (expand_oacc_for): Likewise.
10255 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
10257 PR rtl-optimization/85354
10258 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
10259 * sel-sched.c (sel_global_init): ... here.
10261 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
10264 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
10265 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
10266 mode for PE-COFF targets.
10267 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
10268 (i386_pe_asm_lto_end): Likewise.
10269 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
10270 (TARGET_ASM_LTO_END): Likewise.
10271 * config/i386/winnt.c (saved_debug_info_level): New static variable.
10272 (i386_pe_asm_lto_start): New function.
10273 (i386_pe_asm_lto_end): Likewise.
10275 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
10276 Richard Biener <rguenther@suse.de>
10278 PR middle-end/84955
10279 * lto-streamer-out.c (output_function): Fix CFG loop state before
10281 * omp-expand.c (expand_oacc_for): Handle calls to internal
10282 functions like regular functions.
10284 2018-04-12 Richard Biener <rguenther@suse.de>
10287 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
10288 for the early LTO debug to properly generate references to it
10289 during DIE emission. Do not re-use that for the skeleton for
10291 (dwarf2out_early_finish): Likewise.
10293 2018-04-12 Jakub Jelinek <jakub@redhat.com>
10296 * config/i386/sse.md
10297 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
10298 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
10299 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
10300 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
10301 and output is a reg, avoid creating invalid lowpart subreg, but
10302 instead split into a 512-bit move. Don't split if not AVX512VL,
10303 input is xmm16+ reg and output is a mem.
10304 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
10305 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
10306 xmm16+ reg and output is a mem.
10308 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10310 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
10311 also for flag_dwarf2_cfi_asm.
10313 2018-04-12 Jakub Jelinek <jakub@redhat.com>
10315 PR rtl-optimization/85342
10316 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
10317 a bool scalar var inside of the loop instead. Don't try to update
10318 recog_data.operand after failed apply_change_group.
10320 2018-04-12 Tom de Vries <tom@codesourcery.com>
10323 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
10324 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
10325 array with flexible array member as array without given dimension.
10326 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
10327 argument for undefined param to true.
10329 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
10332 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
10333 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
10334 from PowerPC section.
10335 * config/rs6000/sysv4.opt (mcall-): Improve help text.
10336 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
10337 help text that is too long.
10338 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
10339 help text that is too long.
10340 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
10341 help text that is too long.
10343 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
10345 * config/alpha/alpha.md (stack_probe_internal): Rename
10346 from "probe_stack". Update all callers.
10348 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10350 PR rtl-optimization/84566
10351 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
10352 sched_macro_fuse_insns.
10354 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10357 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
10358 (compute_block_dependences): ... from here.
10360 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10362 PR tree-optimization/85331
10363 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
10364 from int to HOST_WIDE_INT.
10366 2018-04-11 Martin Jambor <mjambor@suse.cz>
10369 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
10370 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
10371 not the same as the source val.
10372 (cgraph_edge_brings_value_p): New parameter.
10373 (gather_edges_for_value): Pass destination value to
10374 cgraph_edge_brings_value_p.
10375 (perhaps_add_new_callers): Likewise.
10376 (get_info_about_necessary_edges): Likewise and exclude values brought
10377 only by self-recursive edges.
10378 (create_specialized_node): Redirect only clones of self-calling edges.
10379 (+self_recursive_pass_through_p): New function.
10380 (find_more_scalar_values_for_callers_subset): Use it.
10381 (find_aggregate_values_for_callers_subset): Likewise.
10382 (known_aggs_to_agg_replacement_list): Removed.
10383 (decide_whether_version_node): Re-calculate known constants for all
10384 remaining context clones.
10386 2018-04-11 Richard Biener <rguenther@suse.de>
10389 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
10390 from early DWARF output.
10391 (dwarf2out_early_finish): Output line info unconditionally into
10392 early DWARF and add reference to it.
10394 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10397 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
10398 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
10399 other than V2DFmode using iptr mode attribute.
10400 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
10402 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10404 PR rtl-optimization/84659
10405 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
10407 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10410 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
10412 (output_loc_list): Pass address of a dummy size variable even in the
10413 locview handling loop.
10414 (index_location_lists): Add comment on why skip_loc_list_entry can't
10417 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
10420 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
10423 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
10426 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
10427 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
10428 and -mstring-compare-inline-limit.
10430 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10433 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
10434 for stack clash protection in a register whenever we need it to be in
10437 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10439 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
10440 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
10442 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10445 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
10447 (mlong-double-): Ditto.
10448 * config/rs6000/sysv4.opt (msdata=): Ditto.
10449 (mtls-size=): Ditto.
10451 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10453 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
10454 erroneous entries for
10455 "vector int vec_ldl (int, long int *)", and
10456 "vector unsigned int vec_ldl (int, unsigned long int *)".
10457 Add comments and entries for
10458 "vector bool char vec_ldl (int, bool char *)",
10459 "vector bool short vec_ldl (int, bool short *)",
10460 "vector bool int vec_ldl (int, bool int *)",
10461 "vector bool long long vec_ldl (int, bool long long *)",
10462 "vector pixel vec_ldl (int, pixel *)",
10463 "vector long long vec_ldl (int, long long *)",
10464 "vector unsigned long long vec_ldl (int, unsigned long long *)".
10465 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
10466 type tree bool_long_long_type_node and correct definition of
10467 bool_V2DI_type_node to make reference to this new type tree.
10468 (rs6000_mangle_type): Replace erroneous reference to
10469 bool_long_type_node with bool_long_long_type_node.
10470 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
10471 comments to emphasize sign distinctions for char and int types and
10472 replace RS6000_BTI_bool_long constant with
10473 RS6000_BTI_bool_long_long constant. Also add comment to restrict
10474 use of RS6000_BTI_pixel.
10475 (bool_long_type_node): Remove this macro definition.
10476 (bool_long_long_type_node): New macro definition
10478 2018-04-10 Jakub Jelinek <jakub@redhat.com>
10480 PR rtl-optimization/85300
10481 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
10482 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
10483 simplify_unary_operation fails.
10485 2018-04-10 Martin Liska <mliska@suse.cz>
10487 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
10488 cgraph_edge and ipa_ref.
10490 2018-04-10 Jakub Jelinek <jakub@redhat.com>
10494 * config/i386/sse.md
10495 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
10496 computation of the VEC_MERGE selector from mask.
10497 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
10498 Fix decoding of the VEC_MERGE selector into mask.
10500 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
10502 PR tree-optimization/85286
10503 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
10505 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
10507 * final.c (final_1): Set insn_last_address as well as
10508 insn_current_address.
10510 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10513 * explow.c (emit_stack_probe): Call validize_mem on memory location
10514 before passing it to gen_probe_stack. Create address operand and
10515 legitimize it for the probe_stack_address case.
10517 2018-04-09 Jan Hubicka <jh@suse.cz>
10520 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
10521 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
10522 * tree.c (free_lang_data_in_type): Fix handling of binfos;
10524 (free_lang_data): Rebuild type inheritance graph.
10526 2018-04-09 Martin Sebor <msebor@redhat.com>
10528 * invoke.texi (-finline-small-functions): Mention other optimization
10530 (-findirect-inlining, -fpartial-inlining): Same.
10531 (-finline-functions-called-once): Same.
10532 (-freorder-blocks-and-partition): Same.
10534 2018-04-09 Jan Hubicka <jh@suse.cz>
10537 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
10538 jumps; choose last target that matches the criteria (i.e.
10539 no partition changes for non-crossing jumps).
10540 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
10541 support for redirecting crossing jumps to non-crossing.
10543 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
10545 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
10546 also for naked functions.
10548 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
10550 * config/arc/arc.md (add_shift): New pattern.
10551 (add_shift2): Likewise.
10552 (sub_shift): Likewise.
10553 (sub_shift_cmp0_noout): Likewise.
10554 (compare_si_ashiftsi): Likewise.
10555 (xbfu_cmp0_noout): New combine pattern.
10556 (xbfu_cmp0"): Likewise.
10557 (movsi_set_cc_insn): Place the predicable variant first.
10558 (commutative_binary_cmp0_noout): Remove clobber.
10559 (commutative_binary_cmp0): New pattern.
10560 (noncommutative_binary_cmp0): Likewise.
10561 (noncommutative_binary_cmp0_noout): Likewise.
10562 (noncommutative_binary_comparison_result_used): Removed.
10563 (rsub_cmp0): New pattern.
10564 (rsub_cmp0_noout): Likewise.
10565 (extzvsi): Changed, keep only meaningful variants.
10566 (SQH, SEZ): New iterators.
10567 (SQH_postfix): New mode attribute.
10568 (SEZ_prefix): New code attribute.
10569 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
10570 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
10571 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
10572 of numerical value.
10573 (noncommutative_operator): Check the availability of barrel
10576 2018-04-09 Richard Biener <rguenther@suse.de>
10578 PR tree-optimization/85284
10579 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
10580 Only use the niter constraining form of simple_iv when the exit
10581 is always executed.
10583 2018-04-09 Tom de Vries <tom@codesourcery.com>
10586 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
10587 (define_expand "*memory_barrier"): New define_expand.
10588 (define_insn "memory_barrier"): New insn.
10590 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10592 PR rtl-optimization/80463
10593 PR rtl-optimization/83972
10594 PR rtl-optimization/83480
10596 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
10597 correct producer for the insn.
10598 (tidy_control_flow): Fixup seqnos in case of debug insns.
10600 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10602 PR rtl-optimization/83913
10604 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
10605 different sched-times when merging exprs.
10607 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10609 PR rtl-optimization/83962
10611 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
10612 tidy_fallthru_edge and tidy_control_flow.
10614 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
10616 PR rtl-optimization/83530
10618 * sel-sched.c (force_next_insn): New global variable.
10619 (remove_insn_for_debug): When force_next_insn is true, also leave only
10620 next insn in the ready list.
10621 (sel_sched_region): When the region wasn't scheduled, make another pass
10622 over it with force_next_insn set to 1.
10624 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
10626 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
10628 * config/nds32/constants.md (unspec_volatile_element): Add enum values
10629 for interrupt control.
10630 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
10631 functions for interrupt control.
10632 * config/nds32/nds32-intrinsic.md: Likewise.
10633 * config/nds32/nds32_intrinsic.h: Likewise.
10634 * config/nds32/nds32.h (nds32_builtins): Likewise.
10636 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
10638 * config/nds32/nds32.c (nds32_init_machine_status,
10639 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
10640 strict_aligned_p field.
10641 (nds32_expand_to_rtl_hook): New function.
10642 (TARGET_EXPAND_TO_RTL_HOOK): Define.
10643 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
10645 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10646 Chung-Ju Wu <jasonwucj@gmail.com>
10648 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
10649 * config/nds32/nds32-n7.md: New file.
10650 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
10651 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
10653 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
10654 * config/nds32/nds32.md (pipeline_model): Add n7.
10655 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
10656 * config/nds32/pipelines.md: Include n7 settings.
10658 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10659 Chung-Ju Wu <jasonwucj@gmail.com>
10661 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
10662 * config/nds32/nds32-e8.md: New file.
10663 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
10664 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
10666 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
10667 * config/nds32/nds32.md (pipeline_model): Add e8.
10668 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
10669 * config/nds32/pipelines.md: Include e8 settings.
10671 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10672 Chung-Ju Wu <jasonwucj@gmail.com>
10674 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
10675 * config/nds32/nds32-n8.md: New file.
10676 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
10677 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
10679 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
10680 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
10681 * config/nds32/nds32.md (pipeline_model): Add n8.
10682 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
10683 * config/nds32/pipelines.md: Include n8 settings.
10685 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
10686 Chung-Ju Wu <jasonwucj@gmail.com>
10688 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
10689 * config/nds32/nds32-n9-2r1w.md: New file.
10690 * config/nds32/nds32-n9-3r2w.md: New file.
10691 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
10692 nds32_register_ports): New or modify for cpu n9.
10693 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
10695 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
10696 * config/nds32/nds32-utils.c: New file.
10697 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
10698 TARGET_MUL_SLOW): Define.
10699 * config/nds32/nds32.md (pipeline_model): New attribute.
10700 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
10701 New options that support cpu n9.
10702 * config/nds32/pipelines.md: Include n9 settings.
10703 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
10705 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
10707 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
10708 information if necessary.
10709 (output_cond_branch_compare_zero): Likewise.
10710 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
10711 (nds32_target_alignment): Refine for alignment.
10712 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
10713 (FUNCTION_BOUNDARY): Modify.
10714 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
10716 * config/nds32/nds32.opt (malways-align, malign-functions): New.
10718 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
10720 * config/nds32/constants.md (unspec_volatile_element): Add values for
10721 TLB operation and data prefetch.
10722 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
10723 functions for TLB operation and data prefetch.
10724 * config/nds32/nds32-intrinsic.md: Likewise.
10725 * config/nds32/nds32_intrinsic.h: Likewise.
10726 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
10727 (nds32_print_operand): Likewise.
10728 * config/nds32/nds32.h (nds32_builtins): Likewise.
10730 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
10731 Andrew Pinski <pinsika@gcc.gnu.org>
10733 PR middle-end/82976
10734 * match.pd: Use constant_boolean_node of correct type instead of
10735 boolean_true_node or boolean_false_node for simplifying
10736 pointer comparisons to zero.
10738 2018-04-07 Jakub Jelinek <jakub@redhat.com>
10740 PR tree-optimization/80021
10741 * tree.c (verify_type_variant): Make error call in verify_variant_match
10742 translatable and remove final full stop.
10744 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10746 * config/nds32/constants.md (unspec_volatile_element): Add
10747 UNSPEC_VOLATILE_EH_RETURN.
10748 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
10749 nds32_output_stack_pop): Support dwarf exception handling process.
10750 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
10751 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
10752 exception handling process.
10753 (nds32_compute_stack_frame): Likewise.
10754 (nds32_return_addr_rtx): Likewise.
10755 (nds32_initial_elimination_offset): Likewise.
10756 (nds32_expand_prologue): Likewise.
10757 (nds32_expand_epilogue): Likewise.
10758 (nds32_dynamic_chain_address): New function.
10759 * config/nds32/nds32.h (machine_function): Add fields for dwarf
10760 exception handling.
10761 (DYNAMIC_CHAIN_ADDRESS): Define.
10762 (EH_RETURN_DATA_REGNO): Define.
10763 (EH_RETURN_STACKADJ_RTX): Define.
10764 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
10765 patterns for dwarf exception handling.
10767 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10769 * config/nds32/nds32.h: Clean up obsolete macros.
10771 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10773 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10774 Add enum values for particular instructions.
10775 * config/nds32/nds32-intrinsic.c: Implementation of expanding
10776 particular intrinsic functions.
10777 * config/nds32/nds32-intrinsic.md: Likewise.
10778 * config/nds32/nds32_intrinsic.h: Likewise.
10779 * config/nds32/nds32.h (nds32_builtins): Likewise.
10780 * config/nds32/nds32.md (type): Add pbsad and pbsada.
10781 (btst, ave): New patterns for particular instructions.
10783 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10785 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10786 Add enum values for atomic load/store and memory sync.
10787 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
10789 * config/nds32/nds32-intrinsic.md: Likewise.
10790 * config/nds32/nds32_intrinsic.h: Likewise.
10791 * config/nds32/nds32.h (nds32_builtins): Likewise.
10793 2018-04-07 Jakub Jelinek <jakub@redhat.com>
10795 PR tree-optimization/85257
10796 * fold-const.c (native_encode_vector): If not all elts could fit
10797 and off is -1, return 0 rather than offset.
10798 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
10799 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
10800 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
10801 adjust buffer in native_interpret_expr call.
10803 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10805 * config/nds32/constants.md (unspec_volatile_element): Add cache
10806 control enum values.
10807 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
10808 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
10809 * config/nds32/nds32.c (nds32_cctl_names): New.
10810 (nds32_print_operand): Handle cache control register names.
10811 * config/nds32/nds32.h (nds32_builtins): New enum values.
10812 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
10814 * config/nds32/nds32.md (type): Add mmu.
10815 * config/nds32/pipelines.md (simple_insn): Add mmu.
10817 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10819 * config/nds32/nds32.md (type): Remove call.
10820 * config/nds32/pipelines.md (simple_insn): Likewise.
10822 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10824 * config/nds32/constants.md (unspec_volatile_element): Add
10825 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
10826 UNSPEC_VOLATILE_FMFCFG.
10827 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
10828 description for fmfcfg and fmfcsr.
10829 (bdesc_1arg): Add fmtcsr.
10830 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
10831 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
10832 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
10833 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
10834 unspec_fmfcfg): New patterns.
10835 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
10836 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
10837 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
10838 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
10839 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
10840 __nds32__fmfcfg): Define.
10842 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10844 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
10845 intrinsic register names.
10846 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
10847 intrinsic register enum values and macros.
10849 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10851 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
10852 for load/store addressing form.
10853 (nds32_print_operand_address): Likewise.
10855 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
10858 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
10859 based on LABEL_REF. Remove useless assertion.
10860 (pic_address_needs_scratch): Fix formatting.
10861 (sparc_legitimize_pic_address): Minor tweaks.
10862 (sparc_delegitimize_address): Adjust assertion accordingly.
10863 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
10864 into symbolic_operand.
10865 (movsi_high_pic_label_ref): Likewise.
10866 (movsi_lo_sum_pic_label_ref): Likewise.
10867 (movdi_pic_label_ref): Likewise.
10868 (movdi_high_pic_label_ref): Likewise.
10869 (movdi_lo_sum_pic_label_ref): Likewise.
10871 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
10873 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
10874 custom LIB_SPEC setup.
10876 2018-04-06 Ruslan Bukin <br@bsdpad.com>
10877 Kito Cheng <kito.cheng@gmail.com>
10879 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
10880 * config/riscv/freebsd.h: New.
10882 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10884 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
10885 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
10888 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10889 Kito Cheng <kito.cheng@gmail.com>
10891 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
10892 nds32_output_call, nds32_symbol_binds_local_p): New functions.
10893 * config/nds32/nds32-protos.h (nds32_output_call,
10894 nds32_output_return): Declare.
10895 * config/nds32/nds32.md: Refine all the call and return patterns.
10897 2018-04-06 Jakub Jelinek <jakub@redhat.com>
10900 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
10901 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
10903 PR rtl-optimization/84872
10904 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
10905 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
10906 EDGE_CROSSING edge.
10908 2018-04-06 Tamar Christina <tamar.christina@arm.com>
10910 * expr.c (copy_blkmode_to_reg): Revert 254862.
10911 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
10913 2018-04-06 Richard Biener <rguenther@suse.de>
10915 PR middle-end/85244
10916 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
10917 after seeing a component reference with an adjacent field. Treat
10918 refs to arrays at struct end of external decls similar to
10919 refs to unconstrained commons.
10921 2018-04-06 Jakub Jelinek <jakub@redhat.com>
10924 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
10925 look through SAVE_EXPRs with non-side-effects argument. Adjust
10927 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
10930 2018-04-06 Richard Biener <rguenther@suse.de>
10932 PR middle-end/85180
10933 * alias.c (find_base_term): New wrapper around find_base_term
10934 unwinding CSELIB_VAL_PTR changes.
10935 (find_base_term): Do not restore CSELIB_VAL_PTR during the
10938 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10940 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
10942 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
10943 constant definitions.
10944 ("nop"): lr 0,0 -> nopr r0
10945 ("nop_lr0", "nop_lr1"): New insn definitions.
10947 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10949 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
10950 NDS32_V3PUSH_AVAILABLE_P macro.
10952 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
10953 Chung-Ju Wu <jasonwucj@gmail.com>
10955 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
10956 (nds32*-*-*): Add float and fpu_config into supported_defaults.
10957 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
10958 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
10959 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
10960 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
10961 * config/nds32/constraints.md: New constraints and checking for hard
10962 float configuration.
10963 * config/nds32/iterators.md: New mode iterator and attribute for hard
10964 float configuration.
10965 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
10967 * config/nds32/nds32-fpu.md: New file.
10968 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
10969 deal with hard float code generation.
10970 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
10972 (abi_type, float_reg_number): New enum type.
10973 * config/nds32/nds32-predicates.c: New predicates for hard float.
10974 * config/nds32/nds32-protos.h: Declare functions for hard float.
10975 * config/nds32/nds32.c: Implementation for hard float configuration.
10976 * config/nds32/nds32.h: Definitions for hard float configuration.
10977 * config/nds32/nds32.md: Include hard float machine description and
10978 modify patterns for hard float configuration.
10979 * config/nds32/nds32.opt: New options for hard float configuration.
10980 * config/nds32/predicates.md: New predicates for hard float
10983 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10985 * common/config/nds32/nds32-common.c
10986 (nds32_option_optimization_table): Enable -mreleax-hint by default.
10988 2018-04-05 Jakub Jelinek <jakub@redhat.com>
10990 PR middle-end/85195
10991 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
10992 CONSTRUCTOR_ELT (ctor, ...)->value.
10994 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
10997 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
10999 2018-04-05 Tom de Vries <tom@codesourcery.com>
11002 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
11005 2018-04-05 Shiva Chen <shiva0217@gmail.com>
11006 Kito Cheng <kito.cheng@gmail.com>
11008 * config/nds32/constraints.md (U33): Fine-tune checking condition.
11009 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
11010 * config/nds32/nds32.h (nds32_16bit_address_type): Add
11011 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
11013 2018-04-05 Shiva Chen <shiva0217@gmail.com>
11014 Kito Cheng <kito.cheng@gmail.com>
11016 * config/nds32/constraints.md (Ufe): New memory constraint.
11017 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
11018 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
11019 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
11021 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
11022 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
11024 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11026 * config/nds32/nds32.md: Use optimize_size in the condition for
11027 alu-shift instructions.
11029 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11031 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
11033 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11035 * config/nds32/nds32.md (negsi2): Refine pattern.
11037 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
11038 Chung-Ju Wu <jasonwucj@gmail.com>
11040 * config/nds32/iterators.md (shift_rotate): New code iterator.
11041 (shift): New code attribute.
11042 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
11043 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
11044 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
11045 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
11046 bit-wise operations.
11047 (andsi3, *andsi3): Ditto.
11048 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
11049 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
11050 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
11051 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
11052 nds32_ior_operand, nds32_xor_operand): New predicates.
11054 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11056 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
11057 (addsi3, subsi3): ... this.
11059 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11061 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
11063 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11065 * config/nds32/nds32.md: Adjust indention.
11067 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
11069 * config/nds32/nds32.md (feature): New attribute.
11071 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11073 * config/nds32/nds32.md (subtype): New attribute.
11075 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
11078 * config/arm/arm-builtins.c (arm_expand_builtin): Change
11079 expansion to perform a bitwise AND of the argument followed by a
11080 boolean negation of the result.
11082 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
11084 PR rtl-optimization/84878
11085 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
11086 the basic block. Assert the use reference is not artificial and that
11087 it has an associated insn.
11089 2018-04-04 Michael Matz <matz@suse.de>
11091 * builtins.c (compute_objsize): Pass correct operand
11092 to array_at_struct_end_p.
11094 2018-04-04 Richard Biener <rguenther@suse.de>
11097 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
11098 from contexts for DINFO_LEVEL_TERSE and below.
11100 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11102 * config/nds32/nds32-doubleword.md (move_<mode>): Require
11103 resiter_operand condition.
11104 * config/nds32/nds32.md (*move<mode>): Ditto.
11106 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11107 Monk Chiang <sh.chiang04@gmail.com>
11109 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
11111 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11113 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
11115 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11116 Kito Cheng <kito.cheng@gmail.com>
11118 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
11119 nds32_cond_code_str, output_cond_branch,
11120 output_cond_branch_compare_zero, nds32_expand_cbranch,
11121 nds32_expand_cstore, nds32_expand_movcc,
11122 nds32_output_cbranchsi4_equality_zero,
11123 nds32_output_cbranchsi4_equality_reg,
11124 nds32_output_cbranchsi4_equality_reg_or_const_int,
11125 nds32_output_cbranchsi4_greater_less_zero: New functions.
11126 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
11127 nds32_expand_cstore, nds32_expand_movcc,
11128 nds32_output_cbranchsi4_equality_zero,
11129 nds32_output_cbranchsi4_equality_reg,
11130 nds32_output_cbranchsi4_equality_reg_or_const_int,
11131 nds32_output_cbranchsi4_greater_less_zero): Declare.
11132 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
11133 nds32_rimm11s_operand): New predicates.
11134 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
11135 * config/nds32/nds32.md: Rewrite all the branch and conditional move
11138 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11140 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
11141 * config/nds32/nds32.md: Ditto.
11142 * config/nds32/pipelines.md: Ditto.
11144 2018-04-04 Richard Biener <rguenther@suse.de>
11146 PR tree-optimization/85168
11147 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
11148 propagating abnormals.
11150 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11152 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
11154 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11155 Kito Cheng <kito.cheng@gmail.com>
11157 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
11158 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
11159 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
11160 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
11161 * config/nds32/nds32.md (sibcall_internal): New.
11162 (sibcall_register): Remove.
11163 (sibcall_immediate): Remove.
11164 (sibcall_value_internal): New.
11165 (sibcall_value_register): Remove.
11166 (sibcall_value_immediate): Remove.
11167 * config/nds32/predicates.md (nds32_general_register_operand): New.
11168 (nds32_call_address_operand): New.
11170 2018-04-03 Jakub Jelinek <jakub@redhat.com>
11172 PR rtl-optimization/85167
11173 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
11174 bb_defs if *split_p, instead preinitialize it to NULL.
11176 PR tree-optimization/85156
11177 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
11178 evaluating the argument multiple times.
11180 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
11182 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
11184 (_mm_cvtpd_ps): Likewise.
11185 (_mm_cvttpd_epi32): Likewise.
11186 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
11187 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
11188 vector, pixel, and bool following altivec.h include.
11190 2018-04-03 Martin Sebor <msebor@redhat.com>
11192 * doc/extend.texi (Common Function Attributes): Clarify.
11193 (const attribute): Likewise.
11194 (pure attribute): Likewise.
11196 2018-04-03 Jakub Jelinek <jakub@redhat.com>
11199 * config/i386/i386.c (ix86_expand_vector_set): Use
11200 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
11202 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
11204 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
11205 instructions when changing rounding bits to preserve precision bits
11206 in the x87 control word.
11208 2018-04-03 Martin Liska <mliska@suse.cz>
11210 PR tree-optimization/82491
11211 * rtl.h (strip_offset_and_add): Replace += suboffset with
11212 poly_uint64 () + suboffset.
11214 2018-03-29 Martin Liska <mliska@suse.cz>
11215 Martin Jambor <mjambor@suse.cz>
11218 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
11219 param_type is not an integral or pointer type.
11221 2018-04-03 Richard Biener <rguenther@suse.de>
11223 * sese.h (recompute_all_dominators): Remove.
11225 2018-04-02 Martin Sebor <msebor@redhat.com>
11227 * doc/invoke.texi (-Wrestrict): Fix typos.
11229 2018-04-02 Jim Wilson <jimw@sifive.com>
11231 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
11232 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
11233 (<optab>di3, <optab>si3_extend): Likewise.
11234 (<optab>si3_mask, <optab>si3_mask_1): New.
11235 (<optab>di3_mask, <optab>di3_mask_1): New.
11236 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
11237 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
11238 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
11240 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
11242 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
11245 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
11247 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
11248 (nds32_canonicalize_comparison): New function.
11250 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
11251 Kito Cheng <kito.cheng@gmail.com>
11252 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11254 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
11255 * config/nds32/constants.md (unspec_volatile_element): Add
11256 UNSPEC_VOLATILE_RELAX_GROUP.
11257 * config/nds32/nds32-relax-opt.c: New file.
11258 * config/nds32/nds32-predicates.c
11259 (nds32_symbol_load_store_p): New function.
11260 * config/nds32/nds32-protos.h
11261 (nds32_symbol_load_store_p): Declare function.
11262 (make_pass_nds32_relax_opt): Declare new rtl pass function.
11263 * config/nds32/nds32.c
11264 (nds32_register_pass): New function to register pass.
11265 (nds32_register_passes): New function to register passes.
11266 * config/nds32/nds32.md (relax_group): New pattern.
11267 * config/nds32/nds32.opt (mrelax-hint): New option.
11268 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
11270 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
11272 * config/nds32/t-nds32: Modify files dependency.
11274 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
11276 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
11277 (PROFILE_HOOK): Define its implementation.
11279 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
11281 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
11282 type and 32-bit size.
11284 2018-04-01 Jakub Jelinek <jakub@redhat.com>
11286 PR middle-end/85090
11287 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
11288 (V_128_256): New mode iterator.
11289 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
11290 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
11291 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
11293 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
11294 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
11296 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
11299 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
11300 NaN inputs correctly.
11302 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
11305 * config/rs6000/vsx.md (??r): New mode attribute.
11306 (*vsx_mov<mode>_64bit): Use it.
11307 (*vsx_mov<mode>_32bit): Likewise.
11309 2018-03-30 Martin Sebor <msebor@redhat.com>
11311 PR tree-optimization/84818
11312 * builtins.c (check_access): Use warning_n.
11314 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11317 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
11319 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
11322 2018-03-30 Julia Koval <julia.koval@intel.com>
11325 * x86-tune.def (movx, partial_reg_dependency): Enable for
11328 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
11330 PR inline-asm/84985
11331 * lra-constraints.c (process_alt_operands): Move setting
11332 this_alternative_matches below.
11334 2018-03-29 Martin Liska <mliska@suse.cz>
11337 * doc/invoke.texi: Document how LTO works with debug info.
11338 Describe auto-load support of binutils. Mention 'x86-64'
11339 as valid option value of -march option.
11341 2018-03-29 Jakub Jelinek <jakub@redhat.com>
11343 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
11346 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
11347 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
11348 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
11351 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
11354 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
11355 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
11356 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
11357 for RS6000_BTM_POWERPC64.
11358 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
11359 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
11360 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
11363 (DIVDEU): Likewise.
11365 2018-03-28 Carl Love <cel@us.ibm.com>
11368 2017-09-27 Carl Love <cel@us.ibm.com>
11370 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
11371 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
11372 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
11375 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11377 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
11378 instead of __vector bool.
11379 (_mm_max_pu8): Likewise.
11380 (_mm_min_pi16): Likewise.
11382 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
11385 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
11386 (DIVWEUO): Likewise.
11387 (DIVDEO): Likewise.
11388 (DIVDEUO): Likewise.
11389 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
11390 DIVWEUO and DIVDEUO.
11391 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
11392 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
11393 (div_extend): Likewise.
11394 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
11396 (__builtin_divweuo): Likewise.
11397 (__builtin_divdeo): Likewise.
11398 (__builtin_divdeuo): Likewise.
11400 2018-03-28 Jakub Jelinek <jakub@redhat.com>
11403 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
11404 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
11406 PR tree-optimization/82004
11407 * gimple-match-head.c (optimize_pow_to_exp): New function.
11408 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
11409 Don't fold to exp if optimize_pow_to_exp is false.
11411 2018-03-28 Martin Liska <mliska@suse.cz>
11414 * calls.c (initialize_argument_information): Fix trailing space.
11415 * common.opt: Fix typo and provide better explanation for
11416 -fsanitize-coverage option.
11417 * config/i386/i386.opt: Fix typo.
11419 2018-03-28 Jakub Jelinek <jakub@redhat.com>
11420 Martin Liska <mliska@suse.cz>
11423 * gimplify.c (asan_poison_variable): Don't do the check for
11424 gimplify_omp_ctxp here.
11425 (gimplify_decl_expr): Do it here.
11426 (gimplify_target_expr): Likewise.
11428 2018-03-28 Martin Liska <mliska@suse.cz>
11431 * config/i386/i386.c (ix86_function_arg_advance): Do not call
11432 chkp_type_bounds_count if MPX is not enabled.
11434 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
11436 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
11438 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
11441 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
11442 function to create the function decl for complex long double
11443 multiply and divide for -mabi=ieeelongdouble.
11444 (init_float128_ieee): Call it.
11446 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
11449 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
11450 -fcf-protection=branch -mibt.
11451 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
11453 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11456 * config/arm/arm.c (arm_valid_symbolic_address): Handle
11457 arm_word_relocations.
11459 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
11462 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
11463 extern array declarations.
11465 2018-03-27 Richard Biener <rguenther@suse.de>
11467 PR middle-end/84067
11468 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
11469 explicit single_use checks.
11471 2018-03-27 Richard Biener <rguenther@suse.de>
11473 PR tree-optimization/85082
11474 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
11477 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11479 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
11480 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
11481 Turn on fasynchronous-unwind-tables and funwind-tables.
11483 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
11486 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
11487 (*bmi_blsr_<mode>_ccz): Ditto.
11489 2018-03-26 Tom de Vries <tom@codesourcery.com>
11491 PR tree-optimization/85063
11492 * omp-general.c (offloading_function_p): New function. Factor out
11494 * omp-offload.c (pass_omp_target_link::gate): ... here.
11495 * omp-general.h (offloading_function_p): Declare.
11496 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
11497 with attribute omp declare target for offloading functions.
11499 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
11501 PR tree-optimization/84005
11502 * tree-data-ref.h (get_base_for_alignment): Declare.
11503 * tree-data-ref.c (get_base_for_alignment_1): New function.
11504 (get_base_for_alignment): Likewise.
11505 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
11506 get_base_for_alignment to find a suitable base object, instead
11507 of always using drb->base_address.
11509 2018-03-23 Jakub Jelinek <jakub@redhat.com>
11511 PR inline-asm/85022
11512 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
11513 known size by default.
11515 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
11517 PR inline-asm/85030
11518 * lra-constraints.c (process_alt_operands): Don't match BLKmode
11519 and non BLKmode operands.
11521 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11524 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
11525 Clean up attributes.
11527 2018-03-23 Richard Biener <rguenther@suse.de>
11530 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
11531 we are going to emit early debug for LTO.
11533 2018-03-23 Jakub Jelinek <jakub@redhat.com>
11535 PR inline-asm/85034
11536 * function.c (match_asm_constraints_1): Don't optimize if input
11537 doesn't satisfy general_operand predicate for output's mode.
11539 PR inline-asm/85022
11540 * alias.c (write_dependence_p): Don't require for x_canonicalized
11541 non-VOIDmode if x has VOIDmode.
11544 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
11545 just don't try to optimize it rather than assert it never happens.
11547 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
11549 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
11550 macro expansions for definition of ST_INTERNAL_<mode> and
11551 LD_INTERNAL_<mode> builtins.
11552 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
11554 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
11556 (altivec_expand_st_builtin): Likewise.
11557 (altivec_expand_builtin): Remove calls to deleted functions.
11558 (rs6000_address_for_altivec): Delete this function.
11559 * config/rs6000/vector.md: Remove expands for
11560 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
11562 2018-03-22 Sudakshina Das <sudi.das@arm.com>
11565 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
11566 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
11567 re-computing once computed.
11568 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
11569 (arm_init_machine_status): Initialize
11570 machine->static_chain_stack_bytes.
11572 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
11575 * doc/extend.texi: Add four new prototypes for vec_ld.
11576 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
11577 definitions for more logical presentation.
11578 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
11579 entries for V1TI variants of __builtin_altivec_ld builtin.
11580 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
11581 handling of V1TI variant of LVX icode pattern.
11582 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
11583 (rs6000_gimple_fold_builtin): Likewise.
11584 (altivec_init_builtins): Add code to define
11585 __builtin_altivec_lvx_v1ti function.
11587 2018-03-22 Jakub Jelinek <jakub@redhat.com>
11589 PR inline-asm/84941
11590 * function.c (match_asm_constraints_1): Don't do the optimization
11591 if input isn't a REG, SUBREG, MEM or constant.
11593 2018-03-22 Tom de Vries <tom@codesourcery.com>
11595 PR tree-optimization/84956
11596 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
11597 bb_has_abnormal_pred.
11599 2018-03-22 Jakub Jelinek <jakub@redhat.com>
11602 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
11603 DECL_INITIAL (decl) to decl at the end.
11604 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
11605 adjust the comment.
11607 2018-03-21 Joseph Myers <joseph@codesourcery.com>
11609 * doc/extend.texi (__builtin_tgmath): Document when complex
11610 integer types are treated as _Complex _Float64.
11612 2018-03-21 Tom de Vries <tom@codesourcery.com>
11614 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
11616 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11618 PR tree-optimization/84960
11619 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
11620 if it is ENTRY block, move them into single succ of ENTRY in that case.
11622 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
11624 PR tree-optimization/84811
11625 * poly-int.h (poly_span_traits): Remove the T3 parameter and
11626 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
11627 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
11628 (known_subrange_p): Update accordingly. Cast each value involved
11629 in the size comparison, rather than casting the result of the
11632 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11634 PR tree-optimization/84982
11635 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
11636 by flipping the least significant bit rather than all bits from
11637 bitpos to bitpos + bitsize - 1.
11639 2018-03-21 Nathan Sidwell <nathan@acm.org>
11641 * doc/extend.texi (Deprecated Features): Remove mention of
11642 long-deleted deprecations.
11644 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11647 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
11648 * configure: Regenerate.
11650 2018-03-21 Tom de Vries <tom@codesourcery.com>
11652 PR tree-optimization/83126
11653 * tree-parloops.c (num_phis): New function.
11654 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
11656 2018-03-21 Nathan Sidwell <nathan@acm.org>
11658 * doc/extend.texi (Deprecated Features): Update deprecated flags,
11659 mention anon-struct/union members and trailing attributes.
11661 2018-03-21 Bin Cheng <bin.cheng@arm.com>
11663 PR tree-optimization/84969
11664 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
11665 builtin memset partitions if they set different rhs values.
11667 2018-03-21 Jakub Jelinek <jakub@redhat.com>
11669 PR rtl-optimization/84989
11670 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
11671 VEC_DUPLICATE with scalar result mode.
11673 2018-03-21 Martin Liska <mliska@suse.cz>
11676 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
11677 not intended return statement.
11679 2018-03-21 Martin Liska <mliska@suse.cz>
11682 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
11683 (chkp_find_bound_slots_1): Limit number of iterations.
11685 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
11688 * Minor grammar fixes for x86 options.
11690 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11693 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
11694 holding REG_CFA_RESTORE notes, instead turn them into a USE.
11696 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
11699 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
11700 (altivec_lvx_<mode>_1op): Likewise.
11701 (altivec_stvx_<mode>_2op): Likewise.
11702 (altivec_stvx_<mode>_1op): Likewise.
11703 (altivec_lvx_<VM2:mode>): New define_expand.
11704 (altivec_stvx_<VM2:mode>): Likewise.
11705 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
11706 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
11707 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
11708 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
11709 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
11710 (rs6000_gen_lvx): Likewise.
11711 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
11712 (altivec_expand_stv_builtin): Likewise.
11713 (altivec_expand_builtin): Likewise.
11714 * config/rs6000/vector.md: Likewise.
11716 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11719 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
11722 2018-03-20 Richard Biener <rguenther@suse.de>
11725 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
11726 sign-conversions as zero, fall back to standard scalar_stmt
11729 2018-03-20 Martin Liska <mliska@suse.cz>
11732 * predict.c (rebuild_frequencies): Handle case when we have
11733 PROFILE_ABSENT, but flag_guess_branch_prob is false.
11735 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11738 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
11739 flag_section_anchors.
11740 * varasm.c (use_blocks_for_decl_p): Remove hack for
11741 dw2_force_const_mem.
11744 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
11746 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
11747 be created, use lowpart_subreg of operands[0] rather than operands[0]
11749 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
11750 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
11751 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
11752 and n constraint instead of aarch64_shift_imm_di and Usd.
11753 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
11754 (*aarch64_<optab>_reg_minus<mode>3): ... this.
11756 2018-03-20 Sudakshina Das <sudi.das@arm.com>
11759 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
11760 to favor GPR over NEON registers.
11761 (<shift>di3_neon): Likewise.
11763 2018-03-20 Tom de Vries <tom@codesourcery.com>
11766 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
11767 (nvptx_process_pars): Emit bar.sync asap and alap.
11769 2018-03-20 Tom de Vries <tom@codesourcery.com>
11772 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
11773 seen_label if seen_label is already set.
11775 2018-03-20 Jakub Jelinek <jakub@redhat.com>
11778 * config/i386/i386.c (fold_builtin_cpu): For features above 31
11779 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
11780 Use 1U instead of 1. Formatting fixes.
11783 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
11784 instead of TREE_TYPE (s1) for the return value.
11786 2018-03-19 Jakub Jelinek <jakub@redhat.com>
11788 PR tree-optimization/84946
11789 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
11790 bitsize + bitsize in poly_uint64 rather than poly_int64.
11793 * dwarf2asm.c: Include fold-const.c.
11794 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
11795 of decl rather than decl itself.
11797 PR rtl-optimization/84643
11798 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
11800 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
11803 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
11804 calling assemble_variable.
11806 2018-03-19 Sudakshina Das <sudi.das@arm.com>
11809 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
11810 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
11812 2018-03-19 Jim Wilson <jimw@sifive.com>
11815 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
11816 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
11817 (riscv_first_stack_step): Likewise.
11818 (riscv_option_override): Use STACK_BOUNDARY instead of
11819 MIN_STACK_BOUNDARY.
11820 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
11821 MIN_STACK_BOUNDARY.
11822 (BIGGEST_ALIGNMENT): Set to 128.
11823 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
11824 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
11827 2018-03-19 Richard Biener <rguenther@suse.de>
11829 PR tree-optimization/84933
11830 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
11831 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
11833 2018-03-19 Richard Biener <rguenther@suse.de>
11835 PR tree-optimization/84859
11836 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
11837 (cond_if_else_store_replacement): Perform sinking operation on
11838 single-store BBs regardless of MAX_STORES_TO_SINK setting.
11839 Generalize what a BB with a single eligible store is.
11841 2018-03-19 Richard Biener <rguenther@suse.de>
11843 PR tree-optimization/84929
11844 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
11845 chrec_is_positive against non-chrec arg.
11847 2018-03-19 Tamar Christina <tamar.christina@arm.com>
11850 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
11852 2018-03-18 Martin Liska <mliska@suse.cz>
11854 PR rtl-optimization/84635
11855 * regrename.c (build_def_use): Use matches_mode only when
11858 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
11860 PR tree-optimization/84913
11861 * tree-vect-loop.c (vectorizable_reduction): Don't try to
11862 vectorize chains of COND_EXPRs.
11864 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11866 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
11868 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11870 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
11872 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11874 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
11876 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
11877 Kito Cheng <kito.cheng@gmail.com>
11879 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
11880 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
11881 (nds32_adjust_reg_alloc_order): New function.
11882 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
11884 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
11886 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
11887 nds32_print_operand, nds32_print_operand_address): Use
11888 HOST_WIDE_INT_PRINT_DEC instead.
11890 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
11892 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
11894 2018-03-17 Jakub Jelinek <jakub@redhat.com>
11897 * config/i386/i386.c (initial_ix86_tune_features,
11898 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
11899 unsigned long long.
11900 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
11901 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
11902 rather than 1u << ix86_tune. Formatting fix.
11903 (ix86_option_override_internal): Change ix86_arch_mask from
11904 unsigned int to unsigned HOST_WIDE_INT, initialize to
11905 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
11906 (ix86_function_specific_restore): Likewise.
11908 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11911 * postreload.c (reload_combine_recognize_pattern): Perform
11912 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
11913 truncate_int_for_mode the result for the destination's mode.
11916 * hsa-gen.c (mem_type_for_type): Fix comment typo.
11917 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
11919 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11922 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
11925 * lra-assigns.c (lra_split_hard_reg_for): Don't use
11926 regno_allocno_class_array and sorted_pseudos.
11927 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
11928 insns where regno is used.
11930 2018-03-16 Martin Liska <mliska@suse.cz>
11933 * multiple_target.c (create_dispatcher_calls): Redirect
11934 reference in the symbol table.
11936 2018-03-16 Martin Liska <mliska@suse.cz>
11939 * multiple_target.c (create_dispatcher_calls): Redirect also
11942 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11946 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
11947 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
11948 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
11950 2018-03-16 Julia Koval <julia.koval@intel.com>
11952 * doc/invoke.texi (Skylake Server): Add CLWB.
11953 Cannonlake): Remove CLWB.
11955 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11957 PR tree-optimization/84841
11958 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
11960 (FLOAT_ONE_CONST_TYPE): Define.
11961 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
11962 (sort_by_operand_rank): Put entries with higher constant_type last
11963 rather than first to match comments.
11965 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
11967 * config/nios2/nios2.md (movsi_internal): Fix thinko in
11970 2018-03-15 Jakub Jelinek <jakub@redhat.com>
11973 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
11974 check and use address of target always.
11976 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
11979 * config/i386/i386.c (indirect_thunk_needed): Update comments.
11980 (indirect_thunk_bnd_needed): Likewise.
11981 (indirect_thunks_used): Likewise.
11982 (indirect_thunks_bnd_used): Likewise.
11983 (indirect_return_needed): New.
11984 (indirect_return_bnd_needed): Likewise.
11985 (output_indirect_thunk_function): Add a bool argument for
11987 (output_indirect_thunk_function): Don't generate alias for
11988 function return thunk.
11989 (ix86_code_end): Call output_indirect_thunk_function to generate
11990 function return thunks.
11991 (ix86_output_function_return): Set indirect_return_bnd_needed
11992 and indirect_return_needed instead of indirect_thunk_bnd_needed
11993 and indirect_thunk_needed.
11995 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
11997 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
11998 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
11999 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
12001 2018-03-15 David Malcolm <dmalcolm@redhat.com>
12002 Paul Hua <paul.hua.gm@gmail.com>
12005 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
12007 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
12009 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
12010 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
12011 resp. SFmode cases.
12013 2018-03-15 Tamar Christina <tamar.christina@arm.com>
12016 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
12017 instead of GET_MODE_SIZE when comparing Units.
12019 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
12022 * varasm.c (hash_section): Return an unchangeble hash value
12023 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
12024 Return !aarch64_can_use_per_function_literal_pools_p ().
12026 2018-03-15 Jakub Jelinek <jakub@redhat.com>
12029 * optabs.c (emit_conditional_move): Pass address of cmode's copy
12030 rather than address of cmode as last argument to prepare_cmp_insn.
12032 2018-03-15 Julia Koval <julia.koval@intel.com>
12034 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
12035 F_AVX512VNNI, F_AVX512BITALG): New.
12037 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
12040 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
12041 insn for floating-point loads and stores.
12043 2018-03-14 Carl Love <cel@us.ibm.com>
12045 * config/rs6000/rs6000-c.c: Add macro definitions for
12046 ALTIVEC_BUILTIN_VEC_PERMXOR.
12047 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
12048 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
12049 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
12050 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
12052 * config/doc/extend.texi: Add prototypes for vec_permxor.
12054 2018-03-14 David Malcolm <dmalcolm@redhat.com>
12057 * diagnostic-show-locus.c (class layout_point): Convert m_line
12058 from int to linenum_type.
12059 (line_span::comparator): Use linenum "compare" function when
12060 comparing line numbers.
12061 (test_line_span): New function.
12062 (layout_range::contains_point): Convert param "row" from int to
12064 (layout_range::intersects_line_p): Likewise.
12065 (layout::will_show_line_p): Likewise.
12066 (layout::print_source_line): Likewise.
12067 (layout::should_print_annotation_line_p): Likewise.
12068 (layout::print_annotation_line): Likewise.
12069 (layout::print_leading_fixits): Likewise.
12070 (layout::annotation_line_showed_range_p): Likewise.
12071 (struct line_corrections): Likewise for field m_row.
12072 (line_corrections::line_corrections): Likewise for param "row".
12073 (layout::print_trailing_fixits): Likewise.
12074 (layout::get_state_at_point): Likewise.
12075 (layout::get_x_bound_for_row): Likewise.
12076 (layout::print_line): Likewise.
12077 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
12078 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
12079 * input.c (selftest::test_linenum_comparisons): New function.
12080 (selftest::input_c_tests): Call it.
12081 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
12082 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
12083 * selftest.h (ASSERT_GT): New macro.
12084 (ASSERT_GT_AT): New macro.
12085 (ASSERT_LT): New macro.
12086 (ASSERT_LT_AT): New macro.
12088 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
12090 PR rtl-optimization/84780
12091 * combine.c (distribute_links): Don't make a link based on pc_rtx.
12093 2018-03-14 Martin Liska <mliska@suse.cz>
12095 * tree.c (record_node_allocation_statistics): Use
12096 get_stats_node_kind.
12097 (get_stats_node_kind): New function extracted from
12098 record_node_allocation_statistics.
12099 (free_node): Use get_stats_node_kind.
12101 2018-03-14 Richard Biener <rguenther@suse.de>
12103 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
12104 that the value-set of ANTIC_IN doesn't grow.
12107 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
12109 (BB_VISITED_WITH_VISITED_SUCCS): New define.
12110 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
12112 2018-03-14 Julia Koval <julia.koval@intel.com>
12114 * config.gcc (icelake-client, icelake-server): New.
12116 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
12117 (initial_ix86_arch_features): Ditto.
12118 (PTA_SKYLAKE): Add SGX.
12119 (PTA_ICELAKE): Remove.
12120 (PTA_ICELAKE_CLIENT): New.
12121 (PTA_ICELAKE_SERVER): New.
12122 (ix86_option_override_internal): Split up icelake on icelake client and
12124 (get_builtin_code_for_version): Ditto.
12125 (fold_builtin_cpu): Ditto.
12126 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
12127 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
12128 * config/i386/i386.h (processor_type): Ditto.
12129 * doc/invoke.texi: Ditto.
12131 2018-03-14 Jakub Jelinek <jakub@redhat.com>
12134 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
12135 INTEGER_CST offset, add it together with bitpos / 8 and
12136 sign extend based on POINTER_SIZE.
12140 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
12143 * config/i386/constraints.md (Yc): New register constraint.
12144 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
12145 Use Yc constraint for alternative 2 of operand 0. Remove
12146 preferred_for_speed attribute.
12148 2018-03-14 Richard Biener <rguenther@suse.de>
12150 PR tree-optimization/84830
12151 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
12152 with the old one to avoid oscillations.
12154 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
12157 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
12159 (assign_by_spills): Return a flag of reload assignment failure.
12160 Do not process the reload assignment failures. Do not spill other
12161 reload pseudos if they has the same reg class. Update n if
12163 (lra_assign): Add a return arg. Set up from the result of
12164 assign_by_spills call.
12165 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12166 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12167 usage_insns if it is not NULL.
12168 (spill_hard_reg_in_range): New function.
12169 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12170 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12171 function prototypes.
12172 (lra_assign): Change prototype.
12173 * lra.c (lra): Add code to deal with fails by splitting hard reg
12176 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
12178 * config/riscv/riscv.opt (mrelax): New option.
12179 * config/riscv/riscv.c (riscv_file_start): Emit ".option
12180 "norelax" when riscv_mrelax is disabled.
12181 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
12183 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12186 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
12187 reassociation for int modes.
12189 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12191 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
12192 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
12194 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
12195 * config/aarch64/aarch64-sve.md
12196 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
12197 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
12198 (*extend<mode><Vwide>2): Rename to...
12199 (aarch64_sve_extend<mode><Vwide>2): ...this.
12200 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
12201 renaming the old pattern to...
12202 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
12204 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
12205 define_expand, renaming the old pattern to...
12206 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
12207 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
12208 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
12209 account when deciding which SVE instruction the optab should use.
12210 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
12212 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12214 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
12215 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
12216 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
12217 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
12218 (tlsdesc_small_<mode>): Turn a define_expand and use
12219 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
12220 (tlsdesc_small_advsimd_<mode>): ...this.
12221 (tlsdesc_small_sve_<mode>): New pattern.
12223 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12225 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
12226 (UNSPEC_UMUL_HIGHPART): New constants.
12227 (MUL_HIGHPART): New int iteraor.
12228 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
12229 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
12231 (*<su>mul<mode>3_highpart): New define_insn.
12233 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
12236 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
12239 2018-03-13 Martin Liska <mliska@suse.cz>
12242 * (sem_item_optimizer::sem_item_optimizer): Initialize new
12244 (sem_item_optimizer::~sem_item_optimizer): Release it.
12245 (sem_item_optimizer::merge_classes): Register variable aliases.
12246 (sem_item_optimizer::fixup_pt_set): New function.
12247 (sem_item_optimizer::fixup_points_to_sets): Likewise.
12248 * ipa-icf.h: Declare new variables and functions.
12250 2018-03-13 Jakub Jelinek <jakub@redhat.com>
12252 PR middle-end/84834
12253 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
12254 integer_pow2p@2 and test integer_pow2p in condition.
12255 (A < 0 ? C : 0): Similarly for @1.
12257 PR middle-end/84831
12258 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
12259 characters starting at p contain '\0' character, don't look beyond
12263 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
12264 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
12267 * reg-stack.c (change_stack): Change update_end var from int to
12268 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
12269 also call set_block_for_insn on the newly added insns and rescan.
12272 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
12273 on the last operand.
12276 * tree.c (stabilize_reference_1): Return save_expr (e) for
12277 STATEMENT_LIST even if it doesn't have side-effects.
12279 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
12281 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
12283 2018-03-12 Renlin Li <renlin.li@arm.com>
12285 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
12286 aarch64_output_scalar_simd_mov_immediate.
12288 2018-03-12 Martin Sebor <msebor@redhat.com>
12290 PR tree-optimization/83456
12291 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
12292 for perfectly overlapping calls to memcpy.
12293 (gimple_fold_builtin_memory_chk): Same.
12294 (gimple_fold_builtin_strcpy): Handle no-warning.
12295 (gimple_fold_builtin_stxcpy_chk): Same.
12296 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
12298 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
12300 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
12301 parameter. Use it for SFmode.
12302 (rs6000_function_arg_advance_1): Adjust.
12303 (rs6000_function_arg): Adjust.
12304 (rs6000_gimplify_va_arg): Pass false for that new parameter.
12306 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
12308 PR rtl-optimization/84169
12309 PR rtl-optimization/84780
12310 * combine.c (can_combine_p): Check for a 2-insn combination whether
12311 the destination register is used between the two insns, too.
12313 2018-03-12 Richard Biener <rguenther@suse.de>
12315 PR tree-optimization/84803
12316 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
12317 for refs DR analysis didn't process.
12319 2018-03-12 Richard Biener <rguenther@suse.de>
12321 PR tree-optimization/84777
12322 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
12323 force-vectorize loops ignore whether we are optimizing for size.
12325 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
12327 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
12328 (TARGET_MD_ASM_ADJUST): Define.
12330 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
12331 Kito Cheng <kito.cheng@gmail.com>
12332 Chung-Ju Wu <jasonwucj@gmail.com>
12334 * config/nds32/nds32.c (nds32_compute_stack_frame,
12335 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
12336 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
12337 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
12338 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
12339 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
12340 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
12341 * config/nds32/nds32.md (prologue, epilogue): Use macro
12342 NDS32_V3PUSH_AVAILABLE_P to do checking.
12344 2018-03-11 Jakub Jelinek <jakub@redhat.com>
12347 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
12348 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
12349 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
12350 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
12351 addition of most attributes on !orig_type_die or the attribute not
12352 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
12354 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12355 Chung-Ju Wu <jasonwucj@gmail.com>
12357 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
12358 __NDS32_VH__ macro.
12359 * config/nds32/nds32.opt (mvh): New option.
12361 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12362 Chung-Ju Wu <jasonwucj@gmail.com>
12364 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
12366 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
12367 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
12370 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12371 Chung-Ju Wu <jasonwucj@gmail.com>
12373 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
12375 * config/nds32/nds32-multiple.md (strlensi): New pattern.
12376 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
12378 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
12379 Kito Cheng <kito.cheng@gmail.com>
12380 Chung-Ju Wu <jasonwucj@gmail.com>
12382 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
12383 UNSPEC_FFMISM and UNSPEC_FLMISM.
12384 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
12385 for ffb, ffmism and flmism.
12386 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
12387 (unspec_ffmism): Ditto.
12388 (unspec_flmism): Ditto.
12389 (nds32_expand_builtin_impl): Check if string extension is available.
12390 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
12391 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
12393 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
12396 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
12399 * lra-assigns.c (assign_by_spills): Return a flag of reload
12400 assignment failure. Do not process the reload assignment
12401 failures. Do not spill other reload pseudos if they has the same
12403 (lra_assign): Add a return arg. Set up from the result of
12404 assign_by_spills call.
12405 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12406 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12407 usage_insns if it is not NULL.
12408 (spill_hard_reg_in_range): New function.
12409 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12410 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12411 function prototypes.
12412 (lra_assign): Change prototype.
12413 * lra.c (lra): Add code to deal with fails by splitting hard reg
12416 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
12419 * config/i386/i386.opt: Replace Enforcment with Enforcement.
12421 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
12424 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
12425 (dw_val_node): Add val_symbolic_view.
12426 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
12427 (symview_upper_bound): New.
12428 (new_line_info_table): Initialize symviews_since_reset.
12429 (dwarf2out_source_line): Count symviews_since_reset and set
12430 symview_upper_bound.
12431 (dw_val_equal_p): Handle symview.
12432 (add_AT_symview): New.
12433 (print_dw_val): Handle symview.
12434 (attr_checksum, attr_checksum_ordered): Likewise.
12435 (same_dw_val_p, size_of_die): Likewise.
12436 (value_format, output_die): Likewise.
12437 (add_high_low_attributes): Use add_AT_symview for entry_view.
12438 (dwarf2out_finish): Reset symview_upper_bound, clear
12441 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
12444 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
12445 Add strict argument and use it.
12446 (rs6000_split_multireg_move): Update for new strict argument.
12447 (mem_operand_gpr): Disallow all non-offsettable addresses.
12448 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
12450 2018-03-09 Jakub Jelinek <jakub@redhat.com>
12453 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
12454 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
12455 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
12458 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
12459 decl, use remap_type if we want to use the type.
12461 2018-03-09 Martin Sebor <msebor@redhat.com>
12463 PR tree-optimization/84526
12464 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
12466 (builtin_access::generic_overlap): Be prepared to handle non-array
12469 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
12471 PR rtl-optimization/84682
12472 * lra-constraints.c (process_address_1): Check is_address flag
12473 for address constraints.
12474 (process_alt_operands): Likewise.
12475 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
12476 preprocess_constraints.
12477 * recog.h (preprocess_constraints): Add oploc parameter.
12479 * recog.c (preprocess_constraints): Test address_operand for
12480 CT_ADDRESS constraints.
12482 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
12485 * lra-assigns.c (assign_by_spills): Return a flag of reload
12486 assignment failure. Do not process the reload assignment
12487 failures. Do not spill other reload pseudos if they has the same
12489 (lra_assign): Add a return arg. Set up from the result of
12490 assign_by_spills call.
12491 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12492 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12493 usage_insns if it is not NULL.
12494 (spill_hard_reg_in_range): New function.
12495 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12496 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12497 function prototypes.
12498 (lra_assign): Change prototype.
12499 * lra.c (lra): Add code to deal with fails by splitting hard reg
12502 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12505 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
12506 Accept complain bool parameter. Only emit errors if it is true.
12507 (arm_parse_cpu_option_name): Likewise.
12508 (arm_target_thumb_only): Adjust callers of the above.
12509 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
12510 prototype to take a default true bool parameter.
12511 (arm_parse_arch_option_name): Likewise.
12513 2018-03-09 David Malcolm <dmalcolm@redhat.com>
12514 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
12518 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
12519 * configure: Regenerate.
12520 * configure.ac ("linker --version-script option"): New.
12521 ("linker soname option"): New.
12523 2018-03-09 Richard Biener <rguenther@suse.de>
12525 PR tree-optimization/84775
12526 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
12527 immediate uses of predicate stmts and mark them modified.
12530 PR tree-optimization/84178
12531 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
12533 (version_loop_for_if_conversion): Delay update_ssa call.
12534 (tree_if_conversion): Delay update_ssa until after predicate
12537 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
12540 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
12541 when the function accesses prior frames.
12543 2018-03-08 Jakub Jelinek <jakub@redhat.com>
12546 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
12547 gen_llsym, otherwise call maybe_gen_llsym.
12549 PR inline-asm/84742
12550 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
12551 has ',' character inside of it.
12553 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12556 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
12557 as clobbering CC_REGNUM.
12559 2018-03-08 Richard Biener <rguenther@suse.de>
12561 PR middle-end/84552
12562 * tree-scalar-evolution.c: Include tree-into-ssa.h.
12563 (follow_copies_to_constant): Do not follow SSA names registered
12566 2018-03-08 Richard Biener <rguenther@suse.de>
12568 PR tree-optimization/84178
12569 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
12571 (version_loop_for_if_conversion): Delay update_ssa call.
12572 (tree_if_conversion): Delay update_ssa until after predicate
12575 2018-03-08 David Malcolm <dmalcolm@redhat.com>
12577 PR tree-optimization/84178
12578 * tree-if-conv.c (release_bb_predicate): Remove the
12579 the assertion that the stmts have NULL use_ops.
12580 Discard the statements, asserting that they haven't
12581 yet been added to a BB.
12583 2018-03-08 Richard Biener <rguenther@suse.de>
12585 PR tree-optimization/84746
12586 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
12587 (phi_translate): Pass in destination ANTIC_OUT set.
12588 (phi_translate_1): Likewise. For a simplified result lookup
12589 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
12590 (phi_translate_set): Adjust.
12591 (do_pre_regular_insertion): Likewise.
12592 (do_pre_partial_partial_insertion): Likewise.
12594 2018-03-08 Martin Liska <mliska@suse.cz>
12596 PR gcov-profile/84735
12597 * doc/gcov.texi: Document usage of profile files.
12598 * gcov-io.h: Document changes in the format.
12600 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
12604 * dwarf2out.c (struct dw_line_info_table): Update comments for
12606 (FORCE_RESET_NEXT_VIEW): New.
12607 (FORCE_RESETTING_VIEW_P): New.
12608 (RESETTING_VIEW_P): Check for -1 too.
12609 (ZERO_VIEW_P): Likewise.
12610 (new_line_info_table): Force-reset next view.
12611 (dwarf2out_begin_function): Likewise.
12612 (dwarf2out_source_line): Simplify zero_view_p initialization.
12613 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
12614 view directly. Omit view when omitting .loc at line 0.
12616 2018-03-08 Jakub Jelinek <jakub@redhat.com>
12618 PR tree-optimization/84740
12619 * tree-switch-conversion.c (process_switch): Call build_constructors
12620 only if info.phi_count is non-zero.
12622 PR tree-optimization/84739
12623 * tree-tailcall.c (find_tail_calls): Check call arguments against
12624 DECL_ARGUMENTS (current_function_decl) rather than
12625 DECL_ARGUMENTS (func) when checking for tail recursion.
12627 2018-03-07 Jakub Jelinek <jakub@redhat.com>
12629 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
12630 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
12631 Volker Reichelt's entry and add entries for people that perform
12632 GCC fuzzy testing and report numerous bugs.
12634 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
12637 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
12638 readonly data in sdata, if that is disabled.
12639 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
12640 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
12641 -mreadonly-in-sdata option.
12643 2018-03-07 Martin Sebor <msebor@redhat.com>
12645 PR tree-optimization/84468
12646 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
12647 basic block when looking for nul assignment.
12649 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
12652 * except.h (output_function_exception_table): Adjust prototype.
12653 * except.c (output_function_exception_table): Remove FNNAME parameter
12654 and add SECTION parameter. Ouput one part of the table at a time.
12655 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
12656 the first part of the exception table and emit unwind directives.
12657 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
12658 (i386_pe_seh_cold_init): Likewise.
12659 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
12660 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
12661 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
12662 (ix86_output_call_insn): Emit a nop in one more case for SEH.
12663 * config/i386/winnt.c: Include except.h.
12664 (struct seh_frame_state): Add reg_offset, after_prologue and
12665 in_cold_section fields.
12666 (i386_pe_seh_end_prologue): Set seh->after_prologue.
12667 (i386_pe_seh_cold_init): New function.
12668 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
12669 to seh->in_cold_section.
12670 (seh_emit_push): Record the offset of the push.
12671 (seh_emit_save): Record the offet of the save.
12672 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
12673 Test seh->after_prologue to disregard the epilogue.
12674 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
12675 (i386_pe_end_cold_function): New function.
12677 2018-03-07 Jakub Jelinek <jakub@redhat.com>
12680 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
12681 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
12684 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
12686 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
12687 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
12689 PR middle-end/84723
12690 * multiple_target.c: Include tree-inline.h and intl.h.
12691 (expand_target_clones): Diagnose and fail if node->definition and
12692 !tree_versionable_function_p (node->decl).
12694 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
12696 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
12698 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
12699 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
12700 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
12702 2018-03-06 Jakub Jelinek <jakub@redhat.com>
12705 * combine.c (try_combine): Use reg_or_subregno instead of handling
12706 just paradoxical SUBREGs and REGs.
12708 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
12710 * config/arc/arc.c (arc_finalize_pic): Remove function.
12711 (arc_must_save_register): We use single base PIC register, remove
12712 checks to save/restore the PIC register.
12713 (arc_expand_prologue): Likewise.
12714 * config/arc/arc-protos.h (arc_set_default_type_attributes):
12716 (arc_verify_short): Likewise.
12717 (arc_attr_type): Likewise.
12718 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
12719 (walk_stores): Likewise.
12720 (arc_address_cost): Make it static.
12721 (arc_verify_short): Likewise.
12722 (branch_dest): Likewise.
12723 (arc_attr_type): Likewise.
12724 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
12725 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
12726 (arc_final_prescan_insn): Remove inserting the nops due to
12727 hardware hazards. It is done in reorg step.
12728 (insn_length_variant_t): Remove.
12729 (insn_length_parameters_t): Likewise.
12730 (arc_insn_length_parameters): Likewise.
12731 (arc_get_insn_variants): Likewise.
12732 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
12734 2018-03-06 Jakub Jelinek <jakub@redhat.com>
12736 PR inline-asm/84683
12737 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
12740 PR tree-optimization/84687
12741 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
12743 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
12745 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12747 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
12748 Rename to ppc_speculation_barrier.
12749 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
12750 __builtin_ppc_speculation_barrier.
12752 2018-03-05 Jakub Jelinek <jakub@redhat.com>
12755 * combine.c (combine_simplify_rtx): Don't try to simplify if
12756 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
12759 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
12761 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
12762 to 32 bytes when compiling for POWER9.
12764 2018-03-05 Jakub Jelinek <jakub@redhat.com>
12767 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
12768 regparm >= 3 with no arg reg available also for calls with
12769 flag_force_indirect_call. Pass decl to ix86_function_regparm.
12772 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
12774 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
12776 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
12779 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
12781 2018-03-05 Richard Biener <rguenther@suse.de>
12783 PR tree-optimization/84486
12784 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
12785 When inserting a __builtin_assume_aligned call set the LHS
12786 SSA name alignment info accordingly.
12788 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
12790 PR tree-optimization/84114
12791 * config/aarch64/aarch64.c (aarch64_reassociation_width)
12792 Avoid reassociation of FLOAT_MODE addition.
12794 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
12796 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
12797 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
12798 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
12799 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
12800 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
12801 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
12802 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
12804 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
12805 __builtin_ia32_wbinvd): New builtins.
12806 (SPECIAL_ARGS2): New.
12807 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
12808 (SPECIAL_ARGS2): New.
12809 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
12810 (ix86_valid_target_attribute_inner_p): Ditto.
12811 (ix86_init_mmx_sse_builtins): Add special_args2.
12812 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
12813 TARGET_WBNOINVD_P): New.
12814 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
12815 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
12816 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
12817 * config/i386/immintrin.h (_wbinvd): New intrinsic.
12818 * config/i386/pconfigintrin.h: New file.
12819 * config/i386/wbnoinvdintrin.h: Ditto.
12820 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
12822 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
12824 2018-03-05 Richard Biener <rguenther@suse.de>
12826 PR tree-optimization/84670
12827 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
12829 (BB_VISITED_WITH_VISITED_SUCCS): New define.
12830 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
12831 (compute_antic_aux): Only assert the number of values in ANTIC_IN
12832 doesn't grow if all successors (recursively) were visited at least
12835 2018-03-05 Richard Biener <rguenther@suse.de>
12837 PR tree-optimization/84650
12838 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
12839 if executed in the loop pipeline.
12841 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
12843 * doc/configfiles.texi (Configuration Files): Move info about
12844 conditionalizing $target-protos.h to...
12845 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
12846 differs from $target-protos.h.
12848 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
12849 Chung-Ju Wu <jasonwucj@gmail.com>
12851 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
12852 * config/nds32/nds32-multiple.md (setmemsi): Define.
12853 * config/nds32/nds32-memory-manipulation.c
12854 (nds32_gen_dup_4_byte_to_word_value): New.
12855 (emit_setmem_word_loop): New.
12856 (emit_setmem_byte_loop): New.
12857 (nds32_expand_setmem_loop): New.
12858 (nds32_expand_setmem_loop_v3m): New.
12859 (nds32_expand_setmem_unroll): New.
12860 (nds32_expand_setmem): New.
12862 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12863 Chung-Ju Wu <jasonwucj@gmail.com>
12865 * config/nds32/nds32-memory-manipulation.c
12866 (nds32_emit_load_store): New.
12867 (nds32_emit_post_inc_load_store): New.
12868 (nds32_emit_mem_move): New.
12869 (nds32_emit_mem_move_block): New.
12870 (nds32_expand_movmemsi_loop_unknown_size): New.
12871 (nds32_expand_movmemsi_loop_known_size): New.
12872 (nds32_expand_movmemsi_loop): New.
12873 (nds32_expand_movmemsi_unroll): New.
12874 (nds32_expand_movmemqi): Rename ...
12875 (nds32_expand_movmemsi): ... to this.
12876 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
12877 (movmemsi): ... to this.
12878 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
12879 (nds32_expand_movmemsi): ... to this.
12881 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12882 Monk Chiang <sh.chiang04@gmail.com>
12883 Chung-Ju Wu <jasonwucj@gmail.com>
12885 * config/nds32/nds32-protos.h
12886 (nds32_expand_load_multiple): New arguments.
12887 (nds32_expand_store_multiple): Ditto.
12888 (nds32_valid_multiple_load_store): Rename ...
12889 (nds32_valid_multiple_load_store_p): ... to this.
12890 * config/nds32/nds32-memory-manipulation.c
12891 (nds32_expand_load_multiple): Refine implementation.
12892 (nds32_expand_store_multiple): Ditto.
12893 * config/nds32/nds32-multiple.md
12894 (load_multiple): Update nds32_expand_load_multiple interface.
12895 (store_multiple): Update nds32_expand_store_multiple interface.
12896 * config/nds32/nds32-predicates.c
12897 (nds32_valid_multiple_load_store): Rename ...
12898 (nds32_valid_multiple_load_store_p): ... to this and refine
12900 * config/nds32/predicates.md
12901 (nds32_load_multiple_and_update_address_operation): New predicate.
12902 (nds32_store_multiple_and_update_address_operation): New predicate.
12904 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12905 Chung-Ju Wu <jasonwucj@gmail.com>
12907 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
12908 (combo): New attribute.
12909 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
12911 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
12913 * config/nds32/nds32.opt: Change -mcmodel= default value.
12915 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
12916 Monk Chiang <sh.chiang04@gmail.com>
12917 Chung-Ju Wu <jasonwucj@gmail.com>
12919 * config/nds32/constants.md (unspec_element): New enum.
12920 * config/nds32/constraints.md (Umw): New constraint.
12921 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
12922 * config/nds32/nds32-intrinsic.md: Likewise.
12923 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
12924 (nds32_valid_smw_lwm_base_p): New.
12925 (nds32_output_smw_single_word): New.
12926 (nds32_output_lmw_single_word): New.
12927 (nds32_expand_unaligned_load): New.
12928 (nds32_expand_unaligned_store): New.
12929 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
12930 (nds32_output_smw_single_word): Declare.
12931 (nds32_output_lmw_single_word): Declare.
12932 (nds32_expand_unaligned_load): Declare.
12933 (nds32_expand_unaligned_store): Declare.
12934 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
12935 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
12936 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
12937 NDS32_BUILTIN_UASTORE_DW.
12938 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
12941 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
12942 Kito Cheng <kito.cheng@gmail.com>
12943 Chung-Ju Wu <jasonwucj@gmail.com>
12945 * config/nds32/nds32-intrinsic.c
12946 (nds32_expand_builtin_null_ftype_reg): Delete.
12947 (nds32_expand_builtin_reg_ftype_imm): Ditto.
12948 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
12949 (nds32_read_argument): New.
12950 (nds32_legitimize_target): Ditto.
12951 (nds32_legitimize_argument): Ditto.
12952 (nds32_check_constant_argument): Ditto.
12953 (nds32_expand_unop_builtin): Ditto.
12954 (nds32_expand_unopimm_builtin): Ditto.
12955 (nds32_expand_binop_builtin): Ditto.
12956 (nds32_builtin_decl_impl): Ditto.
12957 (builtin_description): Ditto.
12958 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
12959 (nds32_init_builtins_impl): Ditto.
12960 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
12961 (nds32_builtin_decl): New.
12962 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
12963 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
12965 2018-03-02 Jeff Law <law@redhat.com>
12967 * reorg.c (stop_search_p): Handle DEBUG_INSN.
12968 (redundant_insn, fill_simple_delay_slots): Likewise.
12969 (fill_slots_from_thread): Likewise.
12970 * resource.c (mark_referenced_resources): Likewise.
12971 (mark_set_resources, find_dead_or_set_registers): Likewise.
12973 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12975 * substring-locations.h (format_warning_va): Formatting fix for
12976 ATTRIBUTE_GCC_DIAG.
12977 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
12979 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
12980 * substring-locations.c: Include intl.h.
12981 (format_warning_va): Turned into small wrapper around
12982 format_warning_n_va, renamed to ...
12983 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
12984 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
12986 (format_warning_at_substring_n): New function.
12987 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
12988 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
12989 format_warning_at_substring with just a shorter name instead of
12990 const function pointer.
12991 (fmtwarn_n): New function.
12992 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
12993 appropriate, get rid of all the fmtstr temporaries, move conditionals
12994 with G_() wrapped string literals directly into fmtwarn arguments,
12995 cast dir.len to (int), formatting fixes.
12997 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
12999 * doc/invoke.texi: Remove "Cilk Plus" references.
13001 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13002 Richard Biener <rguenther@suse.de>
13005 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
13006 for error or warning attributes if CALL_FROM_THUNK_P is set.
13009 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13012 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
13013 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
13016 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
13017 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
13019 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
13020 instead of -1U in last predictors element's probability member.
13022 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
13025 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
13026 arguments if they are comparable.
13028 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
13030 PR tree-optimization/84634
13031 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
13032 masks and masked_loop_p with a single loop_masks, making sure it's
13033 null for bb vectorization.
13035 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
13037 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
13038 (vect_analyze_data_ref_access): Use loop->safe_len rather than
13039 loop->force_vectorize to check whether there is no alias.
13041 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13044 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
13046 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
13048 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
13049 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
13050 instead of a loop around prev_real_insn.
13051 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
13054 PR inline-asm/84625
13055 * config/i386/i386.c (ix86_print_operand): Use conditional
13056 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
13059 2018-03-02 Richard Biener <rguenther@suse.de>
13061 PR tree-optimization/84427
13062 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
13063 (bitmap_set_subtract_values): Rewrite to handle multiple
13066 (prune_clobbered_mems): Likewise.
13067 (phi_translate): Take edge instead of pred/phiblock.
13068 (phi_translate_1): Likewise.
13069 (phi_translate_set): Likewise. Insert all translated
13070 exprs for a value into the set, keeping possibly multiple
13071 expressions per value.
13072 (compute_antic_aux): Adjust for phi_translate changes.
13073 When intersecting union the expressions and prune those
13074 not in the final value set, keeping possibly multiple
13075 expressions per value. Do not use value-insertion
13076 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
13077 all expressions. Add verification that the value-sets
13078 only shrink during iteration.
13079 (compute_partial_antic_aux): Adjust for the phi_translate changes.
13080 (do_pre_regular_insertion): Likewise.
13081 (do_pre_partial_partial_insertion): Likewise.
13083 2018-03-02 Richard Biener <rguenther@suse.de>
13086 * config/darwin.c (saved_debug_info_level): New static global.
13087 (darwin_asm_lto_start): Disable debug info generation for LTO out.
13088 (darwin_asm_lto_end): Restore debug info generation settings.
13090 2018-03-01 Martin Liska <mliska@suse.cz>
13093 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13094 volatile arguments.
13096 2018-03-01 Richard Biener <rguenther@suse.de>
13099 * dwarf2out.c (gen_variable_die): Properly handle late VLA
13100 type annotation with LTO when debug was disabled at compile-time.
13102 2018-03-01 Matthew Fortune <mfortune@gmail.com>
13104 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
13106 (mips_final_postscan_insn): Likewise.
13108 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
13110 PR rtl-optimization/84528
13111 * alias.c (init_alias_target): Add commentary.
13112 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
13113 a unique base value if the frame pointer is not eliminated
13114 to the stack pointer.
13116 2018-03-01 Tom de Vries <tom@codesourcery.com>
13118 PR rtl-optimization/83327
13119 * lra-int.h (hard_regs_spilled_into): Declare.
13120 * lra.c (hard_regs_spilled_into): Define.
13121 (init_reg_info): Init hard_regs_spilled_into.
13122 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
13123 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
13124 (process_bb_lives): Handle hard_regs_spilled_into.
13125 (lra_create_live_ranges_1): Before doing liveness propagation, clear
13126 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
13128 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
13130 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
13131 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
13132 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
13133 * config/rs6000/aix72.h: New file.
13135 2018-02-28 Jakub Jelinek <jakub@redhat.com>
13137 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
13138 instead of warning_at with conditional singular and plural messages
13142 * stor-layout.c (update_alignment_for_field): For
13143 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
13144 && !DECL_PACKED (field), do the alignment update, just use
13145 only desired_align instead of MAX (type_align, desired_align)
13147 (place_field): Don't do known_align < desired_align handling
13148 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
13149 is non-NULL, instead do it after rli->prev_field handling and
13150 only if not within a bitfield word. For DECL_PACKED (field)
13151 use type_align of BITS_PER_UNIT.
13153 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
13155 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
13156 superfluous parentheses and trailing spaces.
13158 2018-02-28 Richard Biener <rguenther@suse.de>
13160 PR tree-optimization/84584
13161 * graphite-scop-detection.c (scop_detection::add_scop): Discard
13162 SCoPs with fake exit edge.
13164 2018-02-28 Martin Liska <mliska@suse.cz>
13167 * timevar.c (timer::print): Fix format to properly print 100%
13170 2018-02-28 Richard Biener <rguenther@suse.de>
13172 PR middle-end/84607
13173 * genmatch.c (capture_info::walk_match): Do not mark
13174 captured expressions without operands as expr_p given
13175 they act more like predicates and should be subject to
13176 "lost tail" side-effect preserving.
13178 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
13180 PR rtl-optimization/81611
13181 * auto-inc-dec.c (attempt_change): Move dead note from
13182 mem_insn if it's the next use of regno
13183 (find_address): Take address use of reg holding
13184 non-incremented value. Add parm to limit search to the named
13186 (merge_in_block): Attempt to use a mem insn that is the next
13187 use of the original regno.
13189 2018-02-27 Martin Sebor <msebor@redhat.com>
13192 * doc/invoke.texi (-Wmissing-attributes): New option.
13193 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
13195 2018-02-27 Martin Sebor <msebor@redhat.com>
13197 PR translation/84207
13198 * diagnostic-core.h (warning_n, error_n, inform_n): Change
13199 n argument to unsigned HOST_WIDE_INT.
13200 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
13201 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
13202 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
13203 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
13205 2018-02-27 Richard Biener <rguenther@suse.de>
13207 PR tree-optimization/84512
13208 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
13209 Do not use the estimate returned from record_stmt_cost for
13210 the scalar iteration cost but sum properly using add_stmt_cost.
13212 2018-02-27 Richard Biener <rguenther@suse.de>
13214 PR tree-optimization/84466
13215 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
13216 Adjust last change to less strictly validate use operands.
13218 2018-02-27 Martin Liska <mliska@suse.cz>
13220 PR gcov-profile/84548
13221 * gcov.c (process_file): Allow partial overlap and consider it
13222 also as group functions.
13223 (output_lines): Properly calculate range of lines for a group.
13225 2018-02-27 Martin Liska <mliska@suse.cz>
13227 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
13228 'ggc' suffixes. Change first column width.
13229 (timer::print): Fix formatting of the column.
13231 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
13233 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
13234 preserve inline entry blocks for the sake of debug inline
13235 entry point markers alone.
13236 (remove_unused_locals): Suggest in comments a better place to
13237 force the preservation of inline entry blocks that are
13238 otherwise unused, but do not preserve them.
13240 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
13242 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
13244 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
13247 * config/i386/constraints.md (Bs): Replace
13248 ix86_indirect_branch_register with
13249 TARGET_INDIRECT_BRANCH_REGISTER.
13251 * config/i386/i386.md (indirect_jump): Likewise.
13252 (tablejump): Likewise.
13253 (*sibcall_memory): Likewise.
13254 (*sibcall_value_memory): Likewise.
13255 Peepholes of indirect call and jump via memory: Likewise.
13256 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
13257 (*sibcall_value_GOT_32): Likewise.
13258 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13259 (GOT_memory_operand): Likewise.
13260 (call_insn_operand): Likewise.
13261 (sibcall_insn_operand): Likewise.
13262 (GOT32_symbol_operand): Likewise.
13263 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
13265 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
13267 PR rtl-optimization/83496
13268 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
13269 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
13270 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
13271 redundant insn, if any.
13272 (relax_delay_slots): Likewise.
13273 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
13275 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
13277 PR tree-optimization/83965
13278 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
13279 that grouped statements are part of a reduction chain. Return
13280 true if the statement is not marked as a reduction itself but
13281 is part of a group.
13282 (vect_recog_dot_prod_pattern): Don't check whether the statement
13283 is part of a group here.
13284 (vect_recog_sad_pattern): Likewise.
13285 (vect_recog_widen_sum_pattern): Likewise.
13287 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
13290 * final.c (rest_of_clean_state): Also look for calls inside sequences.
13292 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
13295 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
13297 (ix86_output_indirect_function_return): New prototype.
13298 (ix86_split_simple_return_pop_internal): Likewise.
13299 * config/i386/i386.c (indirect_return_via_cx): New.
13300 (indirect_return_via_cx_bnd): Likewise.
13301 (indirect_thunk_name): Handle return va CX_REG.
13302 (output_indirect_thunk_function): Create alias for
13303 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
13304 (ix86_output_indirect_jmp): Remove the bool argument.
13305 (ix86_output_indirect_function_return): New function.
13306 (ix86_split_simple_return_pop_internal): Likewise.
13307 * config/i386/i386.md (*indirect_jump): Don't pass false
13308 to ix86_output_indirect_jmp.
13309 (*tablejump_1): Likewise.
13310 (simple_return_pop_internal): Change it to define_insn_and_split.
13311 Call ix86_split_simple_return_pop_internal to split it for
13312 -mfunction-return=.
13313 (simple_return_indirect_internal): Call
13314 ix86_output_indirect_function_return instead of
13315 ix86_output_indirect_jmp.
13317 2018-02-26 Jakub Jelinek <jakub@redhat.com>
13320 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
13321 memset and value initialization afterwards.
13323 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
13325 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
13327 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
13330 * common/config/aarch64/aarch64-common.c
13331 (aarch_option_optimization_table[]): Switch
13332 off fomit-frame-pointer
13334 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
13335 Chung-Ju Wu <jasonwucj@gmail.com>
13337 * config/nds32/nds32-multiple.md (load_multiple): Disallow
13339 (store_multiple): Ditto.
13341 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
13343 * config.gcc: Add --with-cpu support for nds32 target.
13344 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
13345 * config/nds32/nds32.opt: Add -mcpu= option.
13347 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
13349 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
13350 isel=yes): Warn for these deprecated options.
13352 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
13354 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
13355 ISA_2_5_MASKS_EMBEDDED.
13357 2018-02-23 Jakub Jelinek <jakub@redhat.com>
13359 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
13360 p->max as pointers rather than using iterative_hash_expr.
13362 2018-02-23 Carl Love <cel@us.ibm.com>
13364 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
13365 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
13367 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
13368 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
13369 P8V_BUILTIN_VEC_VUNSIGNED2.
13371 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
13374 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
13375 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
13377 * lra-constraints.c (curr_insn_transform): Set up
13378 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
13380 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
13381 * lra-eliminations.c (spill_pseudos): Ditto.
13382 (process_insn_for_elimination): Ditto.
13383 * lra-lives.c (reg_early_clobber_p): Use the new macros.
13384 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
13385 LRA_NON_CLOBBERED_ALT.
13387 2018-02-22 Martin Sebor <msebor@redhat.com>
13389 PR tree-optimization/84480
13390 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
13391 to maybe_diag_stxncpy_trunc. Call it.
13392 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
13393 from gimple_fold_builtin_strcpy. Print inlining stack.
13394 (handle_builtin_stxncpy): Print inlining stack.
13395 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
13397 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
13400 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
13401 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
13402 and -fcheck-pointer-bounds are used together.
13403 (indirect_thunk_prefix): New enum.
13404 (indirect_thunk_need_prefix): New function.
13405 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
13406 "_nt" instead of "_bnd" for NOTRACK prefix.
13407 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
13408 (output_indirect_thunk_function): Likewise.
13410 (ix86_code_end): Update output_indirect_thunk_function calls.
13411 (ix86_output_indirect_branch_via_reg): Replace
13412 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
13413 (ix86_output_indirect_branch_via_push): Likewise.
13414 (ix86_output_function_return): Likewise.
13415 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
13416 incompatible with -fcf-protection=branch and
13417 -fcheck-pointer-bounds.
13419 2018-02-22 Steve Ellcey <sellcey@cavium.com>
13422 * config/aarch64/aarch64.c (aarch64_print_address_internal):
13423 Change gcc_assert call to output_operand_lossage.
13425 2018-02-22 Steve Ellcey <sellcey@cavium.com>
13427 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
13429 2018-02-22 DJ Delorie <dj@redhat.com>
13430 Sebastian Perta <sebastian.perta@renesas.com>
13431 Oleg Endo <olegendo@gcc.gnu.org>
13433 * config/rx/rx.c (rx_rtx_costs): New function.
13434 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
13436 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
13438 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
13440 2018-02-22 Martin Liska <mliska@suse.cz>
13443 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
13444 Add "native" as a possible value.
13446 2018-02-22 Martin Liska <mliska@suse.cz>
13449 * config/i386/i386.c (ix86_option_override_internal):
13450 Add "native" as a possible value for -march and -mtune.
13452 2018-02-22 Jakub Jelinek <jakub@redhat.com>
13455 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
13456 to all type variants.
13458 PR tree-optimization/84503
13459 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
13460 width as info->bitpos + info->bitsize - start.
13461 (merged_store_group::merge_overlapping): Simplify width computation.
13462 (check_no_overlap): New function.
13463 (imm_store_chain_info::try_coalesce_bswap): Compute expected
13464 start + width and last_order of the group, fail if check_no_overlap
13466 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
13467 to group if check_no_overlap fails.
13469 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
13471 * config/rs6000/altivec.md: Delete contraint arguments to
13472 define_expand, define_split, and define_peephole2, and in
13473 define_insn_and_split if always unused.
13474 * config/rs6000/darwin.md: Ditto.
13475 * config/rs6000/dfp.md: Ditto.
13476 * config/rs6000/rs6000.md: Ditto.
13477 * config/rs6000/sync.md: Ditto.
13478 * config/rs6000/vector.md: Ditto.
13479 * config/rs6000/vsx.md: Ditto.
13481 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
13483 * config/rs6000/altivec.md: Write output control strings as braced
13484 blocks instead of double-quoted strings.
13485 * config/rs6000/darwin.md: Ditto.
13486 * config/rs6000/rs6000.md: Ditto.
13487 * config/rs6000/vector.md: Ditto.
13488 * config/rs6000/vsx.md: Ditto.
13490 2018-02-21 Jason Merrill <jason@redhat.com>
13492 PR c++/84314 - ICE with templates and fastcall attribute.
13493 * attribs.c (build_type_attribute_qual_variant): Remove assert.
13495 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
13497 * ipa-cp.c (determine_versionability): Fix comment typos.
13499 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
13502 * ipa-cp.c (determine_versionability): Do not version functions caling
13505 2018-02-21 Martin Liska <mliska@suse.cz>
13508 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
13509 Add "native" as a possible value.
13510 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
13511 the macro when native cpu detection is available.
13513 2018-02-21 Martin Liska <mliska@suse.cz>
13516 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
13517 Add "native" as a possible value.
13518 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
13519 when native cpu detection is available.
13521 2018-02-21 Jakub Jelinek <jakub@redhat.com>
13522 Martin Sebor <msebor@redhat.com>
13524 PR tree-optimization/84478
13525 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
13527 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
13528 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
13529 support which is conservatively correct, for 2 only stay conservative
13530 for maxlen. Formatting and comment capitalization fixes. Add STRICT
13531 argument to the 2 argument get_range_strlen, adjust 6 arg
13532 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
13534 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
13535 (gimple_fold_builtin_strlen): Pass true as last argument to
13538 2018-02-20 Martin Sebor <msebor@redhat.com>
13540 PR middle-end/84095
13541 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
13542 (builtin_memref::set_base_and_offset): Same. Handle inner references.
13543 (builtin_memref::builtin_memref): Factor out parts into
13544 set_base_and_offset and call it.
13546 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
13548 PR middle-end/84406
13549 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
13550 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
13551 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
13552 search at the associated MODE_INT.
13554 2018-02-20 Jeff Law <law@redhat.com>
13556 PR middle-end/82123
13557 PR tree-optimization/81592
13558 PR middle-end/79257
13559 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
13560 for range data rather than using global data.
13561 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
13562 range data rather than using global data.
13563 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
13564 pass it to children as needed.
13565 (struct directive::fmtresult): Similarly.
13566 (struct directive::set_width): Similarly.
13567 (struct directive::set_precision): Similarly.
13568 (format_integer, format_directive, parse_directive): Similarly.
13569 (format_none): Accept unnamed vr_values parameter.
13570 (format_percent, format_floating, format_character): Similarly.
13571 (format_string, format_plain): Similarly.
13572 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
13573 the EVRP range analyzer for range data rather than using global data.
13574 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
13575 gimple-ssa-evrp-analyze.h
13576 (class sprintf_dom_walker): Add after_dom_children member function.
13577 Add evrp_range_analyzer member.
13578 (sprintf_dom_walker::before_dom_children): Call into the EVRP
13579 range analyzer as needed.
13580 (sprintf_dom_walker::after_dom_children): New member function.
13581 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
13583 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
13584 (evrp_range_analyzer::pop_to_marker): Likewise.
13586 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
13588 PR tree-optimization/84419
13589 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
13590 with the required type if its current type is compatible but
13593 2018-02-20 Jakub Jelinek <jakub@redhat.com>
13595 PR middle-end/82004
13596 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
13597 after vectorization.
13599 2018-02-20 Martin Liska <mliska@suse.cz>
13602 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
13603 possible values if we don't have a hint.
13605 2018-02-20 Martin Liska <mliska@suse.cz>
13609 * final.c (shorten_branches): Build align_tab array with one
13611 * opts.c (finish_options): Add alignment option limit check.
13612 (MAX_CODE_ALIGN): Likewise.
13613 (MAX_CODE_ALIGN_VALUE): Likewise.
13614 * doc/invoke.texi: Document maximum allowed option value for
13615 all -falign-* options.
13617 2018-02-19 Jakub Jelinek <jakub@redhat.com>
13620 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
13621 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
13622 * var-tracking.c (emit_note_insn_var_location): Remove all references
13623 to NOTE_INSN_CALL_ARG_LOCATION.
13624 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
13625 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
13626 Use copy_rtx_if_shared.
13627 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
13628 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
13629 (dwarf2out_var_location): Remove handling of
13630 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
13632 * final.c (final_scan_insn): Remove all references to
13633 NOTE_INSN_CALL_ARG_LOCATION.
13634 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
13635 before dumping final insns.
13636 * except.c (emit_note_eh_region_end): Remove all references to
13637 NOTE_INSN_CALL_ARG_LOCATION.
13638 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
13639 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
13640 * config/arc/arc.c (hwloop_optimize): Likewise.
13641 * config/arm/arm.c (create_fix_barrier): Likewise.
13642 * config/s390/s390.c (s390_chunkify_start): Likewise.
13643 * config/sh/sh.c (find_barrier): Likewise.
13644 * config/i386/i386.c (rest_of_insert_endbranch,
13645 ix86_seh_fixup_eh_fallthru): Likewise.
13646 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
13647 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
13648 * config/frv/frv.c (frv_function_prologue): Likewise.
13649 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
13651 (note_outside_basic_block_p): Remove all references to
13652 NOTE_INSN_CALL_ARG_LOCATION.
13653 * gengtype.c (adjust_field_rtx_def): Likewise.
13654 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
13656 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
13657 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
13660 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
13663 PR tree-optimization/84452
13664 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
13665 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
13668 2018-02-19 Martin Liska <mliska@suse.cz>
13671 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
13673 2018-02-19 Martin Liska <mliska@suse.cz>
13674 Richard Sandiford <richard.sandiford@linaro.org>
13676 PR tree-optimization/82491
13677 * gimple-fold.c (get_base_constructor): Make earlier bail out
13680 2018-02-19 Carl Love <cel@us.ibm.com>
13682 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
13683 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
13685 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
13686 P8V_BUILTIN_VEC_NEG.
13688 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
13690 * config/rl78/rl78.md (movdf): New define expand.
13692 2018-02-19 Martin Liska <mliska@suse.cz>
13695 * doc/invoke.texi: Fix typo.
13696 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
13698 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
13700 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
13701 handle rs6000_single_float and rs6000_double_float specially for
13704 2018-02-16 Jeff Law <law@redhat.com>
13706 * config/rx/rx.c (add_pop_cfi_notes): New function.;
13707 (pop_regs): Use it.
13709 2018-02-16 Jakub Jelinek <jakub@redhat.com>
13712 * ipa-inline.c (inline_small_functions): Fix a typo.
13714 2018-02-16 Nathan Sidwell <nathan@acm.org>
13716 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
13718 2018-02-16 Carl Love <cel@us.ibm.com>
13720 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
13721 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
13722 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
13723 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
13724 expansion to P8V_BUILTIN_VEC_FLOAT2.
13726 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
13728 PR rtl-optimization/70023
13729 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
13730 src_regno into account.
13732 2018-02-16 Carl Love <cel@us.ibm.com>
13734 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
13735 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
13736 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
13737 * config/rs6000/rs6000.c: Remove case statements for
13738 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
13739 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
13740 and P9V_BUILTIN_VEC_VINSERT4B.
13741 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
13742 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
13743 * config/rs6000/vsx.md:
13744 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
13747 2018-02-16 Carl Love <cel@us.ibm.com>
13749 * config/rs6000/altivec.h: Add builtin names vec_extract4b
13751 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
13753 * config/rs6000/rs6000-c.c: Add the definitions for
13754 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
13755 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
13756 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
13757 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
13758 definition for insert4b and define insn *insert3b_internal.
13759 * doc/extend.texi: Add documentation for vec_extract4b.
13761 2018-02-16 Nathan Sidwell <nathan@acm.org>
13763 * doc/extend.texi (Backwards Compatibility): Mention friend
13764 injection. Note for-scope is deprecated.
13765 * doc/invoke.texi (-ffriend-injection): Deprecate.
13767 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
13769 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
13770 that moved to I2, also allow destinations that are a paradoxical
13771 subreg (instead of a normal reg).
13773 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
13776 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
13779 2018-02-16 Richard Biener <rguenther@suse.de>
13781 PR tree-optimization/84037
13782 PR tree-optimization/84016
13784 * config/i386/i386.c (ix86_builtin_vectorization_cost):
13785 Adjust vec_construct for the fact we need additional higher latency
13786 128bit inserts for AVX256 and AVX512 vector builds.
13787 (ix86_add_stmt_cost): Scale vector construction cost for
13790 2018-02-16 Richard Biener <rguenther@suse.de>
13792 PR tree-optimization/84417
13793 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
13794 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
13795 (non_rewritable_lvalue_p): Likewise, use poly-ints.
13797 2018-02-16 Martin Liska <mliska@suse.cz>
13800 * internal-fn.def (ASAN_CHECK): Set proper flags.
13801 (ASAN_MARK): Likewise.
13803 2018-02-16 Julia Koval <julia.koval@intel.com>
13805 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
13806 from PTA_CANNONLAKE.
13808 2018-02-16 Jakub Jelinek <jakub@redhat.com>
13811 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
13812 Use ++iter rather than iter++ for std::list iterators.
13813 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
13814 defer deleting them until all nodes in the forest are processed. Do
13815 free even leaf nodes. Change to_process into auto_vec.
13818 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
13819 * vec.h (vec_default_construct): Use memset instead of placement new
13820 if BROKEN_VALUE_INITIALIZATION is defined.
13821 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
13822 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
13825 PR rtl-optimization/83723
13826 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
13827 * lra.c (lra_substitute_pseudo): Likewise. If true, use
13828 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
13830 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
13832 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
13834 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
13836 PR rtl-optimization/81443
13837 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
13838 from inner REGs to paradoxical SUBREGs.
13840 2018-02-16 Richard Biener <rguenther@suse.de>
13842 PR tree-optimization/84399
13843 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
13844 For operands we can analyze at their definition make sure we can
13845 analyze them at each use as well.
13847 2018-02-16 Richard Biener <rguenther@suse.de>
13849 PR tree-optimization/84190
13850 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
13851 volatile accesses if the decl isn't volatile.
13853 2018-02-15 Jason Merrill <jason@redhat.com>
13855 PR c++/84314 - ICE with templates and fastcall attribute.
13856 * attribs.c (build_type_attribute_qual_variant): Don't clobber
13857 TYPE_CANONICAL on an existing type.
13859 2018-02-15 Jakub Jelinek <jakub@redhat.com>
13861 PR tree-optimization/84383
13862 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
13863 dstoff nor call operand_equal_p if dstbase is NULL.
13865 PR tree-optimization/84334
13866 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
13867 also a CONSTANT_CLASS_P, punt.
13869 2018-02-14 Jim Wilson <jimw@sifive.com>
13871 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
13872 first SMALL_OPERAND check. New local min_second_step. Move assert
13873 to where locals are set. Add TARGET_RVC support.
13874 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
13876 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
13878 * doc/invoke.texi: Correct -Wformat-overflow code sample.
13880 2018-02-14 Martin Sebor <msebor@redhat.com>
13882 PR tree-optimization/83698
13883 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
13884 arrays constrain the offset range to their bounds.
13885 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
13886 (builtin_access::overlap): Avoid setting the size of overlap if it's
13888 (maybe_diag_overlap): Also consider arrays when deciding what values
13889 of offsets to include in diagnostics.
13891 2018-02-14 Martin Sebor <msebor@redhat.com>
13894 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
13895 that correspond to the kind of a declaration.
13897 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
13900 * config/pa/pa.md: Load address of PIC label using the linkage table
13901 if the label is nonlocal.
13903 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
13905 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
13906 warning message if user requests -maltivec=be.
13907 * doc/invoke.texi: Document deprecation of -maltivec=be.
13909 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
13912 * config/rs6000/rs6000-c.c: Update definitions for
13913 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
13914 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
13916 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13919 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
13920 add _get_ssp intrinsics. Remove argument from
13921 __builtin_ia32_rdssp[d|q].
13922 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
13923 * config/i386/i386-builtin.def: Remove argument from
13924 __builtin_ia32_rdssp[d|q].
13925 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
13926 ix86_expand_special_args_builtin for _rdssp[d|q].
13927 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
13928 Clear register before usage.
13929 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
13930 Add documentation for new _get_ssp and _inc_ssp intrinsics.
13932 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
13934 PR tree-optimization/84357
13935 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
13936 operand 1 of an ARRAY_REF too.
13938 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
13941 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
13942 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
13944 (set_of_reg): New struct.
13945 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
13946 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
13947 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
13949 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
13950 Split into bitclr, bitset, bitinvert patterns if appropriate.
13951 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
13952 use rx_fuse_in_memory_bitop.
13953 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
13954 to named insn, correct maximum insn length.
13956 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
13959 * machmode.def: Define a complex mode for PARTIAL_INT.
13960 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
13962 * doc/rtl.texi: Document CSPImode.
13963 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
13965 (msp430_hard_regno_nregs_with_padding): Likewise.
13967 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
13970 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
13972 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
13974 PR rtl-optimization/84169
13975 * combine.c (try_combine): New variable split_i2i3. Set it to true if
13976 we generated a parallel as new i3 and we split that to new i2 and i3
13977 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
13978 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
13979 those to i2, not i1. Partially rewrite this scan code.
13981 2018-02-13 Jakub Jelinek <jakub@redhat.com>
13984 * stor-layout.c (place_field): For variable length fields, adjust
13985 offset_align afterwards not just based on the field's alignment,
13986 but also on the size.
13988 PR middle-end/84309
13989 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
13990 of exps and logs in the use_exp2 case.
13992 2018-02-13 Jeff Law <law@redhat.com>
13994 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
13995 entry for "vector".
13997 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
14000 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
14004 * common.opt (gas-loc-support, gas-locview-support): New.
14005 (ginline-points, ginternal-reset-location-views): New.
14006 * doc/invoke.texi: Document them. Use @itemx where intended.
14007 (gvariable-location-views): Adjust.
14008 * target.def (reset_location_view): New.
14009 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
14010 (TARGET_RESET_LOCATION_VIEW): New.
14011 * doc/tm.texi: Rebuilt.
14012 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
14013 (dwarf2out_default_as_locview_support): New.
14014 (output_asm_line_debug_info): Use option variables.
14015 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
14016 (output_loc_list): Likewise.
14017 (add_high_low_attributes): Check option variables.
14018 Don't output entry view attribute in strict mode.
14019 (gen_inlined_subroutine_die): Check option variables.
14020 (dwarf2out_inline_entry): Likewise.
14021 (init_sections_and_labels): Likewise.
14022 (dwarf2out_early_finish): Likewise.
14023 (maybe_reset_location_view): New, from...
14024 (dwarf2out_var_location): ... here. Call it.
14025 * debug.h (dwarf2out_default_as_loc_support): Declare.
14026 (dwarf2out_default_as_locview_support): Declare.
14027 * hooks.c (hook_int_rtx_insn_0): New.
14028 * hooks.h (hook_int_rtx_insn_0): Declare.
14029 * toplev.c (process_options): Take -gas-loc-support and
14030 -gas-locview-support from dwarf2out. Enable
14031 -gvariable-location-views by default only with locview
14032 assembler support. Enable -ginternal-reset-location-views by
14033 default only if the target defines the corresponding hook.
14034 Enable -ginline-points by default if location views are
14035 enabled; force it disabled if statement frontiers are
14037 * tree-inline.c (expand_call_inline): Check option variables.
14038 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
14040 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
14042 PR tree-optimization/84321
14043 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
14044 handling. Also check whether the anti-range contains any values
14045 that satisfy the mask; switch to a VR_RANGE if not.
14047 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
14050 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
14052 2018-02-13 Martin Jambor <mjambor@suse.cz>
14055 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
14056 of call statements, also set location of a load to a temporary.
14058 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
14060 * config/rl78/rl78.c (add_vector_labels): New function.
14061 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
14062 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
14063 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
14064 which checks that no arguments are passed.
14065 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
14066 * doc/extend.texi: Documentation for the new attribute.
14068 2018-02-13 Andreas Schwab <schwab@suse.de>
14070 * config/riscv/linux.h (CPP_SPEC): Define.
14072 2018-02-13 Jakub Jelinek <jakub@redhat.com>
14075 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
14076 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
14077 OPTION_MASK_ISA_AES as first argument to def_builtin_const
14078 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
14079 instead of OPTION_MASK_ISA_PCLMUL as first argument to
14080 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
14081 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
14082 temporarily for AES and PCLMUL builtins.
14084 PR tree-optimization/84339
14085 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
14086 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
14089 PR middle-end/84309
14090 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
14091 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
14092 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
14094 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
14096 * omp-simd-clone.h: New file.
14097 * omp-simd-clone.c: Include omp-simd-clone.h.
14098 (expand_simd_clones): No longer static.
14099 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
14100 cgraph.h and omp-simd-clone.h.
14101 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
14102 (vect_recog_widen_shift_pattern): Formatting fix.
14103 (vect_pattern_recog_1): Don't check optab for calls.
14106 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
14107 operands[2] into a REG before using gen_lowpart on it.
14109 2018-02-12 Jeff Law <law@redhat.com>
14112 * config/sh/sh.c (find_barrier): Consider a sibling call
14115 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
14116 successfully back substituting a reg.
14118 2018-02-12 Richard Biener <rguenther@suse.de>
14120 PR tree-optimization/84037
14121 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
14122 parameter, move visited init to caller.
14123 (vect_slp_analyze_operations): Separate cost from validity
14124 check, initialize visited once for all instances.
14125 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
14127 * tree-vect-stmts.c (vect_model_simple_cost): Make early
14129 (vect_model_promotion_demotion_cost): Likewise.
14130 (vectorizable_bswap): Guard cost modeling with !slp_node
14131 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
14133 (vectorizable_call): Likewise.
14134 (vectorizable_conversion): Likewise.
14135 (vectorizable_assignment): Likewise.
14136 (vectorizable_shift): Likewise.
14137 (vectorizable_operation): Likewise.
14138 (vectorizable_store): Likewise.
14139 (vectorizable_load): Likewise.
14140 (vectorizable_condition): Likewise.
14141 (vectorizable_comparison): Likewise.
14143 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
14146 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
14147 (ASAN_MARK): Fix fnspec to account for return value, change pointer
14148 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
14150 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
14152 PR middle-end/83665
14153 * params.def (inline-min-speedup): Increase from 8 to 15.
14154 (max-inline-insns-auto): Decrease from 40 to 30.
14155 * ipa-split.c (consider_split): Add some buffer for function to
14156 be considered inlining candidate.
14157 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
14160 2018-02-12 Richard Biener <rguenther@suse.de>
14162 PR tree-optimization/84037
14163 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
14164 matched stmts if we cannot swap the non-matched ones.
14166 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
14168 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
14169 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
14170 _mm_maskz_scalef_round_ss): New intrinsics.
14171 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
14172 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
14173 __builtin_ia32_scalefss_round): Remove.
14174 (__builtin_ia32_scalefsd_mask_round,
14175 __builtin_ia32_scalefss_mask_round): New intrinsics.
14176 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
14177 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
14178 ((match_operand:VF_128 2 "<round_nimm_predicate>"
14179 "<round_constraint>")): Changed to ...
14180 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
14181 "<round_scalar_constraint>")): ... this.
14182 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
14183 %0, %1, %2<round_op3>}"): Changed to ...
14184 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
14185 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
14186 %2<round_scalar_mask_op3>}"): ... this.
14187 * config/i386/subst.md (round_scalar_nimm_predicate): New.
14189 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
14191 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
14192 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
14193 (_mm_maskz_sqrt_round_ss): New intrinsics.
14194 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
14195 (__builtin_ia32_sqrtsd_mask_round)
14196 (__builtin_ia32_sqrtss_mask_round): New builtins.
14197 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
14198 (__builtin_ia32_sqrtss_round): Remove.
14199 (__builtin_ia32_sqrtsd_mask_round)
14200 (__builtin_ia32_sqrtss_mask_round): New builtins.
14201 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
14202 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
14203 ((match_operand:VF_128 1 "vector_operand"
14204 "xBm,<round_constraint>")): Changed to ...
14205 ((match_operand:VF_128 1 "vector_operand"
14206 "xBm,<round_scalar_constraint>")): ... this.
14207 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
14208 %0, %2, %<iptr>1<round_op3>}): Changed to ...
14209 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
14210 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
14211 %<iptr>1<round_scalar_mask_op3>}): ... this.
14212 ((set_attr "prefix" "<round_prefix>")): Changed to ...
14213 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
14215 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
14218 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
14219 Cast vec_cmpeq result to correct type.
14220 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
14221 Cast vec_cmpgt result to correct type.
14223 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
14225 * final.c (final_scan_insn_1): Renamed from...
14226 (final_scan_insn): ... this. New wrapper, to recover
14227 seen from the outermost call in recursive ones.
14228 * config/sparc/sparc.c (output_return): Drop seen from call.
14229 (output_sibcall): Likewise.
14230 * config/visium/visium.c (output_branch): Likewise.
14232 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
14234 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
14237 2018-02-10 Alan Modra <amodra@gmail.com>
14240 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
14241 Specify LR as an input.
14243 2018-02-10 Jakub Jelinek <jakub@redhat.com>
14246 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
14247 remove_member_access_dummy_vars): New functions.
14248 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
14249 lower_omp_1, execute_lower_omp): Use them.
14251 PR rtl-optimization/84308
14252 * shrink-wrap.c (spread_components): Release todo vector.
14254 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
14256 PR rtl-optimization/57193
14257 * ira-color.c (struct allocno_color_data): Add member
14258 conflict_allocno_hard_prefs.
14259 (update_conflict_allocno_hard_prefs): New.
14260 (bucket_allocno_compare_func): Add a preference based on
14261 conflict_allocno_hard_prefs.
14262 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
14263 (color_allocnos): Remove a dead code. Initiate
14264 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
14266 2018-02-09 Jakub Jelinek <jakub@redhat.com>
14269 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
14270 constraint from =wa to wa. Avoid a subreg on the output operand,
14271 instead use a pseudo and subreg it in a move.
14272 (p9_xxbrd_<mode>): Changed to ...
14273 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
14274 (p9_xxbrd_v2df): New expander.
14275 (p9_xxbrw_<mode>): Changed to ...
14276 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
14277 (p9_xxbrw_v4sf): New expander.
14279 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
14281 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
14283 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
14286 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
14287 multiply in 32-bit mode.
14288 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
14289 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
14292 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
14294 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
14295 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
14296 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
14297 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
14299 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
14302 * dwarf2out.c (is_trivial_indirect_ref): New function.
14303 (dwarf2out_late_global_decl): Do not generate a location
14304 attribute for variables that have a non-trivial DECL_VALUE_EXPR
14305 and that are not defined in the current unit.
14307 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
14309 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
14310 instead of a libcall for UNORDERED.
14312 2018-02-09 Tamar Christina <tamar.christina@arm.com>
14315 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
14316 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
14318 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14321 * config/s390/s390.c (s390_set_current_function): Invoke
14322 s390_indirect_branch_settings also if fndecl didn't change.
14324 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
14326 * config/rs6000/rs6000.md (blockage): Set length to zero.
14328 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
14330 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
14332 2018-02-09 Jakub Jelinek <jakub@redhat.com>
14335 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
14336 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
14340 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
14341 PARALLEL incoming that failed vt_get_decl_and_offset check.
14343 PR middle-end/84237
14344 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
14345 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
14347 (get_variable_section): For decls in named .bss* sections pass true as
14348 second argument to bss_initializer_p.
14350 2018-02-09 Marek Polacek <polacek@redhat.com>
14351 Jakub Jelinek <jakub@redhat.com>
14354 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
14355 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
14356 Sync some changes from cxx_fold_indirect_ref.
14358 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
14360 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
14362 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
14363 (BLOCK_INLINE_ENTRY_LABEL): New.
14364 (dwarf2out_var_location): Disregard inline entry markers.
14365 (inline_entry_data): New struct.
14366 (inline_entry_data_hasher): New hashtable type.
14367 (inline_entry_data_hasher::hash): New.
14368 (inline_entry_data_hasher::equal): New.
14369 (inline_entry_data_table): New variable.
14370 (add_high_low_attributes): Add DW_AT_entry_pc and
14371 DW_AT_GNU_entry_view attributes if a pending entry is found
14372 in inline_entry_data_table. Add old entry_pc attribute only
14373 if debug nonbinding markers are disabled.
14374 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
14375 markers are enabled.
14376 (block_within_block_p, dwarf2out_inline_entry): New.
14377 (dwarf2out_finish): Check that no entries remained in
14378 inline_entry_data_table.
14379 * final.c (reemit_insn_block_notes): Handle inline entry notes.
14380 (final_scan_insn, notice_source_line): Likewise.
14381 (rest_of_clean_state): Skip inline entry markers.
14382 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
14384 * gimple.c (gimple_build_debug_inline_entry): New.
14385 * gimple.h (enum gimple_debug_subcode): Add
14386 GIMPLE_DEBUG_INLINE_ENTRY.
14387 (gimple_build_debug_inline_entry): Declare.
14388 (gimple_debug_inline_entry_p): New.
14389 (gimple_debug_nonbind_marker_p): Adjust.
14390 * insn-notes.def (INLINE_ENTRY): New.
14391 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
14392 inline entry marker notes.
14393 (print_insn): Likewise.
14394 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
14395 (INSN_DEBUG_MARKER_KIND): Likewise.
14396 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
14397 * tree-inline.c (expand_call_inline): Build and insert
14398 debug_inline_entry stmt.
14399 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
14400 inline entry blocks early, if nonbind markers are enabled.
14401 (dump_scope_block): Dump fragment info.
14402 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
14403 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
14404 (gimple_build_debug_inline_entry): New.
14405 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
14406 Enable/disable inline entry points too.
14407 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
14408 (DEBUG_INSN): Describe inline entry markers.
14410 * common.opt (gvariable-location-views): New.
14411 (gvariable-location-views=incompat5): New.
14412 * config.in: Rebuilt.
14413 * configure: Rebuilt.
14414 * configure.ac: Test assembler for view support.
14415 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
14416 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
14417 * dwarf2out.c (var_loc_view): New typedef.
14418 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
14419 (dwarf2out_locviews_in_attribute): New.
14420 (dwarf2out_locviews_in_loclist): New.
14421 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
14422 (enum dw_line_info_opcode): Add LI_adv_address.
14423 (struct dw_line_info_table): Add view.
14424 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
14425 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
14426 (zero_view_p): New variable.
14427 (ZERO_VIEW_P): New macro.
14428 (output_asm_line_debug_info): New.
14429 (struct var_loc_node): Add view.
14430 (add_AT_view_list, AT_loc_list): New.
14431 (add_var_loc_to_decl): Add view param. Test it against last.
14432 (new_loc_list): Add view params. Record them.
14433 (AT_loc_list_ptr): Handle loc and view lists.
14434 (view_list_to_loc_list_val_node): New.
14435 (print_dw_val): Handle dw_val_class_view_list.
14436 (size_of_die): Likewise.
14437 (value_format): Likewise.
14438 (loc_list_has_views): New.
14439 (gen_llsym): Set vl_symbol too.
14440 (maybe_gen_llsym, skip_loc_list_entry): New.
14441 (dwarf2out_maybe_output_loclist_view_pair): New.
14442 (output_loc_list): Output view list or entries too.
14443 (output_view_list_offset): New.
14444 (output_die): Handle dw_val_class_view_list.
14445 (output_dwarf_version): New.
14446 (output_compilation_unit_header): Use it.
14447 (output_skeleton_debug_sections): Likewise.
14448 (output_rnglists, output_line_info): Likewise.
14449 (output_pubnames, output_aranges): Update version comments.
14450 (output_one_line_info_table): Output view numbers in asm comments.
14451 (dw_loc_list): Determine current endview, pass it to new_loc_list.
14452 Call maybe_gen_llsym.
14453 (loc_list_from_tree_1): Adjust.
14454 (add_AT_location_description): Create view list attribute if
14455 needed, check it's absent otherwise.
14456 (convert_cfa_to_fb_loc_list): Adjust.
14457 (maybe_emit_file): Call output_asm_line_debug_info for test.
14458 (dwarf2out_var_location): Reset views as needed. Precompute
14459 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
14460 attribute. Set view.
14461 (new_line_info_table): Reset next view.
14462 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
14463 (dwarf2out_source_line): Likewise. Output view resets and labels to
14464 the assembler, or select appropriate line info opcodes.
14465 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
14466 (optimize_string_length): Catch it. Adjust.
14467 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
14468 dw_val_class_view_list, and remove it if no longer needed.
14469 (hash_loc_list): Hash view numbers.
14470 (loc_list_hasher::equal): Compare them.
14471 (optimize_location_lists): Check whether a view list symbol is
14472 needed, and whether the locview attribute is present, and
14473 whether they match. Remove the locview attribute if no longer
14475 (index_location_lists): Call skip_loc_list_entry for test.
14476 (dwarf2out_finish): Call output_asm_line_debug_info for test.
14477 Use output_dwarf_version.
14478 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
14479 (struct dw_val_node): Add val_view_list.
14480 * final.c (SEEN_NEXT_VIEW): New.
14481 (set_next_view_needed): New.
14482 (clear_next_view_needed): New.
14483 (maybe_output_next_view): New.
14484 (final_start_function): Rename to...
14485 (final_start_function_1): ... this. Take pointer to FIRST,
14486 add SEEN parameter. Emit param bindings in the initial view.
14487 (final_start_function): Reintroduce SEEN-less interface.
14488 (final): Rename to...
14489 (final_1): ... this. Take SEEN parameter. Output final pending
14490 next view at the end.
14491 (final): Reintroduce seen-less interface.
14492 (final_scan_insn): Output pending next view before switching
14493 sections or ending a block. Mark the next view as needed when
14494 outputting variable locations. Notify debug backend of section
14495 changes, and of location view changes.
14496 (rest_of_handle_final): Adjust.
14497 * toplev.c (process_options): Autodetect value for debug variable
14498 location views option. Warn on incompat5 without -gdwarf-5.
14499 * doc/invoke.texi (gvariable-location-views): New.
14500 (gvariable-location-views=incompat5): New.
14501 (gno-variable-location-views): New.
14503 2018-02-08 David Malcolm <dmalcolm@redhat.com>
14505 PR tree-optimization/84136
14506 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
14507 that the result of find_edge is non-NULL.
14509 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
14512 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
14513 storing integer register in SImode. Fix cost of 256 and 512
14514 byte aligned SSE register store.
14516 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
14518 * config/i386/i386.c (ix86_multiplication_cost): Fix
14519 multiplication cost for TARGET_AVX512DQ.
14521 2018-02-08 Marek Polacek <polacek@redhat.com>
14523 PR tree-optimization/84238
14524 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
14527 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14529 PR tree-optimization/84265
14530 * tree-vect-stmts.c (vectorizable_store): Don't treat
14531 VMAT_CONTIGUOUS accesses as grouped.
14532 (vectorizable_load): Likewise.
14534 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14536 PR tree-optimization/81635
14537 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
14538 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
14539 (test_round_for_mask): New functions.
14540 (wide_int_cc_tests): Call test_round_for_mask.
14541 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
14542 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
14543 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
14544 range returned by get_range_info.
14546 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
14549 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
14550 * symtab.c: Include builtins.h
14551 (symtab_node::output_to_lto_symbol_table_p): Move here
14552 from lto-streamer-out.c:output_symbol_p.
14553 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
14554 (output_symbol_p): Move all logic to symtab.c
14555 (produce_symtab): Update.
14557 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14559 * config/s390/s390-opts.h (enum indirect_branch): Define.
14560 * config/s390/s390-protos.h (s390_return_addr_from_memory)
14561 (s390_indirect_branch_via_thunk)
14562 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
14563 (enum s390_indirect_branch_type): Define.
14564 * config/s390/s390.c (struct s390_frame_layout, struct
14565 machine_function): Remove.
14566 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
14567 (indirect_branch_table_label_no, indirect_branch_table_name):
14569 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
14570 (enum s390_indirect_branch_option): Define.
14571 (s390_return_addr_from_memory): New function.
14572 (s390_handle_string_attribute): New function.
14573 (s390_attribute_table): Add new attribute handler.
14574 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
14575 (s390_indirect_branch_via_thunk): New function.
14576 (s390_indirect_branch_via_inline_thunk): New function.
14577 (s390_function_ok_for_sibcall): When jumping via thunk disallow
14578 sibling call optimization for non z10 compiles.
14579 (s390_emit_call): Force indirect branch target to be a single
14580 register. Add r1 clobber for non-z10 compiles.
14581 (s390_emit_epilogue): Emit return jump via return_use expander.
14582 (s390_reorg): Handle JUMP_INSNs as execute targets.
14583 (s390_option_override_internal): Perform validity checks for the
14584 new command line options.
14585 (s390_indirect_branch_attrvalue): New function.
14586 (s390_indirect_branch_settings): New function.
14587 (s390_set_current_function): Invoke s390_indirect_branch_settings.
14588 (s390_output_indirect_thunk_function): New function.
14589 (s390_code_end): Implement target hook.
14590 (s390_case_values_threshold): Implement target hook.
14591 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
14593 * config/s390/s390.h (struct s390_frame_layout)
14594 (struct machine_function): Move here from s390.c.
14595 (TARGET_INDIRECT_BRANCH_NOBP_RET)
14596 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
14597 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
14598 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
14599 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
14600 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
14601 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
14602 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
14603 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
14604 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
14605 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
14606 (mnemonic attribute): Add values which aren't recognized
14608 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
14609 pattern for branch conversion. Fix mnemonic attribute.
14610 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
14611 indirect branch via thunk if requested.
14612 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
14613 ("*indirect_jump"): Disable for branch conversion using out of
14615 ("indirect_jump_via_thunk<mode>_z10")
14616 ("indirect_jump_via_thunk<mode>")
14617 ("indirect_jump_via_inlinethunk<mode>_z10")
14618 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
14619 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
14620 ("casesi_jump_via_inlinethunk<mode>_z10")
14621 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
14622 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
14623 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
14624 ("*indirect2_jump"): Disable for branch conversion.
14625 ("casesi_jump"): Turn into expander and expand patterns for branch
14627 ("return_use"): New expander.
14628 ("*return"): Emit return via thunk and rename it to ...
14629 ("*return<mode>"): ... this one.
14630 * config/s390/s390.opt: Add new options and and enum for the
14633 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
14635 * lra-constraints.c (match_reload): Unconditionally use
14636 gen_lowpart_SUBREG, rather than selecting between that
14637 and equivalent gen_rtx_SUBREG code.
14639 2018-02-08 Richard Biener <rguenther@suse.de>
14641 PR tree-optimization/84233
14642 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
14643 changed flag instead of boguously re-using phi_inserted.
14645 2018-02-08 Martin Jambor <mjambor@suse.cz>
14647 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
14648 static local variables.
14650 2018-02-08 Richard Biener <rguenther@suse.de>
14652 PR tree-optimization/84278
14653 * tree-vect-stmts.c (vectorizable_store): When looking for
14654 smaller vector types to perform grouped strided loads/stores
14655 make sure the mode is supported by the target.
14656 (vectorizable_load): Likewise.
14658 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
14660 * config/aarch64/aarch64.c (aarch64_components_for_bb):
14661 Increase LDP/STP opportunities by adding adjacent callee-saves.
14663 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
14665 PR rtl-optimization/84068
14666 PR rtl-optimization/83459
14667 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
14669 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
14671 PR tree-optimization/84224
14672 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
14673 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
14674 non-zero arguments.
14676 2018-02-07 Iain Sandoe <iain@codesourcery.com>
14679 * config/rs6000/altivec.md (*restore_world): Remove LR use.
14680 * config/rs6000/predicates.md (restore_world_operation): Adjust op
14681 count, remove one USE.
14683 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
14685 * doc/install.texi (Configuration): Document the
14686 --with-long-double-format={ibm,ieee} PowerPC configuration
14690 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
14691 Convert from define_expand to be define_insn_and_split. Rework
14692 float/double/_Float128 conversions to QI/HI/SImode to work with
14693 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
14694 conversions to QI/HImode types did a store and then a load to
14695 truncate the value. For conversions to VSX registers, don't split
14696 the insn, instead emit the code directly. Use the code iterator
14697 any_fix to combine signed and unsigned conversions.
14698 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
14699 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
14700 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
14701 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
14702 (fix_<mode>di2_hw): Likewise.
14703 (fixuns_<mode>di2_hw): Likewise.
14704 (fix_<mode>si2_hw): Likewise.
14705 (fixuns_<mode>si2_hw): Likewise.
14706 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
14707 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
14708 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
14709 fix<uns>_trunc<SFDF:mode>si2_p8.
14710 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
14711 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
14712 (fix<uns>_<mode>_mem): Likewise.
14713 (fctiw<u>z_<mode>_mem): Likewise.
14714 (fix<uns>_<mode>_mem): Likewise.
14715 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
14716 the register allocator from doing a direct move to the GPRs to do
14717 a store, and instead use the ISA 3.0 store byte/half-word from
14718 vector register instruction. For IEEE 128-bit floating point,
14719 also optimize stores of 32-bit ints.
14720 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
14722 2018-02-07 Alan Hayward <alan.hayward@arm.com>
14724 * genextract.c (push_pathstr_operand): New function to support
14726 (walk_rtx): Call push_pathstr_operand.
14727 (print_path): Support [a-zA-Z].
14729 2018-02-07 Richard Biener <rguenther@suse.de>
14731 PR tree-optimization/84037
14732 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
14733 (cse_and_gimplify_to_preheader): Declare.
14734 (vect_get_place_in_interleaving_chain): Likewise.
14735 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14737 (_loop_vec_info::~_loop_vec_info): Delete it.
14738 (cse_and_gimplify_to_preheader): New function.
14739 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
14740 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
14741 (vectorizable_load): Likewise. For grouped stores always base
14742 the IV on the first element.
14743 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
14744 condition before gimplifying.
14746 2018-02-07 Jakub Jelinek <jakub@redhat.com>
14748 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
14749 *DIV_EXPR and *MOD_EXPR.
14751 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
14754 * config/i386/i386.c (ix86_option_override_internal): Mask out
14755 the CF_SET bit when checking -fcf-protection.
14757 2018-02-07 Tom de Vries <tom@codesourcery.com>
14760 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
14763 2018-02-07 Richard Biener <rguenther@suse.de>
14765 PR tree-optimization/84204
14766 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
14769 PR tree-optimization/84205
14770 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
14771 special-case isl_ast_op_zdiv_r.
14773 PR tree-optimization/84223
14774 * graphite-scop-detection.c (gather_bbs::before_dom_children):
14775 Only add conditions from within the region.
14776 (gather_bbs::after_dom_children): Adjust.
14778 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
14781 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
14782 * config/avr/avr.md: Only post-reload split REG-REG moves if
14783 either register is GENERAL_REG_P.
14785 2018-02-07 Jakub Jelinek <jakub@redhat.com>
14787 PR tree-optimization/84235
14788 * tree-ssa-scopedtables.c
14789 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
14790 if the subtraction is performed in floating point type where NaNs are
14791 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
14792 build 1. Formatting fix.
14794 2018-02-06 Jakub Jelinek <jakub@redhat.com>
14797 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
14798 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
14799 and skip it regardless of bb boundaries. Use CALL_P macro,
14800 don't test INSN_P (insn) together with CALL_P or JUMP_P check
14801 unnecessarily, formatting fix.
14803 2018-02-06 Michael Collison <michael.collison@arm.com>
14805 * config/arm/thumb2.md:
14806 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
14807 (*thumb_mov_notscc): Ditto.
14809 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
14812 * config/rs6000/rs6000.md (su code attribute): Use "u" for
14813 unsigned_fix, not "s".
14815 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14817 * configure.ac (gcc_fn_eh_frame_ro): New function.
14818 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
14819 correct .eh_frame permissions.
14820 * configure: Regenerate.
14822 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
14824 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
14825 irrelevant options.
14827 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14829 * config/rs6000/rs6000.c (rs6000_option_override_internal):
14830 Display warning message for -mno-speculate-indirect-jumps.
14832 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
14834 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
14836 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
14838 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
14840 PR tree-optimization/84225
14841 * tree-eh.c (find_trapping_overflow): Only call
14842 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
14844 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14847 * config/i386/i386.c: Reimplement the check of possible options
14848 -mibt/-mshstk conbination. Change error messages.
14849 * doc/invoke.texi: Fix a typo: remove extra '='.
14851 2018-02-06 Marek Polacek <polacek@redhat.com>
14853 PR tree-optimization/84228
14854 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
14856 2018-02-06 Tamar Christina <tamar.christina@arm.com>
14859 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
14860 emitted arch directives.
14861 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
14862 __ARM_FEATURE_COPROC before changing architectures.
14864 2018-02-06 Richard Biener <rguenther@suse.de>
14866 * config/i386/i386.c (print_reg): Fix typo.
14867 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
14869 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
14871 * configure: Regenerate.
14873 2018-02-05 Martin Sebor <msebor@redhat.com>
14875 PR tree-optimization/83369
14876 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
14879 2018-02-05 Martin Liska <mliska@suse.cz>
14881 * doc/invoke.texi: Cherry-pick upstream r323995.
14883 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
14885 * ira.c (ira_init_register_move_cost): Adjust comment.
14887 2018-02-05 Martin Liska <mliska@suse.cz>
14889 PR gcov-profile/84137
14890 * doc/gcov.texi: Fix typo in documentation.
14892 2018-02-05 Martin Liska <mliska@suse.cz>
14894 PR gcov-profile/83879
14895 * doc/gcov.texi: Document necessity of --dynamic-list-data when
14896 using dlopen functionality.
14898 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
14900 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
14901 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
14902 _mm_maskz_range_ss, _mm_mask_range_round_ss,
14903 _mm_maskz_range_round_ss): New intrinsics.
14904 (__builtin_ia32_rangesd128_round)
14905 (__builtin_ia32_rangess128_round): Remove.
14906 (__builtin_ia32_rangesd128_mask_round,
14907 __builtin_ia32_rangess128_mask_round): New builtins.
14908 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
14909 __builtin_ia32_rangess128_round): Remove.
14910 (__builtin_ia32_rangesd128_mask_round,
14911 __builtin_ia32_rangess128_mask_round): New builtins.
14912 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
14913 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
14914 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
14915 "<round_saeonly_constraint>")): Changed to ...
14916 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
14917 "<round_saeonly_scalar_constraint>")): ... this.
14918 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
14919 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
14920 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
14921 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
14922 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
14924 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
14926 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
14928 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
14929 Remove all values except native, 8540 and 8548.
14931 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
14933 * config/i386/i386.c (ix86_output_function_return): Pass
14934 INVALID_REGNUM, instead of -1, as invalid register number to
14935 indirect_thunk_name and output_indirect_thunk.
14937 2018-02-02 Julia Koval <julia.koval@intel.com>
14939 * config.gcc: Add -march=icelake.
14940 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
14941 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
14942 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
14943 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
14944 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
14945 (processor_target_table): Add icelake.
14946 (ix86_option_override_internal): Handle new PTAs.
14947 (get_builtin_code_for_version): Handle icelake.
14948 (M_INTEL_COREI7_ICELAKE): New.
14949 (fold_builtin_cpu): Handle icelake.
14950 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
14951 * doc/invoke.texi: Add -march=icelake.
14953 2018-02-02 Julia Koval <julia.koval@intel.com>
14955 * config/i386/i386.c (ix86_option_override_internal): Change flags type
14956 to wide_int_bitmask.
14957 * wide-int-bitmask.h: New.
14959 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14962 * config/i386/i386.md: Replace Pmode with word_mode in
14963 builtin_setjmp_setup and builtin_longjmp to support x32.
14965 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
14969 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
14971 (rs6000_supported_cpu_names): New static variable.
14972 (linux_cpu_translation_table): Likewise.
14973 (elf_platform) <cpu>: Define new static variable and use it.
14974 Translate kernel AT_PLATFORM name to canonical name if needed.
14975 Error if platform name is unknown.
14977 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
14980 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
14982 2018-02-01 Jeff Law <law@redhat.com>
14985 * config/i386/i386.c (release_scratch_register_on_entry): Add new
14986 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
14987 the scratch if RELEASE_VIA_POP is false.
14988 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
14989 If we have to save a temporary register, decrement SIZE appropriately.
14990 Pass new arguments to release_scratch_register_on_entry.
14991 (ix86_adjust_stack_and_probe): Likewise.
14992 (ix86_emit_probe_stack_range): Pass new arguments to
14993 release_scratch_register_on_entry.
14995 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
14997 PR rtl-optimization/84157
14998 * combine.c (change_zero_ext): Use REG_P predicate in
14999 front of HARD_REGISTER_P predicate.
15001 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
15003 * config/avr/avr.c (avr_option_override): Move disabling of
15004 -fdelete-null-pointer-checks to...
15005 * common/config/avr/avr-common.c (avr_option_optimization_table):
15008 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15010 PR tree-optimization/81635
15011 * tree-data-ref.c (split_constant_offset_1): For types that
15012 wrap on overflow, try to use range info to prove that wrapping
15015 2018-02-01 Renlin Li <renlin.li@arm.com>
15018 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
15019 TAILCALL_ADDR_REGS.
15020 (aarch64_register_move_cost): Likewise.
15021 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
15022 TAILCALL_ADDR_REGS.
15023 (REG_CLASS_NAMES): Likewise.
15024 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
15025 TAILCALL_ADDR_REGS. Remove IP registers.
15026 * config/aarch64/aarch64.md (Ucs): Update register constraint.
15028 2018-02-01 Richard Biener <rguenther@suse.de>
15030 * domwalk.h (dom_walker::dom_walker): Add additional constructor
15031 for specifying RPO order and allow NULL for that.
15032 * domwalk.c (dom_walker::dom_walker): Likewise.
15033 (dom_walker::walk): Handle NULL RPO order.
15034 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
15036 (rewrite_update_dom_walker): Likewise.
15037 (mark_def_dom_walker): Likewise.
15039 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15041 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
15042 (aarch64_maybe_expand_sve_subreg_move): Declare.
15043 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
15044 * config/aarch64/predicates.md (aarch64_any_register_operand): New
15046 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
15047 that are semantically a reverse operation.
15048 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
15049 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
15050 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
15052 (aarch64_can_change_mode_class): For big-endian, forbid changes
15053 between two SVE modes if they have different element sizes.
15055 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15057 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
15058 the TImode handling for big-endian targets.
15060 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15062 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
15063 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
15065 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
15066 Remove BSWAP handing for big-endian targets and use the form of
15067 LD1RQ appropariate for the mode.
15069 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15071 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
15072 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
15073 duplicated element.
15075 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15078 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
15079 check for operands that need to go through aarch64_sve_reload_be.
15081 2018-02-01 Jakub Jelinek <jakub@redhat.com>
15083 PR tree-optimization/81661
15084 PR tree-optimization/84117
15085 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
15086 * tree-eh.c: Include gimplify.h.
15087 (find_trapping_overflow, replace_trapping_overflow,
15088 rewrite_to_non_trapping_overflow): New functions.
15089 * tree-vect-loop.c: Include tree-eh.h.
15090 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
15091 * tree-data-ref.c: Include tree-eh.h.
15092 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
15094 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
15096 PR rtl-optimization/84123
15097 * combine.c (change_zero_ext): Check if hard register satisfies
15098 can_change_dest_mode before calling gen_lowpart_SUBREG.
15100 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
15103 * ira.c (ira_init_register_move_cost): Remove assert.
15105 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15107 PR rtl-optimization/84071
15108 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
15109 * doc/tm.texi: Regenerate.
15111 2018-01-31 Richard Biener <rguenther@suse.de>
15113 PR tree-optimization/84132
15114 * tree-data-ref.c (analyze_miv_subscript): Properly
15115 check whether evolution_function_is_affine_multivariate_p
15116 before calling gcd_of_steps_may_divide_p.
15118 2018-01-31 Julia Koval <julia.koval@intel.com>
15121 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
15122 * config/i386/i386.md (rdpid_rex64) New.
15123 (rdpid): Make 32bit only.
15125 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
15128 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
15129 an IDENTIFIER_NODE for FUNCTION_TYPE's.
15131 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15134 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15136 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15138 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15140 PR rtl-optimization/84071
15141 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
15142 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
15144 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
15146 * config/arc/arc.c (arc_handle_aux_attribute): New function.
15147 (arc_attribute_table): Add 'aux' attribute.
15148 (arc_in_small_data_p): Consider aux like variables.
15149 (arc_is_aux_reg_p): New function.
15150 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
15151 (arc_get_aux_arg): New function.
15152 (prepare_move_operands): Handle aux-register access.
15153 (arc_handle_aux_attribute): New function.
15154 * doc/extend.texi (ARC Variable attributes): Add subsection.
15156 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
15158 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
15159 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
15160 (arc_attribute_table): Add 'uncached' attribute.
15161 (arc_print_operand): Print '.di' flag for uncached memory
15163 (arc_in_small_data_p): Do not consider for small data the uncached
15165 (arc_is_uncached_mem_p): New function.
15166 * config/arc/predicates.md (compact_store_memory_operand): Check
15167 for uncached memory accesses.
15168 (nonvol_nonimm_operand): Likewise.
15169 * doc/extend.texi (ARC Type Attribute): New subsection.
15171 2018-01-31 Jakub Jelinek <jakub@redhat.com>
15174 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
15175 falign-loops=): Add Optimization flag.
15177 2018-01-30 Jeff Law <law@redhat.com>
15180 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
15181 INT_REGISTERS_SAVED. Check it prior to calling
15182 get_scratch_register_on_entry.
15183 (ix86_adjust_stack_and_probe): Similarly.
15184 (ix86_emit_probe_stack_range): Similarly.
15185 (ix86_expand_prologue): Corresponding changes.
15187 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15190 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
15191 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
15193 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
15196 * lra-constraints.c (curr_insn_transform): Process AND in the
15199 2018-01-30 Jakub Jelinek <jakub@redhat.com>
15201 PR rtl-optimization/83986
15202 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
15203 dependence against last_pending_memory_flush in addition to
15204 pending_jump_insns.
15206 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
15208 PR tree-optimization/81611
15209 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
15212 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15215 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
15218 2018-01-30 Richard Biener <rguenther@suse.de>
15219 Jakub Jelinek <jakub@redhat.com>
15221 PR tree-optimization/84111
15222 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
15223 inner loops added during recursion, as they don't have up-to-date
15226 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
15229 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
15230 (can_inline_edge_by_limits_p): ... here.
15231 (can_early_inline_edge_p, check_callers,
15232 update_caller_keys, update_callee_keys, recursive_inlining,
15233 add_new_edges_to_heap, speculation_useful_p,
15234 inline_small_functions,
15235 inline_small_functions, flatten_function,
15236 inline_to_all_callers_1): Update.
15238 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
15240 * profile-count.c (profile_count::combine_with_ipa_count): Handle
15243 2018-01-30 Richard Biener <rguenther@suse.de>
15245 PR tree-optimization/83008
15246 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
15247 invariant and constant vector uses in stmts when they need
15248 more than one stmt.
15250 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15253 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
15254 * configure: Regenerate.
15256 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
15258 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
15260 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
15261 Use gen_rtx_REG rather than gen_lowpart.
15263 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
15265 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
15266 rather than 0 when creating partial subregs.
15268 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
15270 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
15273 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
15276 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
15277 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
15278 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
15279 flags. This restores the settings used before the 2017-07-24.
15280 Turning off pre increment/decrement/modify allows IVOPTS to
15281 optimize DF/SF loops where the index is an int.
15283 2018-01-29 Richard Biener <rguenther@suse.de>
15284 Kelvin Nilsen <kelvin@gcc.gnu.org>
15287 * tree-vect-stmts.c (vectorizable_call): Don't call
15288 targetm.vectorize_builtin_md_vectorized_function if callee is
15291 2018-01-22 Carl Love <cel@us.ibm.com>
15293 * doc/extend.tex: Fix typo in second arg in
15294 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
15296 2018-01-29 Richard Biener <rguenther@suse.de>
15298 PR tree-optimization/84086
15299 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
15300 (flush_ssaname_freelist): When SSA names were released reset
15301 the SCEV hash table.
15303 2018-01-29 Richard Biener <rguenther@suse.de>
15305 PR tree-optimization/84057
15306 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
15307 removed paths when removing edges.
15309 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
15311 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
15312 -mfunction-return=@var{choice}.
15314 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
15316 PR diagnostic/84034
15317 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
15318 Handle CR like TAB.
15319 (layout::print_source_line): Likewise.
15320 (test_get_line_width_without_trailing_whitespace): Add test cases.
15322 2018-01-27 Jakub Jelinek <jakub@redhat.com>
15324 PR middle-end/84040
15325 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
15328 2018-01-26 Jim Wilson <jimw@sifive.com>
15330 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
15332 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
15335 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15337 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
15338 and CMP + SUB-immediate -> SUBS.
15340 2018-01-26 Martin Sebor <msebor@redhat.com>
15342 PR tree-optimization/83896
15343 * tree-ssa-strlen.c (get_string_len): Rename...
15344 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
15345 Avoid assuming length is constant.
15346 (handle_char_store): Use HOST_WIDE_INT for string length.
15348 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
15351 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
15352 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
15354 2018-01-26 Richard Biener <rguenther@suse.de>
15356 PR rtl-optimization/84003
15357 * dse.c (record_store): Only record redundant stores when
15358 the earlier store aliases at least all accesses the later one does.
15360 2018-01-26 Jakub Jelinek <jakub@redhat.com>
15362 PR rtl-optimization/83985
15363 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
15364 REG_CFA_RESTORE insns.
15365 (delete_unmarked_insns): Don't ignore separate shrink wrapping
15366 REG_CFA_RESTORE insns here.
15369 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
15370 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
15372 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15374 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
15375 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
15376 (arc_init): Likewise.
15377 (arc_override_options): Likewise.
15378 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
15380 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
15382 * config/arc/arc.h (TARGET_DBNZ): Define.
15383 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
15384 properly set the tune attribute.
15385 (dbnz): Use TARGET_DBNZ guard.
15386 * config/arc/arc.opt (mtune): Add core3 option.
15388 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15390 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
15391 recognize new pic like addresses.
15392 (arc_delegitimize_address): Clean up.
15394 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15396 * config/arc/arc-arches.def: Option mrf16 valid for all
15398 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
15399 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
15400 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
15401 * config/arc/arc-tables.opt: Regenerate.
15402 * config/arc/arc.c (arc_conditional_register_usage): Handle
15403 reduced register file case.
15404 (arc_file_start): Set must have build attributes.
15405 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
15406 mrf16 option value.
15407 * config/arc/arc.opt (mrf16): Add new option.
15408 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
15409 * config/arc/genmultilib.awk: Handle new mrf16 option.
15410 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
15411 * config/arc/t-multilib: Regenerate.
15412 * doc/invoke.texi (ARC Options): Document mrf16 option.
15414 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15416 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
15417 * config/arc/arc.c (arc_handle_secure_attribute): New function.
15418 (arc_attribute_table): Add 'secure_call' attribute.
15419 (arc_print_operand): Print secure call operand.
15420 (arc_function_ok_for_sibcall): Don't optimize tail calls when
15422 (arc_is_secure_call_p): New function. * config/arc/arc.md
15423 (call_i): Add support for sjli instruction.
15424 (call_value_i): Likewise.
15425 * config/arc/constraints.md (Csc): New constraint.
15427 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15428 John Eric Martin <John.Martin@emmicro-us.com>
15430 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
15431 * config/arc/arc.c (_arc_jli_section): New struct.
15432 (arc_jli_section): New type.
15433 (rc_jli_sections): New static variable.
15434 (arc_handle_jli_attribute): New function.
15435 (arc_attribute_table): Add jli_always and jli_fixed attribute.
15436 (arc_file_end): New function.
15437 (TARGET_ASM_FILE_END): Define.
15438 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
15439 (arc_add_jli_section): New function.
15440 (jli_call_scan): Likewise.
15441 (arc_reorg): Call jli_call_scan.
15442 (arc_output_addsi): Remove 'S' from printing asm operand.
15443 (arc_is_jli_call_p): New function.
15444 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
15446 (movhi_insn): Likewise.
15447 (movsi_insn): Likewise.
15448 (movsi_set_cc_insn): Likewise.
15449 (loadqi_update): Likewise.
15450 (load_zeroextendqisi_update): Likewise.
15451 (load_signextendqisi_update): Likewise.
15452 (loadhi_update): Likewise.
15453 (load_zeroextendhisi_update): Likewise.
15454 (load_signextendhisi_update): Likewise.
15455 (loadsi_update): Likewise.
15456 (loadsf_update): Likewise.
15457 (movsicc_insn): Likewise.
15458 (bset_insn): Likewise.
15459 (bxor_insn): Likewise.
15460 (bclr_insn): Likewise.
15461 (bmsk_insn): Likewise.
15462 (bicsi3_insn): Likewise.
15463 (cmpsi_cc_c_insn): Likewise.
15464 (movsi_ne): Likewise.
15465 (movsi_cond_exec): Likewise.
15466 (clrsbsi2): Likewise.
15467 (norm_f): Likewise.
15477 (call_i): Remove 'S' asm letter, add jli instruction.
15478 (call_value_i): Likewise.
15479 * config/arc/arc.op (mjli-always): New option.
15480 * config/arc/constraints.md (Cji): New constraint.
15481 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
15483 (subsf3_fpx): Likewise.
15484 (mulsf3_fpx): Likewise.
15485 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
15487 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
15488 function attrbutes.
15489 * doc/invoke.texi (ARC): Document mjli-always option.
15491 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
15493 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
15494 avoid addition with 0 and use incw and decw where possible.
15496 2018-01-26 Richard Biener <rguenther@suse.de>
15498 PR tree-optimization/81082
15499 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
15500 association if it requires casting to unsigned.
15501 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
15502 from fold_plusminus_mult_expr to catch important cases late when
15503 range info is available.
15505 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15507 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
15508 * configure.ac (hidden_linkonce): New test.
15509 * configure: Regenerate.
15510 * config.in: Regenerate.
15512 2018-01-26 Julia Koval <julia.koval@intel.com>
15514 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
15515 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
15516 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
15517 _mm_mask_bitshuffle_epi64_mask): Fix type.
15518 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
15519 USI_FTYPE_V4DI_V4DI_USI): Remove.
15520 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
15521 __builtin_ia32_vpshufbitqmb256_mask,
15522 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
15523 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
15524 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
15526 2018-01-26 Alan Modra <amodra@gmail.com>
15529 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
15530 UNSPEC_VBPERMQ. Sort other unspecs.
15532 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
15534 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
15536 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
15538 PR middle-end/83055
15539 * predict.c (drop_profile): Do not push/pop cfun; update also
15541 (handle_missing_profiles): Fix logic looking for zero profiles.
15543 2018-01-25 Jakub Jelinek <jakub@redhat.com>
15545 PR middle-end/83977
15546 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
15547 on functions with #pragma omp declare simd or functions with simd
15549 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
15550 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
15551 Remove trailing \n from warning_at calls.
15553 2018-01-25 Tom de Vries <tom@codesourcery.com>
15556 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
15557 for neutered workers.
15559 2018-01-24 Joseph Myers <joseph@codesourcery.com>
15562 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
15563 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
15565 2018-01-24 Jeff Law <law@redhat.com>
15568 * i386.c (get_probe_interval): Move to earlier point.
15569 (ix86_compute_frame_layout): If -fstack-clash-protection and
15570 the frame is larger than the probe interval, then use pushes
15571 to save registers rather than reg->mem moves.
15572 (ix86_expand_prologue): Remove conditional for int_registers_saved
15575 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
15578 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
15579 min/max for never referenced object.
15581 2018-01-24 Jakub Jelinek <jakub@redhat.com>
15583 PR middle-end/83977
15584 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
15586 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
15587 attributes from DECL_ATTRIBUTES (decl) without affecting
15588 DECL_ATTRIBUTES (current_function_decl).
15589 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
15590 functions with non-NULL DECL_ABSTRACT_ORIGIN.
15592 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
15594 PR tree-optimization/83979
15595 * fold-const.c (fold_comparison): Use constant_boolean_node
15596 instead of boolean_{true,false}_node.
15598 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
15600 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
15603 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15605 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15606 Simplify the clause that sets the length attribute.
15607 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15608 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
15609 clause that sets the length attribute.
15610 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15612 2018-01-24 Tom de Vries <tom@codesourcery.com>
15615 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
15616 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
15617 Add strict parameter.
15618 (prevent_branch_around_nothing): Insert dummy insn between branch to
15619 label and label with no ptx insn inbetween.
15620 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
15622 2018-01-24 Tom de Vries <tom@codesourcery.com>
15625 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
15626 for neutered threads in warp.
15627 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
15629 2018-01-24 Richard Biener <rguenther@suse.de>
15631 PR tree-optimization/83176
15632 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
15635 2018-01-24 Richard Biener <rguenther@suse.de>
15637 PR tree-optimization/82819
15638 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
15639 code generating pluses that are no-ops in the target precision.
15641 2018-01-24 Richard Biener <rguenther@suse.de>
15643 PR middle-end/84000
15644 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
15646 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15648 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
15649 to merge probabilities.
15650 * predict.c (probably_never_executed): Also mark as cold functions
15651 with global 0 profile and guessed local profile.
15652 * profile-count.c (profile_probability::combine_with_count): New
15654 * profile-count.h (profile_probability::operator*,
15655 profile_probability::operator*=, profile_probability::operator/,
15656 profile_probability::operator/=): Reduce precision to adjusted
15657 and set value to guessed on contradictory divisions.
15658 (profile_probability::combine_with_freq): Remove.
15659 (profile_probability::combine_wiht_count): Declare.
15660 (profile_count::force_nonzero):: Set to adjusted.
15661 (profile_count::probability_in):: Set quality to adjusted.
15662 * tree-ssa-tail-merge.c (replace_block_by): Use
15663 combine_with_count.
15665 2018-01-23 Andrew Waterman <andrew@sifive.com>
15666 Jim Wilson <jimw@sifive.com>
15668 * config/riscv/riscv.c (riscv_stack_boundary): New.
15669 (riscv_option_override): Set riscv_stack_boundary. Handle
15670 riscv_preferred_stack_boundary_arg.
15671 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
15672 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
15673 (STACK_BOUNDARY): Set to riscv_stack_boundary.
15674 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
15675 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
15676 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
15678 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
15681 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
15682 of struct ix86_frame.
15683 (ix86_expand_epilogue): Likewise. Add a local variable for
15684 the reg_save_offset field in struct ix86_frame.
15686 2018-01-23 Bin Cheng <bin.cheng@arm.com>
15688 PR tree-optimization/82604
15689 * tree-loop-distribution.c (enum partition_kind): New enum item
15690 PKIND_PARTIAL_MEMSET.
15691 (partition_builtin_p): Support above new enum item.
15692 (generate_code_for_partition): Ditto.
15693 (compute_access_range): Differentiate cases that equality can be
15694 proven at all loops, the innermost loops or no loops.
15695 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
15696 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
15697 (finalize_partitions, distribute_loop): Don't fuse partition of
15698 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
15699 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
15700 parloop is enabled.
15702 2018-01-23 Martin Liska <mliska@suse.cz>
15704 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
15705 order to ignore the predictor.
15706 (PRED_POLYMORPHIC_CALL): Likewise.
15707 (PRED_RECURSIVE_CALL): Likewise.
15709 2018-01-23 Martin Liska <mliska@suse.cz>
15711 * tree-profile.c (tree_profiling): Print function header to
15712 aware reader which function we are working on.
15713 * value-prof.c (gimple_find_values_to_profile): Do not print
15714 not interesting value histograms.
15716 2018-01-23 Martin Liska <mliska@suse.cz>
15718 * profile-count.h (enum profile_quality): Add
15719 profile_uninitialized as the first value. Do not number values
15720 as they are zero based.
15721 (profile_count::verify): Update sanity check.
15722 (profile_probability::verify): Likewise.
15724 2018-01-23 Nathan Sidwell <nathan@acm.org>
15726 * doc/invoke.texi (ffor-scope): Deprecate.
15728 2018-01-23 David Malcolm <dmalcolm@redhat.com>
15730 PR tree-optimization/83510
15731 * domwalk.c (set_all_edges_as_executable): New function.
15732 (dom_walker::dom_walker): Convert bool param
15733 "skip_unreachable_blocks" to enum reachability. Move setup of
15734 edge flags to set_all_edges_as_executable and only do it when
15735 reachability is REACHABLE_BLOCKS.
15736 * domwalk.h (enum dom_walker::reachability): New enum.
15737 (dom_walker::dom_walker): Convert bool param
15738 "skip_unreachable_blocks" to enum reachability.
15739 (set_all_edges_as_executable): New decl.
15740 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
15741 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
15743 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
15744 but converting true to REACHABLE_BLOCKS.
15745 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
15747 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
15748 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
15749 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
15751 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
15752 if check_all_array_refs will be called.
15754 2018-01-23 David Malcolm <dmalcolm@redhat.com>
15756 * tree.c (selftest::test_location_wrappers): Add more test
15759 2018-01-23 David Malcolm <dmalcolm@redhat.com>
15761 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
15762 (selftest::test_bit_in_range): Likewise.
15764 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
15767 * doc/sourcebuild.texi (vect_float): Say that the selector
15768 only describes the situation when -funsafe-math-optimizations is on.
15769 (vect_float_strict): Document.
15771 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
15773 PR tree-optimization/83965
15774 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
15775 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
15776 instead of checking only for a reduction.
15777 (vect_recog_widen_sum_pattern): Likewise.
15779 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15781 * predict.c (probably_never_executed): Only use precise profile info.
15782 (compute_function_frequency): Skip after inlining hack since we now
15783 have quality checking.
15785 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
15787 * profile-count.h (profile_probability::very_unlikely,
15788 profile_probability::unlikely, profile_probability::even): Set
15789 precision to guessed.
15791 2018-01-23 Richard Biener <rguenther@suse.de>
15793 PR tree-optimization/83963
15794 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
15795 Properly terminate dominator walk when crossing the exit edge not
15796 when visiting its source block.
15798 2018-01-23 Jakub Jelinek <jakub@redhat.com>
15801 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
15802 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
15804 2018-01-22 Jakub Jelinek <jakub@redhat.com>
15806 PR tree-optimization/83957
15807 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
15808 semicolon after for body surrounded by braces.
15810 PR tree-optimization/83081
15811 * profile-count.h (profile_probability::split): New method.
15812 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
15813 Use profile_probability::split.
15814 (do_compare_rtx_and_jump): Fix adjustment of probabilities
15815 when splitting a single conditional jump into 2.
15817 2018-01-22 David Malcolm <dmalcolm@redhat.com>
15819 PR tree-optimization/69452
15820 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
15823 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15825 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
15826 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
15827 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
15829 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15831 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
15833 * config/rl78/rl78.md (movdi): New define_expand.
15834 * config/rl78/rl78.c (rl78_split_movdi): New function.
15836 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
15839 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
15841 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
15842 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
15843 128-bit to produce an UNSPEC move to get the double word with the
15844 signbit and then a shift directly to do signbit.
15845 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
15846 implementation with a new version that just does either a direct
15847 move or a regular move. Move memory interface to separate insns.
15848 Move insns so they are next to the expander.
15849 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
15850 with signbit move. Split big and little endian case.
15851 (signbit<mode>2_dm_mem_le): Likewise.
15852 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
15853 (signbit<mode>2_dm2): Likewise.
15855 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15857 * config/rl78/rl78.md (anddi3): New define_expand.
15859 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15861 * config/rl78/rl78.md (umindi3): New define_expand.
15863 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15865 * config/rl78/rl78.md (smindi3): New define_expand.
15867 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15869 * config/rl78/rl78.md (smaxdi3): New define_expand.
15871 2018-01-22 Carl Love <cel@us.ibm.com>
15873 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
15874 LVX_V1TI): Add macro expansion.
15875 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
15876 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
15877 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
15878 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
15879 Change check to determine if the instruction is a byte reversing
15880 entry. Fix typo in comment.
15881 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
15882 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
15883 Add def_builtin calls for new builtins.
15884 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
15885 Add define_insn expansion.
15887 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15889 * config/rl78/rl78.md (umaxdi3): New define_expand.
15891 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15893 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
15894 for non-QImode registers.
15896 2018-01-22 Richard Biener <rguenther@suse.de>
15898 PR tree-optimization/83963
15899 * graphite-scop-detection.c (scop_detection::get_sese): Delay
15900 including the loop exit block.
15901 (scop_detection::merge_sese): Likewise.
15902 (scop_detection::add_scop): Do it here instead.
15904 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15906 * doc/sourcebuild.texi (arm_softfloat): Document.
15908 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
15911 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
15912 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
15913 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
15915 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15916 David Edelsohn <dje.gcc@gmail.com>
15919 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15920 Change "crset eq" to "crset 2".
15921 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15922 (*call_indirect_aix<mode>_nospec): Likewise.
15923 (*call_value_indirect_aix<mode>_nospec): Likewise.
15924 (*call_indirect_elfv2<mode>_nospec): Likewise.
15925 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
15926 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
15927 change assembly output from . to $.
15928 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15929 (indirect_jump<mode>_nospec): Change assembly output from . to $.
15930 (*tablejump<mode>_internal1_nospec): Likewise.
15932 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
15935 * config/sh/sh_optimize_sett_clrt.cc:
15936 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
15938 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
15940 PR tree-optimization/83940
15941 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
15942 offset_dt to vect_constant_def rather than vect_unknown_def_type.
15943 (vect_check_load_store_mask): Add a mask_dt_out parameter and
15944 use it to pass back the definition type.
15945 (vect_check_store_rhs): Likewise rhs_dt_out.
15946 (vect_build_gather_load_calls): Add a mask_dt argument and use
15947 it instead of a call to vect_is_simple_use.
15948 (vectorizable_store): Update calls to vect_check_load_store_mask
15949 and vect_check_store_rhs. Use the dt returned by the latter instead
15950 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
15951 instead of calls to vect_is_simple_use. Pass the scalar rather
15952 than the vector operand to vect_is_simple_use when handling
15953 second and subsequent copies of an rhs value.
15954 (vectorizable_load): Update calls to vect_check_load_store_mask
15955 and vect_build_gather_load_calls. Use the cached mask_dt and
15956 gs_info.offset_dt instead of calls to vect_is_simple_use.
15958 2018-01-20 Jakub Jelinek <jakub@redhat.com>
15960 PR middle-end/83945
15961 * tree-emutls.c: Include gimplify.h.
15962 (lower_emutls_2): New function.
15963 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
15964 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
15965 it before further processing.
15968 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
15969 UINTVAL (trueop1) instead of INTVAL (op1).
15971 2018-01-19 Jakub Jelinek <jakub@redhat.com>
15975 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
15976 INCOMING_FRAME_SP_OFFSET if not defined.
15977 (scan_trace): Add ENTRY argument. If true and
15978 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
15979 emit a note to adjust the CFA offset.
15980 (create_cfi_notes): Adjust scan_trace callers.
15981 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
15982 INCOMING_FRAME_SP_OFFSET in the CIE.
15983 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
15984 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
15986 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
15987 * doc/tm.texi: Regenerated.
15989 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15991 PR rtl-optimization/83147
15992 * lra-constraints.c (remove_inheritance_pseudos): Use
15993 lra_substitute_pseudo_within_insn.
15995 2018-01-19 Tom de Vries <tom@codesourcery.com>
15996 Cesar Philippidis <cesar@codesourcery.com>
15999 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
16001 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
16004 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
16005 spaces for function labels.
16007 2018-01-19 Martin Liska <mliska@suse.cz>
16009 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
16010 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
16011 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
16012 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
16013 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
16014 (PRED_CONST_RETURN): Change from 69 to 65.
16015 (PRED_NULL_RETURN): Change from 91 to 71.
16016 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
16017 (PRED_LOOP_GUARD): Change from 66 to 73.
16019 2018-01-19 Martin Liska <mliska@suse.cz>
16021 * predict.c (predict_insn_def): Add new assert.
16022 (struct branch_predictor): Change type to signed integer.
16023 (test_prediction_value_range): Amend test to cover
16024 PROB_UNINITIALIZED.
16025 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
16026 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
16027 (PRED_LOOP_ITERATIONS_MAX): Likewise.
16028 (PRED_LOOP_IV_COMPARE): Likewise.
16029 * predict.h (PROB_UNINITIALIZED): Define new constant.
16031 2018-01-19 Martin Liska <mliska@suse.cz>
16033 * predict.c (dump_prediction): Add new format for
16034 analyze_brprob.py script which is enabled with -details
16036 * profile-count.h (precise_p): New function.
16038 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
16040 PR tree-optimization/83922
16041 * tree-vect-loop.c (vect_verify_full_masking): Return false if
16042 there are no statements that need masking.
16043 (vect_active_double_reduction_p): New function.
16044 (vect_analyze_loop_operations): Use it when handling phis that
16045 are not in the loop header.
16047 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
16049 PR tree-optimization/83914
16050 * tree-vect-loop.c (vectorizable_induction): Don't convert
16051 init_expr or apply the peeling adjustment for inductions
16052 that are nested within the vectorized loop.
16054 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16056 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
16059 2018-01-18 Jakub Jelinek <jakub@redhat.com>
16063 * function.h (gimplify_parameters): Add gimple_seq * argument.
16064 * function.c: Include gimple.h and options.h.
16065 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
16066 for the added local temporaries if needed.
16067 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
16068 if there are any parameter cleanups, wrap whole body into a
16069 try/finally with the cleanups.
16071 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
16074 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
16075 Use GET_MODE_CLASS for scalar floating point.
16077 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
16081 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
16082 Fix call of call_cgraph_insertion_hooks.
16084 2018-01-18 Martin Sebor <msebor@redhat.com>
16086 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
16088 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
16091 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
16094 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
16097 * common.opt: (-ffile-prefix-map): New option.
16098 * opts.c (common_handle_option): Defer it.
16099 * opts-global.c (handle_common_deferred_options): Handle it.
16100 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
16101 * file-prefix-map.h: New file.
16102 (remap_debug_filename, add_debug_prefix_map): ...here.
16103 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
16104 * final.c (debug_prefix_map, add_debug_prefix_map
16105 remap_debug_filename): Move to...
16106 * file-prefix-map.c: New file.
16107 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
16108 generalize, get rid of alloca(), use strrchr() instead of strchr().
16109 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
16110 Implement in terms of add_prefix_map().
16111 (remap_macro_filename, remap_debug_filename): Implement in term of
16113 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
16114 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
16115 * dbxout.c: Include file-prefix-map.h.
16116 * varasm.c: Likewise.
16117 * vmsdbgout.c: Likewise.
16118 * xcoffout.c: Likewise.
16119 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
16120 * doc/cppopts.texi (-fmacro-prefix-map): Document.
16121 * doc/invoke.texi (-ffile-prefix-map): Document.
16122 (-fdebug-prefix-map): Update description.
16124 2018-01-18 Martin Liska <mliska@suse.cz>
16126 * config/i386/i386.c (indirect_thunk_name): Document that also
16128 (output_indirect_thunk): Document why both instructions
16129 (pause and lfence) are generated.
16131 2018-01-18 Richard Biener <rguenther@suse.de>
16133 PR tree-optimization/83887
16134 * graphite-scop-detection.c
16135 (scop_detection::get_nearest_dom_with_single_entry): Remove.
16136 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
16137 (scop_detection::merge_sese): Re-implement with a flood-fill
16138 algorithm that properly finds a SESE region if it exists.
16140 2018-01-18 Jakub Jelinek <jakub@redhat.com>
16143 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
16144 pointer_diff optimizations use view_convert instead of convert.
16146 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16148 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
16149 Generate different code for -mno-speculate-indirect-jumps.
16150 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
16151 (*call_indirect_aix<mode>): Disable for
16152 -mno-speculate-indirect-jumps.
16153 (*call_indirect_aix<mode>_nospec): New define_insn.
16154 (*call_value_indirect_aix<mode>): Disable for
16155 -mno-speculate-indirect-jumps.
16156 (*call_value_indirect_aix<mode>_nospec): New define_insn.
16157 (*sibcall_nonlocal_sysv<mode>): Generate different code for
16158 -mno-speculate-indirect-jumps.
16159 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
16161 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
16163 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
16164 long double type, set the flags for noting the default long double
16165 type, even if we don't pass or return a long double type.
16167 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
16170 * ipa-inline.c (flatten_function): Do not overwrite final inlining
16173 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
16175 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
16176 support for merge[hl].
16177 (fold_mergehl_helper): New helper function.
16178 (tree-vector-builder.h): New #include for tree_vector_builder usage.
16179 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
16180 (altivec_vmrglw_direct): Add xxmrglw insn.
16182 2018-01-17 Andrew Waterman <andrew@sifive.com>
16184 * config/riscv/riscv.c (riscv_conditional_register_usage): If
16185 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
16187 2018-01-17 David Malcolm <dmalcolm@redhat.com>
16190 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
16191 call the lto_location_cache before reading the
16192 DECL_SOURCE_LOCATION of the types.
16194 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
16195 Richard Sandiford <richard.sandiford@linaro.org>
16197 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
16198 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
16199 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
16200 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
16201 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
16203 * config/aarch64/constraints.md (aarch64_movti_operand):
16205 * config/aarch64/predicates.md (Uti): Add new constraint.
16207 2018-01-17 Carl Love <cel@us.ibm.com>
16209 * config/rs6000/vsx.md (define_expand xl_len_r,
16210 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
16211 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
16213 (define_expand, define_insn): Move the shift left from the
16214 define_insn to the define_expand for lxvl and stxvl instructions.
16215 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
16216 and XL_LEN_R definitions to PURE.
16218 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
16220 * config/i386/i386.c (indirect_thunk_name): Declare regno
16221 as unsigned int. Compare regno with INVALID_REGNUM.
16222 (output_indirect_thunk): Ditto.
16223 (output_indirect_thunk_function): Ditto.
16224 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
16225 in the call to output_indirect_thunk_function.
16227 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
16229 PR middle-end/83884
16230 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
16231 rather than the size of inner_type to determine the stack slot size
16232 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
16234 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
16237 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
16240 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
16242 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
16243 endian Linux systems to optionally enable multilibs for selecting
16244 the long double type if the user configured an explicit type.
16245 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
16246 have no long double multilibs if not defined.
16247 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
16248 warn if the user used -mabi={ieee,ibm}longdouble and we built
16249 multilibs for long double.
16250 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
16251 appropriate multilib option.
16252 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
16254 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
16255 for building long double multilibs.
16256 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
16258 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
16260 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
16263 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
16265 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
16268 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
16271 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
16274 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
16276 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
16277 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
16279 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
16281 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
16282 different rtl trees depending on TARGET_64BIT.
16283 (rs6000_gen_lvx): Likewise.
16285 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
16287 * config/visium/visium.md (nop): Tweak comment.
16288 (hazard_nop): Likewise.
16290 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16292 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
16293 -mspeculate-indirect-jumps.
16294 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
16295 for -mno-speculate-indirect-jumps.
16296 (*call_indirect_elfv2<mode>_nospec): New define_insn.
16297 (*call_value_indirect_elfv2<mode>): Disable for
16298 -mno-speculate-indirect-jumps.
16299 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
16300 (indirect_jump): Emit different RTL for
16301 -mno-speculate-indirect-jumps.
16302 (*indirect_jump<mode>): Disable for
16303 -mno-speculate-indirect-jumps.
16304 (*indirect_jump<mode>_nospec): New define_insn.
16305 (tablejump): Emit different RTL for
16306 -mno-speculate-indirect-jumps.
16307 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
16308 (tablejumpsi_nospec): New define_expand.
16309 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
16310 (tablejumpdi_nospec): New define_expand.
16311 (*tablejump<mode>_internal1): Disable for
16312 -mno-speculate-indirect-jumps.
16313 (*tablejump<mode>_internal1_nospec): New define_insn.
16314 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
16317 2018-01-16 Artyom Skrobov tyomitch@gmail.com
16319 * caller-save.c (insert_save): Drop unnecessary parameter. All
16322 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16323 Richard Biener <rguenth@suse.de>
16326 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
16327 return early, inline manually is_gimple_sizepos. Make sure if we
16328 call gimplify_expr we don't end up with a gimple constant.
16329 * tree.c (variably_modified_type_p): Don't return true for
16330 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
16331 * gimplify.h (is_gimple_sizepos): Remove.
16333 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16335 PR tree-optimization/83857
16336 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
16337 vectorizable_live_operation for pure SLP statements.
16338 (vectorizable_live_operation): Handle PHIs.
16340 2018-01-16 Richard Biener <rguenther@suse.de>
16342 PR tree-optimization/83867
16343 * tree-vect-stmts.c (vect_transform_stmt): Precompute
16344 nested_in_vect_loop_p since the scalar stmt may get invalidated.
16346 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16349 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
16350 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
16351 If off is not INTEGER_CST, issue a may not be aligned warning
16352 rather than isn't aligned. Use isn%'t rather than isn't.
16353 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
16355 <case MULT_EXPR>: Improve the case when bottom and one of the
16356 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
16357 operand, in that case check if the other operand is multiple of
16358 bottom divided by the INTEGER_CST operand.
16360 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16363 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
16364 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
16365 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
16366 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
16367 * config/pa/pa.c (pa_function_arg_advance): Likewise.
16368 (pa_function_arg, pa_arg_partial_bytes): Likewise.
16369 (pa_function_arg_size): New function.
16371 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16373 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
16374 in a separate statement.
16376 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16378 PR tree-optimization/83847
16379 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
16380 group gathers and scatters.
16382 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16384 PR rtl-optimization/86620
16385 * params.def (max-sched-ready-insns): Bump minimum value to 1.
16387 PR rtl-optimization/83213
16388 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
16389 to last if both are JUMP_INSNs.
16391 PR tree-optimization/83843
16392 * gimple-ssa-store-merging.c
16393 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
16394 store_immediate_info for bswap/nop orig_stores.
16396 2018-01-15 Andrew Waterman <andrew@sifive.com>
16398 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
16400 <UDIV>: Increase cost if !TARGET_DIV.
16402 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
16404 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
16405 (define_attr "cr_logical_3op"): New.
16406 (cceq_ior_compare): Adjust.
16407 (cceq_ior_compare_complement): Adjust.
16408 (*cceq_rev_compare): Adjust.
16409 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
16410 (is_cracked_insn): Adjust.
16411 (insn_must_be_first_in_group): Adjust.
16412 * config/rs6000/40x.md: Adjust.
16413 * config/rs6000/440.md: Adjust.
16414 * config/rs6000/476.md: Adjust.
16415 * config/rs6000/601.md: Adjust.
16416 * config/rs6000/603.md: Adjust.
16417 * config/rs6000/6xx.md: Adjust.
16418 * config/rs6000/7450.md: Adjust.
16419 * config/rs6000/7xx.md: Adjust.
16420 * config/rs6000/8540.md: Adjust.
16421 * config/rs6000/cell.md: Adjust.
16422 * config/rs6000/e300c2c3.md: Adjust.
16423 * config/rs6000/e500mc.md: Adjust.
16424 * config/rs6000/e500mc64.md: Adjust.
16425 * config/rs6000/e5500.md: Adjust.
16426 * config/rs6000/e6500.md: Adjust.
16427 * config/rs6000/mpc.md: Adjust.
16428 * config/rs6000/power4.md: Adjust.
16429 * config/rs6000/power5.md: Adjust.
16430 * config/rs6000/power6.md: Adjust.
16431 * config/rs6000/power7.md: Adjust.
16432 * config/rs6000/power8.md: Adjust.
16433 * config/rs6000/power9.md: Adjust.
16434 * config/rs6000/rs64.md: Adjust.
16435 * config/rs6000/titan.md: Adjust.
16437 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16439 * config/i386/predicates.md (indirect_branch_operand): Rewrite
16440 ix86_indirect_branch_register logic.
16442 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16444 * config/i386/constraints.md (Bs): Update
16445 ix86_indirect_branch_register check. Don't check
16446 ix86_indirect_branch_register with GOT_memory_operand.
16448 * config/i386/predicates.md (GOT_memory_operand): Don't check
16449 ix86_indirect_branch_register here.
16450 (GOT32_symbol_operand): Likewise.
16452 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16454 * config/i386/predicates.md (constant_call_address_operand):
16455 Rewrite ix86_indirect_branch_register logic.
16456 (sibcall_insn_operand): Likewise.
16458 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16460 * config/i386/constraints.md (Bs): Replace
16461 ix86_indirect_branch_thunk_register with
16462 ix86_indirect_branch_register.
16464 * config/i386/i386.md (indirect_jump): Likewise.
16465 (tablejump): Likewise.
16466 (*sibcall_memory): Likewise.
16467 (*sibcall_value_memory): Likewise.
16468 Peepholes of indirect call and jump via memory: Likewise.
16469 * config/i386/i386.opt: Likewise.
16470 * config/i386/predicates.md (indirect_branch_operand): Likewise.
16471 (GOT_memory_operand): Likewise.
16472 (call_insn_operand): Likewise.
16473 (sibcall_insn_operand): Likewise.
16474 (GOT32_symbol_operand): Likewise.
16476 2018-01-15 Jakub Jelinek <jakub@redhat.com>
16478 PR middle-end/83837
16479 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
16480 type rather than type addr's type points to.
16481 (expand_omp_atomic_mutex): Likewise.
16482 (expand_omp_atomic): Likewise.
16484 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16487 * config/i386/i386.c (output_indirect_thunk_function): Use
16488 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
16489 for __x86_return_thunk.
16491 2018-01-15 Richard Biener <rguenther@suse.de>
16493 PR middle-end/83850
16494 * expmed.c (extract_bit_field_1): Fix typo.
16496 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16499 * config/arm/iterators.md (VF): New mode iterator.
16500 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
16501 Remove integer-related logic from pattern.
16502 (neon_vabd<mode>_3): Likewise.
16504 2018-01-15 Jakub Jelinek <jakub@redhat.com>
16506 PR middle-end/82694
16507 * common.opt (fstrict-overflow): No longer an alias.
16508 (fwrapv-pointer): New option.
16509 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
16510 also for pointer types based on flag_wrapv_pointer.
16511 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
16512 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
16513 opts->x_flag_wrapv got set.
16514 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
16515 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
16516 POINTER_TYPE_OVERFLOW_UNDEFINED.
16517 * match.pd: Likewise in address comparison pattern.
16518 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
16520 2018-01-15 Richard Biener <rguenther@suse.de>
16523 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
16524 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
16525 Reset type names to their identifier if their TYPE_DECL doesn't
16526 have linkage (and thus is used for ODR and devirt).
16527 (save_debug_info_for_decl): Remove.
16528 (save_debug_info_for_type): Likewise.
16529 (add_tree_to_fld_list): Adjust.
16530 * tree-pretty-print.c (dump_generic_node): Make dumping of
16531 type names more robust.
16533 2018-01-15 Richard Biener <rguenther@suse.de>
16535 * BASE-VER: Bump to 8.0.1.
16537 2018-01-14 Martin Sebor <msebor@redhat.com>
16540 * builtins.c (check_access): Avoid warning when the no-warning bit
16543 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
16545 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
16546 * ira-color (allocno_hard_regs_compare): Likewise.
16548 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
16551 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
16552 Use .pushsection/.popsection.
16554 2018-01-14 Martin Sebor <msebor@redhat.com>
16557 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
16559 2018-01-14 Jakub Jelinek <jakub@redhat.com>
16561 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
16562 entry from extra_headers.
16563 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
16564 extra_headers, make the list bitwise identical to the i?86-*-* one.
16566 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16568 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
16569 -mcmodel=large with -mindirect-branch=thunk,
16570 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
16571 -mfunction-return=thunk-extern.
16572 * doc/invoke.texi: Document -mcmodel=large is incompatible with
16573 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
16574 -mfunction-return=thunk and -mfunction-return=thunk-extern.
16576 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16578 * config/i386/i386.c (print_reg): Print the name of the full
16579 integer register without '%'.
16580 (ix86_print_operand): Handle 'V'.
16581 * doc/extend.texi: Document 'V' modifier.
16583 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16585 * config/i386/constraints.md (Bs): Disallow memory operand for
16586 -mindirect-branch-register.
16588 * config/i386/predicates.md (indirect_branch_operand): Likewise.
16589 (GOT_memory_operand): Likewise.
16590 (call_insn_operand): Likewise.
16591 (sibcall_insn_operand): Likewise.
16592 (GOT32_symbol_operand): Likewise.
16593 * config/i386/i386.md (indirect_jump): Call convert_memory_address
16594 for -mindirect-branch-register.
16595 (tablejump): Likewise.
16596 (*sibcall_memory): Likewise.
16597 (*sibcall_value_memory): Likewise.
16598 Disallow peepholes of indirect call and jump via memory for
16599 -mindirect-branch-register.
16600 (*call_pop): Replace m with Bw.
16601 (*call_value_pop): Likewise.
16602 (*sibcall_pop_memory): Replace m with Bs.
16603 * config/i386/i386.opt (mindirect-branch-register): New option.
16604 * doc/invoke.texi: Document -mindirect-branch-register option.
16606 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16608 * config/i386/i386-protos.h (ix86_output_function_return): New.
16609 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
16610 set function_return_type.
16611 (indirect_thunk_name): Add ret_p to indicate thunk for function
16613 (output_indirect_thunk_function): Pass false to
16614 indirect_thunk_name.
16615 (ix86_output_indirect_branch_via_reg): Likewise.
16616 (ix86_output_indirect_branch_via_push): Likewise.
16617 (output_indirect_thunk_function): Create alias for function
16618 return thunk if regno < 0.
16619 (ix86_output_function_return): New function.
16620 (ix86_handle_fndecl_attribute): Handle function_return.
16621 (ix86_attribute_table): Add function_return.
16622 * config/i386/i386.h (machine_function): Add
16623 function_return_type.
16624 * config/i386/i386.md (simple_return_internal): Use
16625 ix86_output_function_return.
16626 (simple_return_internal_long): Likewise.
16627 * config/i386/i386.opt (mfunction-return=): New option.
16628 (indirect_branch): Mention -mfunction-return=.
16629 * doc/extend.texi: Document function_return function attribute.
16630 * doc/invoke.texi: Document -mfunction-return= option.
16632 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
16634 * config/i386/i386-opts.h (indirect_branch): New.
16635 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
16636 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
16637 with local indirect jump when converting indirect call and jump.
16638 (ix86_set_indirect_branch_type): New.
16639 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
16640 (indirectlabelno): New.
16641 (indirect_thunk_needed): Likewise.
16642 (indirect_thunk_bnd_needed): Likewise.
16643 (indirect_thunks_used): Likewise.
16644 (indirect_thunks_bnd_used): Likewise.
16645 (INDIRECT_LABEL): Likewise.
16646 (indirect_thunk_name): Likewise.
16647 (output_indirect_thunk): Likewise.
16648 (output_indirect_thunk_function): Likewise.
16649 (ix86_output_indirect_branch_via_reg): Likewise.
16650 (ix86_output_indirect_branch_via_push): Likewise.
16651 (ix86_output_indirect_branch): Likewise.
16652 (ix86_output_indirect_jmp): Likewise.
16653 (ix86_code_end): Call output_indirect_thunk_function if needed.
16654 (ix86_output_call_insn): Call ix86_output_indirect_branch if
16656 (ix86_handle_fndecl_attribute): Handle indirect_branch.
16657 (ix86_attribute_table): Add indirect_branch.
16658 * config/i386/i386.h (machine_function): Add indirect_branch_type
16659 and has_local_indirect_jump.
16660 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
16662 (tablejump): Likewise.
16663 (*indirect_jump): Use ix86_output_indirect_jmp.
16664 (*tablejump_1): Likewise.
16665 (simple_return_indirect_internal): Likewise.
16666 * config/i386/i386.opt (mindirect-branch=): New option.
16667 (indirect_branch): New.
16670 (thunk-inline): Likewise.
16671 (thunk-extern): Likewise.
16672 * doc/extend.texi: Document indirect_branch function attribute.
16673 * doc/invoke.texi: Document -mindirect-branch= option.
16675 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
16678 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
16680 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
16682 * ipa-inline.c (want_inline_small_function_p): Return false if
16683 inlining has already failed with CIF_FINAL_ERROR.
16684 (update_caller_keys): Call want_inline_small_function_p before
16686 (update_callee_keys): Likewise.
16688 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
16690 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
16692 (rs6000_quadword_masked_address_p): Likewise.
16693 (quad_aligned_load_p): Likewise.
16694 (quad_aligned_store_p): Likewise.
16695 (const_load_sequence_p): Add comment to describe the outer-most loop.
16696 (mimic_memory_attributes_and_flags): New function.
16697 (rs6000_gen_stvx): Likewise.
16698 (replace_swapped_aligned_store): Likewise.
16699 (rs6000_gen_lvx): Likewise.
16700 (replace_swapped_aligned_load): Likewise.
16701 (replace_swapped_load_constant): Capitalize argument name in
16702 comment describing this function.
16703 (rs6000_analyze_swaps): Add a third pass to search for vector loads
16704 and stores that access quad-word aligned addresses and replace
16705 with stvx or lvx instructions when appropriate.
16706 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
16707 New function prototype.
16708 (rs6000_quadword_masked_address_p): Likewise.
16709 (rs6000_gen_lvx): Likewise.
16710 (rs6000_gen_stvx): Likewise.
16711 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
16712 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
16713 when memory address is aligned.
16714 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
16715 this split to select lvx instruction when memory address is aligned.
16716 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
16717 instruction when memory address is aligned.
16718 (*vsx_le_perm_load_v16qi): Likewise.
16719 (four unnamed splitters): Modify to select the stvx instruction
16720 when memory is aligned.
16722 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
16724 * predict.c (determine_unlikely_bbs): Handle correctly BBs
16725 which appears in the queue multiple times.
16727 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16728 Alan Hayward <alan.hayward@arm.com>
16729 David Sherwood <david.sherwood@arm.com>
16731 * tree-vectorizer.h (vec_lower_bound): New structure.
16732 (_loop_vec_info): Add check_nonzero and lower_bounds.
16733 (LOOP_VINFO_CHECK_NONZERO): New macro.
16734 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
16735 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
16736 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
16737 fields. Make seg_len the distance travelled, not including the
16739 (dr_direction_indicator): Declare.
16740 (dr_zero_step_indicator): Likewise.
16741 (dr_known_forward_stride_p): Likewise.
16742 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
16744 (runtime_alias_check_p): Allow runtime alias checks with
16746 (operator ==): Compare access_size and align.
16747 (prune_runtime_alias_test_list): Rework for new distinction between
16748 the access_size and seg_len.
16749 (create_intersect_range_checks_index): Likewise. Cope with polynomial
16751 (get_segment_min_max): New function.
16752 (create_intersect_range_checks): Use it.
16753 (dr_step_indicator): New function.
16754 (dr_direction_indicator): Likewise.
16755 (dr_zero_step_indicator): Likewise.
16756 (dr_known_forward_stride_p): Likewise.
16757 * tree-loop-distribution.c (data_ref_segment_size): Return
16758 DR_STEP * (niters - 1).
16759 (compute_alias_check_pairs): Update call to the dr_with_seg_len
16761 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
16762 (vect_preserves_scalar_order_p): New function, split out from...
16763 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
16764 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
16765 (vect_vfa_access_size): New function.
16766 (vect_vfa_align): Likewise.
16767 (vect_compile_time_alias): Take access_size_a and access_b arguments.
16768 (dump_lower_bound): New function.
16769 (vect_check_lower_bound): Likewise.
16770 (vect_small_gap_p): Likewise.
16771 (vectorizable_with_step_bound_p): Likewise.
16772 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
16773 depencies if the vectorization factor is 1. Convert the checks
16774 for nonzero steps into checks on the bounds of DR_STEP. Try using
16775 a bunds check for variable steps if the minimum required step is
16776 relatively small. Update calls to the dr_with_seg_len
16777 constructor and to vect_compile_time_alias.
16778 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
16780 (vect_loop_versioning): Call it.
16781 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
16783 (vect_estimate_min_profitable_iters): Account for any bounds checks.
16785 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16786 Alan Hayward <alan.hayward@arm.com>
16787 David Sherwood <david.sherwood@arm.com>
16789 * doc/sourcebuild.texi (vect_scatter_store): Document.
16790 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
16792 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
16794 * genopinit.c (main): Add supports_vec_scatter_store and
16795 supports_vec_scatter_store_cached to target_optabs.
16796 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
16797 IFN_MASK_SCATTER_STORE.
16798 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
16800 * internal-fn.h (internal_store_fn_p): Declare.
16801 (internal_fn_stored_value_index): Likewise.
16802 * internal-fn.c (scatter_store_direct): New macro.
16803 (expand_scatter_store_optab_fn): New function.
16804 (direct_scatter_store_optab_supported_p): New macro.
16805 (internal_store_fn_p): New function.
16806 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
16807 IFN_MASK_SCATTER_STORE.
16808 (internal_fn_mask_index): Likewise.
16809 (internal_fn_stored_value_index): New function.
16810 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
16811 for scatter stores.
16812 * optabs-query.h (supports_vec_scatter_store_p): Declare.
16813 * optabs-query.c (supports_vec_scatter_store_p): New function.
16814 * tree-vectorizer.h (vect_get_store_rhs): Declare.
16815 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
16816 true for scatter stores.
16817 (vect_gather_scatter_fn_p): Handle scatter stores too.
16818 (vect_check_gather_scatter): Consider using scatter stores if
16819 supports_vec_scatter_store_p.
16820 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
16821 scatter stores too.
16822 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
16823 internal_fn_stored_value_index.
16824 (check_load_store_masking): Handle scatter stores too.
16825 (vect_get_store_rhs): Make public.
16826 (vectorizable_call): Use internal_store_fn_p.
16827 (vectorizable_store): Handle scatter store internal functions.
16828 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
16829 when deciding whether the end of the group has been reached.
16830 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
16831 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
16832 (mask_scatter_store<mode>): New insns.
16834 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16835 Alan Hayward <alan.hayward@arm.com>
16836 David Sherwood <david.sherwood@arm.com>
16838 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
16839 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
16840 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
16842 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
16843 Use vect_truncate_gather_scatter_offset if we can't treat the
16844 operation as a normal gather load or scatter store.
16845 (get_group_load_store_type): Take the gather_scatter_info
16846 as argument. Try using a gather load or scatter store for
16847 single-element groups.
16848 (get_load_store_type): Update calls to get_group_load_store_type
16849 and vect_use_strided_gather_scatters_p.
16851 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16852 Alan Hayward <alan.hayward@arm.com>
16853 David Sherwood <david.sherwood@arm.com>
16855 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
16856 optional tree argument.
16857 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
16859 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
16860 but continue to use the current value as a fallback.
16861 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
16862 to compare the updates.
16863 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
16864 (get_load_store_type): Use it when handling a strided access.
16865 (vect_get_strided_load_store_ops): New function.
16866 (vect_get_data_ptr_increment): Likewise.
16867 (vectorizable_load): Handle strided gather loads. Always pass
16868 a step to vect_create_data_ref_ptr and bump_vector_ptr.
16870 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16871 Alan Hayward <alan.hayward@arm.com>
16872 David Sherwood <david.sherwood@arm.com>
16874 * doc/md.texi (gather_load@var{m}): Document.
16875 (mask_gather_load@var{m}): Likewise.
16876 * genopinit.c (main): Add supports_vec_gather_load and
16877 supports_vec_gather_load_cached to target_optabs.
16878 * optabs-tree.c (init_tree_optimization_optabs): Use
16879 ggc_cleared_alloc to allocate target_optabs.
16880 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
16881 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
16883 * internal-fn.h (internal_load_fn_p): Declare.
16884 (internal_gather_scatter_fn_p): Likewise.
16885 (internal_fn_mask_index): Likewise.
16886 (internal_gather_scatter_fn_supported_p): Likewise.
16887 * internal-fn.c (gather_load_direct): New macro.
16888 (expand_gather_load_optab_fn): New function.
16889 (direct_gather_load_optab_supported_p): New macro.
16890 (direct_internal_fn_optab): New function.
16891 (internal_load_fn_p): Likewise.
16892 (internal_gather_scatter_fn_p): Likewise.
16893 (internal_fn_mask_index): Likewise.
16894 (internal_gather_scatter_fn_supported_p): Likewise.
16895 * optabs-query.c (supports_at_least_one_mode_p): New function.
16896 (supports_vec_gather_load_p): Likewise.
16897 * optabs-query.h (supports_vec_gather_load_p): Declare.
16898 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
16899 and memory_type field.
16900 (NUM_PATTERNS): Bump to 15.
16901 * tree-vect-data-refs.c: Include internal-fn.h.
16902 (vect_gather_scatter_fn_p): New function.
16903 (vect_describe_gather_scatter_call): Likewise.
16904 (vect_check_gather_scatter): Try using internal functions for
16905 gather loads. Recognize existing calls to a gather load function.
16906 (vect_analyze_data_refs): Consider using gather loads if
16907 supports_vec_gather_load_p.
16908 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
16909 (vect_get_gather_scatter_offset_type): Likewise.
16910 (vect_convert_mask_for_vectype): Likewise.
16911 (vect_add_conversion_to_patterm): Likewise.
16912 (vect_try_gather_scatter_pattern): Likewise.
16913 (vect_recog_gather_scatter_pattern): New pattern recognizer.
16914 (vect_vect_recog_func_ptrs): Add it.
16915 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
16916 internal_fn_mask_index and internal_gather_scatter_fn_p.
16917 (check_load_store_masking): Take the gather_scatter_info as an
16918 argument and handle gather loads.
16919 (vect_get_gather_scatter_ops): New function.
16920 (vectorizable_call): Check internal_load_fn_p.
16921 (vectorizable_load): Likewise. Handle gather load internal
16923 (vectorizable_store): Update call to check_load_store_masking.
16924 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
16925 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
16926 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
16927 (aarch64_gather_scale_operand_d): New predicates.
16928 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
16929 (mask_gather_load<mode>): New insns.
16931 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16932 Alan Hayward <alan.hayward@arm.com>
16933 David Sherwood <david.sherwood@arm.com>
16935 * optabs.def (fold_left_plus_optab): New optab.
16936 * doc/md.texi (fold_left_plus_@var{m}): Document.
16937 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
16938 * internal-fn.c (fold_left_direct): Define.
16939 (expand_fold_left_optab_fn): Likewise.
16940 (direct_fold_left_optab_supported_p): Likewise.
16941 * fold-const-call.c (fold_const_fold_left): New function.
16942 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
16943 * tree-parloops.c (valid_reduction_p): New function.
16944 (gather_scalar_reductions): Use it.
16945 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
16946 (vect_finish_replace_stmt): Declare.
16947 * tree-vect-loop.c (fold_left_reduction_fn): New function.
16948 (needs_fold_left_reduction_p): New function, split out from...
16949 (vect_is_simple_reduction): ...here. Accept reductions that
16950 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
16951 (vect_force_simple_reduction): Also store the reduction type in
16952 the assignment's STMT_VINFO_REDUC_TYPE.
16953 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
16954 (merge_with_identity): New function.
16955 (vect_expand_fold_left): Likewise.
16956 (vectorize_fold_left_reduction): Likewise.
16957 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
16958 scalar phi in place for it. Check for target support and reject
16959 cases that would reassociate the operation. Defer the transform
16960 phase to vectorize_fold_left_reduction.
16961 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
16962 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
16963 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
16965 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16967 * tree-if-conv.c (predicate_mem_writes): Remove redundant
16968 call to ifc_temp_var.
16970 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16971 Alan Hayward <alan.hayward@arm.com>
16972 David Sherwood <david.sherwood@arm.com>
16974 * target.def (legitimize_address_displacement): Take the original
16975 offset as a poly_int.
16976 * targhooks.h (default_legitimize_address_displacement): Update
16978 * targhooks.c (default_legitimize_address_displacement): Likewise.
16979 * doc/tm.texi: Regenerate.
16980 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
16981 as an argument, moving assert of ad->disp == ad->disp_term to...
16982 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
16983 Try calling targetm.legitimize_address_displacement before expanding
16984 the address rather than afterwards, and adjust for the new interface.
16985 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
16986 Match the new hook interface. Handle SVE addresses.
16987 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
16988 new hook interface.
16990 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16992 * Makefile.in (OBJS): Add early-remat.o.
16993 * target.def (select_early_remat_modes): New hook.
16994 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
16995 * doc/tm.texi: Regenerate.
16996 * targhooks.h (default_select_early_remat_modes): Declare.
16997 * targhooks.c (default_select_early_remat_modes): New function.
16998 * timevar.def (TV_EARLY_REMAT): New timevar.
16999 * passes.def (pass_early_remat): New pass.
17000 * tree-pass.h (make_pass_early_remat): Declare.
17001 * early-remat.c: New file.
17002 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
17004 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
17006 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17007 Alan Hayward <alan.hayward@arm.com>
17008 David Sherwood <david.sherwood@arm.com>
17010 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
17011 vfm1 with a bound_epilog parameter.
17012 (vect_do_peeling): Update calls accordingly, and move the prologue
17013 call earlier in the function. Treat the base bound_epilog as 0 for
17014 fully-masked loops and retain vf - 1 for other loops. Add 1 to
17015 this base when peeling for gaps.
17016 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
17017 with fully-masked loops.
17018 (vect_estimate_min_profitable_iters): Handle the single peeled
17019 iteration in that case.
17021 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17022 Alan Hayward <alan.hayward@arm.com>
17023 David Sherwood <david.sherwood@arm.com>
17025 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
17026 single-element interleaving even if the size is not a power of 2.
17027 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
17028 accesses for single-element interleaving if the group size is
17031 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17032 Alan Hayward <alan.hayward@arm.com>
17033 David Sherwood <david.sherwood@arm.com>
17035 * doc/md.texi (fold_extract_last_@var{m}): Document.
17036 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
17037 * optabs.def (fold_extract_last_optab): New optab.
17038 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
17039 * internal-fn.c (fold_extract_direct): New macro.
17040 (expand_fold_extract_optab_fn): Likewise.
17041 (direct_fold_extract_optab_supported_p): Likewise.
17042 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
17043 * tree-vect-loop.c (vect_model_reduction_cost): Handle
17044 EXTRACT_LAST_REDUCTION.
17045 (get_initial_def_for_reduction): Do not create an initial vector
17046 for EXTRACT_LAST_REDUCTION reductions.
17047 (vectorizable_reduction): Leave the scalar phi in place for
17048 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
17049 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
17050 epilogue code for EXTRACT_LAST_REDUCTION and defer the
17051 transform phase to vectorizable_condition.
17052 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
17054 (vect_finish_stmt_generation): ...here.
17055 (vect_finish_replace_stmt): New function.
17056 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
17057 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
17059 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
17061 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17062 Alan Hayward <alan.hayward@arm.com>
17063 David Sherwood <david.sherwood@arm.com>
17065 * doc/md.texi (extract_last_@var{m}): Document.
17066 * optabs.def (extract_last_optab): New optab.
17067 * internal-fn.def (EXTRACT_LAST): New internal function.
17068 * internal-fn.c (cond_unary_direct): New macro.
17069 (expand_cond_unary_optab_fn): Likewise.
17070 (direct_cond_unary_optab_supported_p): Likewise.
17071 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
17072 loops using EXTRACT_LAST.
17073 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
17074 (extract_last_<mode>): ...this optab.
17075 (vec_extract<mode><Vel>): Update accordingly.
17077 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17078 Alan Hayward <alan.hayward@arm.com>
17079 David Sherwood <david.sherwood@arm.com>
17081 * target.def (empty_mask_is_expensive): New hook.
17082 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
17083 * doc/tm.texi: Regenerate.
17084 * targhooks.h (default_empty_mask_is_expensive): Declare.
17085 * targhooks.c (default_empty_mask_is_expensive): New function.
17086 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
17087 if the target says that empty masks are expensive.
17088 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
17090 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
17092 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17093 Alan Hayward <alan.hayward@arm.com>
17094 David Sherwood <david.sherwood@arm.com>
17096 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
17097 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
17098 (vect_use_loop_mask_for_alignment_p): New function.
17099 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
17100 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
17101 niters_skip argument. Make sure that the first niters_skip elements
17102 of the first iteration are inactive.
17103 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
17104 Update call to vect_set_loop_masks_directly.
17105 (get_misalign_in_elems): New function, split out from...
17106 (vect_gen_prolog_loop_niters): ...here.
17107 (vect_update_init_of_dr): Take a code argument that specifies whether
17108 the adjustment should be added or subtracted.
17109 (vect_update_init_of_drs): Likewise.
17110 (vect_prepare_for_masked_peels): New function.
17111 (vect_do_peeling): Skip prologue peeling if we're using a mask
17112 instead. Update call to vect_update_inits_of_drs.
17113 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17115 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
17116 alignment. Do not include the number of peeled iterations in
17117 the minimum threshold in that case.
17118 (vectorizable_induction): Adjust the start value down by
17119 LOOP_VINFO_MASK_SKIP_NITERS iterations.
17120 (vect_transform_loop): Call vect_prepare_for_masked_peels.
17121 Take the number of skipped iterations into account when calculating
17123 * tree-vect-stmts.c (vect_gen_while_not): New function.
17125 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17126 Alan Hayward <alan.hayward@arm.com>
17127 David Sherwood <david.sherwood@arm.com>
17129 * doc/sourcebuild.texi (vect_fully_masked): Document.
17130 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
17131 default value to 0.
17132 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
17134 (vect_analyze_loop_2): ...here. Don't check the vectorization
17135 factor against the number of loop iterations if the loop is
17138 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17139 Alan Hayward <alan.hayward@arm.com>
17140 David Sherwood <david.sherwood@arm.com>
17142 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
17143 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
17144 (dump_groups): Update accordingly.
17145 (iv_use::mem_type): New member variable.
17146 (address_p): New function.
17147 (record_use): Add a mem_type argument and initialize the new
17149 (record_group_use): Add a mem_type argument. Use address_p.
17150 Remove obsolete null checks of base_object. Update call to record_use.
17151 (find_interesting_uses_op): Update call to record_group_use.
17152 (find_interesting_uses_cond): Likewise.
17153 (find_interesting_uses_address): Likewise.
17154 (get_mem_type_for_internal_fn): New function.
17155 (find_address_like_use): Likewise.
17156 (find_interesting_uses_stmt): Try find_address_like_use before
17157 calling find_interesting_uses_op.
17158 (addr_offset_valid_p): Use the iv mem_type field as the type
17159 of the addressed memory.
17160 (add_autoinc_candidates): Likewise.
17161 (get_address_cost): Likewise.
17162 (split_small_address_groups_p): Use address_p.
17163 (split_address_groups): Likewise.
17164 (add_iv_candidate_for_use): Likewise.
17165 (autoinc_possible_for_pair): Likewise.
17166 (rewrite_groups): Likewise.
17167 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
17168 (determine_group_iv_cost): Update after split of USE_ADDRESS.
17169 (get_alias_ptr_type_for_ptr_address): New function.
17170 (rewrite_use_address): Rewrite address uses in calls that were
17171 identified by find_address_like_use.
17173 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17174 Alan Hayward <alan.hayward@arm.com>
17175 David Sherwood <david.sherwood@arm.com>
17177 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
17179 * gimple-expr.h (is_gimple_addressable: Likewise.
17180 * gimple-expr.c (is_gimple_address): Likewise.
17181 * internal-fn.c (expand_call_mem_ref): New function.
17182 (expand_mask_load_optab_fn): Use it.
17183 (expand_mask_store_optab_fn): Likewise.
17185 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17186 Alan Hayward <alan.hayward@arm.com>
17187 David Sherwood <david.sherwood@arm.com>
17189 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
17190 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
17191 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
17192 (cond_umax@var{mode}): Document.
17193 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
17194 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
17195 (cond_umin_optab, cond_umax_optab): New optabs.
17196 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
17197 (COND_IOR, COND_XOR): New internal functions.
17198 * internal-fn.h (get_conditional_internal_fn): Declare.
17199 * internal-fn.c (cond_binary_direct): New macro.
17200 (expand_cond_binary_optab_fn): Likewise.
17201 (direct_cond_binary_optab_supported_p): Likewise.
17202 (get_conditional_internal_fn): New function.
17203 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
17204 Cope with reduction statements that are vectorized as calls rather
17206 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
17207 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
17208 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
17209 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
17210 (UNSPEC_COND_EOR): New unspecs.
17211 (optab): Add mappings for them.
17212 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
17213 (sve_int_op, sve_fp_op): New int attributes.
17215 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17216 Alan Hayward <alan.hayward@arm.com>
17217 David Sherwood <david.sherwood@arm.com>
17219 * optabs.def (while_ult_optab): New optab.
17220 * doc/md.texi (while_ult@var{m}@var{n}): Document.
17221 * internal-fn.def (WHILE_ULT): New internal function.
17222 * internal-fn.h (direct_internal_fn_supported_p): New override
17223 that takes two types as argument.
17224 * internal-fn.c (while_direct): New macro.
17225 (expand_while_optab_fn): New function.
17226 (convert_optab_supported_p): Likewise.
17227 (direct_while_optab_supported_p): New macro.
17228 * wide-int.h (wi::udiv_ceil): New function.
17229 * tree-vectorizer.h (rgroup_masks): New structure.
17230 (vec_loop_masks): New typedef.
17231 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
17232 and fully_masked_p.
17233 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
17234 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
17235 (vect_max_vf): New function.
17236 (slpeel_make_loop_iterate_ntimes): Delete.
17237 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
17238 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
17239 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
17240 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
17241 internal-fn.h, stor-layout.h and optabs-query.h.
17242 (vect_set_loop_mask): New function.
17243 (add_preheader_seq): Likewise.
17244 (add_header_seq): Likewise.
17245 (interleave_supported_p): Likewise.
17246 (vect_maybe_permute_loop_masks): Likewise.
17247 (vect_set_loop_masks_directly): Likewise.
17248 (vect_set_loop_condition_masked): Likewise.
17249 (vect_set_loop_condition_unmasked): New function, split out from
17250 slpeel_make_loop_iterate_ntimes.
17251 (slpeel_make_loop_iterate_ntimes): Rename to..
17252 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
17253 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
17254 (vect_do_peeling): Update call accordingly.
17255 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
17257 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17258 mask_compare_type, can_fully_mask_p and fully_masked_p.
17259 (release_vec_loop_masks): New function.
17260 (_loop_vec_info): Use it to free the loop masks.
17261 (can_produce_all_loop_masks_p): New function.
17262 (vect_get_max_nscalars_per_iter): Likewise.
17263 (vect_verify_full_masking): Likewise.
17264 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
17265 retries, and free the mask rgroups before retrying. Check loop-wide
17266 reasons for disallowing fully-masked loops. Make the final decision
17267 about whether use a fully-masked loop or not.
17268 (vect_estimate_min_profitable_iters): Do not assume that peeling
17269 for the number of iterations will be needed for fully-masked loops.
17270 (vectorizable_reduction): Disable fully-masked loops.
17271 (vectorizable_live_operation): Likewise.
17272 (vect_halve_mask_nunits): New function.
17273 (vect_double_mask_nunits): Likewise.
17274 (vect_record_loop_mask): Likewise.
17275 (vect_get_loop_mask): Likewise.
17276 (vect_transform_loop): Handle the case in which the final loop
17277 iteration might handle a partial vector. Call vect_set_loop_condition
17278 instead of slpeel_make_loop_iterate_ntimes.
17279 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
17280 (check_load_store_masking): New function.
17281 (prepare_load_store_mask): Likewise.
17282 (vectorizable_store): Handle fully-masked loops.
17283 (vectorizable_load): Likewise.
17284 (supportable_widening_operation): Use vect_halve_mask_nunits for
17286 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
17287 (vect_gen_while): New function.
17288 * config/aarch64/aarch64.md (umax<mode>3): New expander.
17289 (aarch64_uqdec<mode>): New insn.
17291 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17292 Alan Hayward <alan.hayward@arm.com>
17293 David Sherwood <david.sherwood@arm.com>
17295 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
17296 (reduc_xor_scal_optab): New optabs.
17297 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
17298 (reduc_xor_scal_@var{m}): Document.
17299 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
17300 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
17301 internal functions.
17302 * fold-const-call.c (fold_const_call): Handle them.
17303 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
17304 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
17305 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
17306 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
17307 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
17308 (UNSPEC_XORV): New unspecs.
17309 (optab): Add entries for them.
17310 (BITWISEV): New int iterator.
17311 (bit_reduc_op): New int attributes.
17313 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17314 Alan Hayward <alan.hayward@arm.com>
17315 David Sherwood <david.sherwood@arm.com>
17317 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
17318 * internal-fn.def (VEC_SHL_INSERT): New internal function.
17319 * optabs.def (vec_shl_insert_optab): New optab.
17320 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
17321 (duplicate_and_interleave): Likewise.
17322 * tree-vect-loop.c: Include internal-fn.h.
17323 (neutral_op_for_slp_reduction): New function, split out from
17324 get_initial_defs_for_reduction.
17325 (get_initial_def_for_reduction): Handle option 2 for variable-length
17326 vectors by loading the neutral value into a vector and then shifting
17327 the initial value into element 0.
17328 (get_initial_defs_for_reduction): Replace the code argument with
17329 the neutral value calculated by neutral_op_for_slp_reduction.
17330 Use gimple_build_vector for constant-length vectors.
17331 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
17332 but the first group_size elements have a neutral value.
17333 Use duplicate_and_interleave otherwise.
17334 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
17335 Update call to get_initial_defs_for_reduction. Handle SLP
17336 reductions for variable-length vectors by creating one vector
17337 result for each scalar result, with the elements associated
17338 with other scalar results stubbed out with the neutral value.
17339 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
17340 Require IFN_VEC_SHL_INSERT for double reductions on
17341 variable-length vectors, or SLP reductions that have
17342 a neutral value. Require can_duplicate_and_interleave_p
17343 support for variable-length unchained SLP reductions if there
17344 is no neutral value, such as for MIN/MAX reductions. Also require
17345 the number of vector elements to be a multiple of the number of
17346 SLP statements when doing variable-length unchained SLP reductions.
17347 Update call to vect_create_epilog_for_reduction.
17348 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
17349 and remove initial values.
17350 (duplicate_and_interleave): Make public.
17351 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
17352 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
17354 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17355 Alan Hayward <alan.hayward@arm.com>
17356 David Sherwood <david.sherwood@arm.com>
17358 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
17359 (can_duplicate_and_interleave_p): New function.
17360 (vect_get_and_check_slp_defs): Take the vector of statements
17361 rather than just the current one. Remove excess parentheses.
17362 Restriction rejectinon of vect_constant_def and vect_external_def
17363 for variable-length vectors to boolean types, or types for which
17364 can_duplicate_and_interleave_p is false.
17365 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
17366 (duplicate_and_interleave): New function.
17367 (vect_get_constant_vectors): Use gimple_build_vector for
17368 constant-length vectors and suitable variable-length constant
17369 vectors. Use duplicate_and_interleave for other variable-length
17370 vectors. Don't defer the update when inserting new statements.
17372 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17373 Alan Hayward <alan.hayward@arm.com>
17374 David Sherwood <david.sherwood@arm.com>
17376 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
17377 min_profitable_iters doesn't go negative.
17379 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17380 Alan Hayward <alan.hayward@arm.com>
17381 David Sherwood <david.sherwood@arm.com>
17383 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
17384 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
17385 * optabs.def (vec_mask_load_lanes_optab): New optab.
17386 (vec_mask_store_lanes_optab): Likewise.
17387 * internal-fn.def (MASK_LOAD_LANES): New internal function.
17388 (MASK_STORE_LANES): Likewise.
17389 * internal-fn.c (mask_load_lanes_direct): New macro.
17390 (mask_store_lanes_direct): Likewise.
17391 (expand_mask_load_optab_fn): Handle masked operations.
17392 (expand_mask_load_lanes_optab_fn): New macro.
17393 (expand_mask_store_optab_fn): Handle masked operations.
17394 (expand_mask_store_lanes_optab_fn): New macro.
17395 (direct_mask_load_lanes_optab_supported_p): Likewise.
17396 (direct_mask_store_lanes_optab_supported_p): Likewise.
17397 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
17399 (vect_load_lanes_supported): Likewise.
17400 * tree-vect-data-refs.c (strip_conversion): New function.
17401 (can_group_stmts_p): Likewise.
17402 (vect_analyze_data_ref_accesses): Use it instead of checking
17403 for a pair of assignments.
17404 (vect_store_lanes_supported): Take a masked_p parameter.
17405 (vect_load_lanes_supported): Likewise.
17406 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
17407 vect_store_lanes_supported and vect_load_lanes_supported.
17408 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
17409 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
17410 parameter. Don't allow gaps for masked accesses.
17411 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
17412 and vect_load_lanes_supported.
17413 (get_load_store_type): Take a masked_p parameter and update
17414 call to get_group_load_store_type.
17415 (vectorizable_store): Update call to get_load_store_type.
17416 Handle IFN_MASK_STORE_LANES.
17417 (vectorizable_load): Update call to get_load_store_type.
17418 Handle IFN_MASK_LOAD_LANES.
17420 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17421 Alan Hayward <alan.hayward@arm.com>
17422 David Sherwood <david.sherwood@arm.com>
17424 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
17426 * config/aarch64/aarch64-protos.h
17427 (aarch64_sve_struct_memory_operand_p): Declare.
17428 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
17429 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
17430 (VPRED, vpred): Handle SVE structure modes.
17431 * config/aarch64/constraints.md (Utx): New constraint.
17432 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
17433 (aarch64_sve_struct_nonimmediate_operand): New predicates.
17434 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
17435 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
17436 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
17437 structure modes. Split into pieces after RA.
17438 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
17439 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
17441 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
17442 SVE structure modes.
17443 (aarch64_classify_address): Likewise.
17444 (sizetochar): Move earlier in file.
17445 (aarch64_print_operand): Handle SVE register lists.
17446 (aarch64_array_mode): New function.
17447 (aarch64_sve_struct_memory_operand_p): Likewise.
17448 (TARGET_ARRAY_MODE): Redefine.
17450 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17451 Alan Hayward <alan.hayward@arm.com>
17452 David Sherwood <david.sherwood@arm.com>
17454 * target.def (array_mode): New target hook.
17455 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
17456 * doc/tm.texi: Regenerate.
17457 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
17458 * hooks.c (hook_optmode_mode_uhwi_none): New function.
17459 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
17460 targetm.array_mode.
17461 * stor-layout.c (mode_for_array): Likewise. Support polynomial
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 * fold-const.c (fold_binary_loc): Check the argument types
17469 rather than the result type when testing for a vector operation.
17471 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17473 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
17474 * doc/tm.texi: Regenerate.
17476 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17477 Alan Hayward <alan.hayward@arm.com>
17478 David Sherwood <david.sherwood@arm.com>
17480 * doc/invoke.texi (-msve-vector-bits=): Document new option.
17481 (sve): Document new AArch64 extension.
17482 * doc/md.texi (w): Extend the description of the AArch64
17483 constraint to include SVE vectors.
17484 (Upl, Upa): Document new AArch64 predicate constraints.
17485 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
17487 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
17488 (msve-vector-bits=): New option.
17489 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
17490 SVE when these are disabled.
17491 (sve): New extension.
17492 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
17493 modes. Adjust their number of units based on aarch64_sve_vg.
17494 (MAX_BITSIZE_MODE_ANY_MODE): Define.
17495 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
17496 aarch64_addr_query_type.
17497 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
17498 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
17499 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
17500 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
17501 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
17502 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
17503 (aarch64_simd_imm_zero_p): Delete.
17504 (aarch64_check_zero_based_sve_index_immediate): Declare.
17505 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
17506 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
17507 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
17508 (aarch64_sve_float_mul_immediate_p): Likewise.
17509 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
17510 rather than an rtx.
17511 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
17512 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
17513 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
17514 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
17515 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
17516 (aarch64_regmode_natural_size): Likewise.
17517 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
17518 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
17520 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
17521 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
17522 for VG and the SVE predicate registers.
17523 (V_ALIASES): Add a "z"-prefixed alias.
17524 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
17525 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
17526 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
17527 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
17528 (REG_CLASS_NAMES): Add entries for them.
17529 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
17530 and the predicate registers.
17531 (aarch64_sve_vg): Declare.
17532 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
17533 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
17534 (REGMODE_NATURAL_SIZE): Define.
17535 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
17537 * config/aarch64/aarch64.c: Include cfgrtl.h.
17538 (simd_immediate_info): Add a constructor for series vectors,
17539 and an associated step field.
17540 (aarch64_sve_vg): New variable.
17541 (aarch64_dbx_register_number): Handle VG and the predicate registers.
17542 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
17543 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
17544 (VEC_ANY_DATA, VEC_STRUCT): New constants.
17545 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
17546 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
17547 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
17548 (aarch64_get_mask_mode): New functions.
17549 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
17550 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
17551 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
17552 predicate modes and predicate registers. Explicitly restrict
17553 GPRs to modes of 16 bytes or smaller. Only allow FP registers
17554 to store a vector mode if it is recognized by
17555 aarch64_classify_vector_mode.
17556 (aarch64_regmode_natural_size): New function.
17557 (aarch64_hard_regno_caller_save_mode): Return the original mode
17559 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
17560 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
17561 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
17562 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
17564 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
17565 does not overlap dest if the function is frame-related. Handle
17567 (aarch64_split_add_offset): New function.
17568 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
17569 them aarch64_add_offset.
17570 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
17571 and update call to aarch64_sub_sp.
17572 (aarch64_add_cfa_expression): New function.
17573 (aarch64_expand_prologue): Pass extra temporary registers to the
17574 functions above. Handle the case in which we need to emit new
17575 DW_CFA_expressions for registers that were originally saved
17576 relative to the stack pointer, but now have to be expressed
17577 relative to the frame pointer.
17578 (aarch64_output_mi_thunk): Pass extra temporary registers to the
17580 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
17581 IP0 and IP1 values for SVE frames.
17582 (aarch64_expand_vec_series): New function.
17583 (aarch64_expand_sve_widened_duplicate): Likewise.
17584 (aarch64_expand_sve_const_vector): Likewise.
17585 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
17586 Handle SVE constants. Use emit_move_insn to move a force_const_mem
17587 into the register, rather than emitting a SET directly.
17588 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
17589 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
17590 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
17591 (offset_9bit_signed_scaled_p): New functions.
17592 (aarch64_replicate_bitmask_imm): New function.
17593 (aarch64_bitmask_imm): Use it.
17594 (aarch64_cannot_force_const_mem): Reject expressions involving
17595 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
17596 (aarch64_classify_index): Handle SVE indices, by requiring
17597 a plain register index with a scale that matches the element size.
17598 (aarch64_classify_address): Handle SVE addresses. Assert that
17599 the mode of the address is VOIDmode or an integer mode.
17600 Update call to aarch64_classify_symbol.
17601 (aarch64_classify_symbolic_expression): Update call to
17602 aarch64_classify_symbol.
17603 (aarch64_const_vec_all_in_range_p): New function.
17604 (aarch64_print_vector_float_operand): Likewise.
17605 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
17606 "vN" for FP registers with SVE modes. Handle (const ...) vectors
17607 and the FP immediates 1.0 and 0.5.
17608 (aarch64_print_address_internal): Handle SVE addresses.
17609 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
17610 (aarch64_regno_regclass): Handle predicate registers.
17611 (aarch64_secondary_reload): Handle big-endian reloads of SVE
17613 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
17614 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
17615 (aarch64_convert_sve_vector_bits): New function.
17616 (aarch64_override_options): Use it to handle -msve-vector-bits=.
17617 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
17618 rather than an rtx.
17619 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
17620 Handle SVE vector and predicate modes. Accept VL-based constants
17621 that need only one temporary register, and VL offsets that require
17622 no temporary registers.
17623 (aarch64_conditional_register_usage): Mark the predicate registers
17624 as fixed if SVE isn't available.
17625 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
17626 Return true for SVE vector and predicate modes.
17627 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
17628 rather than an unsigned int. Handle SVE modes.
17629 (aarch64_preferred_simd_mode): Update call accordingly. Handle
17631 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
17633 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
17634 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
17635 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
17636 (aarch64_sve_float_mul_immediate_p): New functions.
17637 (aarch64_sve_valid_immediate): New function.
17638 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
17639 Explicitly reject structure modes. Check for INDEX constants.
17640 Handle PTRUE and PFALSE constants.
17641 (aarch64_check_zero_based_sve_index_immediate): New function.
17642 (aarch64_simd_imm_zero_p): Delete.
17643 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
17644 vector modes. Accept constants in the range of CNT[BHWD].
17645 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
17646 ask for an Advanced SIMD mode.
17647 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
17648 (aarch64_simd_vector_alignment): Handle SVE predicates.
17649 (aarch64_vectorize_preferred_vector_alignment): New function.
17650 (aarch64_simd_vector_alignment_reachable): Use it instead of
17652 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
17653 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
17655 (MAX_VECT_LEN): Delete.
17656 (expand_vec_perm_d): Add a vec_flags field.
17657 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
17658 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
17659 (aarch64_evpc_ext): Don't apply a big-endian lane correction
17661 (aarch64_evpc_rev): Rename to...
17662 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
17663 (aarch64_evpc_rev_global): New function.
17664 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
17665 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
17667 (aarch64_evpc_sve_tbl): New function.
17668 (aarch64_expand_vec_perm_const_1): Update after rename of
17669 aarch64_evpc_rev. Handle SVE permutes too, trying
17670 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
17671 than aarch64_evpc_tbl.
17672 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
17673 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
17674 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
17675 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
17676 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
17677 (aarch64_expand_sve_vcond): New functions.
17678 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
17679 of aarch64_vector_mode_p.
17680 (aarch64_dwarf_poly_indeterminate_value): New function.
17681 (aarch64_compute_pressure_classes): Likewise.
17682 (aarch64_can_change_mode_class): Likewise.
17683 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
17684 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
17685 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
17686 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
17687 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
17688 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
17689 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
17690 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
17692 (Dn, Dl, Dr): Accept const as well as const_vector.
17693 (Dz): Likewise. Compare against CONST0_RTX.
17694 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
17695 of "vector" where appropriate.
17696 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
17697 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
17698 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
17699 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
17700 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
17701 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
17702 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
17703 (v_int_equiv): Extend to SVE modes.
17704 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
17706 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
17707 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
17708 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
17709 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
17710 (SVE_COND_FP_CMP): New int iterators.
17711 (perm_hilo): Handle the new unpack unspecs.
17712 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
17714 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
17715 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
17716 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
17717 (aarch64_equality_operator, aarch64_constant_vector_operand)
17718 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
17719 (aarch64_sve_nonimmediate_operand): Likewise.
17720 (aarch64_sve_general_operand): Likewise.
17721 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
17722 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
17723 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
17724 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
17725 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
17726 (aarch64_sve_float_arith_immediate): Likewise.
17727 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
17728 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
17729 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
17730 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
17731 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
17732 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
17733 (aarch64_sve_float_arith_operand): Likewise.
17734 (aarch64_sve_float_arith_with_sub_operand): Likewise.
17735 (aarch64_sve_float_mul_operand): Likewise.
17736 (aarch64_sve_vec_perm_operand): Likewise.
17737 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
17738 (aarch64_mov_operand): Accept const_poly_int and const_vector.
17739 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
17740 as well as const_vector.
17741 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
17742 in file. Use CONST0_RTX and CONSTM1_RTX.
17743 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
17744 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
17745 Use aarch64_simd_imm_zero.
17746 * config/aarch64/aarch64-sve.md: New file.
17747 * config/aarch64/aarch64.md: Include it.
17748 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
17749 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
17750 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
17751 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
17752 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
17753 (sve): New attribute.
17754 (enabled): Disable instructions with the sve attribute unless
17756 (movqi, movhi): Pass CONST_POLY_INT operaneds through
17757 aarch64_expand_mov_immediate.
17758 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
17759 CNT[BHSD] immediates.
17760 (movti): Split CONST_POLY_INT moves into two halves.
17761 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
17762 Split additions that need a temporary here if the destination
17763 is the stack pointer.
17764 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
17765 (*add<mode>3_poly_1): New instruction.
17766 (set_clobber_cc): New expander.
17768 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17770 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
17771 parameter and use it instead of GET_MODE_SIZE (innermode). Use
17772 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
17773 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
17774 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
17775 Change innermode from fixed_mode_size to machine_mode.
17776 (simplify_subreg): Update call accordingly. Handle a constant-sized
17777 subreg of a variable-length CONST_VECTOR.
17779 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17780 Alan Hayward <alan.hayward@arm.com>
17781 David Sherwood <david.sherwood@arm.com>
17783 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
17784 (add_offset_to_base): New function, split out from...
17785 (create_mem_ref): ...here. When handling a scale other than 1,
17786 check first whether the address is valid without the offset.
17787 Add it into the base if so, leaving the index and scale as-is.
17789 2018-01-12 Jakub Jelinek <jakub@redhat.com>
17792 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
17793 fold_for_warn before checking if arg2 is INTEGER_CST.
17795 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
17797 * config/rs6000/predicates.md (load_multiple_operation): Delete.
17798 (store_multiple_operation): Delete.
17799 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
17800 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
17801 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
17802 guarded by TARGET_STRING.
17803 (rs6000_output_load_multiple): Delete.
17804 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
17805 OPTION_MASK_STRING / TARGET_STRING handling.
17806 (print_operand) <'N', 'O'>: Add comment that these are unused now.
17807 (const rs6000_opt_masks) <"string">: Change mask to 0.
17808 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
17809 (MASK_STRING): Delete.
17810 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
17812 (load_multiple): Delete.
17819 (store_multiple): Delete.
17826 (movmemsi_8reg): Delete.
17827 (corresponding unnamed define_insn): Delete.
17828 (movmemsi_6reg): Delete.
17829 (corresponding unnamed define_insn): Delete.
17830 (movmemsi_4reg): Delete.
17831 (corresponding unnamed define_insn): Delete.
17832 (movmemsi_2reg): Delete.
17833 (corresponding unnamed define_insn): Delete.
17834 (movmemsi_1reg): Delete.
17835 (corresponding unnamed define_insn): Delete.
17836 * config/rs6000/rs6000.opt (mno-string): New.
17837 (mstring): Replace by deprecation warning stub.
17838 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
17840 2018-01-12 Jakub Jelinek <jakub@redhat.com>
17842 * regrename.c (regrename_do_replace): If replacing the same
17843 reg multiple times, try to reuse last created gen_raw_REG.
17846 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
17847 main to workaround a bug in GDB.
17849 2018-01-12 Tom de Vries <tom@codesourcery.com>
17852 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
17854 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
17856 PR rtl-optimization/80481
17857 * ira-color.c (get_cap_member): New function.
17858 (allocnos_conflict_by_live_ranges_p): Use it.
17859 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
17860 (setup_slot_coalesced_allocno_live_ranges): Ditto.
17862 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
17865 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
17866 (*saddl_se_1): Ditto.
17867 (*ssubsi_1): Ditto.
17868 (*ssubl_se_1): Ditto.
17870 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17872 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
17873 rather than wi::to_widest for DR_INITs.
17874 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
17875 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
17876 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
17878 (vect_analyze_group_access_1): Note that here.
17880 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17882 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
17883 polynomial type sizes.
17885 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17887 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
17888 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
17889 (gimple_add_tmp_var): Likewise.
17891 2018-01-12 Martin Liska <mliska@suse.cz>
17893 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
17894 (gimple_alloc_sizes): Likewise.
17895 (dump_gimple_statistics): Use PRIu64 in printf format.
17896 * gimple.h: Change uint64_t to int.
17898 2018-01-12 Martin Liska <mliska@suse.cz>
17900 * tree-core.h: Use uint64_t instead of int.
17901 * tree.c (tree_node_counts): Likewise.
17902 (tree_node_sizes): Likewise.
17903 (dump_tree_statistics): Use PRIu64 in printf format.
17905 2018-01-12 Martin Liska <mliska@suse.cz>
17907 * Makefile.in: As qsort_chk is implemented in vec.c, add
17908 vec.o to linkage of gencfn-macros.
17909 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
17910 passing the info to record_node_allocation_statistics.
17911 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
17913 * ggc-common.c (struct ggc_usage): Add operator== and use
17914 it in operator< and compare function.
17915 * mem-stats.h (struct mem_usage): Likewise.
17916 * vec.c (struct vec_usage): Remove operator< and compare
17917 function. Can be simply inherited.
17919 2018-01-12 Martin Jambor <mjambor@suse.cz>
17922 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
17923 * tree-ssa-math-opts.c: Include domwalk.h.
17924 (convert_mult_to_fma_1): New function.
17925 (fma_transformation_info): New type.
17926 (fma_deferring_state): Likewise.
17927 (cancel_fma_deferring): New function.
17928 (result_of_phi): Likewise.
17929 (last_fma_candidate_feeds_initial_phi): Likewise.
17930 (convert_mult_to_fma): Added deferring logic, split actual
17931 transformation to convert_mult_to_fma_1.
17932 (math_opts_dom_walker): New type.
17933 (math_opts_dom_walker::after_dom_children): New method, body moved
17934 here from pass_optimize_widening_mul::execute, added deferring logic
17936 (pass_optimize_widening_mul::execute): Moved most of code to
17937 math_opts_dom_walker::after_dom_children.
17938 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
17939 * config/i386/i386.c (ix86_option_override_internal): Added
17940 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
17942 2018-01-12 Richard Biener <rguenther@suse.de>
17945 * dwarf2out.c (gen_variable_die): Do not reset old_die for
17946 inline instance vars.
17948 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
17951 * config/rx/rx.c (rx_is_restricted_memory_address):
17952 Handle SUBREG case.
17954 2018-01-12 Richard Biener <rguenther@suse.de>
17956 PR tree-optimization/80846
17957 * target.def (split_reduction): New target hook.
17958 * targhooks.c (default_split_reduction): New function.
17959 * targhooks.h (default_split_reduction): Declare.
17960 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
17961 target requests first reduce vectors by combining low and high
17963 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
17964 (get_vectype_for_scalar_type_and_size): Export.
17965 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
17966 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
17967 * doc/tm.texi: Regenerate.
17968 * config/i386/i386.c (ix86_split_reduction): Implement
17969 TARGET_VECTORIZE_SPLIT_REDUCTION.
17971 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17974 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
17975 in PIC mode except for TARGET_VXWORKS_RTP.
17976 * config/sparc/sparc.c: Include cfgrtl.h.
17977 (TARGET_INIT_PIC_REG): Define.
17978 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
17979 (sparc_pic_register_p): New predicate.
17980 (sparc_legitimate_address_p): Use it.
17981 (sparc_legitimize_pic_address): Likewise.
17982 (sparc_delegitimize_address): Likewise.
17983 (sparc_mode_dependent_address_p): Likewise.
17984 (gen_load_pcrel_sym): Remove 4th parameter.
17985 (load_got_register): Adjust call to above. Remove obsolete stuff.
17986 (sparc_expand_prologue): Do not call load_got_register here.
17987 (sparc_flat_expand_prologue): Likewise.
17988 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
17989 (sparc_use_pseudo_pic_reg): New function.
17990 (sparc_init_pic_reg): Likewise.
17991 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
17992 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
17994 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
17996 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
17997 Add item for branch_cost.
17999 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
18001 PR rtl-optimization/83565
18002 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
18003 not extend the result to a larger mode for rotate operations.
18004 (num_sign_bit_copies1): Likewise.
18006 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18009 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
18011 Use values-Xc.o for -pedantic.
18012 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
18014 2018-01-12 Martin Liska <mliska@suse.cz>
18017 * ipa-devirt.c (final_warning_record::grow_type_warnings):
18019 (possible_polymorphic_call_targets): Use it.
18020 (ipa_devirt): Likewise.
18022 2018-01-12 Martin Liska <mliska@suse.cz>
18024 * profile-count.h (enum profile_quality): Use 0 as invalid
18025 enum value of profile_quality.
18027 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
18029 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
18030 -mext-string options.
18032 2018-01-12 Richard Biener <rguenther@suse.de>
18034 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
18035 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
18036 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
18038 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
18040 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
18042 * configure.ac (--with-long-double-format): Add support for the
18043 configuration option to change the default long double format on
18045 * config.gcc (powerpc*-linux*-*): Likewise.
18046 * configure: Regenerate.
18047 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
18048 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
18049 used without modification.
18051 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18053 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
18054 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
18055 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
18056 MISC_BUILTIN_SPEC_BARRIER.
18057 (rs6000_init_builtins): Likewise.
18058 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
18060 (speculation_barrier): New define_insn.
18061 * doc/extend.texi: Document __builtin_speculation_barrier.
18063 2018-01-11 Jakub Jelinek <jakub@redhat.com>
18066 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
18067 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
18068 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
18070 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
18071 integral modes instead of "ss" and "sd".
18072 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
18073 vectors with 32-bit and 64-bit elements.
18074 (vecdupssescalarmodesuffix): New mode attribute.
18075 (vec_dup<mode>): Use it.
18077 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
18080 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
18081 frame if argument is passed on stack.
18083 2018-01-11 Jakub Jelinek <jakub@redhat.com>
18086 * ree.c (combine_reaching_defs): Optimize also
18087 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
18088 reg2=any_extend(exp); reg1=reg2;, formatting fix.
18090 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
18092 PR middle-end/83189
18093 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
18095 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
18097 PR middle-end/83718
18098 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
18099 after they are computed.
18101 2018-01-11 Bin Cheng <bin.cheng@arm.com>
18103 PR tree-optimization/83695
18104 * gimple-loop-linterchange.cc
18105 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
18106 reset cached scev information after interchange.
18107 (pass_linterchange::execute): Remove call to scev_reset_htab.
18109 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18111 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
18112 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
18113 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
18114 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
18115 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
18116 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
18117 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
18118 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
18119 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
18120 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
18121 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
18122 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
18123 (V_lane_reg): Likewise.
18124 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
18126 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
18127 (vfmal_lane_low<mode>_intrinsic,
18128 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
18129 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
18130 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
18131 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
18132 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
18133 vfmsl_lane_high<mode>_intrinsic): New define_insns.
18135 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18137 * config/arm/arm-cpus.in (fp16fml): New feature.
18138 (ALL_SIMD): Add fp16fml.
18139 (armv8.2-a): Add fp16fml as an option.
18140 (armv8.3-a): Likewise.
18141 (armv8.4-a): Add fp16fml as part of fp16.
18142 * config/arm/arm.h (TARGET_FP16FML): Define.
18143 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
18145 * config/arm/arm-modes.def (V2HF): Define.
18146 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
18147 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
18148 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
18149 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
18150 vfmsl_low, vfmsl_high): New set of builtins.
18151 * config/arm/iterators.md (PLUSMINUS): New code iterator.
18152 (vfml_op): New code attribute.
18153 (VFMLHALVES): New int iterator.
18154 (VFML, VFMLSEL): New mode attributes.
18155 (V_reg): Define mapping for V2HF.
18156 (V_hi, V_lo): New mode attributes.
18157 (VF_constraint): Likewise.
18158 (vfml_half, vfml_half_selector): New int attributes.
18159 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
18161 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
18162 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
18164 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
18165 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
18166 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
18167 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
18169 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
18170 Document new effective target and option set.
18172 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18174 * config/arm/arm-cpus.in (armv8_4): New feature.
18175 (ARMv8_4a): New fgroup.
18176 (armv8.4-a): New arch.
18177 * config/arm/arm-tables.opt: Regenerate.
18178 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
18179 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
18180 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
18181 Add matching rules for -march=armv8.4-a and extensions.
18182 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
18184 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
18187 * config/rx/rx.md (BW): New mode attribute.
18188 (sync_lock_test_and_setsi): Add mode suffix to insn output.
18190 2018-01-11 Richard Biener <rguenther@suse.de>
18192 PR tree-optimization/83435
18193 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
18194 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
18195 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
18197 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18198 Alan Hayward <alan.hayward@arm.com>
18199 David Sherwood <david.sherwood@arm.com>
18201 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
18203 (aarch64_classify_address): Initialize it. Track polynomial offsets.
18204 (aarch64_print_address_internal): Use it to check for a zero offset.
18206 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18207 Alan Hayward <alan.hayward@arm.com>
18208 David Sherwood <david.sherwood@arm.com>
18210 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
18211 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
18212 Return a poly_int64 rather than a HOST_WIDE_INT.
18213 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
18214 rather than a HOST_WIDE_INT.
18215 * config/aarch64/aarch64.h (aarch64_frame): Protect with
18216 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
18217 hard_fp_offset, frame_size, initial_adjust, callee_offset and
18218 final_offset from HOST_WIDE_INT to poly_int64.
18219 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
18220 to_constant when getting the number of units in an Advanced SIMD
18222 (aarch64_builtin_vectorized_function): Check for a constant number
18224 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
18226 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
18227 attribute instead of GET_MODE_NUNITS.
18228 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
18229 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
18230 GET_MODE_SIZE for fixed-size registers.
18231 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
18232 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
18233 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
18234 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
18235 (aarch64_print_operand, aarch64_print_address_internal)
18236 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
18237 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
18238 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
18239 Handle polynomial GET_MODE_SIZE.
18240 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
18241 wider than SImode without modification.
18242 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
18243 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
18244 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
18245 passing and returning SVE modes.
18246 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
18247 rather than GEN_INT.
18248 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
18249 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
18250 (aarch64_allocate_and_probe_stack_space): Likewise.
18251 (aarch64_layout_frame): Cope with polynomial offsets.
18252 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
18253 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
18254 polynomial offsets.
18255 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
18256 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
18257 poly_int64 rather than a HOST_WIDE_INT.
18258 (aarch64_get_separate_components, aarch64_process_components)
18259 (aarch64_expand_prologue, aarch64_expand_epilogue)
18260 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
18261 (aarch64_anchor_offset): New function, split out from...
18262 (aarch64_legitimize_address): ...here.
18263 (aarch64_builtin_vectorization_cost): Handle polynomial
18264 TYPE_VECTOR_SUBPARTS.
18265 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
18267 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
18268 number of elements from the PARALLEL rather than the mode.
18269 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
18270 rather than GET_MODE_BITSIZE.
18271 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
18272 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
18273 (aarch64_expand_vec_perm_const_1): Handle polynomial
18274 d->perm.length () and d->perm elements.
18275 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
18276 Apply to_constant to d->perm elements.
18277 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
18278 polynomial CONST_VECTOR_NUNITS.
18279 (aarch64_move_pointer): Take amount as a poly_int64 rather
18281 (aarch64_progress_pointer): Avoid temporary variable.
18282 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
18283 the mode attribute instead of GET_MODE.
18285 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18286 Alan Hayward <alan.hayward@arm.com>
18287 David Sherwood <david.sherwood@arm.com>
18289 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
18290 x exists before using it.
18291 (aarch64_add_constant_internal): Rename to...
18292 (aarch64_add_offset_1): ...this. Replace regnum with separate
18293 src and dest rtxes. Handle the case in which they're different,
18294 including when the offset is zero. Replace scratchreg with an rtx.
18295 Use 2 additions if there is no spare register into which we can
18296 move a 16-bit constant.
18297 (aarch64_add_constant): Delete.
18298 (aarch64_add_offset): Replace reg with separate src and dest
18299 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
18300 Use aarch64_add_offset_1.
18301 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
18302 an rtx rather than an int. Take the delta as a poly_int64
18303 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
18304 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
18305 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
18306 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
18307 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
18308 and aarch64_add_sp.
18309 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
18310 aarch64_add_constant.
18312 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18314 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
18315 Use scalar_float_mode.
18317 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18319 * config/aarch64/aarch64-simd.md
18320 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
18321 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
18322 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
18323 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
18324 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
18325 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
18326 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
18327 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
18328 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
18329 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
18331 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18334 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
18335 targ_options->x_arm_arch_string is non NULL.
18337 2018-01-11 Tamar Christina <tamar.christina@arm.com>
18339 * config/aarch64/aarch64.h
18340 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
18342 2018-01-11 Sudakshina Das <sudi.das@arm.com>
18345 * expmed.c (emit_store_flag_force): Swap if const op0
18346 and change VOIDmode to mode of op0.
18348 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18350 PR rtl-optimization/83761
18351 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
18352 than bytes to mode_for_size.
18354 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18356 PR middle-end/83189
18357 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
18358 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
18361 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18363 PR middle-end/83575
18364 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
18365 when in layout mode.
18366 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
18367 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
18370 2018-01-10 Michael Collison <michael.collison@arm.com>
18372 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
18373 * config/aarch64/aarch64-option-extension.def: Add
18374 AARCH64_OPT_EXTENSION of 'fp16fml'.
18375 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18376 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
18377 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
18378 * config/aarch64/constraints.md (Ui7): New constraint.
18379 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
18380 (VFMLA_SEL_W): Ditto.
18383 (VFMLA16_LOW): New int iterator.
18384 (VFMLA16_HIGH): Ditto.
18385 (UNSPEC_FMLAL): New unspec.
18386 (UNSPEC_FMLSL): Ditto.
18387 (UNSPEC_FMLAL2): Ditto.
18388 (UNSPEC_FMLSL2): Ditto.
18389 (f16mac): New code attribute.
18390 * config/aarch64/aarch64-simd-builtins.def
18391 (aarch64_fmlal_lowv2sf): Ditto.
18392 (aarch64_fmlsl_lowv2sf): Ditto.
18393 (aarch64_fmlalq_lowv4sf): Ditto.
18394 (aarch64_fmlslq_lowv4sf): Ditto.
18395 (aarch64_fmlal_highv2sf): Ditto.
18396 (aarch64_fmlsl_highv2sf): Ditto.
18397 (aarch64_fmlalq_highv4sf): Ditto.
18398 (aarch64_fmlslq_highv4sf): Ditto.
18399 (aarch64_fmlal_lane_lowv2sf): Ditto.
18400 (aarch64_fmlsl_lane_lowv2sf): Ditto.
18401 (aarch64_fmlal_laneq_lowv2sf): Ditto.
18402 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
18403 (aarch64_fmlalq_lane_lowv4sf): Ditto.
18404 (aarch64_fmlsl_lane_lowv4sf): Ditto.
18405 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
18406 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
18407 (aarch64_fmlal_lane_highv2sf): Ditto.
18408 (aarch64_fmlsl_lane_highv2sf): Ditto.
18409 (aarch64_fmlal_laneq_highv2sf): Ditto.
18410 (aarch64_fmlsl_laneq_highv2sf): Ditto.
18411 (aarch64_fmlalq_lane_highv4sf): Ditto.
18412 (aarch64_fmlsl_lane_highv4sf): Ditto.
18413 (aarch64_fmlalq_laneq_highv4sf): Ditto.
18414 (aarch64_fmlsl_laneq_highv4sf): Ditto.
18415 * config/aarch64/aarch64-simd.md:
18416 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
18417 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
18418 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
18419 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
18420 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
18421 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
18422 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
18423 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
18424 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
18425 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
18426 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
18427 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
18428 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
18429 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
18430 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
18431 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
18432 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
18433 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
18434 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
18435 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
18436 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
18437 (vfmlsl_low_u32): Ditto.
18438 (vfmlalq_low_u32): Ditto.
18439 (vfmlslq_low_u32): Ditto.
18440 (vfmlal_high_u32): Ditto.
18441 (vfmlsl_high_u32): Ditto.
18442 (vfmlalq_high_u32): Ditto.
18443 (vfmlslq_high_u32): Ditto.
18444 (vfmlal_lane_low_u32): Ditto.
18445 (vfmlsl_lane_low_u32): Ditto.
18446 (vfmlal_laneq_low_u32): Ditto.
18447 (vfmlsl_laneq_low_u32): Ditto.
18448 (vfmlalq_lane_low_u32): Ditto.
18449 (vfmlslq_lane_low_u32): Ditto.
18450 (vfmlalq_laneq_low_u32): Ditto.
18451 (vfmlslq_laneq_low_u32): Ditto.
18452 (vfmlal_lane_high_u32): Ditto.
18453 (vfmlsl_lane_high_u32): Ditto.
18454 (vfmlal_laneq_high_u32): Ditto.
18455 (vfmlsl_laneq_high_u32): Ditto.
18456 (vfmlalq_lane_high_u32): Ditto.
18457 (vfmlslq_lane_high_u32): Ditto.
18458 (vfmlalq_laneq_high_u32): Ditto.
18459 (vfmlslq_laneq_high_u32): Ditto.
18460 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
18461 (AARCH64_FL_FOR_ARCH8_4): New.
18462 (AARCH64_ISA_F16FML): New ISA flag.
18463 (TARGET_F16FML): New feature flag for fp16fml.
18464 (doc/invoke.texi): Document new fp16fml option.
18466 2018-01-10 Michael Collison <michael.collison@arm.com>
18468 * config/aarch64/aarch64-builtins.c:
18469 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
18470 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18471 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
18472 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
18473 (AARCH64_ISA_SHA3): New ISA flag.
18474 (TARGET_SHA3): New feature flag for sha3.
18475 * config/aarch64/iterators.md (sha512_op): New int attribute.
18476 (CRYPTO_SHA512): New int iterator.
18477 (UNSPEC_SHA512H): New unspec.
18478 (UNSPEC_SHA512H2): Ditto.
18479 (UNSPEC_SHA512SU0): Ditto.
18480 (UNSPEC_SHA512SU1): Ditto.
18481 * config/aarch64/aarch64-simd-builtins.def
18482 (aarch64_crypto_sha512hqv2di): New builtin.
18483 (aarch64_crypto_sha512h2qv2di): Ditto.
18484 (aarch64_crypto_sha512su0qv2di): Ditto.
18485 (aarch64_crypto_sha512su1qv2di): Ditto.
18486 (aarch64_eor3qv8hi): Ditto.
18487 (aarch64_rax1qv2di): Ditto.
18488 (aarch64_xarqv2di): Ditto.
18489 (aarch64_bcaxqv8hi): Ditto.
18490 * config/aarch64/aarch64-simd.md:
18491 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
18492 (aarch64_crypto_sha512su0qv2di): Ditto.
18493 (aarch64_crypto_sha512su1qv2di): Ditto.
18494 (aarch64_eor3qv8hi): Ditto.
18495 (aarch64_rax1qv2di): Ditto.
18496 (aarch64_xarqv2di): Ditto.
18497 (aarch64_bcaxqv8hi): Ditto.
18498 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
18499 (vsha512h2q_u64): Ditto.
18500 (vsha512su0q_u64): Ditto.
18501 (vsha512su1q_u64): Ditto.
18502 (veor3q_u16): Ditto.
18503 (vrax1q_u64): Ditto.
18504 (vxarq_u64): Ditto.
18505 (vbcaxq_u16): Ditto.
18506 * config/arm/types.md (crypto_sha512): New type attribute.
18507 (crypto_sha3): Ditto.
18508 (doc/invoke.texi): Document new sha3 option.
18510 2018-01-10 Michael Collison <michael.collison@arm.com>
18512 * config/aarch64/aarch64-builtins.c:
18513 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
18514 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18515 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
18516 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
18517 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
18518 (AARCH64_ISA_SM4): New ISA flag.
18519 (TARGET_SM4): New feature flag for sm4.
18520 * config/aarch64/aarch64-simd-builtins.def
18521 (aarch64_sm3ss1qv4si): Ditto.
18522 (aarch64_sm3tt1aq4si): Ditto.
18523 (aarch64_sm3tt1bq4si): Ditto.
18524 (aarch64_sm3tt2aq4si): Ditto.
18525 (aarch64_sm3tt2bq4si): Ditto.
18526 (aarch64_sm3partw1qv4si): Ditto.
18527 (aarch64_sm3partw2qv4si): Ditto.
18528 (aarch64_sm4eqv4si): Ditto.
18529 (aarch64_sm4ekeyqv4si): Ditto.
18530 * config/aarch64/aarch64-simd.md:
18531 (aarch64_sm3ss1qv4si): Ditto.
18532 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
18533 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
18534 (aarch64_sm4eqv4si): Ditto.
18535 (aarch64_sm4ekeyqv4si): Ditto.
18536 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
18537 (sm3part_op): Ditto.
18538 (CRYPTO_SM3TT): Ditto.
18539 (CRYPTO_SM3PART): Ditto.
18540 (UNSPEC_SM3SS1): New unspec.
18541 (UNSPEC_SM3TT1A): Ditto.
18542 (UNSPEC_SM3TT1B): Ditto.
18543 (UNSPEC_SM3TT2A): Ditto.
18544 (UNSPEC_SM3TT2B): Ditto.
18545 (UNSPEC_SM3PARTW1): Ditto.
18546 (UNSPEC_SM3PARTW2): Ditto.
18547 (UNSPEC_SM4E): Ditto.
18548 (UNSPEC_SM4EKEY): Ditto.
18549 * config/aarch64/constraints.md (Ui2): New constraint.
18550 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
18551 * config/arm/types.md (crypto_sm3): New type attribute.
18552 (crypto_sm4): Ditto.
18553 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
18554 (vsm3tt1aq_u32): Ditto.
18555 (vsm3tt1bq_u32): Ditto.
18556 (vsm3tt2aq_u32): Ditto.
18557 (vsm3tt2bq_u32): Ditto.
18558 (vsm3partw1q_u32): Ditto.
18559 (vsm3partw2q_u32): Ditto.
18560 (vsm4eq_u32): Ditto.
18561 (vsm4ekeyq_u32): Ditto.
18562 (doc/invoke.texi): Document new sm4 option.
18564 2018-01-10 Michael Collison <michael.collison@arm.com>
18566 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
18567 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
18568 (AARCH64_FL_FOR_ARCH8_4): New.
18569 (AARCH64_FL_V8_4): New flag.
18570 (doc/invoke.texi): Document new armv8.4-a option.
18572 2018-01-10 Michael Collison <michael.collison@arm.com>
18574 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18575 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
18576 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
18577 * config/aarch64/aarch64-option-extension.def: Add
18578 AARCH64_OPT_EXTENSION of 'sha2'.
18579 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
18580 (crypto): Disable sha2 and aes if crypto disabled.
18581 (crypto): Enable aes and sha2 if enabled.
18582 (simd): Disable sha2 and aes if simd disabled.
18583 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
18585 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
18586 (TARGET_SHA2): New feature flag for sha2.
18587 (TARGET_AES): New feature flag for aes.
18588 * config/aarch64/aarch64-simd.md:
18589 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
18590 conditional on TARGET_AES.
18591 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
18592 (aarch64_crypto_sha1hsi): Make pattern conditional
18594 (aarch64_crypto_sha1hv4si): Ditto.
18595 (aarch64_be_crypto_sha1hv4si): Ditto.
18596 (aarch64_crypto_sha1su1v4si): Ditto.
18597 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
18598 (aarch64_crypto_sha1su0v4si): Ditto.
18599 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
18600 (aarch64_crypto_sha256su0v4si): Ditto.
18601 (aarch64_crypto_sha256su1v4si): Ditto.
18602 (doc/invoke.texi): Document new aes and sha2 options.
18604 2018-01-10 Martin Sebor <msebor@redhat.com>
18606 PR tree-optimization/83781
18607 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
18610 2018-01-11 Martin Sebor <msebor@gmail.com>
18611 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18613 PR tree-optimization/83501
18614 PR tree-optimization/81703
18616 * tree-ssa-strlen.c (get_string_cst): Rename...
18617 (get_string_len): ...to this. Handle global constants.
18618 (handle_char_store): Adjust.
18620 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
18621 Jim Wilson <jimw@sifive.com>
18623 * config/riscv/riscv-protos.h (riscv_output_return): New.
18624 * config/riscv/riscv.c (struct machine_function): New naked_p field.
18625 (riscv_attribute_table, riscv_output_return),
18626 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
18627 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
18628 (riscv_compute_frame_info): Only compute frame->mask if not a naked
18630 (riscv_expand_prologue): Add early return for naked function.
18631 (riscv_expand_epilogue): Likewise.
18632 (riscv_function_ok_for_sibcall): Return false for naked function.
18633 (riscv_set_current_function): New.
18634 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
18635 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
18636 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
18637 * doc/extend.texi (RISC-V Function Attributes): New.
18639 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
18641 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
18642 check for 128-bit long double before checking TCmode.
18643 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
18644 128-bit long doubles before checking TFmode or TCmode.
18645 (FLOAT128_IBM_P): Likewise.
18647 2018-01-10 Martin Sebor <msebor@redhat.com>
18649 PR tree-optimization/83671
18650 * builtins.c (c_strlen): Unconditionally return zero for the empty
18652 Use -Warray-bounds for warnings.
18653 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
18654 for non-constant array indices with COMPONENT_REF, arrays of
18655 arrays, and pointers to arrays.
18656 (gimple_fold_builtin_strlen): Determine and set length range for
18657 non-constant character arrays.
18659 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
18661 PR middle-end/81897
18662 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
18665 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
18667 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
18669 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
18672 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
18673 VECTOR_MEM_ALTIVEC_OR_VSX_P.
18674 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
18675 indexed_or_indirect_operand predicate.
18676 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
18677 (*vsx_le_perm_load_v8hi): Likewise.
18678 (*vsx_le_perm_load_v16qi): Likewise.
18679 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
18680 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
18681 (*vsx_le_perm_store_v8hi): Likewise.
18682 (*vsx_le_perm_store_v16qi): Likewise.
18683 (eight unnamed splitters): Likewise.
18685 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
18687 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
18688 * config/rs6000/emmintrin.h: Likewise.
18689 * config/rs6000/mmintrin.h: Likewise.
18690 * config/rs6000/xmmintrin.h: Likewise.
18692 2018-01-10 David Malcolm <dmalcolm@redhat.com>
18695 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
18697 * tree.c (tree_nop_conversion): Return true for location wrapper
18699 (maybe_wrap_with_location): New function.
18700 (selftest::check_strip_nops): New function.
18701 (selftest::test_location_wrappers): New function.
18702 (selftest::tree_c_tests): Call it.
18703 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
18704 (maybe_wrap_with_location): New decl.
18705 (EXPR_LOCATION_WRAPPER_P): New macro.
18706 (location_wrapper_p): New inline function.
18707 (tree_strip_any_location_wrapper): New inline function.
18709 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
18712 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
18713 stack_realign_offset for the largest alignment of stack slot
18715 (ix86_find_max_used_stack_alignment): New function.
18716 (ix86_finalize_stack_frame_flags): Use it. Set
18717 max_used_stack_alignment if we don't realign stack.
18718 * config/i386/i386.h (machine_function): Add
18719 max_used_stack_alignment.
18721 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
18723 * config/arm/arm.opt (-mbranch-cost): New option.
18724 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
18727 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
18730 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
18731 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
18733 2018-01-10 Richard Biener <rguenther@suse.de>
18736 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
18737 early out so it also covers the case where we have a non-NULL
18740 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
18742 PR tree-optimization/83753
18743 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
18744 for non-strided grouped accesses if the number of elements is 1.
18746 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18749 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
18750 * i386.h (TARGET_USE_GATHER): Define.
18751 * x86-tune.def (X86_TUNE_USE_GATHER): New.
18753 2018-01-10 Martin Liska <mliska@suse.cz>
18756 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
18757 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
18759 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
18760 CLEANUP_NO_PARTITIONING is not set.
18762 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
18764 * doc/rtl.texi: Remove documentation of (const ...) wrappers
18765 for vectors, as a partial revert of r254296.
18766 * rtl.h (const_vec_p): Delete.
18767 (const_vec_duplicate_p): Don't test for vector CONSTs.
18768 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
18769 * expmed.c (make_tree): Likewise.
18772 * common.md (E, F): Use CONSTANT_P instead of checking for
18774 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
18775 checking for CONST_VECTOR.
18777 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18779 PR middle-end/83575
18780 * predict.c (force_edge_cold): Handle in more sane way edges
18781 with no prediction.
18783 2018-01-09 Carl Love <cel@us.ibm.com>
18785 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
18787 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
18788 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
18789 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
18790 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
18791 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
18792 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
18793 * config/rs6000/rs6000-protos.h: Add extern defition for
18794 rs6000_generate_float2_double_code.
18795 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
18797 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
18798 (float2_v2df): Add define_expand.
18800 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
18803 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
18804 op_mode in the force_to_mode call.
18806 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18808 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
18809 instead of checking each element individually.
18810 (aarch64_evpc_uzp): Likewise.
18811 (aarch64_evpc_zip): Likewise.
18812 (aarch64_evpc_ext): Likewise.
18813 (aarch64_evpc_rev): Likewise.
18814 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
18815 instead of checking each element individually. Return true without
18817 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
18818 whether all selected elements come from the same input, instead of
18819 checking each element individually. Remove calls to gen_rtx_REG,
18820 start_sequence and end_sequence and instead assert that no rtl is
18823 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18825 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
18826 order of HIGH and CONST checks.
18828 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18830 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
18831 if the destination isn't an SSA_NAME.
18833 2018-01-09 Richard Biener <rguenther@suse.de>
18835 PR tree-optimization/83668
18836 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
18838 (canonicalize_loop_form): ... here, renamed from ...
18839 (canonicalize_loop_closed_ssa_form): ... this and amended to
18840 swap successor edges for loop exit blocks to make us use
18841 the RPO order we need for initial schedule generation.
18843 2018-01-09 Joseph Myers <joseph@codesourcery.com>
18845 PR tree-optimization/64811
18846 * match.pd: When optimizing comparisons with Inf, avoid
18847 introducing or losing exceptions from comparisons with NaN.
18849 2018-01-09 Martin Liska <mliska@suse.cz>
18852 * asan.c (shadow_mem_size): Add gcc_assert.
18854 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
18856 Don't save registers in main().
18859 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
18860 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
18861 * config/avr/avr.c (avr_set_current_function): Don't error if
18862 naked, OS_task or OS_main are specified at the same time.
18863 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
18865 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
18867 * common/config/avr/avr-common.c (avr_option_optimization_table):
18868 Switch on -mmain-is-OS_task for optimizing compilations.
18870 2018-01-09 Richard Biener <rguenther@suse.de>
18872 PR tree-optimization/83572
18873 * graphite.c: Include cfganal.h.
18874 (graphite_transform_loops): Connect infinite loops to exit
18875 and remove fake edges at the end.
18877 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18879 * ipa-inline.c (edge_badness): Revert accidental checkin.
18881 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18884 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
18885 symbols; not inline clones.
18887 2018-01-09 Jakub Jelinek <jakub@redhat.com>
18890 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
18891 hard registers. Formatting fixes.
18893 PR preprocessor/83722
18894 * gcc.c (try_generate_repro): Pass
18895 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
18896 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
18899 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
18900 Kito Cheng <kito.cheng@gmail.com>
18902 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
18903 (riscv_leaf_function_p): Delete.
18904 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
18906 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18908 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
18910 (do_ifelse): New function.
18911 (do_isel): New function.
18912 (do_sub3): New function.
18913 (do_add3): New function.
18914 (do_load_mask_compare): New function.
18915 (do_overlap_load_compare): New function.
18916 (expand_compare_loop): New function.
18917 (expand_block_compare): Call expand_compare_loop() when appropriate.
18918 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
18919 option description.
18920 (-mblock-compare-inline-loop-limit): New option.
18922 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18925 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
18926 Reverse order of second and third operands in first alternative.
18927 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
18928 of first and second elements in UNSPEC_VPERMR vector.
18929 (altivec_expand_vec_perm_le): Likewise.
18931 2018-01-08 Jeff Law <law@redhat.com>
18933 PR rtl-optimizatin/81308
18934 * tree-switch-conversion.c (cfg_altered): New file scoped static.
18935 (process_switch): If group_case_labels makes a change, then set
18937 (pass_convert_switch::execute): If a switch is converted, then
18938 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
18940 PR rtl-optimization/81308
18941 * recog.c (split_all_insns): Conditionally cleanup the CFG after
18944 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
18946 PR target/83663 - Revert r255946
18947 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
18948 generation for cases where splatting a value is not useful.
18949 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
18950 across a vec_duplicate and a paradoxical subreg forming a vector
18951 mode to a vec_concat.
18953 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18955 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
18956 -march=armv8.3-a variants.
18957 * config/arm/t-multilib: Likewise.
18958 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
18960 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18962 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
18964 (cceq_ior_compare_complement): Give it a name so I can use it, and
18965 change boolean_or_operator predicate to boolean_operator so it can
18966 be used to generate a crand.
18967 (eqne): New code iterator.
18968 (bd/bd_neg): New code_attrs.
18969 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
18970 a single define_insn.
18971 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
18972 decrement (bdnzt/bdnzf/bdzt/bdzf).
18973 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
18974 with the new names of the branch decrement patterns, and added the
18975 names of the branch decrement conditional patterns.
18977 2018-01-08 Richard Biener <rguenther@suse.de>
18979 PR tree-optimization/83563
18980 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
18983 2018-01-08 Richard Biener <rguenther@suse.de>
18985 PR middle-end/83713
18986 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
18988 2018-01-08 Richard Biener <rguenther@suse.de>
18990 PR tree-optimization/83685
18991 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
18992 references to abnormals.
18994 2018-01-08 Richard Biener <rguenther@suse.de>
18997 * dwarf2out.c (output_indirect_strings): Handle empty
18998 skeleton_debug_str_hash.
18999 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
19001 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
19003 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
19004 (emit_store_direct): Likewise.
19005 (arc_trampoline_adjust_address): Likewise.
19006 (arc_asm_trampoline_template): New function.
19007 (arc_initialize_trampoline): Use asm_trampoline_template.
19008 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
19009 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
19010 * config/arc/arc.md (flush_icache): Delete pattern.
19012 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
19014 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
19015 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
19018 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
19021 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
19022 by not USED_FOR_TARGET.
19023 (make_pass_resolve_sw_modes): Likewise.
19025 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
19027 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
19030 2018-01-08 Richard Biener <rguenther@suse.de>
19032 PR middle-end/83580
19033 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
19035 2018-01-08 Richard Biener <rguenther@suse.de>
19037 PR middle-end/83517
19038 * match.pd ((t * 2) / 2) -> t): Add missing :c.
19040 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
19042 PR middle-end/81897
19043 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
19044 basic blocks with a small number of successors.
19045 (convert_control_dep_chain_into_preds): Improve handling of
19047 (dump_predicates): Split apart into...
19048 (dump_pred_chain): ...here...
19049 (dump_pred_info): ...and here.
19050 (can_one_predicate_be_invalidated_p): Add debugging printfs.
19051 (can_chain_union_be_invalidated_p): Improve check for invalidation
19053 (uninit_uses_cannot_happen): Avoid unnecessary if
19054 convert_control_dep_chain_into_preds yielded nothing.
19056 2018-01-06 Martin Sebor <msebor@redhat.com>
19058 PR tree-optimization/83640
19059 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
19060 subtracting negative offset from size.
19061 (builtin_access::overlap): Adjust offset bounds of the access to fall
19062 within the size of the object if possible.
19064 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
19066 PR rtl-optimization/83699
19067 * expmed.c (extract_bit_field_1): Restrict the vector usage of
19068 extract_bit_field_as_subreg to cases in which the extracted
19069 value is also a vector.
19071 * lra-constraints.c (process_alt_operands): Test for the equivalence
19072 substitutions when detecting a possible reload cycle.
19074 2018-01-06 Jakub Jelinek <jakub@redhat.com>
19077 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
19078 by default if flag_selective_schedling{,2}. Formatting fixes.
19080 PR rtl-optimization/83682
19081 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
19082 if it has non-VECTOR_MODE element mode.
19083 (vec_duplicate_p): Likewise.
19085 PR middle-end/83694
19086 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
19087 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
19089 2018-01-05 Jakub Jelinek <jakub@redhat.com>
19092 * config/i386/i386-builtin.def
19093 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
19094 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
19095 Require also OPTION_MASK_ISA_AVX512F in addition to
19096 OPTION_MASK_ISA_GFNI.
19097 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
19098 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
19099 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
19100 to OPTION_MASK_ISA_GFNI.
19101 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
19102 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
19103 OPTION_MASK_ISA_AVX512BW.
19104 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
19105 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
19106 addition to OPTION_MASK_ISA_GFNI.
19107 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
19108 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
19109 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
19110 to OPTION_MASK_ISA_GFNI.
19111 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
19112 a requirement for all ISAs rather than any of them with a few
19114 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
19116 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
19117 bitmasks to be enabled with 3 exceptions, instead of requiring any
19118 enabled ISA with lots of exceptions.
19119 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
19120 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
19121 Change avx512bw in isa attribute to avx512f.
19122 * config/i386/sgxintrin.h: Add license boilerplate.
19123 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
19124 to __AVX512F__ and __AVX512VL to __AVX512VL__.
19125 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
19126 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
19128 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
19129 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
19130 temporarily sse2 rather than sse if not enabled already.
19133 * config/i386/sse.md (VI248_VLBW): Rename to ...
19134 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
19135 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
19136 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
19137 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
19138 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
19139 mode iterator instead of VI248_VLBW.
19141 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
19143 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
19144 (record_modified): Skip clobbers; add debug output.
19145 (param_change_prob): Use sreal frequencies.
19147 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
19149 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
19150 punt for user-aligned variables.
19152 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
19154 * tree-chrec.c (chrec_contains_symbols): Return true for
19157 2018-01-05 Sudakshina Das <sudi.das@arm.com>
19160 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
19161 of (x|y) == x for BICS pattern.
19163 2018-01-05 Jakub Jelinek <jakub@redhat.com>
19165 PR tree-optimization/83605
19166 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
19167 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
19170 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
19172 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
19173 * config/epiphany/rtems.h: New file.
19175 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19176 Uros Bizjak <ubizjak@gmail.com>
19179 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
19180 QIreg_operand instead of register_operand predicate.
19181 * config/i386/i386.c (ix86_rop_should_change_byte_p,
19182 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
19183 comments instead of -fmitigate[-_]rop.
19185 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19188 * cgraphunit.c (symbol_table::compile): Switch to text_section
19189 before calling assembly_start debug hook.
19190 * run-rtl-passes.c (run_rtl_passes): Likewise.
19193 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19195 * tree-vrp.c (extract_range_from_binary_expr_1): Check
19196 range_int_cst_p rather than !symbolic_range_p before calling
19197 extract_range_from_multiplicative_op_1.
19199 2018-01-04 Jeff Law <law@redhat.com>
19201 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
19202 redundant test in assertion.
19204 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19206 * doc/rtl.texi: Document machine_mode wrapper classes.
19208 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19210 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
19211 using tree_to_uhwi.
19213 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19215 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
19216 the VEC_PERM_EXPR fold to fail.
19218 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19221 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
19222 to switched_sections.
19224 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19227 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
19228 test for d.testing.
19230 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
19233 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
19234 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
19236 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19239 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
19240 is BLKmode and bitpos not zero or mode change is needed.
19242 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19245 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
19248 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
19251 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
19252 instead of MULT rtx. Update all corresponding splitters.
19253 (*saddl_se): Ditto.
19254 (*ssub<modesuffix>): Ditto.
19255 (*ssubl_se): Ditto.
19256 (*cmp_sadd_di): Update split patterns.
19257 (*cmp_sadd_si): Ditto.
19258 (*cmp_sadd_sidi): Ditto.
19259 (*cmp_ssub_di): Ditto.
19260 (*cmp_ssub_si): Ditto.
19261 (*cmp_ssub_sidi): Ditto.
19262 * config/alpha/predicates.md (const23_operand): New predicate.
19263 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
19264 Look for ASHIFT, not MULT inner operand.
19265 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
19267 2018-01-04 Martin Liska <mliska@suse.cz>
19269 PR gcov-profile/83669
19270 * gcov.c (output_intermediate_file): Add version to intermediate
19272 * doc/gcov.texi: Document new field 'version' in intermediate
19273 file format. Fix location of '-k' option of gcov command.
19275 2018-01-04 Martin Liska <mliska@suse.cz>
19278 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
19280 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19282 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
19284 2018-01-03 Martin Sebor <msebor@redhat.com>
19286 PR tree-optimization/83655
19287 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
19288 checking calls with invalid arguments.
19290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19292 * tree-vect-stmts.c (vect_get_store_rhs): New function.
19293 (vectorizable_mask_load_store): Delete.
19294 (vectorizable_call): Return false for masked loads and stores.
19295 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
19296 instead of gimple_assign_rhs1.
19297 (vectorizable_load): Handle IFN_MASK_LOAD.
19298 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
19300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19302 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
19304 (vectorizable_mask_load_store): ...here.
19305 (vectorizable_load): ...and here.
19307 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19309 * tree-vect-stmts.c (vect_build_all_ones_mask)
19310 (vect_build_zero_merge_argument): New functions, split out from...
19311 (vectorizable_load): ...here.
19313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19315 * tree-vect-stmts.c (vect_check_store_rhs): New function,
19317 (vectorizable_mask_load_store): ...here.
19318 (vectorizable_store): ...and here.
19320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19322 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
19324 (vectorizable_mask_load_store): ...here.
19326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19328 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
19329 (vect_model_store_cost): Take a vec_load_store_type instead of a
19331 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
19332 (vect_model_store_cost): Take a vec_load_store_type instead of a
19334 (vectorizable_mask_load_store): Update accordingly.
19335 (vectorizable_store): Likewise.
19336 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
19338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19340 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
19341 IFN_MASK_LOAD calls here rather than...
19342 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
19344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19345 Alan Hayward <alan.hayward@arm.com>
19346 David Sherwood <david.sherwood@arm.com>
19348 * expmed.c (extract_bit_field_1): For vector extracts,
19349 fall back to extract_bit_field_as_subreg if vec_extract
19352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19353 Alan Hayward <alan.hayward@arm.com>
19354 David Sherwood <david.sherwood@arm.com>
19356 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
19357 they are variable or constant sized.
19358 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
19359 slots for constant-sized data.
19361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19362 Alan Hayward <alan.hayward@arm.com>
19363 David Sherwood <david.sherwood@arm.com>
19365 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
19366 handling COND_EXPRs with boolean comparisons, try to find a better
19367 basis for the mask type than the boolean itself.
19369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19371 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
19372 is calculated and how it can be overridden.
19373 * genmodes.c (max_bitsize_mode_any_mode): New variable.
19374 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
19376 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
19379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19380 Alan Hayward <alan.hayward@arm.com>
19381 David Sherwood <david.sherwood@arm.com>
19383 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
19384 Remove the mode argument.
19385 (aarch64_simd_valid_immediate): Remove the mode and inverse
19387 * config/aarch64/iterators.md (bitsize): New iterator.
19388 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
19389 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
19390 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
19391 aarch64_simd_valid_immediate.
19392 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
19393 (aarch64_reg_or_bic_imm): Likewise.
19394 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
19395 with an insn_type enum and msl with a modifier_type enum.
19396 Replace element_width with a scalar_mode. Change the shift
19397 to unsigned int. Add constructors for scalar_float_mode and
19398 scalar_int_mode elements.
19399 (aarch64_vect_float_const_representable_p): Delete.
19400 (aarch64_can_const_movi_rtx_p)
19401 (aarch64_simd_scalar_immediate_valid_for_move)
19402 (aarch64_simd_make_constant): Update call to
19403 aarch64_simd_valid_immediate.
19404 (aarch64_advsimd_valid_immediate_hs): New function.
19405 (aarch64_advsimd_valid_immediate): Likewise.
19406 (aarch64_simd_valid_immediate): Remove mode and inverse
19407 arguments. Rewrite to use the above. Use const_vec_duplicate_p
19408 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
19409 and aarch64_float_const_representable_p on the result.
19410 (aarch64_output_simd_mov_immediate): Remove mode argument.
19411 Update call to aarch64_simd_valid_immediate and use of
19412 simd_immediate_info.
19413 (aarch64_output_scalar_simd_mov_immediate): Update call
19416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19417 Alan Hayward <alan.hayward@arm.com>
19418 David Sherwood <david.sherwood@arm.com>
19420 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
19421 (mode_nunits): Likewise CONST_MODE_NUNITS.
19422 * machmode.def (ADJUST_NUNITS): Document.
19423 * genmodes.c (mode_data::need_nunits_adj): New field.
19424 (blank_mode): Update accordingly.
19425 (adj_nunits): New variable.
19426 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
19428 (emit_mode_size_inline): Set need_bytesize_adj for all modes
19429 listed in adj_nunits.
19430 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
19431 listed in adj_nunits. Don't emit case statements for such modes.
19432 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
19433 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
19434 nothing if adj_nunits is nonnull.
19435 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
19436 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
19437 (emit_mode_fbit): Update use of print_maybe_const_decl.
19438 (emit_move_size): Likewise. Treat the array as non-const
19440 (emit_mode_adjustments): Handle adj_nunits.
19442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19444 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
19445 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
19446 (VECTOR_MODES): Use it.
19447 (make_vector_modes): Take the prefix as an argument.
19449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19450 Alan Hayward <alan.hayward@arm.com>
19451 David Sherwood <david.sherwood@arm.com>
19453 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
19454 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
19455 for MODE_VECTOR_BOOL.
19456 * machmode.def (VECTOR_BOOL_MODE): Document.
19457 * genmodes.c (VECTOR_BOOL_MODE): New macro.
19458 (make_vector_bool_mode): New function.
19459 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
19461 * lto-streamer-in.c (lto_input_mode_table): Likewise.
19462 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
19464 * stor-layout.c (int_mode_for_mode): Likewise.
19465 * tree.c (build_vector_type_for_mode): Likewise.
19466 * varasm.c (output_constant_pool_2): Likewise.
19467 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
19468 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
19469 for MODE_VECTOR_BOOL.
19470 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
19471 of mode class checks.
19472 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
19473 instead of a list of mode class checks.
19474 (expand_vector_scalar_condition): Likewise.
19475 (type_for_widest_vector_mode): Handle BImode as an inner mode.
19477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19478 Alan Hayward <alan.hayward@arm.com>
19479 David Sherwood <david.sherwood@arm.com>
19481 * machmode.h (mode_size): Change from unsigned short to
19483 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
19484 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
19485 or if measurement_type is not polynomial.
19486 (fixed_size_mode::includes_p): Check for constant-sized modes.
19487 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
19488 return a poly_uint16 rather than an unsigned short.
19489 (emit_mode_size): Change the type of mode_size from unsigned short
19490 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
19491 (emit_mode_adjustments): Cope with polynomial vector sizes.
19492 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19494 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19496 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
19497 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
19498 * caller-save.c (setup_save_areas): Likewise.
19499 (replace_reg_with_saved_mem): Likewise.
19500 * calls.c (emit_library_call_value_1): Likewise.
19501 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
19502 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
19503 (gen_lowpart_for_combine): Likewise.
19504 * convert.c (convert_to_integer_1): Likewise.
19505 * cse.c (equiv_constant, cse_insn): Likewise.
19506 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
19507 (cselib_subst_to_values): Likewise.
19508 * dce.c (word_dce_process_block): Likewise.
19509 * df-problems.c (df_word_lr_mark_ref): Likewise.
19510 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
19511 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
19512 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
19513 (rtl_for_decl_location): Likewise.
19514 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
19515 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
19516 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
19517 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
19518 (expand_expr_real_1): Likewise.
19519 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
19520 (pad_below): Likewise.
19521 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
19522 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
19523 * ira.c (get_subreg_tracking_sizes): Likewise.
19524 * ira-build.c (ira_create_allocno_objects): Likewise.
19525 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
19526 (ira_sort_regnos_for_alter_reg): Likewise.
19527 * ira-costs.c (record_operand_costs): Likewise.
19528 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
19529 (resolve_simple_move): Likewise.
19530 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
19531 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
19532 (lra_constraints): Likewise.
19533 (CONST_POOL_OK_P): Reject variable-sized modes.
19534 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
19535 (add_pseudo_to_slot, lra_spill): Likewise.
19536 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19537 * optabs-query.c (get_best_extraction_insn): Likewise.
19538 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
19539 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
19540 (expand_mult_highpart, valid_multiword_target_p): Likewise.
19541 * recog.c (offsettable_address_addr_space_p): Likewise.
19542 * regcprop.c (maybe_mode_change): Likewise.
19543 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
19544 * regrename.c (build_def_use): Likewise.
19545 * regstat.c (dump_reg_info): Likewise.
19546 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
19547 (find_reloads, find_reloads_subreg_address): Likewise.
19548 * reload1.c (eliminate_regs_1): Likewise.
19549 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
19550 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
19551 (simplify_binary_operation_1, simplify_subreg): Likewise.
19552 * targhooks.c (default_function_arg_padding): Likewise.
19553 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
19554 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
19555 (verify_gimple_assign_ternary): Likewise.
19556 * tree-inline.c (estimate_move_cost): Likewise.
19557 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19558 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
19559 (get_address_cost_ainc): Likewise.
19560 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
19561 (vect_supportable_dr_alignment): Likewise.
19562 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
19563 (vectorizable_reduction): Likewise.
19564 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
19565 (vectorizable_operation, vectorizable_load): Likewise.
19566 * tree.c (build_same_sized_truth_vector_type): Likewise.
19567 * valtrack.c (cleanup_auto_inc_dec): Likewise.
19568 * var-tracking.c (emit_note_insn_var_location): Likewise.
19569 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
19570 (ADDR_VEC_ALIGN): Likewise.
19572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19573 Alan Hayward <alan.hayward@arm.com>
19574 David Sherwood <david.sherwood@arm.com>
19576 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
19578 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
19579 or if measurement_type is polynomial.
19580 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
19581 * combine.c (make_extraction): Likewise.
19582 * dse.c (find_shift_sequence): Likewise.
19583 * dwarf2out.c (mem_loc_descriptor): Likewise.
19584 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
19585 (extract_bit_field, extract_low_bits): Likewise.
19586 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
19587 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
19588 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
19589 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
19590 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
19591 * reload.c (find_reloads): Likewise.
19592 * reload1.c (alter_reg): Likewise.
19593 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
19594 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
19595 * tree-if-conv.c (predicate_mem_writes): Likewise.
19596 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19597 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
19598 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
19599 * valtrack.c (dead_debug_insert_temp): Likewise.
19600 * varasm.c (mergeable_constant_section): Likewise.
19601 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
19603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19604 Alan Hayward <alan.hayward@arm.com>
19605 David Sherwood <david.sherwood@arm.com>
19607 * expr.c (expand_assignment): Cope with polynomial mode sizes
19608 when assigning to a CONCAT.
19610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19611 Alan Hayward <alan.hayward@arm.com>
19612 David Sherwood <david.sherwood@arm.com>
19614 * machmode.h (mode_precision): Change from unsigned short to
19616 (mode_to_precision): Return a poly_uint16 rather than an unsigned
19618 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
19619 or if measurement_type is not polynomial.
19620 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
19621 in which the mode is already known to be a scalar_int_mode.
19622 * genmodes.c (emit_mode_precision): Change the type of mode_precision
19623 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
19625 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19626 for GET_MODE_PRECISION.
19627 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19628 for GET_MODE_PRECISION.
19629 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
19631 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
19632 (expand_field_assignment, make_extraction): Likewise.
19633 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
19634 (get_last_value): Likewise.
19635 * convert.c (convert_to_integer_1): Likewise.
19636 * cse.c (cse_insn): Likewise.
19637 * expr.c (expand_expr_real_1): Likewise.
19638 * lra-constraints.c (simplify_operand_subreg): Likewise.
19639 * optabs-query.c (can_atomic_load_p): Likewise.
19640 * optabs.c (expand_atomic_load): Likewise.
19641 (expand_atomic_store): Likewise.
19642 * ree.c (combine_reaching_defs): Likewise.
19643 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
19644 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
19645 * tree.h (type_has_mode_precision_p): Likewise.
19646 * ubsan.c (instrument_si_overflow): Likewise.
19648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19649 Alan Hayward <alan.hayward@arm.com>
19650 David Sherwood <david.sherwood@arm.com>
19652 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
19653 polynomial numbers of units.
19654 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
19655 (valid_vector_subparts_p): New function.
19656 (build_vector_type): Remove temporary shim and take the number
19657 of units as a poly_uint64 rather than an int.
19658 (build_opaque_vector_type): Take the number of units as a
19659 poly_uint64 rather than an int.
19660 * tree.c (build_vector_from_ctor): Handle polynomial
19661 TYPE_VECTOR_SUBPARTS.
19662 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
19663 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
19664 (build_vector_from_val): If the number of units is variable,
19665 use build_vec_duplicate_cst for constant operands and
19666 VEC_DUPLICATE_EXPR otherwise.
19667 (make_vector_type): Remove temporary is_constant ().
19668 (build_vector_type, build_opaque_vector_type): Take the number of
19669 units as a poly_uint64 rather than an int.
19670 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
19672 * cfgexpand.c (expand_debug_expr): Likewise.
19673 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
19674 (store_constructor, expand_expr_real_1): Likewise.
19675 (const_scalar_mask_from_tree): Likewise.
19676 * fold-const-call.c (fold_const_reduction): Likewise.
19677 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
19678 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
19679 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
19680 (fold_relational_const): Likewise.
19681 (native_interpret_vector): Likewise. Change the size from an
19682 int to an unsigned int.
19683 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
19684 TYPE_VECTOR_SUBPARTS.
19685 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
19686 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
19687 duplicating a non-constant operand into a variable-length vector.
19688 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
19689 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
19690 * ipa-icf.c (sem_variable::equals): Likewise.
19691 * match.pd: Likewise.
19692 * omp-simd-clone.c (simd_clone_subparts): Likewise.
19693 * print-tree.c (print_node): Likewise.
19694 * stor-layout.c (layout_type): Likewise.
19695 * targhooks.c (default_builtin_vectorization_cost): Likewise.
19696 * tree-cfg.c (verify_gimple_comparison): Likewise.
19697 (verify_gimple_assign_binary): Likewise.
19698 (verify_gimple_assign_ternary): Likewise.
19699 (verify_gimple_assign_single): Likewise.
19700 * tree-pretty-print.c (dump_generic_node): Likewise.
19701 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19702 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
19703 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
19704 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
19705 (vect_shift_permute_load_chain): Likewise.
19706 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
19707 (expand_vector_condition, optimize_vector_constructor): Likewise.
19708 (lower_vec_perm, get_compute_type): Likewise.
19709 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
19710 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
19711 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
19712 (vect_recog_mask_conversion_pattern): Likewise.
19713 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
19714 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
19715 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19716 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
19717 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
19718 (vectorizable_shift, vectorizable_operation, vectorizable_store)
19719 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
19720 (supportable_widening_operation): Likewise.
19721 (supportable_narrowing_operation): Likewise.
19722 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
19724 * varasm.c (output_constant): Likewise.
19726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19727 Alan Hayward <alan.hayward@arm.com>
19728 David Sherwood <david.sherwood@arm.com>
19730 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
19731 so that both the length == 3 and length != 3 cases set up their
19732 own permute vectors. Add comments explaining why we know the
19733 number of elements is constant.
19734 (vect_permute_load_chain): Likewise.
19736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19737 Alan Hayward <alan.hayward@arm.com>
19738 David Sherwood <david.sherwood@arm.com>
19740 * machmode.h (mode_nunits): Change from unsigned char to
19742 (ONLY_FIXED_SIZE_MODES): New macro.
19743 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
19744 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
19745 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
19747 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
19748 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
19749 or if measurement_type is not polynomial.
19750 * genmodes.c (ZERO_COEFFS): New macro.
19751 (emit_mode_nunits_inline): Make mode_nunits_inline return a
19753 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
19754 Use ZERO_COEFFS when emitting initializers.
19755 * data-streamer.h (bp_pack_poly_value): New function.
19756 (bp_unpack_poly_value): Likewise.
19757 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
19758 for GET_MODE_NUNITS.
19759 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
19760 for GET_MODE_NUNITS.
19761 * tree.c (make_vector_type): Remove temporary shim and make
19762 the real function take the number of units as a poly_uint64
19763 rather than an int.
19764 (build_vector_type_for_mode): Handle polynomial nunits.
19765 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
19766 * emit-rtl.c (const_vec_series_p_1): Likewise.
19767 (gen_rtx_CONST_VECTOR): Likewise.
19768 * fold-const.c (test_vec_duplicate_folding): Likewise.
19769 * genrecog.c (validate_pattern): Likewise.
19770 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
19771 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
19772 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
19773 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
19774 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
19775 * rtlanal.c (subreg_get_info): Likewise.
19776 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19777 (vect_grouped_load_supported): Likewise.
19778 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19779 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
19780 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19781 (simplify_const_unary_operation, simplify_binary_operation_1)
19782 (simplify_const_binary_operation, simplify_ternary_operation)
19783 (test_vector_ops_duplicate, test_vector_ops): Likewise.
19784 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
19785 instead of CONST_VECTOR_NUNITS.
19786 * varasm.c (output_constant_pool_2): Likewise.
19787 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
19788 explicit-encoded elements in the XVEC for variable-length vectors.
19790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19792 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
19794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19795 Alan Hayward <alan.hayward@arm.com>
19796 David Sherwood <david.sherwood@arm.com>
19798 * coretypes.h (fixed_size_mode): Declare.
19799 (fixed_size_mode_pod): New typedef.
19800 * builtins.h (target_builtins::x_apply_args_mode)
19801 (target_builtins::x_apply_result_mode): Change type to
19802 fixed_size_mode_pod.
19803 * builtins.c (apply_args_size, apply_result_size, result_vector)
19804 (expand_builtin_apply_args_1, expand_builtin_apply)
19805 (expand_builtin_return): Update accordingly.
19807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19809 * cse.c (hash_rtx_cb): Hash only the encoded elements.
19810 * cselib.c (cselib_hash_rtx): Likewise.
19811 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
19812 CONST_VECTOR encoding.
19814 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19815 Jeff Law <law@redhat.com>
19818 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
19819 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
19820 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
19821 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
19824 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
19825 explicitly probe *sp in a noreturn function if there were any callee
19826 register saves or frame pointer is needed.
19828 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19831 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
19832 BLKmode for ternary, binary or unary expressions.
19835 * var-tracking.c (delete_vta_debug_insn): New inline function.
19836 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
19837 insns from get_insns () to NULL instead of each bb separately.
19838 Use delete_vta_debug_insn. No longer static.
19839 (vt_debug_insns_local, variable_tracking_main_1): Adjust
19840 delete_vta_debug_insns callers.
19841 * rtl.h (delete_vta_debug_insns): Declare.
19842 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
19843 instead of variable_tracking_main.
19845 2018-01-03 Martin Sebor <msebor@redhat.com>
19847 PR tree-optimization/83603
19848 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
19849 arguments past the endof the argument list in functions declared
19850 without a prototype.
19851 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
19852 Avoid checking when arguments are null.
19854 2018-01-03 Martin Sebor <msebor@redhat.com>
19857 * doc/extend.texi (attribute const): Fix a typo.
19858 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
19859 issuing -Wsuggest-attribute for void functions.
19861 2018-01-03 Martin Sebor <msebor@redhat.com>
19863 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
19864 offset_int::from instead of wide_int::to_shwi.
19865 (maybe_diag_overlap): Remove assertion.
19866 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
19867 * gimple-ssa-sprintf.c (format_directive): Same.
19868 (parse_directive): Same.
19869 (sprintf_dom_walker::compute_format_length): Same.
19870 (try_substitute_return_value): Same.
19872 2018-01-03 Jeff Law <law@redhat.com>
19874 PR middle-end/83654
19875 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
19876 non-constant residual for zero at runtime and avoid probing in
19877 that case. Reorganize code for trailing problem to mirror handling
19880 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19882 PR tree-optimization/83501
19883 * tree-ssa-strlen.c (get_string_cst): New.
19884 (handle_char_store): Call get_string_cst.
19886 2018-01-03 Martin Liska <mliska@suse.cz>
19888 PR tree-optimization/83593
19889 * tree-ssa-strlen.c: Include tree-cfg.h.
19890 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
19891 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
19892 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
19894 (strlen_dom_walker::before_dom_children): Call
19895 gimple_purge_dead_eh_edges. Dump tranformation with details
19897 (strlen_dom_walker::before_dom_children): Update call by adding
19898 new argument cleanup_eh.
19899 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
19901 2018-01-03 Martin Liska <mliska@suse.cz>
19904 * cif-code.def (VARIADIC_THUNK): New enum value.
19905 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
19908 2018-01-03 Jan Beulich <jbeulich@suse.com>
19910 * sse.md (mov<mode>_internal): Tighten condition for when to use
19911 vmovdqu<ssescalarsize> for TI and OI modes.
19913 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19915 Update copyright years.
19917 2018-01-03 Martin Liska <mliska@suse.cz>
19920 * ipa-visibility.c (function_and_variable_visibility): Skip
19921 functions with noipa attribure.
19923 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19925 * gcc.c (process_command): Update copyright notice dates.
19926 * gcov-dump.c (print_version): Ditto.
19927 * gcov.c (print_version): Ditto.
19928 * gcov-tool.c (print_version): Ditto.
19929 * gengtype.c (create_file): Ditto.
19930 * doc/cpp.texi: Bump @copying's copyright year.
19931 * doc/cppinternals.texi: Ditto.
19932 * doc/gcc.texi: Ditto.
19933 * doc/gccint.texi: Ditto.
19934 * doc/gcov.texi: Ditto.
19935 * doc/install.texi: Ditto.
19936 * doc/invoke.texi: Ditto.
19938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19940 * vector-builder.h (vector_builder::m_full_nelts): Change from
19941 unsigned int to poly_uint64.
19942 (vector_builder::full_nelts): Update prototype accordingly.
19943 (vector_builder::new_vector): Likewise.
19944 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
19945 (vector_builder::operator ==): Likewise.
19946 (vector_builder::finalize): Likewise.
19947 * int-vector-builder.h (int_vector_builder::int_vector_builder):
19948 Take the number of elements as a poly_uint64 rather than an
19950 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
19951 from unsigned int to poly_uint64.
19952 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
19953 (vec_perm_indices::new_vector): Likewise.
19954 (vec_perm_indices::length): Likewise.
19955 (vec_perm_indices::nelts_per_input): Likewise.
19956 (vec_perm_indices::input_nelts): Likewise.
19957 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
19958 number of elements per input as a poly_uint64 rather than an
19959 unsigned int. Use the original encoding for variable-length
19960 vectors, rather than clamping each individual element.
19961 For the second and subsequent elements in each pattern,
19962 clamp the step and base before clamping their sum.
19963 (vec_perm_indices::series_p): Handle polynomial element counts.
19964 (vec_perm_indices::all_in_range_p): Likewise.
19965 (vec_perm_indices_to_tree): Likewise.
19966 (vec_perm_indices_to_rtx): Likewise.
19967 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19968 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
19969 (tree_vector_builder::new_binary_operation): Handle polynomial
19970 element counts. Return false if we need to know the number
19971 of elements at compile time.
19972 * fold-const.c (fold_vec_perm): Punt if the number of elements
19973 isn't known at compile time.
19975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19977 * vec-perm-indices.h (vec_perm_builder): Change element type
19978 from HOST_WIDE_INT to poly_int64.
19979 (vec_perm_indices::element_type): Update accordingly.
19980 (vec_perm_indices::clamp): Handle polynomial element_types.
19981 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19982 (vec_perm_indices::all_in_range_p): Likewise.
19983 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
19985 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
19986 polynomial vec_perm_indices element types.
19987 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
19988 * fold-const.c (fold_vec_perm): Likewise.
19989 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
19990 * tree-vect-generic.c (lower_vec_perm): Likewise.
19991 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19992 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
19993 element type to HOST_WIDE_INT.
19995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19996 Alan Hayward <alan.hayward@arm.com>
19997 David Sherwood <david.sherwood@arm.com>
19999 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
20000 rather than an int. Use plus_constant.
20001 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
20002 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
20004 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20005 Alan Hayward <alan.hayward@arm.com>
20006 David Sherwood <david.sherwood@arm.com>
20008 * calls.c (emit_call_1, expand_call): Change struct_value_size from
20009 a HOST_WIDE_INT to a poly_int64.
20011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20012 Alan Hayward <alan.hayward@arm.com>
20013 David Sherwood <david.sherwood@arm.com>
20015 * calls.c (load_register_parameters): Cope with polynomial
20016 mode sizes. Require a constant size for BLKmode parameters
20017 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
20018 forces a parameter to be padded at the lsb end in order to
20019 fill a complete number of words, require the parameter size
20020 to be ordered wrt UNITS_PER_WORD.
20022 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20023 Alan Hayward <alan.hayward@arm.com>
20024 David Sherwood <david.sherwood@arm.com>
20026 * reload1.c (spill_stack_slot_width): Change element type
20027 from unsigned int to poly_uint64_pod.
20028 (alter_reg): Treat mode sizes as polynomial.
20030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20031 Alan Hayward <alan.hayward@arm.com>
20032 David Sherwood <david.sherwood@arm.com>
20034 * reload.c (complex_word_subreg_p): New function.
20035 (reload_inner_reg_of_subreg, push_reload): Use it.
20037 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20038 Alan Hayward <alan.hayward@arm.com>
20039 David Sherwood <david.sherwood@arm.com>
20041 * lra-constraints.c (process_alt_operands): Reject matched
20042 operands whose sizes aren't ordered.
20043 (match_reload): Refer to this check here.
20045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20046 Alan Hayward <alan.hayward@arm.com>
20047 David Sherwood <david.sherwood@arm.com>
20049 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
20050 that the mode size is in the set {1, 2, 4, 8, 16}.
20052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20053 Alan Hayward <alan.hayward@arm.com>
20054 David Sherwood <david.sherwood@arm.com>
20056 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
20057 Use plus_constant instead of gen_rtx_PLUS.
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 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
20064 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
20065 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
20066 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
20067 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
20068 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
20069 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
20070 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
20071 * config/i386/i386.c (ix86_push_rounding): ...this new function.
20072 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
20074 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
20075 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
20076 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
20077 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
20078 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
20079 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
20080 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
20081 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
20082 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
20083 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
20085 * expr.c (emit_move_resolve_push): Treat the input and result
20086 of PUSH_ROUNDING as a poly_int64.
20087 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
20088 (emit_push_insn): Likewise.
20089 * lra-eliminations.c (mark_not_eliminable): Likewise.
20090 * recog.c (push_operand): Likewise.
20091 * reload1.c (elimination_effects): Likewise.
20092 * rtlanal.c (nonzero_bits1): Likewise.
20093 * calls.c (store_one_arg): Likewise. Require the padding to be
20094 known at compile time.
20096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20097 Alan Hayward <alan.hayward@arm.com>
20098 David Sherwood <david.sherwood@arm.com>
20100 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
20101 Use plus_constant instead of gen_rtx_PLUS.
20103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20104 Alan Hayward <alan.hayward@arm.com>
20105 David Sherwood <david.sherwood@arm.com>
20107 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
20108 rather than an int.
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 (expand_expr_real_1): Use tree_to_poly_uint64
20115 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
20116 via stack temporaries. Treat the mode size as polynomial too.
20118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20119 Alan Hayward <alan.hayward@arm.com>
20120 David Sherwood <david.sherwood@arm.com>
20122 * expr.c (expand_expr_real_2): When handling conversions involving
20123 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
20124 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
20125 as a poly_uint64 too.
20127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20128 Alan Hayward <alan.hayward@arm.com>
20129 David Sherwood <david.sherwood@arm.com>
20131 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
20133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20134 Alan Hayward <alan.hayward@arm.com>
20135 David Sherwood <david.sherwood@arm.com>
20137 * combine.c (can_change_dest_mode): Handle polynomial
20138 REGMODE_NATURAL_SIZE.
20139 * expmed.c (store_bit_field_1): Likewise.
20140 * expr.c (store_constructor): Likewise.
20141 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
20142 and polynomial REGMODE_NATURAL_SIZE.
20143 (gen_lowpart_common): Likewise.
20144 * reginfo.c (record_subregs_of_mode): Likewise.
20145 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
20152 numbers of elements.
20154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20155 Alan Hayward <alan.hayward@arm.com>
20156 David Sherwood <david.sherwood@arm.com>
20158 * match.pd: Cope with polynomial numbers of vector elements.
20160 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20161 Alan Hayward <alan.hayward@arm.com>
20162 David Sherwood <david.sherwood@arm.com>
20164 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
20165 in a POINTER_PLUS_EXPR.
20167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20168 Alan Hayward <alan.hayward@arm.com>
20169 David Sherwood <david.sherwood@arm.com>
20171 * omp-simd-clone.c (simd_clone_subparts): New function.
20172 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
20173 (ipa_simd_modify_function_body): Likewise.
20175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20176 Alan Hayward <alan.hayward@arm.com>
20177 David Sherwood <david.sherwood@arm.com>
20179 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
20180 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
20181 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
20182 (expand_vector_condition, vector_element): Likewise.
20183 (subparts_gt): New function.
20184 (get_compute_type): Use subparts_gt.
20185 (count_type_subparts): Delete.
20186 (expand_vector_operations_1): Use subparts_gt instead of
20187 count_type_subparts.
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-data-refs.c (vect_no_alias_p): Replace with...
20194 (vect_compile_time_alias): ...this new function. Do the calculation
20195 on poly_ints rather than trees.
20196 (vect_prune_runtime_alias_test_list): Update call accordingly.
20198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20199 Alan Hayward <alan.hayward@arm.com>
20200 David Sherwood <david.sherwood@arm.com>
20202 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
20204 (vect_schedule_slp_instance): Likewise.
20206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20207 Alan Hayward <alan.hayward@arm.com>
20208 David Sherwood <david.sherwood@arm.com>
20210 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
20211 constant and extern definitions for variable-length vectors.
20212 (vect_get_constant_vectors): Note that the number of units
20213 is known to be constant.
20215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20216 Alan Hayward <alan.hayward@arm.com>
20217 David Sherwood <david.sherwood@arm.com>
20219 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
20220 of units as polynomial. Choose between WIDE and NARROW based
20223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20224 Alan Hayward <alan.hayward@arm.com>
20225 David Sherwood <david.sherwood@arm.com>
20227 * tree-vect-stmts.c (simd_clone_subparts): New function.
20228 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
20230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20231 Alan Hayward <alan.hayward@arm.com>
20232 David Sherwood <david.sherwood@arm.com>
20234 * tree-vect-stmts.c (vectorizable_call): Treat the number of
20235 vectors as polynomial. Use build_index_vector for
20236 IFN_GOMP_SIMD_LANE.
20238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20239 Alan Hayward <alan.hayward@arm.com>
20240 David Sherwood <david.sherwood@arm.com>
20242 * tree-vect-stmts.c (get_load_store_type): Treat the number of
20243 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
20244 for variable-length vectors.
20245 (vectorizable_mask_load_store): Treat the number of units as
20246 polynomial, asserting that it is constant if the condition has
20247 already been enforced.
20248 (vectorizable_store, vectorizable_load): Likewise.
20250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20251 Alan Hayward <alan.hayward@arm.com>
20252 David Sherwood <david.sherwood@arm.com>
20254 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
20255 of units as polynomial. Punt if we can't tell at compile time
20256 which vector contains the final result.
20258 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20259 Alan Hayward <alan.hayward@arm.com>
20260 David Sherwood <david.sherwood@arm.com>
20262 * tree-vect-loop.c (vectorizable_induction): Treat the number
20263 of units as polynomial. Punt on SLP inductions. Use an integer
20264 VEC_SERIES_EXPR for variable-length integer reductions. Use a
20265 cast of such a series for variable-length floating-point
20268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20269 Alan Hayward <alan.hayward@arm.com>
20270 David Sherwood <david.sherwood@arm.com>
20272 * tree.h (build_index_vector): Declare.
20273 * tree.c (build_index_vector): New function.
20274 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
20275 of units as polynomial, forcibly converting it to a constant if
20276 vectorizable_reduction has already enforced the condition.
20277 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
20278 to create a {1,2,3,...} vector.
20279 (vectorizable_reduction): Treat the number of units as polynomial.
20280 Choose vectype_in based on the largest scalar element size rather
20281 than the smallest number of units. Enforce the restrictions
20284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20285 Alan Hayward <alan.hayward@arm.com>
20286 David Sherwood <david.sherwood@arm.com>
20288 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
20289 number of units as polynomial.
20291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20292 Alan Hayward <alan.hayward@arm.com>
20293 David Sherwood <david.sherwood@arm.com>
20295 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
20296 * target.def (autovectorize_vector_sizes): Return the vector sizes
20297 by pointer, using vector_sizes rather than a bitmask.
20298 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
20299 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
20300 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
20302 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
20303 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
20304 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
20305 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
20306 * omp-general.c (omp_max_vf): Likewise.
20307 * omp-low.c (omp_clause_aligned_alignment): Likewise.
20308 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
20309 * tree-vect-loop.c (vect_analyze_loop): Likewise.
20310 * tree-vect-slp.c (vect_slp_bb): Likewise.
20311 * doc/tm.texi: Regenerate.
20312 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
20314 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
20315 the vector size as a poly_uint64 rather than an unsigned int.
20316 (current_vector_size): Change from an unsigned int to a poly_uint64.
20317 (get_vectype_for_scalar_type): Update accordingly.
20318 * tree.h (build_truth_vector_type): Take the size and number of
20319 units as a poly_uint64 rather than an unsigned int.
20320 (build_vector_type): Add a temporary overload that takes
20321 the number of units as a poly_uint64 rather than an unsigned int.
20322 * tree.c (make_vector_type): Likewise.
20323 (build_truth_vector_type): Take the number of units as a poly_uint64
20324 rather than an unsigned int.
20326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20327 Alan Hayward <alan.hayward@arm.com>
20328 David Sherwood <david.sherwood@arm.com>
20330 * target.def (get_mask_mode): Take the number of units and length
20331 as poly_uint64s rather than unsigned ints.
20332 * targhooks.h (default_get_mask_mode): Update accordingly.
20333 * targhooks.c (default_get_mask_mode): Likewise.
20334 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
20335 * doc/tm.texi: Regenerate.
20337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20338 Alan Hayward <alan.hayward@arm.com>
20339 David Sherwood <david.sherwood@arm.com>
20341 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
20342 * omp-general.c (omp_max_vf): Likewise.
20343 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
20344 (expand_omp_simd): Handle polynomial safelen.
20345 * omp-low.c (omplow_simd_context): Add a default constructor.
20346 (omplow_simd_context::max_vf): Change from int to poly_uint64.
20347 (lower_rec_simd_input_clauses): Update accordingly.
20348 (lower_rec_input_clauses): Likewise.
20350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20351 Alan Hayward <alan.hayward@arm.com>
20352 David Sherwood <david.sherwood@arm.com>
20354 * tree-vectorizer.h (vect_nunits_for_cost): New function.
20355 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
20356 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
20357 (vect_analyze_slp_cost): Likewise.
20358 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
20359 (vect_model_load_cost): Likewise.
20361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20362 Alan Hayward <alan.hayward@arm.com>
20363 David Sherwood <david.sherwood@arm.com>
20365 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
20366 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
20367 from an unsigned int * to a poly_uint64_pod *.
20368 (calculate_unrolling_factor): New function.
20369 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
20371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20372 Alan Hayward <alan.hayward@arm.com>
20373 David Sherwood <david.sherwood@arm.com>
20375 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
20376 from an unsigned int to a poly_uint64.
20377 (_loop_vec_info::slp_unrolling_factor): Likewise.
20378 (_loop_vec_info::vectorization_factor): Change from an int
20380 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
20381 (vect_get_num_vectors): New function.
20382 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
20383 (vect_get_num_copies): Use vect_get_num_vectors.
20384 (vect_analyze_data_ref_dependences): Change max_vf from an int *
20385 to an unsigned int *.
20386 (vect_analyze_data_refs): Change min_vf from an int * to a
20388 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
20389 than an unsigned HOST_WIDE_INT.
20390 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
20391 (vect_analyze_data_ref_dependence): Change max_vf from an int *
20392 to an unsigned int *.
20393 (vect_analyze_data_ref_dependences): Likewise.
20394 (vect_compute_data_ref_alignment): Handle polynomial vf.
20395 (vect_enhance_data_refs_alignment): Likewise.
20396 (vect_prune_runtime_alias_test_list): Likewise.
20397 (vect_shift_permute_load_chain): Likewise.
20398 (vect_supportable_dr_alignment): Likewise.
20399 (dependence_distance_ge_vf): Take the vectorization factor as a
20400 poly_uint64 rather than an unsigned HOST_WIDE_INT.
20401 (vect_analyze_data_refs): Change min_vf from an int * to a
20403 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
20404 vfm1 as a poly_uint64 rather than an int. Make the same change
20405 for the returned bound_scalar.
20406 (vect_gen_vector_loop_niters): Handle polynomial vf.
20407 (vect_do_peeling): Likewise. Update call to
20408 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
20409 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
20411 * tree-vect-loop.c (vect_determine_vectorization_factor)
20412 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
20413 (vect_get_known_peeling_cost): Likewise.
20414 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
20415 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
20416 (vect_transform_loop): Likewise. Use the lowest possible VF when
20417 updating the upper bounds of the loop.
20418 (vect_min_worthwhile_factor): Make static. Return an unsigned int
20419 rather than an int.
20420 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
20421 polynomial unroll factors.
20422 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
20423 (vect_make_slp_decision): Likewise.
20424 (vect_supported_load_permutation_p): Likewise, and polynomial
20426 (vect_analyze_slp_cost): Handle polynomial vf.
20427 (vect_slp_analyze_node_operations): Likewise.
20428 (vect_slp_analyze_bb_1): Likewise.
20429 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
20430 than an unsigned HOST_WIDE_INT.
20431 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
20432 (vectorizable_load): Handle polynomial vf.
20433 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
20435 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
20437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20438 Alan Hayward <alan.hayward@arm.com>
20439 David Sherwood <david.sherwood@arm.com>
20441 * match.pd: Handle bit operations involving three constants
20442 and try to fold one pair.
20444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20446 * tree-vect-loop-manip.c: Include gimple-fold.h.
20447 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
20448 niters_maybe_zero parameters. Handle other cases besides a step of 1.
20449 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
20450 Add a path that uses a step of VF instead of 1, but disable it
20452 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
20453 and niters_no_overflow parameters. Update calls to
20454 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
20455 Create a new SSA name if the latter choses to use a ste other
20456 than zero, and return it via niters_vector_mult_vf_var.
20457 * tree-vect-loop.c (vect_transform_loop): Update calls to
20458 vect_do_peeling, vect_gen_vector_loop_niters and
20459 slpeel_make_loop_iterate_ntimes.
20460 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
20461 (vect_gen_vector_loop_niters): Update declarations after above changes.
20463 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
20465 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
20466 128-bit round to integer instructions.
20467 (ceil<mode>2): Likewise.
20468 (btrunc<mode>2): Likewise.
20469 (round<mode>2): Likewise.
20471 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20473 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
20474 unaligned VSX load/store on P8/P9.
20475 (expand_block_clear): Allow the use of unaligned VSX
20476 load/store on P8/P9.
20478 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20480 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
20482 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
20483 swap associated with both a load and a store.
20485 2018-01-02 Andrew Waterman <andrew@sifive.com>
20487 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
20488 * config/riscv/riscv.md (clear_cache): Use it.
20490 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
20492 * web.c: Remove out-of-date comment.
20494 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20496 * expr.c (fixup_args_size_notes): Check that any existing
20497 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
20498 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
20499 (emit_single_push_insn): ...here.
20501 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20503 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
20504 (const_vector_encoded_nelts): New function.
20505 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
20506 (const_vector_int_elt, const_vector_elt): Declare.
20507 * emit-rtl.c (const_vector_int_elt_1): New function.
20508 (const_vector_elt): Likewise.
20509 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
20510 of CONST_VECTOR_ELT.
20512 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20514 * expr.c: Include rtx-vector-builder.h.
20515 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
20516 directly on the tree encoding.
20517 (const_vector_from_tree): Likewise.
20518 * optabs.c: Include rtx-vector-builder.h.
20519 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
20520 sequence of "u" values.
20521 * vec-perm-indices.c: Include rtx-vector-builder.h.
20522 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
20523 directly on the vec_perm_indices encoding.
20525 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20527 * doc/rtl.texi (const_vector): Describe new encoding scheme.
20528 * Makefile.in (OBJS): Add rtx-vector-builder.o.
20529 * rtx-vector-builder.h: New file.
20530 * rtx-vector-builder.c: Likewise.
20531 * rtl.h (rtx_def::u2): Add a const_vector field.
20532 (CONST_VECTOR_NPATTERNS): New macro.
20533 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
20534 (CONST_VECTOR_DUPLICATE_P): Likewise.
20535 (CONST_VECTOR_STEPPED_P): Likewise.
20536 (CONST_VECTOR_ENCODED_ELT): Likewise.
20537 (const_vec_duplicate_p): Check for a duplicated vector encoding.
20538 (unwrap_const_vec_duplicate): Likewise.
20539 (const_vec_series_p): Check for a non-duplicated vector encoding.
20540 Say that the function only returns true for integer vectors.
20541 * emit-rtl.c: Include rtx-vector-builder.h.
20542 (gen_const_vec_duplicate_1): Delete.
20543 (gen_const_vector): Call gen_const_vec_duplicate instead of
20544 gen_const_vec_duplicate_1.
20545 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
20546 (gen_const_vec_duplicate): Use rtx_vector_builder.
20547 (gen_const_vec_series): Likewise.
20548 (gen_rtx_CONST_VECTOR): Likewise.
20549 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
20550 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
20551 Build a new vector rather than modifying a CONST_VECTOR in-place.
20552 (handle_special_swappables): Update call accordingly.
20553 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
20554 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
20555 Build a new vector rather than modifying a CONST_VECTOR in-place.
20556 (handle_special_swappables): Update call accordingly.
20558 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20560 * simplify-rtx.c (simplify_const_binary_operation): Use
20561 CONST_VECTOR_ELT instead of XVECEXP.
20563 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20565 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
20566 the selector elements to be different from the data elements
20567 if the selector is a VECTOR_CST.
20568 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
20569 ssizetype for the selector.
20571 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20573 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
20574 before testing each element individually.
20575 * tree-vect-generic.c (lower_vec_perm): Likewise.
20577 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20579 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
20580 * selftest-run-tests.c (selftest::run_tests): Call it.
20581 * vector-builder.h (vector_builder::operator ==): New function.
20582 (vector_builder::operator !=): Likewise.
20583 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
20584 (vec_perm_indices::all_from_input_p): New function.
20585 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
20586 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
20587 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
20588 instead of reading the VECTOR_CST directly. Detect whether both
20589 vector inputs are the same before constructing the vec_perm_indices,
20590 and update the number of inputs argument accordingly. Use the
20591 utility functions added above. Only construct sel2 if we need to.
20593 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20595 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
20596 the broadcast of the low byte.
20597 (expand_mult_highpart): Use an explicit encoding for the permutes.
20598 * optabs-query.c (can_mult_highpart_p): Likewise.
20599 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
20600 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20601 (vectorizable_bswap): Likewise.
20602 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
20603 explicit encoding for the power-of-2 permutes.
20604 (vect_permute_store_chain): Likewise.
20605 (vect_grouped_load_supported): Likewise.
20606 (vect_permute_load_chain): Likewise.
20608 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20610 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
20611 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
20612 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
20613 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
20614 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
20615 (vect_gen_perm_mask_any): Likewise.
20617 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20619 * int-vector-builder.h: New file.
20620 * vec-perm-indices.h: Include int-vector-builder.h.
20621 (vec_perm_indices): Redefine as an int_vector_builder.
20622 (auto_vec_perm_indices): Delete.
20623 (vec_perm_builder): Redefine as a stand-alone class.
20624 (vec_perm_indices::vec_perm_indices): New function.
20625 (vec_perm_indices::clamp): Likewise.
20626 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
20627 (vec_perm_indices::new_vector): New function.
20628 (vec_perm_indices::new_expanded_vector): Update for new
20629 vec_perm_indices class.
20630 (vec_perm_indices::rotate_inputs): New function.
20631 (vec_perm_indices::all_in_range_p): Operate directly on the
20632 encoded form, without computing elided elements.
20633 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
20634 encoding. Update for new vec_perm_indices class.
20635 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
20636 the given vec_perm_builder.
20637 (expand_vec_perm_var): Update vec_perm_builder constructor.
20638 (expand_mult_highpart): Use vec_perm_builder instead of
20639 auto_vec_perm_indices.
20640 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
20641 vec_perm_indices instead of auto_vec_perm_indices. Use a single
20642 or double series encoding as appropriate.
20643 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
20644 vec_perm_indices instead of auto_vec_perm_indices.
20645 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20646 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20647 (vect_permute_store_chain): Likewise.
20648 (vect_grouped_load_supported): Likewise.
20649 (vect_permute_load_chain): Likewise.
20650 (vect_shift_permute_load_chain): Likewise.
20651 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
20652 (vect_transform_slp_perm_load): Likewise.
20653 (vect_schedule_slp_instance): Likewise.
20654 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20655 (vectorizable_mask_load_store): Likewise.
20656 (vectorizable_bswap): Likewise.
20657 (vectorizable_store): Likewise.
20658 (vectorizable_load): Likewise.
20659 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
20660 vec_perm_indices instead of auto_vec_perm_indices. Use
20661 tree_to_vec_perm_builder to read the vector from a tree.
20662 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
20663 vec_perm_builder instead of a vec_perm_indices.
20664 (have_whole_vector_shift): Use vec_perm_builder and
20665 vec_perm_indices instead of auto_vec_perm_indices. Leave the
20666 truncation to calc_vec_perm_mask_for_shift.
20667 (vect_create_epilog_for_reduction): Likewise.
20668 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
20669 from auto_vec_perm_indices to vec_perm_indices.
20670 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
20671 instead of changing individual elements.
20672 (aarch64_vectorize_vec_perm_const): Use new_vector to install
20673 the vector in d.perm.
20674 * config/arm/arm.c (expand_vec_perm_d::perm): Change
20675 from auto_vec_perm_indices to vec_perm_indices.
20676 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
20677 instead of changing individual elements.
20678 (arm_vectorize_vec_perm_const): Use new_vector to install
20679 the vector in d.perm.
20680 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
20681 Update vec_perm_builder constructor.
20682 (rs6000_expand_interleave): Likewise.
20683 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
20684 (rs6000_expand_interleave): Likewise.
20686 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20688 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
20689 to qimode could truncate the indices.
20690 * optabs.c (expand_vec_perm_var): Likewise.
20692 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20694 * Makefile.in (OBJS): Add vec-perm-indices.o.
20695 * vec-perm-indices.h: New file.
20696 * vec-perm-indices.c: Likewise.
20697 * target.h (vec_perm_indices): Replace with a forward class
20699 (auto_vec_perm_indices): Move to vec-perm-indices.h.
20700 * optabs.h: Include vec-perm-indices.h.
20701 (expand_vec_perm): Delete.
20702 (selector_fits_mode_p, expand_vec_perm_var): Declare.
20703 (expand_vec_perm_const): Declare.
20704 * target.def (vec_perm_const_ok): Replace with...
20705 (vec_perm_const): ...this new hook.
20706 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
20707 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
20708 * doc/tm.texi: Regenerate.
20709 * optabs.def (vec_perm_const): Delete.
20710 * doc/md.texi (vec_perm_const): Likewise.
20711 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
20712 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
20713 expand_vec_perm for constant permutation vectors. Assert that
20714 the mode of variable permutation vectors is the integer equivalent
20715 of the mode that is being permuted.
20716 * optabs-query.h (selector_fits_mode_p): Declare.
20717 * optabs-query.c: Include vec-perm-indices.h.
20718 (selector_fits_mode_p): New function.
20719 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
20720 is defined, instead of checking whether the vec_perm_const_optab
20721 exists. Use targetm.vectorize.vec_perm_const instead of
20722 targetm.vectorize.vec_perm_const_ok. Check whether the indices
20723 fit in the vector mode before using a variable permute.
20724 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
20725 vec_perm_indices instead of an rtx.
20726 (expand_vec_perm): Replace with...
20727 (expand_vec_perm_const): ...this new function. Take the selector
20728 as a vec_perm_indices rather than an rtx. Also take the mode of
20729 the selector. Update call to shift_amt_for_vec_perm_mask.
20730 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
20731 Use vec_perm_indices::new_expanded_vector to expand the original
20732 selector into bytes. Check whether the indices fit in the vector
20733 mode before using a variable permute.
20734 (expand_vec_perm_var): Make global.
20735 (expand_mult_highpart): Use expand_vec_perm_const.
20736 * fold-const.c: Includes vec-perm-indices.h.
20737 * tree-ssa-forwprop.c: Likewise.
20738 * tree-vect-data-refs.c: Likewise.
20739 * tree-vect-generic.c: Likewise.
20740 * tree-vect-loop.c: Likewise.
20741 * tree-vect-slp.c: Likewise.
20742 * tree-vect-stmts.c: Likewise.
20743 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
20745 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
20746 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
20747 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
20748 (aarch64_vectorize_vec_perm_const): ...this new function.
20749 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
20750 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20751 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
20752 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
20753 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
20754 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20755 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
20757 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
20758 check for NEON modes.
20759 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
20760 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
20761 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
20762 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
20764 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
20765 the old VEC_PERM_CONST conditions.
20766 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
20767 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
20768 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
20769 (ia64_vectorize_vec_perm_const_ok): Merge into...
20770 (ia64_vectorize_vec_perm_const): ...this new function.
20771 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
20772 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
20773 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
20774 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
20775 * config/mips/mips.c (mips_expand_vec_perm_const)
20776 (mips_vectorize_vec_perm_const_ok): Merge into...
20777 (mips_vectorize_vec_perm_const): ...this new function.
20778 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
20779 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
20780 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
20781 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
20782 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
20783 (rs6000_expand_vec_perm_const): Delete.
20784 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
20786 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20787 (altivec_expand_vec_perm_const_le): Take each operand individually.
20788 Operate on constant selectors rather than rtxes.
20789 (altivec_expand_vec_perm_const): Likewise. Update call to
20790 altivec_expand_vec_perm_const_le.
20791 (rs6000_expand_vec_perm_const): Delete.
20792 (rs6000_vectorize_vec_perm_const_ok): Delete.
20793 (rs6000_vectorize_vec_perm_const): New function.
20794 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
20795 an element count and rtx array.
20796 (rs6000_expand_extract_even): Update call accordingly.
20797 (rs6000_expand_interleave): Likewise.
20798 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
20799 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
20800 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
20801 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
20802 (rs6000_expand_vec_perm_const): Delete.
20803 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
20804 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20805 (altivec_expand_vec_perm_const_le): Take each operand individually.
20806 Operate on constant selectors rather than rtxes.
20807 (altivec_expand_vec_perm_const): Likewise. Update call to
20808 altivec_expand_vec_perm_const_le.
20809 (rs6000_expand_vec_perm_const): Delete.
20810 (rs6000_vectorize_vec_perm_const_ok): Delete.
20811 (rs6000_vectorize_vec_perm_const): New function. Remove stray
20812 reference to the SPE evmerge intructions.
20813 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
20814 an element count and rtx array.
20815 (rs6000_expand_extract_even): Update call accordingly.
20816 (rs6000_expand_interleave): Likewise.
20817 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
20818 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
20820 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20822 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20824 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
20825 vector mode and that that mode matches the mode of the data
20827 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
20828 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
20829 directly using expand_vec_perm_1 when forcing selectors into
20831 (expand_vec_perm_var): New function, split out from expand_vec_perm.
20833 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20835 * optabs-query.h (can_vec_perm_p): Delete.
20836 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
20837 * optabs-query.c (can_vec_perm_p): Split into...
20838 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
20839 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
20840 particular selector is valid.
20841 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20842 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20843 (vect_grouped_load_supported): Likewise.
20844 (vect_shift_permute_load_chain): Likewise.
20845 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
20846 (vect_transform_slp_perm_load): Likewise.
20847 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20848 (vectorizable_bswap): Likewise.
20849 (vect_gen_perm_mask_checked): Likewise.
20850 * fold-const.c (fold_ternary_loc): Likewise. Don't take
20851 implementations of variable permutation vectors into account
20852 when deciding which selector to use.
20853 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
20854 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
20855 with a false third argument.
20856 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
20857 to test whether the constant selector is valid and can_vec_perm_var_p
20858 to test whether a variable selector is valid.
20860 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20862 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
20863 * optabs-query.c (can_vec_perm_p): Likewise.
20864 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
20865 instead of vec_perm_indices.
20866 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
20867 (vect_gen_perm_mask_checked): Likewise,
20868 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
20869 (vect_gen_perm_mask_checked): Likewise,
20871 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20873 * optabs-query.h (qimode_for_vec_perm): Declare.
20874 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
20875 (qimode_for_vec_perm): ...this new function.
20876 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
20878 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20880 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
20881 does not have a conditional at the top.
20883 2018-01-02 Richard Biener <rguenther@suse.de>
20885 * ipa-inline.c (big_speedup_p): Fix expression.
20887 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
20890 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
20893 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
20897 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
20898 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
20899 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
20900 cond_taken_branch_cost 3->4.
20902 2018-01-01 Jakub Jelinek <jakub@redhat.com>
20904 PR tree-optimization/83581
20905 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
20906 TODO_cleanup_cfg if any changes have been made.
20908 PR middle-end/83608
20909 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
20910 convert_modes if target mode has the right side, but different mode
20913 PR middle-end/83609
20914 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
20915 last argument when extracting from CONCAT. If either from_real or
20916 from_imag is NULL, use expansion through memory. If result is not
20917 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
20918 the parts directly to inner mode, if even that fails, use expansion
20921 PR middle-end/83623
20922 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
20923 check for bswap in mode rather than HImode and use that in expand_unop
20926 Copyright (C) 2018 Free Software Foundation, Inc.
20928 Copying and distribution of this file, with or without modification,
20929 are permitted in any medium without royalty provided the copyright
20930 notice and this notice are preserved.