1 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
2 Bo Zhou <zbo.zhou@hisilicon.com>
4 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
5 * config/aarch64/aarch64-tune.md: Regenerated.
6 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
7 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
8 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
10 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
11 Julian Brown <julian@codesourcery.com>
13 * builtins.c (get_builtin_sync_mem): Handle address spaces.
15 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
17 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
18 if the call takes a static chain.
20 2018-09-21 Martin Liska <mliska@suse.cz>
22 * auto-profile.c (autofdo_source_profile::read): Do not
24 (read_profile): Do not add working sets.
25 (read_autofdo_file): Remove sum_all.
26 (afdo_callsite_hot_enough_for_early_inline): Remove const
28 * coverage.c (struct counts_entry): Remove gcov_summary.
29 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
30 do not support GCOV_TAG_PROGRAM_SUMMARY.
31 (get_coverage_counts): Remove summary and expected
33 * coverage.h (get_coverage_counts): Likewise.
34 * doc/gcov-dump.texi: Remove -w option.
35 * gcov-dump.c (dump_working_sets): Remove.
36 (main): Do not support '-w' option.
37 (print_usage): Likewise.
38 (tag_summary): Likewise.
39 * gcov-io.c (gcov_write_summary): Do not dump
41 (gcov_read_summary): Likewise.
42 (gcov_histo_index): Remove.
43 (gcov_histogram_merge): Likewise.
44 (compute_working_sets): Likewise.
45 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
47 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
48 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
49 (GCOV_HISTOGRAM_SIZE): Remove.
50 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
51 (struct gcov_summary): Simplify rapidly just
52 to runs and sum_max fields.
53 (gcov_histo_index): Remove.
54 (NUM_GCOV_WORKING_SETS): Likewise.
55 (compute_working_sets): Likewise.
56 * gcov-tool.c (print_overlap_usage_message): Remove
58 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
59 (output_lines): Remove program related line.
60 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
61 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
63 (input_profile_summary): Do not read it.
64 (merge_profile_summaries): And do not merge it.
65 (input_symtab): Do not call removed function.
66 * modulo-sched.c (sms_schedule): Do not print sum_max.
67 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
68 removed when histogram method was invented.
69 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
71 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
73 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
74 and dump selected value.
75 * profile.c (add_working_set): Remove.
76 (get_working_sets): Likewise.
77 (find_working_set): Likewise.
78 (get_exec_counts): Do not work with working sets.
79 (read_profile_edge_counts): Do not inform as sum_max is removed.
80 (compute_branch_probabilities): Likewise.
81 (compute_value_histograms): Remove argument for call of
83 * profile.h: Do not make gcov_summary const.
85 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
87 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
89 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
91 PR tree-optimization/86990
92 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
93 Check that the entire merged store group is made of constants only for
96 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
98 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
99 (VTABLE_VERIFICATION_SPEC): Likewise.
100 (SANITIZER_EARLY_SPEC): Likewise.
101 (SANITIZER_SPEC): Likewise.
102 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
103 * doc/invoke.texi (Link Options): Document -r.
105 2018-09-20 Richard Biener <rguenther@suse.de>
108 * gimplify.c (gimplify_expr): Retain alignment of
109 addressable lvalue in dereference.
111 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
114 * configure.ac: Check for .loc is_stmt support.
115 * configure, config.in: Rebuilt.
116 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
119 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
121 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
122 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
125 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
127 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
129 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
132 2018-09-20 Marek Polacek <polacek@redhat.com>
134 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
136 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
138 PR tree-optimization/87288
139 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
140 into account when determining PEELING_FOR_NITERS.
142 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
144 PR tree-optimization/86877
145 * tree-vect-loop.c (vect_analyze_loop_2): Call
146 vect_verify_datarefs_alignment.
148 2018-09-19 Marek Polacek <polacek@redhat.com>
150 * doc/invoke.texi: Document -Wclass-conversion.
152 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
154 * config/pa/pa.c (pa_adjust_priority): Delete.
155 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
157 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
158 (atomic_storehi): Likewise.
159 (atomic_storesi): Likewise.
160 (atomic_loaddi): Restore compare and swap exchange loop code.
162 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
164 PR rtl-optimization/86902
165 * combine.c (try_combine): When changing the CC mode used, don't change
166 an unrelated mode in other_insn to that new CC mode.
168 2018-09-19 David Malcolm <dmalcolm@redhat.com>
170 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
171 with %T in place of calls to dump_generic_expr.
172 (prune_runtime_alias_test_list): Likewise.
173 (create_runtime_alias_checks): Likewise.
174 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
175 (vect_analyze_data_ref_dependence): Likewise.
176 (vect_slp_analyze_data_ref_dependence): Likewise.
177 (vect_record_base_alignment): Likewise. Use %G in place of call
179 (vect_compute_data_ref_alignment): Likewise.
180 (verify_data_ref_alignment): Likewise.
181 (vect_find_same_alignment_drs): Likewise.
182 (vect_analyze_group_access_1): Likewise.
183 (vect_analyze_data_ref_accesses): Likewise.
184 (dependence_distance_ge_vf): Likewise.
185 (dump_lower_bound): Likewise.
186 (vect_prune_runtime_alias_test_list): Likewise.
187 (vect_find_stmt_data_reference): Likewise.
188 (vect_analyze_data_refs): Likewise.
189 (vect_create_addr_base_for_vector_ref): Likewise.
190 (vect_create_data_ref_ptr): Likewise.
191 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
192 (vect_can_advance_ivs_p): Likewise.
193 (vect_update_ivs_after_vectorizer): Likewise.
194 (vect_gen_prolog_loop_niters): Likewise.
195 (vect_prepare_for_masked_peels): Likewise.
196 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
197 (vect_determine_vectorization_factor): Likewise.
198 (vect_is_simple_iv_evolution): Likewise.
199 (vect_analyze_scalar_cycles_1): Likewise.
200 (vect_analyze_loop_operations): Likewise.
201 (report_vect_op): Likewise.
202 (vect_is_slp_reduction): Likewise.
203 (check_reduction_path): Likewise.
204 (vect_is_simple_reduction): Likewise.
205 (vect_create_epilog_for_reduction): Likewise.
206 (vect_finalize_reduction:): Likewise.
207 (vectorizable_induction): Likewise.
208 (vect_transform_loop_stmt): Likewise.
209 (vect_transform_loop): Likewise.
210 (optimize_mask_stores): Likewise.
211 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
212 (vect_split_statement): Likewise.
213 (vect_recog_over_widening_pattern): Likewise.
214 (vect_recog_average_pattern): Likewise.
215 (vect_determine_min_output_precision_1): Likewise.
216 (vect_determine_precisions_from_range): Likewise.
217 (vect_determine_precisions_from_users): Likewise.
218 (vect_mark_pattern_stmts): Likewise.
219 (vect_pattern_recog_1): Likewise.
220 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
221 (vect_record_max_nunits): Likewise.
222 (vect_build_slp_tree_1): Likewise.
223 (vect_build_slp_tree_2): Likewise.
224 (vect_print_slp_tree): Likewise.
225 (vect_analyze_slp_instance): Likewise.
226 (vect_detect_hybrid_slp_stmts): Likewise.
227 (vect_detect_hybrid_slp_1): Likewise.
228 (vect_slp_analyze_operations): Likewise.
229 (vect_slp_analyze_bb_1): Likewise.
230 (vect_transform_slp_perm_load): Likewise.
231 (vect_schedule_slp_instance): Likewise.
232 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
233 (vect_mark_stmts_to_be_vectorized): Likewise.
234 (vect_init_vector_1): Likewise.
235 (vect_get_vec_def_for_operand): Likewise.
236 (vect_finish_stmt_generation_1): Likewise.
237 (vect_check_load_store_mask): Likewise.
238 (vectorizable_call): Likewise.
239 (vectorizable_conversion): Likewise.
240 (vectorizable_operation): Likewise.
241 (vectorizable_load): Likewise.
242 (vect_analyze_stmt): Likewise.
243 (vect_is_simple_use): Likewise.
244 (vect_get_vector_types_for_stmt): Likewise.
245 (vect_get_mask_type_for_stmt): Likewise.
246 * tree-vectorizer.c (increase_alignment): Likewise.
248 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
250 * doc/rtl.texi: Adjust vec_select description.
251 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
252 non-constant selectors.
254 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
256 * config/aarch64/aarch64-protos.h
257 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
258 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
259 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
260 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
261 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
262 (AARCH64_FL_PROFILE): Move index so flags are ordered.
263 (AARCH64_ISA_RCPC8_4): New flag.
264 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
265 to aarch64_offset_9bit_signed_unscaled_p.
266 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
268 * config/aarch64/constraints.md (Ust): New constraint.
269 * config/aarch64/predicates.md.
270 (aarch64_9bit_offset_memory_operand): New predicate.
271 (aarch64_rcpc_memory_operand): New predicate.
273 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
275 PR rtl-optimization/87361
276 * rtlanal.c (nonzero_bits1): Revert accidental change.
278 2018-09-19 Richard Biener <rguenther@suse.de>
280 PR tree-optimization/87349
281 PR tree-optimization/87342
282 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
284 2018-09-18 Marek Polacek <polacek@redhat.com>
286 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
287 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
289 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
291 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
293 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
295 PR rtl-optimization/86882
296 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
298 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
300 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
301 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
303 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
306 * doc/invoke.texi (Link Options): Fix formatting and grammar.
308 2018-09-18 Richard Biener <rguenther@suse.de>
311 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
312 (compute_samebase_partition_bases): Likewise.
313 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
314 (gimple_can_coalesce_p): Simplify.
316 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
318 Handle a library implementation of ffs calling __builtin_ffs.
319 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
320 (mmix_init_libfuncs): New function: make __builtin_ffs expand
323 2018-09-17 David Malcolm <dmalcolm@redhat.com>
325 * diagnostic-show-locus.c (class layout_range): Add field
327 (layout_range::layout_range): Add "original_idx" param and use it
328 to initialize new field.
329 (make_range): Use 0 for original_idx.
330 (layout::layout): Pass in index to calls to
331 maybe_add_location_range.
332 (layout::maybe_add_location_range): Add param "original_idx" and
333 pass it on to layout_range.
334 (layout::print_any_labels): Pass on range->m_original_idx to
336 (gcc_rich_location::add_location_if_nearby): Use 0 for
338 * gcc-rich-location.h (text_range_label::get_text): Update for new
340 (range_label_for_type_mismatch::get_text): Likewise.
342 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
344 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
346 2018-09-17 David Malcolm <dmalcolm@redhat.com>
348 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
349 format_string_diagnostic_t.
350 (fmtwarn_n): Likewise.
351 * substring-locations.c
352 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
353 (format_warning_n_va): Convert to...
354 (format_string_diagnostic_t::emit_warning_n_va): ...this.
355 (format_warning_va): Convert to...
356 (format_string_diagnostic_t::emit_warning_va): ...this.
357 (format_warning_at_substring): Convert to...
358 (format_string_diagnostic_t::emit_warning): ...this.
359 (format_warning_at_substring_n): Convert to...
360 (format_string_diagnostic_t::emit_warning_n): ...this.
361 * substring-locations.h (class format_string_diagnostic_t): New
363 (format_warning_va): Convert to
364 format_string_diagnostic_t::emit_warning_va.
365 (format_warning_n_va): Convert to
366 format_string_diagnostic_t::emit_warning_n_va.
367 (format_warning_at_substring): Convert to
368 format_string_diagnostic_t::emit_warning.
369 (format_warning_at_substring_n): Convert to
370 format_string_diagnostic_t::emit_warning_n.
372 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
373 Bernd Schmidt <bernds_cb1@t-online.de>
375 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
378 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
380 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
381 operand 0 predicate to nonimmediate operand.
382 (rint<mode>2_frndint): Remove insn pattern.
383 (rint<mode>2): Change operand 1 predicate to general_operand.
384 Extend operand 1 to XFmode and generate rintxf2 insn.
385 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
386 Do not use X87MODEF mode macro.
387 (frndintxf2_<rounding>_i387): Rename from
388 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
389 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
390 to XFmode and generate significandxf3 insn.
392 2018-09-17 Richard Biener <rguenther@suse.de>
394 PR tree-optimization/87328
395 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
396 visiting unexecutable backedges when not iterating.
397 (do_rpo_vn): Mark all edges not executable even when not
400 2018-09-17 Martin Jambor <mjambor@suse.cz>
403 * doc/invoke.texi (Warning Options): Likewise.
405 2018-09-17 Richard Biener <rguenther@suse.de>
407 PR tree-optimization/87301
408 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
409 clean EH info from leftover copy assignments.
411 2018-09-17 Martin Liska <mliska@suse.cz>
413 PR gcov-profile/85871
414 * gcov.c (output_intermediate_file): Fix out of bounds
417 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
419 * config/arc/arc.c: Object attributes for core4 not reflected
421 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
424 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
426 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
428 2018-09-17 Martin Liska <mliska@suse.cz>
430 * doc/gcov.texi: Document new option --use-hotness-colors.
431 * gcov.c (struct source_info): Declare new field.
432 (source_info::source_info): Set default for maximum_count.
433 (print_usage): Add new -q option.
434 (process_args): Process it.
435 (accumulate_line_info): Save src->maximum_count.
436 (output_line_beginning): Make color line number if
437 flag_use_hotness_colors is set.
438 (output_line_details): Pass default argument value.
439 (output_lines): Pass src->maximum_count.
441 2018-09-17 Martin Liska <mliska@suse.cz>
443 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
444 Use processor_names table.
445 * config/i386/i386.c (ix86_default_align): Use
446 processor_cost_table for alignment values.
447 (ix86_option_override_internal): Use processor_names.
448 (ix86_function_specific_print): Likewise.
449 * config/i386/i386.h (struct processor_costs):
450 Add alignment values.
451 (struct ptt): Remove and replace with const char *.
452 * config/i386/x86-tune-costs.h (struct processor_costs):
453 Declare default alignments for all costs.
455 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
457 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
458 symbolics or VR_VARYING ranges for ABS_EXPR.
459 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
460 when range will wrap.
462 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
465 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
466 before and after a JUMP_TABLE_DATA.
468 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
471 * dojump.c (do_compare_and_jump): Canonicalize function pointers
472 when one operand is a function pointer. Use POINTER_TYPE_P and
473 FUNC_OR_METHOD_TYPE_P.
474 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
475 * fold-const.c (build_range_check): Likewise.
476 * match.pd (simple_comparison): Likewise.
478 2018-09-14 David Malcolm <dmalcolm@redhat.com>
481 * spellcheck.c (get_edit_distance_cutoff): New function.
482 (selftest::test_edit_distance_unit_test_oneway): Rename to...
483 (selftest::test_get_edit_distance_one_way): ...this.
484 (selftest::test_get_edit_distance_unit): Rename to...
485 (selftest::test_get_edit_distance_both_ways): ...this.
486 (selftest::test_edit_distances): Move tests to this new function,
487 and test some more pairs of strings. Update for above renaming.
488 (selftest::get_old_cutoff): New function.
489 (selftest::test_get_edit_distance_cutoff): New function.
490 (selftest::assert_suggested_for): New function.
491 (ASSERT_SUGGESTED_FOR): New macro.
492 (selftest::assert_not_suggested_for): New function.
493 (ASSERT_NOT_SUGGESTED_FOR): New macro.
494 (selftest::test_suggestions): New function.
495 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
496 tests to selftest::test_edit_distances and call it. Add calls to
497 selftest::test_get_edit_distance_cutoff and
498 selftest::test_suggestions.
499 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
500 (best_match::consider): Replace hard-coded cutoff calculation with
502 (best_match::get_cutoff): New declaration.
503 (best_match::get_best_meaningful_candidate): Likewise.
505 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
507 * builtins.c (fold_builtin_strlen): Remove TODO comment.
509 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
512 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
514 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
515 terminated string literal.
517 2018-09-14 Martin Sebor <msebor@redhat.com>
519 * builtins.c (unterminated_array): Handle ARRAY_REF.
520 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
521 * builtins.h (unterminated_array): Declare extern.
522 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
524 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
527 2018-09-14 Martin Sebor <msebor@redhat.com>
528 Jeff Law <law@redhat.com>
530 * builtins.c (unterminated_array): New.
531 (expand_builtin_strcpy): Adjust.
532 (expand_builtin_strcpy_args): Detect unterminated arrays.
533 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
535 * gimple-fold.h (get_maxval_strlen): Add argument.
536 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
538 * gimple-fold.c (get_range_strlen): Add argument.
539 (get_maxval_strlen): Adjust.
540 * gimple-fold.h (get_range_strlen): Add argument.
542 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
544 * config/i386/movdirintrin.h: Fix copyright year.
546 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
548 * reg-stack.c: Include regs.h.
549 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
550 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
551 FIRST_STACK_REG, not DFmode.
552 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
553 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
554 (change stack): Default register mode to the reg_raw_mode of
555 FIRST_STACK_REG, not DFmode.
556 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
557 (*swapxf): Rename from swapxf.
559 2018-09-14 Carl Love <cel@us.ibm.com>
561 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
562 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
564 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
567 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
570 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
573 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
575 2018-09-14 Jason Merrill <jason@redhat.com>
577 Fix --enable-gather-detailed-mem-stats.
578 * hash-table.c (hash_table_usage): Change from variable to function.
579 * hash-table.h: Adjust.
580 * Makefile.in: Add missing dependencies on hash-table.h.
582 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
584 PR tree-optimization/87259
586 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
587 execute_cse_reciprocals_1 has tried transforming.
589 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
591 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
592 VR_VARYING for PLUS/MINUS_EXPR.
594 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
596 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
599 2018-09-14 Richard Biener <rguenther@suse.de>
602 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
603 bits for the merged partition.
605 2018-09-13 Martin Sebor <msebor@redhat.com>
606 Bernd Edlinger <bernd.edlinger@hotmail.de>
608 * builtins.h (c_srlen): Add argument.
609 * builtins.c (warn_string_no_nul): New function.
610 (c_strlen): Add argument and use it. Update recursive calls.
611 Pass DECL argument to string_constant to get info on non
612 terminated strings. Update *NONSTR as needed.
613 (fold_builtin_strlen): Add argument to calls to c_strlen.
614 Warn for unterminated arrays.
615 (warn_string_no_null): Add prototype.
616 * expr.c (string_constant): Update arguments. Update recursive
617 calls appropriately. Detect missing NUL terminator and outermost
618 declaration its missing in.
619 Improve checks for arrays with nonzero lower bound or elements
620 that are not a single byte. Simplify offset computation.
621 Simplify checks for non-NUL terminated strings.
622 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
623 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
625 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
627 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
629 * fold-const.c (c_getstr): Fix function comment. Remove unused third
630 argument. Fix range checks.
631 * fold-const.h (c_getstr): Adjust protoype.
632 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
633 string is constant but contains no NUL byte.
635 * expr.c (string_constant): Adjust function comment.
636 Remove bogus check for zero termination.
638 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
640 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
641 (get_constant_size): Don't make STRING_CSTs larger than they are.
642 (check_string_literal): New check function for STRING_CSTs.
643 (output_constant): Use it.
645 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
648 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
650 2018-09-13 Richard Biener <rguenther@suse.de>
652 PR tree-optimization/87263
653 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
654 (struct unwind_state): Add max_rpo field.
655 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
656 Compute max_rpo, the max RPO number a block can be backwards reached
657 from. Re-write non-iterating mode to a RPO ordered worklist approach,
658 separating it from the iterating mode.
660 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
662 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
663 (rfs_decision): New scheduling decision.
665 2018-09-13 Richard Biener <rguenther@suse.de>
668 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
669 (vn_nary_op_insert_pieces_predicated): Do not write useless
670 valid_dominated_by_p entry outside of the allocated storage.
672 2018-09-13 Omar Sandoval <osandov@osandov.com>
673 Tom de Vries <tdevries@suse.de>
676 * dwarf2out.c (is_c): New function.
677 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
679 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
682 * config/aarch64/aarch64.md (*aarch64_bfxil):
684 * config/aarch64/constraints.md (Ulc): Define.
685 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
687 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
690 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
692 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
693 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
694 aarch64_layout_frame call.
695 (aarch64_expand_epilogue): Likewise.
696 (aarch64_initial_elimination_offset): Likewise.
697 (aarch64_get_separate_components): Likewise.
698 (aarch64_use_return_insn_p): Likewise.
699 (aarch64_layout_frame): Remove unneeded check.
701 2018-09-13 Jakub Jelinek <jakub@redhat.com>
703 * configure.ac: Only append
704 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
705 gcc_config_arguments if it was never reconfigured or last reconfigure
706 was with different arguments.
707 * configure: Regenerated.
709 2018-09-13 Jakub Jelinek <jakub@redhat.com>
710 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
713 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
714 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
716 2018-09-13 Jakub Jelinek <jakub@redhat.com>
718 PR tree-optimization/87287
719 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
720 X % C == 0 to X % (unsigned) C == 0 optimization to ...
721 * match.pd (X % C == 0): ... here. New optimization.
723 2018-09-12 Jakub Jelinek <jakub@redhat.com>
726 * expr.h (maybe_optimize_mod_cmp): Declare.
727 * expr.c (mod_inv): New function.
728 (maybe_optimize_mod_cmp): New function.
729 (do_store_flag): Use it.
730 * cfgexpand.c (expand_gimple_cond): Likewise.
732 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
733 Julian Brown <julian@codesourcery.com>
736 * gimplify.c (gimplify_scan_omp_clauses): Set
737 target_firstprivatize_array_bases in OpenACC parallel and kernels
738 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
739 OpenACC data regions.
741 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
743 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
744 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
746 2018-09-12 Richard Biener <rguenther@suse.de>
748 PR tree-optimization/87280
749 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
750 edge but unreachable target.
751 (do_rpo_vn): For conservatively handling a PHI only mark
752 the backedge executable but not the block reachable.
754 2018-09-12 Richard Biener <rguenther@suse.de>
756 PR tree-optimization/87266
757 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
760 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
762 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
764 ("trunc<BFP:mode><DFP_ALL:mode>2")
765 ("trunc<DFP_ALL:mode><BFP:mode>2")
766 ("extend<BFP:mode><DFP_ALL:mode>2")
767 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
768 according to the target operand type.
770 2018-09-12 Jakub Jelinek <jakub@redhat.com>
771 Andreas Krebbel <krebbel@linux.ibm.com>
773 PR tree-optimization/86844
774 * gimple-ssa-store-merging.c
775 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
776 there are any overlapping stores in between them, make sure they are
777 also coalesced or we give up completely.
779 2018-09-12 Jakub Jelinek <jakub@redhat.com>
782 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
783 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
785 2018-09-12 Tom de Vries <tdevries@suse.de>
787 * common.opt (gdescribe-dies): Add option.
788 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
789 attribute for artifical and nameless decls.
790 (dwarf2out_register_external_die): Add description attribute to
791 external reference die.
792 (add_desc_attribute): New functions.
793 (gen_subprogram_die): Add description attribute to
794 DW_TAG_call_site_parameter.
795 * tree-pretty-print.c (print_generic_expr_to_str): New function.
796 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
797 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
799 (@item -gdescribe-dies): Add.
801 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
803 * tree-vrp.c (vrp_shift_undefined_p): Remove.
804 (extract_range_from_binary_expr_1: Call
805 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
806 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
809 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
811 * gimple-ssa-warn-alloca.c
812 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
813 field for ALLOCA_BOUND_*_LARGE.
815 2018-09-11 Nathan Sidwell <nathan@acm.org>
817 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
819 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
821 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
822 for clobbers. Remove obsolete comment.
824 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
826 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
827 mpxchk, mpxld and mpxst types.
828 (define_attr length_immediate): Remove all processing of mpx types.
829 (define_attr prefix_0f): Ditto.
830 (define_attr memory): Ditto.
832 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
834 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
835 (log<mode>2): Change operand 1 predicate to general_operand.
836 Extend operand 1 to XFmode and generate logxf3 insn.
837 (log10<mode>2): Change operand 1 predicate to general_operand.
838 Extend operand 1 to XFmode and generate log10xf3 insn.
839 (log2<mode>2): Change operand 1 predicate to general_operand.
840 Extend operand 1 to XFmode and generate log2xf3 insn.
841 (fyl2xp1_extend<mode>xf3_i387): Remove.
842 (log1p<mode>2): Change operand 1 predicate to general_operand.
843 Extend operand 1 to XFmode and generate log1pxf3 insn.
844 (fxtract_extend<mode>xf3_i387): Remove.
845 (logb<mode>2): Change operand 1 predicate to general_operand.
846 Extend operand 1 to XFmode and generate logbxf3 insn.
847 (ilogb<mode>2): Change operand 1 predicate to general_operand.
848 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
849 (significand<mode>2): Change operand 1 predicate to general_operand.
850 Extend operand 1 to XFmode and generate significandxf3 insn.
852 2018-09-11 Nathan Sidwell <nathan@acm.org>
854 * gcc.c (perror_with_name, pfatal_with_name): Delete.
855 (load_specs): Use fatal_error.
856 (DELETE_IF_ORDINARY, process_command): Use error.
857 (execute, run_attempt): Use fatal_error.
859 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
861 * diagnostic-core.h (sorry_at): New prototype.
862 * diagnostic.c (sorry_at): New function.
864 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
866 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
867 by zero as VR_UNDEFINED.
869 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
871 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
872 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
873 (<sincos>mode2): New expander.
874 (sincos_extend<mode>xf3_i387): Remove insn pattern.
875 (sincos -> sin, cos splitters): Remove splitter patterns.
876 (sincos<mode>3): Change operand 2 predicate to general_operand.
877 Extend operand 2 to XFmode and generate sincosxf3 insn.
878 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
879 Change operand 3 predicate to const1_operand.
880 (fptan_extend<mode>xf4_i387): Remove insn pattern.
881 (tanxf2): Update operands in the call to fptanxf4_i387.
882 (tan<mode>2): Change operand 1 predicate to general_operand.
883 Extend operand 1 to XFmode and generate tanxf3 insn.
884 (atan2xf3): Rename from *fpatanxf3_i387.
885 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
886 (atan2xf3): Remove expander.
887 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
888 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
889 (atan<mode>2): Change operand 1 predicate to general_operand.
890 Extend operand 1 to XFmode and generate atanxf3 insn.
892 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
894 * config/i386/i386.md (x87/SSE constant load splitter): Use
895 memory_operand instead of nonimmediate_operand for input operand
898 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
900 * config/i386/i386.md (float partial SSE register stall splitter): Move
901 splitter near its instruction pattern.
902 (float_extend partial SSE register stall splitter): Ditto.
903 (float_truncate partial SSE register stall splitter): Ditto.
905 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
908 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
909 to speculation_safe_value_not_needed.
912 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
913 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
914 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
915 leaf_function_p, instead use has_hard_reg_initial_val.
917 2018-09-09 Nathan Sidwell <nathan@acm.org>
919 * gcc.h (pfatal_with_name): Don't declare here.
920 * gcc.c (pfatal_with_name): Make static.
922 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
924 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
927 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
929 PR rtl-optimization/85458
930 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
931 priority hook to reduce the priority of EXPR.
933 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
935 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
936 DImode for x87 on 32bit targets. Conditionally disable x87 modes
937 with X87_ENABLE_FLOAT. Remove preparation code.
938 (*float<SWI48:mode><MODEF:mode>2): Rename from
939 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
940 math using "enabled" attribute.
941 (*floatdi<MODEF:mode>2_i387): Rename from
942 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
943 enable for 32bit targets only.
944 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
946 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
947 as operand 1 predicate. Rewrite as define_insn_and_split.
948 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
950 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
952 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
953 to fallthru to FLOAT case.
955 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
958 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
959 around folding of vec_sl to handle out of range shift values.
961 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
963 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
964 Update callers to gen_fix_trunc<mode>_i387_fisttp
965 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
966 nonimmediate_operand.
967 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
968 and corresponding splitters.
969 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
970 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
971 (fix_truncdi_i387_with_temp): Remove insn pattern
972 and corresponding splitters.
973 (fix_trunc<mode>_i387): Change operand 0 predicate to
974 nonimmediate_operand.
975 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
976 and corresponding splitters.
977 (*fistdi2_1): Remove.
979 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
980 (lrintxfdi2): Remove expander. Reimplement as define_insn.
981 (*fist<mode>2_1): Remove.
982 (fist<mode>2): Ditto.
983 (fist<mode>2_with_temp): Remove insn pattern and corresponding
985 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
986 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
987 (fistdi2_<rounding>): Change operand 0 predicate to
988 nonimmediate_operand.
989 (fistdi2_<rounding>_with_temp): Remove insn pattern
990 and corresponding splitters.
991 (fist<mode>2_<rounding>): Change operand 0 predicate to
992 nonimmediate_operand.
993 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
994 and corresponding splitters.
996 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
998 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
1000 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
1003 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
1005 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
1006 early gimple folding of vec_splat().
1007 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
1008 * gimple-fold.h: Add an extern define for tree_vec_extract().
1010 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
1012 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
1013 wrappers around TREE_TYPE comparisons.
1015 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
1018 * config/s390/predicates.md: Add nonsym_memory_operand.
1019 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
1020 contains a SYMBOL_REF, load it into an intermediate pseudo.
1021 (s390_emit_compare_and_swap): Legitimize operand.
1022 * config/s390/s390.md: Use the new nonsym_memory_operand
1023 with UNSPECV_CAS patterns.
1025 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
1028 * config/s390/s390-passes.def: New file.
1029 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
1031 (make_pass_s390_early_mach): Add declaration.
1032 * config/s390/s390.c (make_pass_s390_early_mach):
1033 (s390_option_override): Remove dynamic registration.
1034 * config/s390/t-s390: Add s390-passes.def.
1036 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
1038 * config/s390/s390.c (s390_decompose_constant_pool_ref):
1039 Remove UNSPEC_LTREL_BASE check.
1040 (annotate_constant_pool_refs): Likewise.
1041 (find_constant_pool_ref): Likewise.
1042 (find_ltrel_base): Removed.
1043 (replace_ltrel_base): Removed.
1044 (s390_mainpool_finish): Remove replace_ltrel_base call.
1045 (s390_chunkify_start): Remove pending LTREL_BASE logic.
1046 (s390_chunkify_finish): Remove replace_ltrel_base call.
1047 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
1049 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
1052 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1053 to speculation_safe_value_not_needed.
1055 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
1056 Bernd Schmidt <bernds_cb1@t-online.de>
1058 * config/nvptx/nvptx-opts.h: New file.
1059 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
1060 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
1062 (TARGET_SM35): New macro.
1063 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
1065 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
1067 (misa=): New option.
1068 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
1070 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
1072 * config/i386/i386.md (truncdfsf2): Remove expander.
1073 (truncdfsf2_with_temp): Ditto.
1074 (truncxf<mode>2): Ditto.
1075 (*truncdfsf_fast_mixed): Remove insn pattern.
1076 (*truncdfsf_fast_i387): Ditto.
1077 (*truncdfsf_mixed): Ditto.
1078 (*truncdfsf_i387): Ditto.
1079 (*truncdfsf2_i387_1): Ditto.
1080 (*truncxfsf2_mixed): Ditto.
1081 (*truncxfdf2_mixed): Ditto.
1082 (*truncxf<mode>2_i387_noop): Ditto. Update callers
1083 to call gen_truncxf<mode>2 instead.
1084 (*truncxf<mode>2_i387): Remove.
1085 (reg->reg splitters): Remove splitter pattern.
1086 (reg->mem splitters): Ditto.
1088 (truncdfsf2): New insn pattern.
1089 (truncxf<mode>2): Ditto.
1091 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1093 * tree-ssa-math-opts.c (is_mult_by): New function.
1094 (is_square_of): Use the above.
1095 (optimize_recip_sqrt): New function.
1096 (pass_cse_reciprocals::execute): Use the above.
1098 2018-09-05 Richard Biener <rguenther@suse.de>
1101 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
1102 to zero-init the emplaced vec.
1104 2018-09-05 Martin Liska <mliska@suse.cz>
1106 PR tree-optimization/87205
1107 * tree-switch-conversion.c (pass_lower_switch::execute):
1108 Group cases for switch statements.
1110 2018-09-05 Richard Biener <rguenther@suse.de>
1112 PR tree-optimization/87217
1113 * tree-ssa-sccvn.c (vuse_valueize): New.
1114 (vn_reference_lookup_pieces): Use it.
1115 (vn_reference_lookup): Likewise.
1117 2018-09-05 Nathan Sidwell <nathan@acm.org>
1120 * stor-layout.c (place_field): Scan forwards to check last
1121 bitfield when ms_bitfield_placement is in effect.
1123 2018-09-05 Richard Biener <rguenther@suse.de>
1126 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
1129 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
1130 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
1132 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
1133 * config.gcc (extra_objs): Build it.
1134 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
1136 * config/aarch64/aarch64-passes.def
1137 (pass_tag_collision_avoidance): New pass.
1138 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
1139 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
1140 (aarch64_classify_address): Remove static qualifier.
1141 (aarch64_address_info, aarch64_address_type): Move to...
1142 * config/aarch64/aarch64-protos.h: ... here.
1143 (make_pass_tag_collision_avoidance): New function.
1144 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
1147 2018-09-05 Martin Liska <mliska@suse.cz>
1149 * doc/gcov.texi: Update documentation of humar
1151 * gcov.c (format_count): Print one decimal place, it provides
1152 more fine number of situations like '1G' vs. '1.4G'.
1154 2018-09-05 Martin Liska <mliska@suse.cz>
1157 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
1158 * optc-gen.awk: Allow 'Var' for Deprecated options in order
1159 to generate a MASK value.
1161 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
1164 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
1165 if frame pointer isn't used.
1166 (compute_frame_pointer_to_fb_displacement): Likewise.
1168 2018-09-04 Jakub Jelinek <jakub@redhat.com>
1171 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
1172 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
1173 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
1174 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
1175 and OPTION_MASK_ISA_XSAVEC_UNSET.
1177 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
1179 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
1180 XOR operations in NAND case.
1182 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
1184 * wide-int-range.cc (wide_int_range_convert): New.
1185 * wide-int-range.h (wide_int_range_convert): New.
1186 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
1187 code into wide_int_range_convert.
1188 (extract_range_into_wide_ints): Do not munge anti range constants
1189 into the entire domain. Just return the range back.
1191 2018-09-04 Martin Liska <mliska@suse.cz>
1193 * genmatch.c (output_line_directive): Add new argument
1195 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
1197 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
1199 * doc/invoke.texi (Option Summary): Add whitespace.
1201 * doc/invoke.texi (Option Summary): Add -Waligned-new.
1203 2018-09-04 Richard Biener <rguenther@suse.de>
1205 PR tree-optimization/87211
1206 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
1207 backedge value we're supposed to treat as VARYING also number
1208 the PHI to VARYING in case it got a different value-number already.
1210 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
1212 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
1213 (extract_range_from_binary_expr_1): Do not call
1214 vrp_can_optimize_bit_op.
1215 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
1217 (wide_int_range_get_mask_and_bounds): New.
1218 (wide_int_range_optimize_bit_op): New.
1219 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
1220 (wide_int_range_bit_and): Same.
1221 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
1222 (wide_int_range_optimize_bit_op): New.
1223 (wide_int_range_get_mask_and_bounds): New.
1225 2018-09-04 Richard Biener <rguenther@suse.de>
1227 PR tree-optimization/87176
1228 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
1229 variable. When value-numbering a virtual PHI node make sure
1230 to not value-number to the backedge value.
1232 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
1234 * doc/extend.texi (Long Long, Hex Floats): Document support for
1235 long long and hex floats in more recent versions of ISO C++.
1237 2018-09-03 Richard Biener <rguenther@suse.de>
1239 PR tree-optimization/87177
1240 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
1243 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1245 * bb-reorder.c (edge_order): Convert to C-qsort-style
1246 tri-state comparator.
1247 (reorder_basic_blocks_simple): Change std::stable_sort to gcc_stablesort.
1249 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1251 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
1252 tri-state comparator.
1253 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
1255 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1257 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
1258 (mergesort): ... here as maximum count for using netsort.
1259 (gcc_qsort): Set nlim to 3 if stable sort is requested.
1260 (gcc_stablesort): New.
1261 * system.h (gcc_stablesort): Declare.
1263 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1265 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
1266 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
1267 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
1269 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
1271 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
1272 lxsdx and stxsdx alternatives.
1273 (*mov<mode>_hardfloat64): Ditto.
1274 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
1276 2018-09-03 Richard Biener <rguenther@suse.de>
1278 PR tree-optimization/87200
1279 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
1282 2018-09-03 Martin Liska <mliska@suse.cz>
1284 PR tree-optimization/87201
1285 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
1286 Fix parenthesis in an expression.
1288 2018-09-03 Richard Biener <rguenther@suse.de>
1290 PR tree-optimization/87197
1291 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
1292 visited. CSE the VN_INFO hashtable lookup.
1294 PR tree-optimization/87169
1295 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
1296 iterating make sure there's no extra backedges from irreducible
1297 regions feeding the header. Mark the destination block
1300 2018-09-03 Martin Liska <mliska@suse.cz>
1303 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
1304 * common/common-targhooks.c (default_get_valid_option_values):
1306 * common/common-targhooks.h (default_get_valid_option_values):
1308 * common/config/i386/i386-common.c: Move processor_target_table
1310 (ix86_get_valid_option_values): New function.
1311 (TARGET_GET_VALID_OPTION_VALUES): New macro.
1312 * config/i386/i386.c (struct ptt): Move to i386-common.c.
1313 (PTA_*): Move all defined masks into i386-common.c.
1314 (ix86_function_specific_restore): Use new processor_cost_table.
1315 * config/i386/i386.h (struct ptt): Moved from i386.c.
1316 (struct pta): Likewise.
1317 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
1318 * doc/tm.texi.in: Likewise.
1319 * opt-suggestions.c (option_proposer::suggest_option):
1320 Pass prefix to build_option_suggestions.
1321 (option_proposer::get_completions): Likewise.
1322 (option_proposer::build_option_suggestions): Use the new target
1324 * opts.c (struct option_help_tuple): New struct.
1325 (print_filtered_help): Use the new target hook.
1327 2018-09-03 Martin Liska <mliska@suse.cz>
1330 * predict.c (set_even_probabilities): Add likely_edges
1331 argument and handle cases where we have precisely one
1333 (combine_predictions_for_bb): Catch also likely_edges.
1334 (tree_predict_by_opcode): Handle gswitch statements.
1335 * tree-cfg.h (find_case_label_for_value): New declaration.
1336 (find_taken_edge_switch_expr): Likewise.
1337 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
1338 Find pivot in decision tree based on probabily, not by number of
1341 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
1343 * doc/standards.texi (Standards): Update Objective-C reference.
1345 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1347 * doc/install.texi (Prerequisites): Update link for MPC.
1349 2018-09-01 Michael Matz <matz@suse.de>
1351 PR tree-optimization/87074
1352 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
1353 PHIs for outer-loop uses.
1355 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1357 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
1358 * doc/invoke.texi (C Dialect Options): Ditto.
1360 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1362 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
1364 2018-08-31 Richard Biener <rguenther@suse.de>
1366 PR tree-optimization/87168
1367 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
1368 (rpo_elim::eliminate_avail): When OP was not visited it must
1371 2018-08-31 David Malcolm <dmalcolm@redhat.com>
1373 * tree-vrp.c (copy_value_range): Convert param "from" from
1374 "value_range *" to "const value_range *".
1375 (range_is_null): Likewise for param "vr".
1376 (range_int_cst_p): Likewise.
1377 (range_int_cst_singleton_p): Likewise.
1378 (symbolic_range_p): Likewise.
1379 (value_ranges_intersect_p): Likewise for both params.
1380 (value_range_nonnegative_p): Likewise for param "vr".
1381 (value_range_constant_singleton): Likewise.
1382 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
1383 (extract_range_into_wide_ints): Likewise for param "vr".
1384 (extract_range_from_multiplicative_op): Likewise for params "vr0"
1386 (vrp_can_optimize_bit_op): Likewise.
1387 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
1389 (extract_range_from_unary_expr): Likewise.
1390 (debug_value_range): Likewise for param "vr".
1391 (value_range::dump): Add "const" qualifier.
1392 (vrp_prop::check_array_ref): Convert local "vr" from
1393 "value_range *" to "const value_range *".
1394 (vrp_prop::check_mem_ref): Likewise.
1395 (vrp_prop::visit_stmt): Likewise for local "old_vr".
1396 (vrp_intersect_ranges_1): Likewise for param "vr_1".
1397 (vrp_intersect_ranges): Likewise.
1398 (simplify_stmt_for_jump_threading): Likewise for local "vr".
1399 (vrp_prop::vrp_finalize): Likewise.
1400 * tree-vrp.h (value_range::dump): Add "const" qualifier.
1401 (vrp_intersect_ranges): Add "const" qualifier to params as above.
1402 (extract_range_from_unary_expr): Likewise.
1403 (value_range_constant_singleton): Likewise.
1404 (symbolic_range_p): Likewise.
1405 (copy_value_range): Likewise.
1406 (extract_range_from_binary_expr_1): Likewise.
1407 (range_int_cst_p): Likewise.
1408 (vrp_set_zero_nonzero_bits): Likewise.
1409 (range_int_cst_singleton_p): Likewise.
1411 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
1413 * config/aarch64/arm_neon.h (vabsd_s64): New.
1414 (vnegd_s64): Likewise.
1416 2018-08-31 Martin Jambor <mjambor@suse.cz>
1418 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
1420 2018-08-31 Martin Liska <mliska@suse.cz>
1422 * ipa-icf.c (sem_item::add_type): Use
1423 sem_item::m_type_hash_cache.
1424 * ipa-icf.h: Move the cache from sem_item_optimizer
1427 2018-08-31 Nathan Sidwell <nathan@acm.org>
1429 * doc/extend.texi (Backwards Compatibility): Remove implicit
1430 extern C leeway of () being (...).
1432 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1434 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
1436 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
1440 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
1442 2018-08-31 Jakub Jelinek <jakub@redhat.com>
1445 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
1446 gen_int_mode. Formatting fixes.
1448 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
1450 * target.def (custom_function_descriptors): Improve documentation.
1451 * doc/tm.texi.in (Trampolines): Expand discussion of function
1452 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
1453 beginning of the section.
1454 * doc/tm.texi: Regenerated.
1456 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
1458 * cfg.h (class auto_edge_flag): Spell out the template-id of the
1459 base class in the initializer list. This is a workaround for
1460 building with older GCC.
1461 (class auto_bb_flag): Likewise.
1463 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
1465 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
1466 (altivec_vcmpequ<VI_char>_p): Remove star.
1467 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
1469 (expand_strncmp_vec_sequence): New function.
1470 (emit_final_str_compare_vec): New function.
1471 (expand_strn_compare): Add support for vector strncmp.
1472 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
1473 length specification to bytes.
1474 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
1475 (vcmpnezb_p): New pattern.
1476 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
1477 for option -mstring-compare-inline-limit.
1479 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
1481 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
1482 (PTA_SKYLAKE): Add PTA_AES.
1483 (PTA_GOLDMONT): Likewise.
1485 2018-08-29 Jan Hubicka <jh@suse.cz>
1488 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
1489 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
1491 2018-08-29 Jan Hubicka <jh@suse.cz>
1493 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
1495 (hash_tree): Do not visit TYPE_STUB_DECL.
1496 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
1497 stream TYPE_STUB_DECL.
1498 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
1499 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
1500 after free_lang_data.
1501 (type_in_anonymous_namespace_p): Likewise.
1503 2018-08-29 Jan Hubicka <jh@suse.cz>
1505 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1506 comment that it has to be even number.
1507 (class sreal): Change m_sig type to int32_t.
1508 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1509 int64_t for temporary calculations.
1510 (sreal_verify_basics): Drop one bit from minimum and maximum.
1512 2018-08-30 Richard Biener <rguenther@suse.de>
1514 PR tree-optimization/87147
1515 * tree-ssa-sccvn.c (SSA_VISITED): New function.
1516 (visit_phi): When the degenerate result is from the backedge and
1517 we didn't visit its definition yet drop to VARYING.
1518 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
1520 2018-08-29 Jan Hubicka <jh@suse.cz>
1522 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
1524 (hash_tree): Likewise.
1526 2018-08-29 Jan Hubicka <jh@suse.cz>
1528 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
1529 and TYPE_NEXT_REF_TO.
1531 2018-08-29 Jan Hubicka <jh@suse.cz>
1533 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1534 comment that it has to be even number.
1535 (class sreal): Change m_sig type to int32_t.
1536 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1537 int64_t for temporary calculations.
1538 (sreal_verify_basics): Drop one bit from minimum and maximum.
1540 2018-08-30 Tamar Christina <tamar.christina@arm.com>
1542 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
1544 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
1547 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
1548 if to_add is negative.
1550 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
1553 * builtins.c (c_strlen): Improve range checks.
1555 2018-08-29 Martin Sebor <msebor@redhat.com>
1556 Jeff Law <law@redhat.com>
1558 PR tree-optimization/86714
1559 PR tree-optimization/86711
1560 * builtins.c (c_strlen): Add arguments to call to string_constant.
1561 * expr.c (string_constant): Add argument. Detect missing nul
1562 terminator and outermost declaration it's missing in.
1563 * expr.h (string_constant): Add argument.
1564 * fold-const.c (read_from_constant_string): Add arguments to call to
1566 (c_getstr): Likewise.
1567 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1569 * tree-ssa-strlen.c (get_stridx): Likewise.
1571 2018-08-29 Jan Hubicka <jh@suse.cz>
1573 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
1574 Do not stream DECL_VINDEX.
1575 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
1576 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
1577 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
1579 2018-08-29 Richard Biener <rguenther@suse.de>
1581 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
1582 virtual operands that are not default defs to honor region
1584 (rpo_vn_valueize): Remove ineffective code here.
1586 2018-08-29 Richard Biener <rguenther@suse.de>
1588 PR tree-optimization/87132
1589 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
1590 when skipping defs reachable over backedges.
1592 2018-08-29 Richard Biener <rguenther@suse.de>
1594 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
1595 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
1596 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
1597 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
1598 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
1599 (vn_reference_lookup_3): Remove use of const_parms.
1600 (free_rpo_vn): Do not free const_parms.
1601 (do_rpo_vn): Do not call init_const_parms.
1602 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
1603 SSA_NAME_POINTS_TO_READONLY_MEMORY.
1604 (call_may_clobber_ref_p_1): Likewise.
1606 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
1609 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
1611 (-ftree-scev-cprop): Document.
1613 2018-08-29 Jan Hubicka <jh@suse.cz>
1615 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
1617 (sreal constructor): Update.
1618 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
1619 sreal:operator/): Update.
1621 2018-08-29 Martin Liska <mliska@suse.cz>
1623 * tree-switch-conversion.c (switch_conversion::expand):
1624 Strenghten assumption about gswitch statements.
1626 2018-08-29 Richard Biener <rguenther@suse.de>
1628 PR tree-optimization/87117
1629 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
1630 re-value-number released SSA VDEFs.
1632 2018-08-29 Richard Biener <rguenther@suse.de>
1634 PR tree-optimization/87126
1635 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
1637 2018-08-28 Jim Wilson <jimw@sifive.com>
1639 * config/riscv/pic.md: Rewrite.
1640 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
1642 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
1643 (SOFTF, default_load, softload, softstore): New.
1645 2018-08-28 Jeff Law <law@redhat.com>
1647 * fold-const.c (fold_binary_loc): Remove recently added assert.
1649 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
1651 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
1652 to OP parmeter of generated function.
1654 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
1656 PR tree-optimization/87009
1657 * match.pd: Add boolean optimizations.
1659 2018-08-28 Martin Sebor <msebor@redhat.com>
1662 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
1663 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
1664 (pass_walloca::gate): Use it.
1665 (alloca_call_type): Same.
1666 (pass_walloca::execute): Same.
1667 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
1669 2018-08-28 David Malcolm <dmalcolm@redhat.com>
1671 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
1672 GCC_VERSION for usage of "__gcc_dump_printf__" format from
1675 2018-08-28 Richard Biener <rguenther@suse.de>
1677 PR tree-optimization/87124
1678 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
1679 constants before looking up avail.
1681 2018-08-28 Jakub Jelinek <jakub@redhat.com>
1684 * calls.c (maybe_warn_nonstring_arg): Punt early if
1685 warn_stringop_overflow is zero. Don't call get_range_strlen
1686 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
1687 Swap comparison operands to have constants on rhs. Only use
1688 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
1689 increment lenrng[0].
1691 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
1693 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
1694 use of tree_to_shwi. Remove duplicated test for the size being
1695 a whole number of bytes.
1697 2018-08-28 Richard Biener <rguenther@suse.de>
1699 PR tree-optimization/87117
1700 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
1701 Handle removed stmt without LHS (GIMPLE_NOP).
1703 2018-08-28 Richard Biener <rguenther@suse.de>
1705 PR tree-optimization/87117
1706 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
1707 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
1709 2018-08-28 Richard Biener <rguenther@suse.de>
1711 PR tree-optimization/87117
1712 * tree-ssa-pre.c (compute_avail): Do not make expressions
1713 with predicated values available.
1714 (get_expr_value_id): Assert we do not run into predicated value
1717 2018-08-28 Richard Biener <rguenther@suse.de>
1719 PR tree-optimization/87117
1720 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
1721 get virtual operands.
1722 (get_expr_operands): Handle STRING_CST like other decls.
1724 2018-08-28 Martin Liska <mliska@suse.cz>
1726 * tree.h: Update documentation of fndecl_built_in_p
1730 2018-08-27 Jeff Law <law@redhat.com>
1731 PR tree-optimization/87110
1732 * tree-ssa-dse.c (compute_trims): Handle non-constant
1735 2018-08-27 Martin Sebor <msebor@redhat.com>
1737 PR tree-optimization/86914
1738 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
1740 2018-08-27 Martin Sebor <msebor@redhat.com>
1742 PR tree-optimization/87112
1743 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
1744 the type of the bound argument.
1746 2018-08-27 Jeff Law <law@redhat.com>
1748 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
1749 type does not have a TYPE_SIZE_UNIT.
1751 2018-08-27 Steve Ellcey <sellcey@cavium.com>
1753 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
1754 with include of backend.h.
1756 2018-08-27 Richard Biener <rguenther@suse.de>
1758 PR tree-optimization/86927
1759 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
1760 use const cond reduction code.
1762 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
1764 PR tree-optimization/85758
1765 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
1767 2018-08-27 David Malcolm <dmalcolm@redhat.com>
1770 * diagnostic-show-locus.c (class layout_range): Update for
1771 conversion of show_caret_p to a tri-state.
1772 (layout_range::layout_range): Likewise.
1773 (make_range): Likewise.
1774 (layout::maybe_add_location_range): Likewise.
1775 (layout::should_print_annotation_line_p): Don't show annotation
1776 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
1777 (layout::get_state_at_point): Update for conversion of
1778 show_caret_p to a tri-state. Bail out early for
1779 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
1780 underlining or source colorization.
1781 (gcc_rich_location::add_location_if_nearby): Update for conversion
1782 of show_caret_p to a tri-state.
1783 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
1784 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
1786 (selftest::test_one_liner_labels): Likewise.
1787 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
1788 conversion of show_caret_p to a tri-state.
1789 * pretty-print.c (text_info::set_location): Likewise.
1790 * pretty-print.h (text_info::set_location): Likewise.
1791 * substring-locations.c (format_warning_n_va): Likewise.
1792 * tree-diagnostic.c (default_tree_printer): Likewise.
1793 * tree-pretty-print.c (newline_and_indent): Likewise.
1795 2018-08-27 David Malcolm <dmalcolm@redhat.com>
1798 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
1799 line above for line-insertion fix-it hints.
1800 (selftest::test_fixit_insert_containing_newline): Update the
1801 expected results, and add a test with line-numbering enabled.
1803 2018-08-27 Martin Liska <mliska@suse.cz>
1806 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
1807 params with DECL_HAS_VALUE_EXPR_P.
1809 2018-08-27 Martin Liska <mliska@suse.cz>
1811 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
1812 selected expansion strategy.
1814 2018-08-27 Martin Liska <mliska@suse.cz>
1816 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
1817 * builtins.c (is_builtin_fn): Likewise.
1818 * attribs.c (diag_attr_exclusions): Use new function
1819 fndecl_built_in_p and remove check for FUNCTION_DECL if
1821 (builtin_mathfn_code): Likewise.
1822 (fold_builtin_expect): Likewise.
1823 (fold_call_expr): Likewise.
1824 (fold_builtin_call_array): Likewise.
1825 (fold_call_stmt): Likewise.
1826 (set_builtin_user_assembler_name): Likewise.
1827 (is_simple_builtin): Likewise.
1828 * calls.c (gimple_alloca_call_p): Likewise.
1829 (maybe_warn_nonstring_arg): Likewise.
1830 * cfgexpand.c (expand_call_stmt): Likewise.
1831 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
1832 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
1833 (cgraph_node::verify_node): Likewise.
1834 * cgraphclones.c (build_function_decl_skip_args): Likewise.
1835 (cgraph_node::create_clone): Likewise.
1836 * config/arm/arm.c (arm_insert_attributes): Likewise.
1837 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
1838 * dse.c (scan_insn): Likewise.
1839 * expr.c (expand_expr_real_1): Likewise.
1840 * fold-const.c (operand_equal_p): Likewise.
1841 (fold_binary_loc): Likewise.
1842 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
1843 * gimple-low.c (lower_stmt): Likewise.
1844 * gimple-pretty-print.c (dump_gimple_call): Likewise.
1845 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Likewise.
1846 * gimple.c (gimple_build_call_from_tree): Likewise.
1847 (gimple_call_builtin_p): Likewise.
1848 (gimple_call_combined_fn): Likewise.
1849 * gimplify.c (gimplify_call_expr): Likewise.
1850 (gimple_boolify): Likewise.
1851 (gimplify_modify_expr): Likewise.
1852 (gimplify_addr_expr): Likewise.
1853 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
1854 * ipa-cp.c (determine_versionability): Likewise.
1855 * ipa-fnsummary.c (compute_fn_summary): Likewise.
1856 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
1857 * ipa-split.c (visit_bb): Likewise.
1858 (split_function): Likewise.
1859 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1860 * lto-cgraph.c (input_node): Likewise.
1861 * lto-streamer-out.c (write_symbol): Likewise.
1862 * omp-low.c (setjmp_or_longjmp_p): Likewise.
1863 (lower_omp_1): Likewise.
1864 * predict.c (strip_predict_hints): Likewise.
1865 * print-tree.c (print_node): Likewise.
1866 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
1867 * trans-mem.c (is_tm_irrevocable): Likewise.
1868 (is_tm_load): Likewise.
1869 (is_tm_simple_load): Likewise.
1870 (is_tm_store): Likewise.
1871 (is_tm_simple_store): Likewise.
1872 (is_tm_abort): Likewise.
1873 (tm_region_init_1): Likewise.
1874 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
1875 * tree-cfg.c (verify_gimple_call): Likewise.
1876 (move_stmt_r): Likewise.
1877 (stmt_can_terminate_bb_p): Likewise.
1878 * tree-eh.c (lower_eh_constructs_2): Likewise.
1879 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
1880 * tree-inline.c (remap_gimple_stmt): Likewise.
1881 (copy_bb): Likewise.
1882 (estimate_num_insns): Likewise.
1883 (fold_marked_statements): Likewise.
1884 * tree-sra.c (scan_function): Likewise.
1885 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
1886 (optimize_stack_restore): Likewise.
1887 (pass_fold_builtins::execute): Likewise.
1888 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
1889 (mark_all_reaching_defs_necessary_1): Likewise.
1890 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
1891 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1892 (pass_forwprop::execute): Likewise.
1893 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1894 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
1895 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
1896 * tree-ssa-strlen.c (get_string_length): Likewise.
1897 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
1898 (find_func_aliases_for_call): Likewise.
1899 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
1900 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
1901 * tree-tailcall.c (find_tail_calls): Likewise.
1902 * tree.c (need_assembler_name_p): Likewise.
1903 (free_lang_data_in_decl): Likewise.
1904 (get_call_combined_fn): Likewise.
1905 * ubsan.c (is_ubsan_builtin_p): Likewise.
1906 * varasm.c (incorporeal_function_p): Likewise.
1907 * tree.h (DECL_BUILT_IN): Remove and replace with
1909 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
1910 (fndecl_built_in_p): New.
1912 2018-08-27 Martin Liska <mliska@suse.cz>
1914 PR tree-optimization/86847
1915 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
1916 Dump also subtree probability.
1917 (switch_decision_tree::do_jump_if_equal): New function.
1918 (switch_decision_tree::emit_case_nodes): Handle special
1919 situations in balanced tree that can be emitted much simpler.
1920 Fix calculation of probabilities that happen in tree expansion.
1921 * tree-switch-conversion.h (struct cluster): Add
1923 (struct simple_cluster): Likewise.
1924 (struct case_tree_node): Add new function has_child.
1925 (do_jump_if_equal): New.
1927 2018-08-27 Martin Liska <mliska@suse.cz>
1929 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
1930 Add new argument to bit_test_cluster constructor.
1931 (bit_test_cluster::emit): Set bits really number of values
1933 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
1934 probability argument.
1935 * tree-switch-conversion.h (struct bit_test_cluster):
1936 Add m_handles_entire_switch.
1938 2018-08-27 Martin Liska <mliska@suse.cz>
1940 PR tree-optimization/86702
1941 * tree-switch-conversion.c (jump_table_cluster::emit):
1942 Make probabilities even for values in jump table
1943 according to number of cases handled.
1944 (switch_decision_tree::compute_cases_per_edge): Pass
1945 argument to reset_out_edges_aux function.
1946 (switch_decision_tree::analyze_switch_statement): Likewise.
1947 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
1950 2018-08-27 Martin Liska <mliska@suse.cz>
1952 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
1953 cfun argument explicitly.
1954 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
1955 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
1956 function gimple_switch_default_bb.
1957 (convert_switch_statements):
1959 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
1960 * stmt.c (label_to_block_fn): Use label_to_block and pass
1961 cfun argument explicitly and use gimple_switch_label_bb.
1962 (expand_case): Likewise.
1963 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
1964 cfun argument explicitly. Likewise.
1965 (make_edges_bb): Likewise.
1966 (make_cond_expr_edges): Likewise.
1967 (get_cases_for_edge): Likewise.
1968 (make_gimple_switch_edges): Likewise.
1969 (label_to_block_fn): Likewise.
1970 (label_to_block): Likewise.
1971 (make_goto_expr_edges): Likewise.
1972 (make_gimple_asm_edges): Likewise.
1973 (main_block_label): Likewise.
1974 (group_case_labels_stmt): Likewise.
1975 (find_taken_edge_computed_goto): Likewise.
1976 (find_taken_edge_switch_expr): Likewise.
1977 (gimple_verify_flow_info): Likewise.
1978 (gimple_redirect_edge_and_branch): Likewise.
1979 (gimple_switch_label_bb): New function.
1980 (gimple_switch_default_bb): Likewise.
1981 (gimple_switch_edge): Likewise.
1982 (gimple_switch_default_edge): Likewise.
1983 * tree-cfg.h (label_to_block_fn): Remove and replace ...
1984 (label_to_block): ... with this.
1985 (gimple_switch_label_bb): New.
1986 (gimple_switch_default_bb): Likewise.
1987 (gimple_switch_edge): Likewise.
1988 (gimple_switch_default_edge): Likewise.
1989 * tree-cfgcleanup.c (convert_single_case_switch): Use
1990 new gimple functions and pass new argument to label_to_block.
1991 (cleanup_control_flow_bb):
1992 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
1993 cfun argument explicitly.
1994 (make_eh_edges): Likewise.
1995 (redirect_eh_dispatch_edge): Likewise.
1996 (lower_resx): Likewise.
1997 (lower_eh_dispatch): Likewise.
1998 (maybe_remove_unreachable_handlers): Likewise.
1999 (unsplit_eh): Likewise.
2000 (cleanup_empty_eh): Likewise.
2001 (verify_eh_edges): Likewise.
2002 (verify_eh_dispatch_edge): Likewise.
2003 * tree-ssa-dom.c (record_edge_info): Likewise.
2004 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
2005 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
2006 (thread_through_normal_block): Likewise.
2007 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
2008 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
2009 * tree-switch-conversion.c (switch_conversion::collect): Use new
2011 (switch_conversion::check_final_bb): Likewise.
2012 (switch_conversion::gather_default_values): Pass new argument
2014 (switch_conversion::build_constructors): Likewise.
2015 (switch_decision_tree::compute_cases_per_edge): Use new
2016 gimple_switch_edge function.
2017 (switch_decision_tree::analyze_switch_statement): Pass new argument
2019 (switch_decision_tree::try_switch_expansion): Use
2020 gimple_switch_default_edge.
2021 * tree-vrp.c (find_switch_asserts): Pass new argument
2023 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
2024 (vr_values::simplify_switch_using_ranges): Likewise.
2026 2018-08-27 Richard Biener <rguenther@suse.de>
2028 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
2029 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
2031 * tree-ssa-sccvn.h (struct vn_pval): New structure.
2032 (struct vn_nary_op_s): Add unwind_to member. Add
2033 predicated_values flag and put result into a union together
2034 with a linked list of vn_pval.
2035 (struct vn_ssa_aux): Add name member to make maintaining
2036 a map of SSA name to vn_ssa_aux possible. Remove no longer
2037 needed info, dfsnum, low, visited, on_sccstack, use_processed
2038 and range_info_anti_range_p members.
2039 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
2040 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
2042 (vn_valueize): New global.
2043 (vn_context_bb): Likewise.
2044 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
2045 VN_INFO_PTR_INFO): Remove.
2046 * tree-ssa-sccvn.c: ... (rewrite)
2047 (pass_fre::execute): For -O2+ initialize loops and run
2048 RPO VN in optimistic mode (iterating). For -O1 and -Og
2049 run RPO VN in non-optimistic mode.
2050 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
2051 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
2052 * doc/invoke.texi (sccvn-max-scc-size): Remove.
2053 (rpo-vn-max-loop-depth): Document.
2054 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
2055 when valuezing the VUSE signals we walked out of the region.
2056 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
2057 (phi_translate): Set VN context block to use for availability
2059 (compute_avail): Likewise.
2060 (pre_valueize): New function.
2061 (pass_pre::execute): Adjust to the RPO VN API.
2063 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
2064 (propagate_constants_for_unrolling): Remove.
2065 (tree_unroll_loops_completely): Perform value-numbering
2066 on the unrolled bodies loop parent.
2068 2018-08-27 Richard Biener <rguenther@suse.de>
2070 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
2071 for partial antic compute.
2073 2018-08-27 Jakub Jelinek <jakub@redhat.com>
2075 PR rtl-optimization/87065
2076 * combine.c (simplify_if_then_else): Formatting fix.
2077 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
2079 (known_cond): Don't return const_true_rtx for vector modes. Use
2080 CONST0_RTX instead of const0_rtx. Formatting fixes.
2082 2018-08-27 Martin Liska <mliska@suse.cz>
2084 PR gcov-profile/87069
2085 * gcov.c (process_file): Record files already processed
2086 and warn about a file being processed multiple times.
2088 2018-08-27 Martin Liska <mliska@suse.cz>
2091 * config/aarch64/aarch64.c (aarch64_override_options_internal):
2092 Set default values for x_aarch64_*_string strings.
2093 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
2094 prefix. For -mabi do not print '=ABI' in help and use
2095 <option_value> format for -msve-vector-bits and -moverride
2098 2018-08-26 Jeff Law <law@redhat.com>
2100 * config/mips/frame-header-opt.c: Include "backend.h" rather than
2103 2018-08-26 Marek Polacek <polacek@redhat.com>
2105 PR c++/87029, Implement -Wredundant-move.
2106 * doc/invoke.texi: Document -Wredundant-move.
2108 2018-08-25 Martin Sebor <msebor@redhat.com>
2110 PR tree-optimization/87059
2111 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
2112 to the same type as the other.
2113 * fold-const.c (fold_binary_loc): Assert expectation.
2115 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
2117 * config/darwin.c (machopic_legitimize_pic_address): Clean up
2118 extraneous parentheses, dead code section and formatting.
2120 2018-08-24 David Malcolm <dmalcolm@redhat.com>
2123 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
2124 wide enough for jumps in the line-numbering to be visible.
2125 (layout::print_gap_in_line_numbering): New member function.
2126 (layout::calculate_line_spans): When using line numbering, merge
2127 line spans that are only 1 line apart.
2128 (diagnostic_show_locus): When printing line numbers, show gaps in
2129 line numbering directly, rather than printing headers.
2130 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
2131 line-numbering with multiple line spans.
2132 (selftest::test_fixit_insert_containing_newline_2): Add test of
2133 line-numbering, in which the spans are close enough to be merged.
2135 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
2137 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
2138 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
2139 * tree-vrp.c (range_is_nonnull): Remove.
2140 (range_includes_zero_p): Accept value_range instead of min/max.
2141 (extract_range_from_binary_expr_1): Do not early bail on
2143 Use range_includes_zero_p instead of range_is_nonnull.
2144 (extract_range_from_unary_expr): Use range_includes_zero_p instead
2145 of range_is_nonnull.
2146 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
2147 special case VR_ANTI_RANGE.
2148 (vrp_finalize): Same.
2149 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
2151 (range_is_nonnull): Remove.
2152 * vr-values.c (vrp_stmt_computes_nonzero): Use
2153 range_includes_zero_p instead of range_is_nonnull.
2154 (extract_range_basic): Pass value_range to range_includes_zero_p
2155 instead of range_is_nonnull.
2157 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
2159 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
2160 * emit-rtl.h (rtl_data): Remove return_bnd.
2161 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
2162 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
2163 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
2164 (POINTER_BOUNDS_MODE): Remove definition.
2165 (make_pointer_bounds_mode): Remove.
2166 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
2167 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
2168 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
2169 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
2170 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
2171 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
2172 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
2174 * config/i386/i386-modes.def (BND32, BND64): Remove.
2175 * config/i386/i386.c (dbx_register_map): Remove bound registers.
2176 (dbx64_register_map): Ditto.
2177 (svr4_dbx_register_map): Ditto.
2178 (indirect_thunk_bnd_needed): Remove.
2179 (indirect_thunks_bnd_used): Ditto.
2180 (indirect_return_bnd_needed): Ditto.
2181 (indirect_return_via_cx_bnd): Ditto.
2182 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
2183 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
2184 (output_indirect_thunk): Ditto. Remove need_prefix argument.
2185 (output_indirect_thunk_function): Remove handling of
2186 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
2187 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
2188 (ix86_save_reg): Remove handling of crtl->return_bnd.
2189 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
2190 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
2191 and UNSPEC_BNDLX_ADDR.
2192 (ix86_output_indirect_branch_via_reg): Remove handling of
2193 indirect_thunk_prefix_bnd.
2194 (ix86_output_indirect_branch_via_push): Ditto.
2195 (ix86_output_function_return): Ditto.
2196 (ix86_output_indirect_function_return): Ditto.
2197 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
2198 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
2199 (CALL_USED_REGISTERS): Ditto.
2200 (REG_ALLOC_ORDER): Update for removal of bound registers.
2201 (HI_REGISTER_NAMES): Ditto.
2202 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
2203 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
2204 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
2205 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
2206 (FIRST_PSEUDO_REG): Update.
2207 (BND): Remove mode iterator.
2208 * config/i386/predicates.md (bnd_mem_operator): Remove.
2210 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
2212 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
2215 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
2217 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
2218 the case in which the permute needs only a single element and
2219 repeats for every vector of the result. Extend that case to
2220 handle variable-length vectors.
2221 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
2223 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
2226 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
2227 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
2229 2018-08-24 Richard Biener <rguenther@suse.de>
2231 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
2232 bb_flags_allocated members.
2233 (auto_flag): New RAII class for allocating flags.
2234 (auto_edge_flag): New RAII class for allocating edge flags.
2235 (auto_bb_flag): New RAII class for allocating bb flags.
2236 * cfgloop.c (verify_loop_structure): Allocate temporary edge
2238 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
2239 in favor of temporarily allocated BB flag.
2240 * hsa-brig.c: Re-order includes.
2241 * hsa-dump.c: Likewise.
2242 * hsa-regalloc.c: Likewise.
2243 * print-rtl.c: Likewise.
2244 * profile-count.c: Likewise.
2246 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
2249 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
2252 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
2255 * wide-int-range.cc (wide_int_range_div): Do not ignore result
2256 from wide_int_range_multiplicative_op.
2258 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2260 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
2263 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
2265 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
2268 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
2270 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
2271 full GENERIC expression used for replacement.
2273 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
2275 * tree-vrp.c (abs_extent_range): Remove.
2276 (extract_range_into_wide_ints): Pass wide ints by reference.
2277 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
2278 Pass wide ints by reference in all calls to
2279 extract_range_into_wide_ints.
2280 * wide-int-range.cc (wide_int_range_div): New.
2281 * wide-int-range.h (wide_int_range_div): New.
2282 (wide_int_range_includes_zero_p): New.
2283 (wide_int_range_zero_p): New.
2285 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
2287 * config/aarch64/aarch64.md (arches): New enum.
2288 (arch): New enum attr.
2289 (arch_enabled): New attr.
2290 (enabled): Now uses arch_enabled only.
2291 (simd, sve, fp16): Removed attribute.
2292 (fp): Attr now defined in terms of 'arch'.
2293 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
2294 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
2295 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
2296 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
2297 attributes into 'arch'.
2298 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
2299 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
2300 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
2301 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
2302 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
2303 'simd' attribute into 'arch'.
2304 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
2305 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
2306 Convert use of 'fp' attribute to 'arch'.
2307 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
2308 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
2310 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
2311 (different modes) Merge 'fp' and 'simd' into 'arch'.
2312 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
2315 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
2317 PR rtl-optimization/87026
2318 * expmed.c (canonicalize_comparison): If we can no longer create
2319 pseudoregisters, don't.
2321 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
2324 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
2326 * config/arm/arm.c (speculation_barrier_libfunc): New static
2328 (arm_init_libfuncs): Initialize it.
2329 (arm_emit_speculation_barrier): New function.
2330 * config/arm/arm.md (speculation_barrier): Call
2331 arm_emit_speculation_barrier for architectures that do not have
2333 (speculation_barrier_insn): Only match on Armv7 or later.
2335 2018-08-23 Richard Biener <rguenther@suse.de>
2338 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
2341 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
2343 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
2344 of single-vector TBLs.
2345 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
2348 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
2351 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
2352 aarch64_evpc_tbl guard.
2354 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
2356 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
2359 2018-08-22 Martin Sebor <msebor@redhat.com>
2362 * tree-pretty-print.c (pretty_print_string): Add argument.
2363 (dump_generic_node): Call to pretty_print_string with string size.
2365 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
2367 PR rtl-optimization/86771
2368 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
2369 of two SETs into those two SETs, one to be placed at i2, if that SETs
2370 destination is modified between i2 and i3.
2372 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2374 PR tree-optimization/86725
2375 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
2377 (vect_analyze_scalar_cycles_1): Check it.
2379 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2381 PR tree-optimization/86725
2382 * tree-vect-loop.c (vect_is_simple_reduction): When treating
2383 an outer loop phi as a double reduction, make sure that the
2384 single user of the phi result is an inner loop phi.
2386 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2388 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
2389 grouped stores with gaps to a strided group.
2391 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2393 * tree-vect-stmts.c (get_group_load_store_type)
2394 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
2395 first statement in a group.
2397 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2399 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
2400 the sequence used in gcc/gcc.c.
2402 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2405 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
2408 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2410 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
2411 Darwin10-specific unwinder-shim.
2412 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
2413 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
2414 New to cater for Darwin10 Rosetta.
2416 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2418 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
2419 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
2421 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2426 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
2427 (dwarf2out_switch_text_section): Generate a local label for the
2428 second function sub-section and apply it as the second FDE start
2430 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
2431 second sub-section start.
2433 2018-08-22 Richard Biener <rguenther@suse.de>
2435 PR tree-optimization/86988
2436 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
2438 2018-08-22 Richard Biener <rguenther@suse.de>
2440 PR tree-optimization/86945
2441 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
2443 2018-08-22 Alexandre Oliva <oliva@adacore.com>
2445 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
2446 a comment about how uses of r2 for .sdata2 come about.
2448 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
2450 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
2452 2018-08-21 Marek Polacek <polacek@redhat.com>
2454 PR c++/86981, Implement -Wpessimizing-move.
2455 * doc/invoke.texi: Document -Wpessimizing-move.
2457 2018-08-21 Jan Hubicka <jh@suse.cz>
2459 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
2460 * tree.h (is_redundant_typedef): Remove.
2461 * dwarf2out.c (is_redundant_typedef): Turn into static function.
2463 2018-08-21 Jan Hubicka <jh@suse.cz>
2465 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
2468 2018-08-21 Tamar Christina <tamar.christina@arm.com>
2470 * expmed.c (extract_low_bits): Reject invalid subregs early.
2472 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
2475 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
2478 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2480 * config/vxworks.h: Guard vxworks_asm_out_constructor and
2481 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
2482 * config/vxworks.c: Likewise.
2484 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2486 * config/vxworks.c: Set targetm.have_ctors_dtors
2487 if HAVE_INITFINI_ARRAY_SUPPORT.
2488 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
2489 if HAVE_INITFINI_ARRAY_SUPPORT.
2491 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2493 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
2494 default search path for VxWorks < 7.
2496 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
2498 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
2499 (get_format_string): Refer to c_getstr.
2501 2018-08-21 Tom de Vries <tdevries@suse.de>
2503 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2504 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2505 (debuginfo_early_stop): Declare.
2506 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2507 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2508 (debuginfo_early_stop): New function.
2509 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
2510 and debuginfo_early_stop.
2511 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
2512 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
2513 (general_init): Call debuginfo_early_init.
2514 (finalize): Call debuginfo_fini.
2515 (do_compile): Call debuginfo_init.
2516 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
2518 (@item -fdump-debug, @item -fdump-earlydebug): Add.
2520 2018-08-21 Tom de Vries <tdevries@suse.de>
2522 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
2523 flag_dump_noaddr and flag_dump_unnumbered.
2525 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
2527 * wide-int-range.cc (wide_int_range_abs): New.
2528 (wide_int_range_order_set): Rename from wide_int_range_min_max.
2529 * wide-int-range.h (wide_int_range_abs): New.
2530 (wide_int_range_min_max): New.
2531 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
2532 case to call wide_int_range_abs.
2533 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
2534 (extract_range_from_abs_expr): Delete.
2536 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
2539 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
2540 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
2543 2018-08-20 Nathan Sidwell <nathan@acm.org>
2544 Jeff Law <law@redhat.com>
2546 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
2547 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
2549 2018-08-20 David Malcolm <dmalcolm@redhat.com>
2552 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
2553 (decl_attributes): Likewise.
2554 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
2556 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2557 * diagnostic-core.h (class auto_diagnostic_group): New class.
2558 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
2559 (diagnostic_report_diagnostic): Handle the first diagnostics within
2561 (emit_diagnostic): Add auto_diagnostic_group instance.
2563 (inform_n): Likewise.
2564 (warning): Likewise.
2565 (warning_at): Likewise.
2566 (warning_n): Likewise.
2567 (pedwarn): Likewise.
2568 (permerror): Likewise.
2570 (error_n): Likewise.
2571 (error_at): Likewise.
2573 (fatal_error): Likewise.
2574 (internal_error): Likewise.
2575 (internal_error_no_backtrace): Likewise.
2576 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
2577 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
2578 * diagnostic.h (struct diagnostic_context): Add fields
2579 "diagnostic_group_nesting_depth",
2580 "diagnostic_group_emission_count", "begin_group_cb",
2582 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
2583 Add auto_diagnostic_group instance(s).
2584 (find_explicit_erroneous_behavior): Likewise.
2585 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
2586 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
2587 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
2588 (gimplify_va_arg_expr): Likewise.
2589 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
2590 (HSA_SORRY_AT): Likewise.
2591 * ipa-devirt.c (compare_virtual_tables): Likewise.
2592 (warn_odr): Likewise.
2593 * multiple_target.c (expand_target_clones): Likewise.
2594 * opts-common.c (cmdline_handle_error): Likewise.
2595 * reginfo.c (globalize_reg): Likewise.
2596 * substring-locations.c (format_warning_n_va): Likewise.
2597 * tree-inline.c (expand_call_inline): Likewise.
2598 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
2599 * tree-ssa-loop-niter.c
2600 (do_warn_aggressive_loop_optimizations): Likewise.
2601 * tree-ssa-uninit.c (warn_uninit): Likewise.
2602 * tree.c (warn_deprecated_use): Likewise.
2604 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
2607 * config/i386/i386.md (eh_return): Always update EH return
2608 address in word_mode.
2610 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
2612 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
2613 TARGET_SPLIT_COMPLEX_ARG is defined.
2615 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
2617 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
2619 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
2622 * expr.c (expand_assignment): Assert that bitpos is positive.
2623 (store_field): Likewise
2624 (expand_expr_real_1): Make sure that bitpos is positive.
2625 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
2628 2018-08-20 Nathan Sidwell <nathan@acm.org>
2630 * Makefile.in (CPP_ID_DATA_H): Delete.
2631 (CPP_INTERNAL_H): Don't add it.
2632 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
2633 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
2635 2018-08-20 Richard Biener <rguenther@suse.de>
2637 PR tree-optimization/78655
2638 * tree-vrp.c (extract_range_from_binary_expr_1): Make
2639 pointer + offset nonnull if either operand is nonnull work.
2641 2018-08-20 Tom de Vries <tdevries@suse.de>
2643 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
2644 unless the referenced die describes the added property using
2645 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
2646 Otherwise, add a DW_AT_location to the referenced die.
2648 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
2651 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
2652 register_operand when calling ix86_set_reg_reg_cost.
2653 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
2654 Set *total to 0 for operands that satisfy x86_64_immediate_operand
2655 predicate and to 1 otherwise.
2657 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
2659 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
2660 emit a diagnostic that it is not supported and reset the option.
2661 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
2662 supported and consume the option. (ASM_FINAL_SPEC): New.
2664 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
2666 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
2669 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
2671 C-SKY port: Documentation
2673 * doc/extend.texi (C-SKY Function Attributes): New section.
2674 * doc/invoke.texi (Option Summary): Add C-SKY options.
2675 (C-SKY Options): New section.
2676 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
2678 2018-08-17 Jojo <jijie_rong@c-sky.com>
2679 Huibin Wang <huibin_wang@c-sky.com>
2680 Sandra Loosemore <sandra@codesourcery.com>
2681 Chung-Lin Tang <cltang@codesourcery.com>
2683 C-SKY port: Backend implementation
2685 * config/csky/*: New.
2686 * common/config/csky/*: New.
2688 2018-08-17 Jojo <jijie_rong@c-sky.com>
2689 Huibin Wang <huibin_wang@c-sky.com>
2690 Sandra Loosemore <sandra@codesourcery.com>
2691 Chung-Lin Tang <cltang@codesourcery.com>
2692 Andrew Jenner <andrew@codesourcery.com>
2694 C-SKY port: Configury
2696 * config.gcc (csky-*-*): New.
2697 * configure.ac: Add csky to targets for dwarf2 debug_line support.
2698 * configure: Regenerated.
2700 2018-08-17 David Malcolm <dmalcolm@redhat.com>
2702 * dump-context.h: Include "dumpfile.h".
2703 (dump_context::dump_printf_va): Convert final param from va_list
2704 to va_list *. Convert from ATTRIBUTE_PRINTF to
2705 ATTRIBUTE_GCC_DUMP_PRINTF.
2706 (dump_context::dump_printf_loc_va): Likewise.
2707 * dumpfile.c: Include "stringpool.h".
2708 (make_item_for_dump_printf_va): Delete.
2709 (make_item_for_dump_printf): Delete.
2710 (class dump_pretty_printer): New class.
2711 (dump_pretty_printer::dump_pretty_printer): New ctor.
2712 (dump_pretty_printer::emit_items): New member function.
2713 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
2715 (dump_pretty_printer::emit_item): New member function.
2716 (dump_pretty_printer::stash_item): New member function.
2717 (dump_pretty_printer::format_decoder_cb): New member function.
2718 (dump_pretty_printer::decode_format): New member function.
2719 (dump_context::dump_printf_va): Reimplement in terms of
2720 dump_pretty_printer.
2721 (dump_context::dump_printf_loc_va): Convert final param from va_list
2723 (dump_context::begin_scope): Reimplement call to
2724 make_item_for_dump_printf.
2725 (dump_printf): Update for change to dump_printf_va.
2726 (dump_printf_loc): Likewise.
2727 (selftest::test_capture_of_dump_calls): Convert "stmt" from
2728 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
2729 with %T, %E, and %G.
2730 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
2731 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
2732 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
2733 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
2734 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
2735 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
2736 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
2737 within a dump_printf_loc call to "%wu".
2738 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
2739 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
2740 missing space after "=".
2741 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
2742 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
2743 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
2744 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
2745 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
2746 duplicate "vectorized" from message.
2748 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
2750 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
2751 polyNxK_t element's TYPE_STRING_FLAG.
2753 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
2755 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
2756 (they were unnamed before). Fix comments.
2758 2018-08-17 Nathan Sidwell <nathan@acm.org>
2760 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
2762 2018-08-17 Richard Biener <rguenther@suse.de>
2764 PR tree-optimization/86841
2765 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
2767 2018-08-17 Martin Liska <mliska@suse.cz>
2769 * common.opt: Remove Warn, Init and Report for options with
2770 Ignore/Deprecated flag. Warning is done automatically for
2772 * config/i386/i386.opt: Likewise.
2773 * config/ia64/ia64.opt: Likewise.
2774 * config/rs6000/rs6000.opt: Likewise.
2775 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
2776 Remove usage of flag_check_pointer_bounds.
2777 * lto-wrapper.c (merge_and_complain): Do not handle
2778 OPT_fcheck_pointer_bounds.
2779 (append_compiler_options): Likewise.
2780 * opt-functions.awk: Do not handle Deprecated.
2781 * optc-gen.awk: Check that Var, Report and Init are not
2782 used for an option with Ignore/Deprecated flag.
2783 * opts-common.c (decode_cmdline_option): Do not report
2785 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
2787 * opts.h (struct cl_option): Remove cl_deprecated flag.
2788 (CL_ERR_DEPRECATED): Remove error enum value.
2790 2018-08-17 Richard Biener <rguenther@suse.de>
2793 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
2794 across a va-arg-pack using call adjust its return value accordingly.
2796 2018-08-16 Martin Sebor <msebor@redhat.com>
2798 PR tree-optimization/86853
2799 * gimple-ssa-sprintf.c (struct format_result): Rename member.
2800 (struct fmtresult): Add member and initialize it in ctors.
2801 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
2802 (format_string): Handle %S the same as %ls. Set MAYFAIL.
2803 (format_directive): Set POSUNDER4K when MAYFAIL is set.
2804 (parse_directive): Handle %C same as %c.
2805 (sprintf_dom_walker::compute_format_length): Adjust.
2806 (is_call_safe): Adjust.
2808 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
2810 * builtins.c (c_strlen): Add new parameter eltsize. Use it
2811 for determining how to count the elements.
2812 * builtins.h (c_strlen): Adjust prototype.
2813 * expr.c (string_constant): Add new parameter mem_size.
2814 Set *mem_size appropriately.
2815 * expr.h (string_constant): Adjust protoype.
2816 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
2817 * gimple-fold.h (get_range_strlen): Adjust prototype.
2818 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
2819 (format_string): Call get_string_length with eltsize.
2821 2018-08-16 David Malcolm <dmalcolm@redhat.com>
2823 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
2824 to emit the span, rather than setting it as the prefix.
2826 2018-08-16 David Malcolm <dmalcolm@redhat.com>
2828 * diagnostic-show-locus.c (layout::start_annotation_line): Add
2829 "margin_char" parameter, defaulting to space. Use it in place
2830 of pp_space for the initial part of the margin.
2831 (layout::print_leading_fixits): Use '+' when filling the margin
2832 of line-insertion fix-it hints.
2834 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2836 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
2839 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2841 * config/rs6000/altivec.md: Don't set length attribute to the default
2843 * config/rs6000/darwin.md: Ditto.
2844 * config/rs6000/dfp.md: Ditto.
2845 * config/rs6000/htm.md: Ditto.
2846 * config/rs6000/rs6000.md: Ditto.
2847 * config/rs6000/sync.md: Ditto.
2848 * config/rs6000/vsx.md: Ditto.
2850 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2852 * config/rs6000/altivec.md: Don't set length attribute to the default
2853 value, for branch instructions.
2854 * config/rs6000/darwin.md: Ditto.
2855 * config/rs6000/rs6000.md: Ditto.
2857 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2859 * config/rs6000/rs6000.md (length): Always define as const_int 4.
2860 (unnamed conditional branch define_insn): Set length to 4 or 8
2861 depending on offset.
2862 (<bd>_<mode>): Similar, for alternative 0.
2863 (<bd>tf_<mode>): Ditto.
2865 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2867 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
2869 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
2871 * doc/rtl.texi: Replace old RTX class names with new names.
2874 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
2876 * expmed.h (canonicalize_comparison): New declaration.
2877 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
2878 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
2879 * optabs.c (prepare_cmp_insn): Likewise.
2880 * rtl.h (unsigned_condition_p): New function which checks if a
2881 comparison operator is unsigned.
2883 2018-08-16 Nathan Sidwell <nathan@acm.org>
2885 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
2886 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
2888 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2891 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
2892 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
2893 (mov<mov>): ..this and enable unconditionally.
2895 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2897 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
2899 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
2901 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
2902 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
2903 "Common" with "Target".
2905 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
2907 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
2908 * doc/invoke.texi (mmitigate-rop): Remove.
2909 * config/i386/i386.c: Do not include "regrename.h".
2910 (ix86_rop_should_change_byte_p, reg_encoded_number)
2911 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
2913 (ix86_reorg): Remove call to ix86_mitigate_rop.
2914 * config/i386/i386.md (attr "modrm_class"): Remove.
2915 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
2916 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
2917 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
2919 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
2921 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
2922 allow folding of mergeh() and mergel() for the float and double types.
2923 (fold_mergehl_helper): Rework to handle building a permute tree
2926 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
2928 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
2931 2018-08-15 David Malcolm <dmalcolm@redhat.com>
2933 * common.opt (fdiagnostics-show-labels): New option.
2934 * diagnostic-show-locus.c (class layout_range): Add field
2936 (class layout): Add field "m_show_labels_p".
2937 (layout_range::layout_range): Add param "label" and use it to
2939 (make_range): Pass in NULL for new "label" param of layout_range's
2941 (layout::layout): Initialize m_show_labels_p.
2942 (layout::maybe_add_location_range): Pass in loc_range->m_label
2943 when constructing layout_range instances.
2944 (struct line_label): New struct.
2945 (layout::print_any_labels): New member function.
2946 (layout::print_line): Call it if label-printing is enabled.
2947 (selftest::test_one_liner_labels): New test.
2948 (selftest::test_diagnostic_show_locus_one_liner): Call it.
2949 * diagnostic.c (diagnostic_initialize): Initialize
2950 context->show_labels_p.
2951 * diagnostic.h (struct diagnostic_context): Add field
2953 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
2954 -fno-diagnostics-show-labels.
2955 * dwarf2out.c (gen_producer_string): Add
2956 OPT_fdiagnostics_show_labels to the ignored options.
2957 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
2959 (gcc_rich_location::maybe_add_expr): Likewise.
2960 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
2961 label" param, defaulting to NULL.
2962 (gcc_rich_location::add_expr): Add "label" param.
2963 (gcc_rich_location::maybe_add_expr): Likewise.
2964 (class text_range_label): New class.
2965 (class range_label_for_type_mismatch): New class.
2966 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
2967 of format_warning_va.
2968 (fmtwarn_n): Likewise for new params of format_warning_n_va.
2969 * lto-wrapper.c (merge_and_complain): Add
2970 OPT_fdiagnostics_show_labels to the "pick one setting" options.
2971 (append_compiler_options): Likewise to the dropped options.
2972 (append_diag_options): Likewise to the passed-on options.
2973 * opts.c (common_handle_option): Handle the new option.
2974 * selftest-diagnostic.c
2975 (test_diagnostic_context::test_diagnostic_context): Enable
2977 * substring-locations.c: Include "gcc-rich-location.h".
2978 (format_warning_n_va): Add "fmt_label" and "param_label" params
2979 and use them as appropriate.
2980 (format_warning_va): Add "fmt_label" and "param_label" params,
2981 passing them on to format_warning_n_va.
2982 (format_warning_at_substring): Likewise.
2983 (format_warning_at_substring_n): Likewise.
2984 * substring-locations.h (format_warning_va): Add "fmt_label" and
2985 "param_label" params.
2986 (format_warning_n_va): Likewise.
2987 (format_warning_at_substring): Likewise.
2988 (format_warning_at_substring_n): Likewise.
2989 * toplev.c (general_init): Initialize global_dc->show_labels_p.
2991 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
2994 * builtins.c (expand_builtin_memcmp): Do not expand the call
2995 when overflow is detected.
2997 2018-08-15 Martin Sebor <msebor@redhat.com>
2999 PR tree-optimization/71625
3000 * config/aarch64/aarch64-builtins.c
3001 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
3003 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
3005 * config/s390/s390.c (s390_reorg): Remove loop.
3007 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
3010 (darwin_function_switched_text_sections): Delete.
3011 * gcc/config/darwin.h
3012 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
3014 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
3017 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
3018 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
3019 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
3021 2018-08-15 Martin Liska <mliska@suse.cz>
3023 PR tree-optimization/86925
3024 * predict.c (expr_expected_value_1): When taking
3025 later predictor, assign also probability.
3026 Use fold_build2_initializer_loc in order to fold
3027 the expression in -frounding-math.
3029 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
3031 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
3033 (expand_vec_perm_1): Try the new method.
3035 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
3038 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
3039 Check whether lra_live_max_point is 0 before dividing.
3041 2018-08-14 Martin Sebor <msebor@redhat.com>
3043 PR tree-optimization/86650
3044 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
3045 (vrp_prop::check_mem_ref): Same.
3047 2018-08-13 Liu Hao <lh_mouse@126.com>
3049 * pretty-print.c (eat_esc_sequence): Swap the foreground and
3050 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
3051 and clear it thereafter, as it only works for DBCS.
3053 2018-08-13 Liu Hao <lh_mouse@126.com>
3055 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
3056 handle returned by _get_osf_handle().
3058 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3060 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
3061 for folding vec_perm.
3063 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3065 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
3066 Add support for gimple-folding of vec_pack() and vec_unpack()
3069 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3071 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
3072 vec_xst variants to the list.
3073 (rs6000_gimple_fold_builtin): Add support for folding unaligned
3074 vector loads and stores.
3076 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
3078 * config.gcc (rs6000-ibm-aix4.x): Delete.
3079 (rs6000-ibm-aix5.1): Delete.
3080 (rs6000-ibm-aix5.2): Delete.
3081 (rs6000-ibm-aix5.3): Delete.
3082 * config/rs6000/aix43.h: Delete.
3083 * config/rs6000/aix51.h: Delete.
3084 * config/rs6000/aix52.h: Delete.
3085 * config/rs6000/t-aix43: Delete.
3087 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
3089 * config/s390/s390.c (s390_decompose_constant_pool_ref):
3091 (s390_decompose_address): Factor out constant pool ref
3094 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3096 * config/nds32/nds32-predicates.c
3097 (nds32_can_use_bclr_p): Change return type as bool.
3098 (nds32_can_use_bset_p): Ditto.
3099 (nds32_can_use_btgl_p): Ditto.
3100 (nds32_can_use_bitci_p): Ditto.
3101 * config/nds32/nds32-protos.h
3102 (nds32_can_use_bclr_p): Change declaration.
3103 (nds32_can_use_bset_p): Ditto.
3104 (nds32_can_use_btgl_p): Ditto.
3105 (nds32_can_use_bitci_p): Ditto.
3107 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3109 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
3110 Support -msched-prolog-epilog option.
3111 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
3113 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3115 * common/config/nds32/nds32-common.c
3116 (nds32_option_optimization_table): Enalbe -malways-align.
3118 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3120 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
3122 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
3123 OPT_misr_secure_ case.
3124 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
3125 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
3126 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
3128 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
3129 (nds32_isr_info): New field security_level.
3130 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
3131 * config/nds32/nds32.md (return_internal): Consider critical attribute.
3132 * config/nds32/nds32.opt (misr-secure): New option.
3133 * config/nds32/nds32_init.inc: New file.
3134 * config/nds32/nds32_isr.h: New file.
3136 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
3138 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
3139 Update comment for atomic instructions.
3140 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
3141 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
3143 (atomic_loaddi): Revise fence expansion to only emit fence prior to
3144 load for __ATOMIC_SEQ_CST model.
3145 (atomic_loaddi_1): Remove float register target.
3146 (atomic_storedi): Handle CONST_INT values.
3147 (atomic_storedi_1): Remove float register source. Add special case
3149 (memory_barrier): New expander and insn.
3151 2018-08-11 Jakub Jelinek <jakub@redhat.com>
3153 PR tree-optimization/86835
3154 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
3155 new_stmt after def_gsi, make sure to insert new_square_stmt after
3156 that stmt, not 2 stmts before it.
3158 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
3161 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
3164 2018-08-10 Martin Liska <mliska@suse.cz>
3167 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
3169 * builtins.c (expand_builtin_expect_with_probability):
3171 (expand_builtin_expect_with_probability): New function.
3172 (build_builtin_expect_predicate): Add new argumnet probability
3173 for BUILT_IN_EXPECT_WITH_PROBABILITY.
3174 (fold_builtin_expect):
3177 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
3178 * builtins.h (fold_builtin_expect): Set new argument.
3179 * doc/extend.texi: Document __builtin_expect_with_probability.
3180 * doc/invoke.texi: Likewise.
3181 * gimple-fold.c (gimple_fold_call): Pass new argument.
3182 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
3183 also BUILT_IN_EXPECT_WITH_PROBABILITY.
3184 * predict.c (get_predictor_value): New function.
3185 (expr_expected_value): Add new argument probability. Assume
3186 that predictor and probability are always non-null.
3187 (expr_expected_value_1): Likewise. For __builtin_expect and
3188 __builtin_expect_with_probability set probability. Handle
3189 combination in binary expressions.
3190 (tree_predict_by_opcode): Simplify code by simply calling
3191 get_predictor_value.
3192 (pass_strip_predict_hints::execute): Add handling of
3193 BUILT_IN_EXPECT_WITH_PROBABILITY.
3194 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
3196 * tree.h (DECL_BUILT_IN_P): New function.
3198 2018-08-10 Martin Liska <mliska@suse.cz>
3200 PR tree-optimization/85799
3201 * passes.def: Add argument for pass_strip_predict_hints.
3202 * predict.c (class pass_strip_predict_hints): Add new argument
3204 (strip_predictor_early): New function.
3205 (pass_strip_predict_hints::execute): Call the function to
3207 (strip_predict_hints): New function.
3208 * predict.def: Fix comment.
3210 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
3212 * Makefile.in: Clarify which tm.texi to copy over to assert the
3213 right to grant a GFDL license for all.
3215 2018-08-09 Jeff Law <law@redhat.com>
3217 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
3220 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
3222 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
3225 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
3227 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
3228 reductions for variable-length vectors.
3230 2018-08-09 David Malcolm <dmalcolm@redhat.com>
3233 * common.opt (fdiagnostics-show-line-numbers): New option.
3234 * diagnostic-show-locus.c (class layout): Add fields
3235 "m_show_line_numbers_p" and "m_linenum_width";
3236 (num_digits): New function.
3237 (test_num_digits): New function.
3238 (layout::layout): Initialize new fields. Update m_x_offset
3239 logic to handle any left margin.
3240 (layout::print_source_line): Print line number when requested.
3241 (layout::start_annotation_line): New member function.
3242 (layout::print_annotation_line): Call it.
3243 (layout::print_leading_fixits): Likewise.
3244 (layout::print_trailing_fixits): Likewise. Update calls to
3245 move_to_column for new parameter.
3246 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
3247 it to potentially call start_annotation_line.
3248 (layout::show_ruler): Call start_annotation_line.
3249 (selftest::test_line_numbers_multiline_range): New selftest.
3250 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
3251 and selftest::test_line_numbers_multiline_range.
3252 * diagnostic.c (diagnostic_initialize): Initialize
3253 show_line_numbers_p.
3254 * diagnostic.h (struct diagnostic_context): Add field
3255 "show_line_numbers_p".
3256 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
3257 -fno-diagnostics-show-line-numbers.
3258 * dwarf2out.c (gen_producer_string): Add
3259 OPT_fdiagnostics_show_line_numbers to the ignored options.
3260 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
3261 one setting" options.
3262 (append_compiler_options): Likewise to the dropped options.
3263 (append_diag_options): Likewise to the passed-on options.
3264 * opts.c (common_handle_option): Handle the new option.
3265 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
3267 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
3269 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
3270 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
3271 third argument of type "const signed char" to descriptions of
3272 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
3273 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
3274 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
3275 __builtin_bcdsub_ov functions.
3277 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
3279 PR tree-optimization/86858
3280 * tree-vect-loop.c (vect_is_simple_reduction): Restore
3281 flow_bb_inside_loop_p calls.
3283 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
3285 PR tree-optimization/86871
3286 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
3287 instead of gimple_assign_lhs.
3289 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
3292 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
3293 register constraint to operand 0.
3294 (add<mode>3_carryinC): Likewise.
3295 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
3297 2018-08-09 Martin Liska <mliska@suse.cz>
3300 * common.opt: Remove extra line.
3302 2018-08-09 Martin Liska <mliska@suse.cz>
3304 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
3305 at the end of a line, make first letter capital and end up
3306 a sentence with a dot.
3307 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
3308 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
3309 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
3310 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
3311 (PARAM_MAX_ISL_OPERATIONS): Likewise.
3312 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
3313 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
3314 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
3315 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
3316 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
3317 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
3318 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
3319 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
3320 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
3321 (PARAM_TREE_REASSOC_WIDTH): Likewise.
3322 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
3323 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
3324 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
3326 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
3329 * config/s390/s390.c (s390_option_override_internal): Reduce the
3330 stack-clash-protection-probe-interval param if it would be too big
3333 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
3336 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
3337 * config/m68k/m68k.c (handle_move_double): Don't call
3338 m68k_final_prescan_insn.
3339 (m68k_adjust_decorated_operand): Renamed from
3340 m68k_final_prescan_insn, remove first and third operand and
3342 (print_operand): Call it.
3343 (print_operand_address): Call it.
3345 2018-08-08 Nathan Sidwell <nathan@acm.org>
3347 * diagnostic.c (diagnostic_report_current_module): Use
3348 linemap_included_from & linemap_included_from_linemap.
3350 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
3352 * config/aarch64/aarch64-cores.def: Add phecda core.
3353 * config/aarch64/aarch64-tune.md: Regenerate.
3354 * doc/invoke.texi: Add phecda core.
3356 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
3359 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
3361 * config/s390/s390.md ("movti"): Add more alternatives for
3362 constant to GPR copies.
3364 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
3366 * config/s390/s390.c: Fix whitespace damage throughout the file.
3367 * config/s390/s390.h: Likewise.
3368 * config/s390/tpf.h: Likewise.
3370 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
3372 * config/s390/s390.c (s390_loadrelative_operand_p):
3373 Remove TARGET_CPU_ZARCH usages.
3374 (s390_rtx_costs): Likewise.
3375 (s390_legitimate_constant_p): Likewise.
3376 (s390_cannot_force_const_mem): Likewise.
3377 (legitimate_reload_constant_p): Likewise.
3378 (s390_preferred_reload_class): Likewise.
3379 (legitimize_pic_address): Likewise.
3380 (legitimize_tls_address): Likewise.
3381 (s390_split_branches): Removed.
3382 (s390_add_execute): Removed.
3383 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
3384 (s390_mainpool_start): Likewise.
3385 (s390_mainpool_finish): Likewise.
3386 (s390_mainpool_cancel): Removed.
3387 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
3388 (s390_chunkify_cancel): Likewise.
3389 (s390_return_addr_rtx): Likewise.
3390 (s390_register_info): Remove split_branches_pending_p uages.
3391 (s390_optimize_register_info): Likewise.
3392 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
3393 split_branches_pending_p usages.
3394 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
3395 (s390_load_got): Likewise.
3396 (s390_expand_split_stack_prologue): Likewise.
3397 (output_asm_nops): Likewise.
3398 (s390_function_profiler): Likewise.
3399 (s390_emit_call): Likewise.
3400 (s390_conditional_register_usage): Likewise.
3401 (s390_optimize_prologue): Likewise.
3402 (s390_reorg): Remove TARGET_CPU_ZARCH and
3403 split_branches_pending_p usages.
3404 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
3406 (s390_output_indirect_thunk_function): Likewise.
3407 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
3408 (TARGET_CPU_ZARCH_P): Removed.
3409 (struct machine_function): Remove split_branches_pending_p.
3410 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
3412 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
3414 * common/config/s390/s390-common.c (processor_flags_table):
3416 * config.gcc: Remove with_arch/with_tune support.
3417 * config/s390/2064.md: Remove cpu attribute comparisons.
3418 * config/s390/driver-native.c (s390_host_detect_local_cpu):
3420 * config/s390/linux.h (ASM_SPEC):
3421 Remove -march support.
3422 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
3423 Use a table to get an arch level.
3424 * config/s390/s390-opts.h (enum processor_type):
3426 * config/s390/s390.c
3427 (processor_table): Remove entries, add arch_level values.
3428 (s390_issue_rate): Remove cases.
3429 (s390_option_override): Adjust
3430 s390_option_override_internal() call.
3431 (s390_option_override_internal): Remove deprecation warning.
3432 (s390_valid_target_attribute_tree): Adjust
3433 s390_option_override_internal() call.
3434 * config/s390/s390.h (struct s390_processor):
3435 Share with s390-c.c, add arch_level field.
3436 * config/s390/s390.md:
3437 Remove occurrences in cpu attribute.
3438 * config/s390/s390.opt: Remove -march/-mtune support.
3439 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
3440 * doc/invoke.texi: Remove deprecation warning.
3442 2018-08-08 Luis Machado <luis.machado@linaro.org>
3444 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
3446 (qdf24xx_tunings): Set vector cost structure to
3447 qdf24xx_vector_cost.
3449 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
3450 <register_sextend>: Set to 3.
3452 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
3455 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
3456 * config/aarch64/aarch64-simd.md
3457 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
3458 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
3459 (aarch64_frecpx<mode>): ...this new pattern.
3460 * config/aarch64/aarch64-simd-builtins.def: Remove comment
3461 about aarch64_frecp<FRECP:frecp_suffix><mode>.
3463 2018-08-07 Martin Liska <mliska@suse.cz>
3466 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
3467 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
3468 * predict.def (PRED_MALLOC_NONNULL): New predictor.
3469 * doc/extend.texi: Document that malloc attribute adds
3472 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
3475 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3476 Define to speculation_safe_value_not_needed.
3478 2018-08-06 Jeff Law <law@redhat.com>
3480 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
3481 the vr_values instance to cprop_into_stmt.
3482 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
3483 (cprop_operand): Also query EVRP to determine if OP is a constant.
3485 2018-08-06 Nathan Sidwell <nathan@acm.org>
3487 * diagnostic.c (diagnostic_report_current_module): Reroll
3488 included-at loop. Translate text.
3490 2018-08-06 David Malcolm <dmalcolm@redhat.com>
3492 * function-tests.c (selftest::test_expansion_to_rtl): Call
3493 free_after_compilation.
3495 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3497 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
3499 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
3501 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
3502 loops with memory block operations from getting unrolled.
3504 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
3507 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3508 Define to speculation_safe_value_not_needed.
3510 2018-08-06 Jeff Law <law@redhat.com>
3512 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
3515 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3518 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
3519 with all enabled __intN types.
3521 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
3523 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3525 * alias.c (record_set): Check for clobber high.
3526 * cfgexpand.c (expand_gimple_stmt): Likewise.
3527 * combine-stack-adj.c (single_set_for_csa): Likewise.
3528 * combine.c (find_single_use_1): Likewise.
3529 (set_nonzero_bits_and_sign_copies): Likewise.
3530 (get_combine_src_dest): Likewise.
3531 (is_parallel_of_n_reg_sets): Likewise.
3532 (try_combine): Likewise.
3533 (record_dead_and_set_regs_1): Likewise.
3534 (reg_dead_at_p_1): Likewise.
3535 (reg_dead_at_p): Likewise.
3536 * dce.c (deletable_insn_p): Likewise.
3537 (mark_nonreg_stores_1): Likewise.
3538 (mark_nonreg_stores_2): Likewise.
3539 * df-scan.c (df_find_hard_reg_defs): Likewise.
3540 (df_uses_record): Likewise.
3541 (df_get_call_refs): Likewise.
3542 * dwarf2out.c (mem_loc_descriptor): Likewise.
3543 * haifa-sched.c (haifa_classify_rtx): Likewise.
3544 * ira-build.c (create_insn_allocnos): Likewise.
3545 * ira-costs.c (scan_one_insn): Likewise.
3546 * ira.c (equiv_init_movable_p): Likewise.
3547 (rtx_moveable_p): Likewise.
3548 (interesting_dest_for_shprep): Likewise.
3549 * jump.c (mark_jump_label_1): Likewise.
3550 * postreload-gcse.c (record_opr_changes): Likewise.
3551 * postreload.c (reload_cse_simplify): Likewise.
3552 (struct reg_use): Add source expr.
3553 (reload_combine): Check for clobber high.
3554 (reload_combine_note_use): Likewise.
3555 (reload_cse_move2add): Likewise.
3556 (move2add_note_store): Likewise.
3557 * print-rtl.c (print_pattern): Likewise.
3558 * recog.c (decode_asm_operands): Likewise.
3559 (store_data_bypass_p): Likewise.
3560 (if_test_bypass_p): Likewise.
3561 * regcprop.c (kill_clobbered_value): Likewise.
3562 (kill_set_value): Likewise.
3563 * reginfo.c (reg_scan_mark_refs): Likewise.
3564 * reload1.c (maybe_fix_stack_asms): Likewise.
3565 (eliminate_regs_1): Likewise.
3566 (elimination_effects): Likewise.
3567 (mark_not_eliminable): Likewise.
3568 (scan_paradoxical_subregs): Likewise.
3569 (forget_old_reloads_1): Likewise.
3570 * reorg.c (find_end_label): Likewise.
3571 (try_merge_delay_insns): Likewise.
3572 (redundant_insn): Likewise.
3573 (own_thread_p): Likewise.
3574 (fill_simple_delay_slots): Likewise.
3575 (fill_slots_from_thread): Likewise.
3576 (dbr_schedule): Likewise.
3577 * resource.c (update_live_status): Likewise.
3578 (mark_referenced_resources): Likewise.
3579 (mark_set_resources): Likewise.
3580 * rtl.c (copy_rtx): Likewise.
3581 * rtlanal.c (reg_referenced_p): Likewise.
3582 (single_set_2): Likewise.
3583 (noop_move_p): Likewise.
3584 (note_stores): Likewise.
3585 * sched-deps.c (sched_analyze_reg): Likewise.
3586 (sched_analyze_insn): Likewise.
3588 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3590 * cse.c (invalidate_reg): New function extracted from...
3591 (invalidate): ...here.
3592 (canonicalize_insn): Check for clobber high.
3593 (invalidate_from_clobbers): invalidate clobber highs.
3594 (invalidate_from_sets_and_clobbers): Likewise.
3595 (count_reg_usage): Check for clobber high.
3596 (insn_live_p): Likewise.
3597 * cselib.c (cselib_expand_value_rtx_1):Likewise.
3598 (cselib_invalidate_regno): Check for clobber in setter.
3599 (cselib_invalidate_rtx): Pass through setter.
3600 (cselib_invalidate_rtx_note_stores):
3601 (cselib_process_insn): Check for clobber high.
3602 * cselib.h (cselib_invalidate_rtx): Add operand.
3604 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3606 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
3607 (mark_not_eliminable): Likewise.
3608 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
3609 * lra-lives.c (process_bb_lives): Check for clobber high.
3610 * lra.c (new_insn_reg): Remember clobber highs.
3611 (collect_non_operand_hard_regs): Check for clobber high.
3612 (lra_set_insn_recog_data): Likewise.
3613 (add_regs_to_insn_regno_info): Likewise.
3614 (lra_update_insn_regno_info): Likewise.
3616 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3618 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
3619 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
3621 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3623 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
3624 (copy_insn_1): Likewise.
3625 (gen_hard_reg_clobber_high): New gen function.
3626 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
3627 * genemit.c (gen_exp): Likewise.
3628 (gen_emit_seq): Pass through info.
3629 (gen_insn): Check for CLOBBER_HIGH.
3630 (gen_expand): Pass through info.
3631 (gen_split): Likewise.
3632 (output_add_clobbers): Likewise.
3633 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
3634 (remove_clobbers): Likewise.
3635 * rtl.h (gen_hard_reg_clobber_high): New declaration.
3637 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3639 * doc/rtl.texi (clobber_high): Add.
3640 (parallel): Add in clobber high
3641 * rtl.c (rtl_check_failed_code3): Add function.
3642 * rtl.def (CLOBBER_HIGH): Add expression.
3643 * rtl.h (RTL_CHECKC3): Add macro.
3644 (rtl_check_failed_code3): Add declaration.
3645 (XC3EXP): Add macro.
3647 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
3650 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
3651 cfun->machine->max_used_stack_alignment if needed.
3653 2018-08-04 Martin Sebor <msebor@redhat.com>
3655 PR tree-optimization/86571
3656 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
3659 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
3662 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3665 2018-08-03 Jeff Law <law@redhat.com>
3668 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3669 Define to speculation_safe_value_not_needed.
3671 2018-08-03 David Malcolm <dmalcolm@redhat.com>
3673 * doc/gcov.texi (-x): Remove duplicate "to".
3674 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
3675 (-Wif-not-aligned): Remove duplicate "is".
3676 (-flto): Remove duplicate "the".
3677 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
3678 duplicate "v5.00.b".
3679 (MSP430 Options): Remove duplicate "and" from the description
3680 of "-mgprel-sec=regexp".
3681 (x86 Options): Remove duplicate copies of "vmldLog102" and
3682 vmlsLog104 from description of "-mveclibabi=type".
3684 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
3686 * internal-fn.h (first_commutative_argument): Declare.
3687 * internal-fn.c (first_commutative_argument): New function.
3688 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
3689 restrictions for pattern statements. Use first_commutative_argument
3690 to look for commutative operands in calls to internal functions.
3692 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
3694 * Makefile.in (wide-int-range.o): New.
3695 * tree-vrp.c: Move all the wide_int_* functions to...
3696 * wide-int-range.cc: ...here.
3697 * tree-vrp.h: Move all the wide_int_* prototypes to...
3698 * wide-int-range.h: ...here.
3700 2018-08-03 Tom de Vries <tdevries@suse.de>
3702 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
3704 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
3705 * except.c (output_function_exception_table): Do early exit if
3706 targetm_common.except_unwind_info (&global_options) == UI_NONE.
3708 2018-08-03 Martin Liska <mliska@suse.cz>
3710 * predict.c (dump_prediction): Change to 2 digits
3713 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
3715 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
3717 (falkor_am_1_gtov_gtov): ... a new insn reservation.
3719 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
3721 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
3722 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
3723 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
3724 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
3725 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
3727 2018-08-02 David Malcolm <dmalcolm@redhat.com>
3729 * diagnostic-show-locus.c (diagnostic_show_locus): Use
3730 pp_take_prefix when saving the existing prefix.
3731 * diagnostic.c (diagnostic_append_note): Likewise.
3732 * langhooks.c (lhd_print_error_function): Likewise.
3733 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
3734 param's type. Free the existing prefix.
3735 (pp_take_prefix): New function.
3736 (pretty_printer::pretty_printer): Drop the prefix parameter.
3737 Rename the length parameter to match the comment.
3738 (pretty_printer::~pretty_printer): Free the prefix.
3739 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
3741 (struct pretty_printer): Drop the "const" from "prefix" field's
3742 type and clarify memory management.
3743 (pp_set_prefix): Drop the "const" from the 2nd param.
3744 (pp_take_prefix): New decl.
3746 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
3748 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
3749 for word_mode_ok here instead of passing as argument.
3750 (expand_block_compare): Change select_block_compare_mode() call.
3751 (expand_strncmp_gpr_sequence): New function.
3752 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
3754 2018-08-02 Jeff Law <law@redhat.com>
3757 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3758 Define to speculation_safe_value_not_needed.
3761 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3762 Define to speculation_safe_value_not_needed.
3764 2018-08-02 Tom de Vries <tdevries@suse.de>
3767 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
3768 function. Return UI_TARGET unconditionally.
3769 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
3770 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
3772 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3774 * genemit.c (print_overload_test): Fix typo.
3776 2018-08-02 Richard Biener <rguenther@suse.de>
3778 PR tree-optimization/86816
3779 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
3780 which checks for value availability before querying it.
3781 (gvn_uses_equal): Use it.
3782 (same_succ_hash): Likewise.
3783 (gimple_equal_p): Likewise.
3785 2018-08-02 Nick Clifton <nickc@redhat.com>
3788 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3789 Define to speculation_safe_value_not_needed.
3792 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3793 Define to speculation_safe_value_not_needed.
3796 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3797 Define to speculation_safe_value_not_needed.
3800 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3801 Define to speculation_safe_value_not_needed.
3804 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3805 Define to speculation_safe_value_not_needed.
3808 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3809 Define to speculation_safe_value_not_needed.
3812 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3813 Define to speculation_safe_value_not_needed.
3816 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3817 Define to speculation_safe_value_not_needed.
3820 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
3821 speculation_safe_value_not_needed.
3824 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
3825 to speculation_safe_value_not_needed.
3827 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3829 * doc/md.texi: Expand the documentation of instruction names
3830 to mention port-local uses. Document '@' in pattern names.
3831 * read-md.h (overloaded_instance, overloaded_name): New structs.
3833 (md_reader::handle_overloaded_name): New member function.
3834 (md_reader::get_overloads): Likewise.
3835 (md_reader::m_first_overload): New member variable.
3836 (md_reader::m_next_overload_ptr): Likewise.
3837 (md_reader::m_overloads_htab): Likewise.
3838 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
3839 m_next_overload_ptr and m_overloads_htab.
3840 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
3841 (get_mode_token, get_code_token, get_int_token): New functions.
3842 (map_attr_string): Add an optional argument that passes back
3843 the associated iterator.
3844 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
3845 (md_reader::handle_overloaded_name, add_overload_instance): New
3847 (apply_iterators): Handle '@' names. Report an error if '@'
3848 is used without iterators.
3849 (initialize_iterators): Initialize the new iterator_group fields.
3850 * genopinit.c (handle_overloaded_code_for)
3851 (handle_overloaded_gen): New functions.
3852 (main): Use them to print declarations of maybe_code_for_* and
3853 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
3854 * genemit.c (print_overload_arguments, print_overload_test)
3855 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
3856 (main): Use it to print definitions of maybe_code_for_* and
3857 maybe_gen_* functions.
3858 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
3859 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
3860 instead of explicit mode checks.
3861 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
3862 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
3863 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
3864 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
3865 (aarch64_expand_compare_and_swap): Likewise
3866 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
3867 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
3868 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
3869 (aarch64_constant_pool_reload_icode): Delete.
3870 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
3871 instead of aarch64_constant_pool_reload_icode. Use
3872 code_for_aarch64_reload_mov instead of explicit mode checks.
3873 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
3874 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
3875 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
3876 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
3877 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
3878 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
3879 (aarch64_atomic_load_op_code): Delete.
3880 (aarch64_emit_atomic_load_op): Likewise.
3881 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
3882 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
3883 instead of aarch64_emit_atomic_load_op.
3884 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
3885 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
3886 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
3887 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
3888 character before the pattern name.
3889 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
3890 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
3891 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
3892 (aarch64_frecps<mode>): Likewise.
3893 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
3894 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
3895 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
3896 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
3897 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
3899 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3901 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
3902 Allow HFmode constants if TARGET_FP_F16INST.
3904 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
3907 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
3908 No longer check last store for clobber of address register.
3910 2018-08-02 Martin Liska <mliska@suse.cz>
3912 PR gcov-profile/86817
3913 * gcov.c (process_all_functions): New function.
3915 (process_file): Move functions processing to
3916 process_all_functions.
3918 2018-08-02 David Malcolm <dmalcolm@redhat.com>
3920 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
3921 "const" to the "gimple *" and "rtx_insn *" parameters.
3922 * dumpfile.h (dump_user_location_t::dump_user_location_t):
3924 (dump_location_t::dump_location_t): Likewise.
3926 2018-08-01 Martin Sebor <msebor@redhat.com>
3928 PR tree-optimization/86650
3929 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
3930 rather than a "gcall *". Directly pass the data of interest
3931 to percent_K_format, rather than building a temporary CALL_EXPR
3933 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
3934 (gimple_fold_builtin_strncat): Adjust.
3935 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
3936 gcall* argument with gimple*.
3937 * gimple-ssa-warn-restrict.c (check_call): Same.
3938 (wrestrict_dom_walker::before_dom_children): Same.
3939 (builtin_access::builtin_access): Same.
3940 (check_bounds_or_overlap): Same
3941 (maybe_diag_overlap): Same.
3942 (maybe_diag_offset_bounds): Same.
3943 * tree-diagnostic.c (default_tree_printer): Move usage of
3944 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
3946 * tree-pretty-print.c (percent_K_format): Add argument.
3947 * tree-pretty-print.h: Add argument.
3948 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
3949 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
3950 (maybe_diag_stxncpy_trunc): Same.
3951 (handle_builtin_stxncpy): Same.
3952 (handle_builtin_strcat): Same.
3954 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3956 * match.pd: Optimise pointer range checks.
3958 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3960 PR tree-optimization/86758
3961 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
3962 to remove pattern statements.
3964 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3966 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
3967 result of dfs_enumerate_from when constructing stmt_vec_infos,
3968 instead of additionally calling get_loop_body.
3970 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3972 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
3974 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
3975 When creating an iv, assert that the step is not known to be zero.
3976 (vect_setup_realignment): Update call accordingly.
3977 * tree-vect-stmts.c (vectorizable_store): Likewise.
3978 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
3980 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3982 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
3983 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
3984 (vectorizable_reduction): Likewise.
3985 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3986 (vect_detect_hybrid_slp_stmts): Likewise.
3987 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
3989 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
3991 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
3992 (wide_int_set_zero_nonzero_bits): ...this.
3993 (zero_nonzero_bits_from_vr): Rename to...
3994 (vrp_set_zero_nonzero_bits): ...this.
3995 (extract_range_from_multiplicative_op_1): Abstract wide int
3997 (wide_int_range_multiplicative_op): ...here.
3998 (extract_range_from_binary_expr_1): Extract wide int binary
3999 operations into their own functions.
4000 (wide_int_range_lshift): New.
4001 (wide_int_range_can_optimize_bit_op): New.
4002 (wide_int_range_shift_undefined_p): New.
4003 (wide_int_range_bit_xor): New.
4004 (wide_int_range_bit_ior): New.
4005 (wide_int_range_bit_and): New.
4006 (wide_int_range_trunc_mod): New.
4007 (extract_range_into_wide_ints): New.
4008 (vrp_shift_undefined_p): New.
4009 (extract_range_from_multiplicative_op): New.
4010 (vrp_can_optimize_bit_op): New.
4011 * tree-vrp.h (value_range::dump): New.
4012 (wide_int_range_multiplicative_op): New.
4013 (wide_int_range_lshift):New.
4014 (wide_int_range_shift_undefined_p): New.
4015 (wide_int_range_bit_xor): New.
4016 (wide_int_range_bit_ior): New.
4017 (wide_int_range_bit_and): New.
4018 (wide_int_range_trunc_mod): New.
4019 (zero_nonzero_bits_from_bounds): Rename to...
4020 (wide_int_set_zero_nonzero_bits): ...this.
4021 (zero_nonzero_bits_from_vr): Rename to...
4022 (vrp_set_zero_nonzero_bits): ...this.
4023 (range_easy_mask_min_max): Rename to...
4024 (wide_int_range_can_optimize_bit_op): this.
4025 * vr-values.c (simplify_bit_ops_using_ranges): Rename
4026 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
4028 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4030 * tree-vectorizer.h (vect_orig_stmt): New function.
4031 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
4032 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
4033 (vect_create_epilog_for_reduction): Likewise.
4034 (vectorizable_live_operation): Likewise.
4035 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4036 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
4037 * tree-vect-stmts.c (vectorizable_call): Likewise.
4038 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
4040 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4042 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
4044 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
4045 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
4046 (vect_transform_loop): Likewise.
4047 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
4049 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4051 * tree-vectorizer.h (vect_schedule_slp): Return void.
4052 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
4053 (vect_schedule_slp): Likewise.
4055 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4057 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
4059 (vect_transform_loop): Update calls accordingly. Schedule SLP
4060 instances before the main loop, if any exist.
4062 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4064 PR tree-optimization/86749
4065 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
4066 If the lhs is used in a COND_EXPR, check that it is being used
4067 as the "then" or "else" value.
4069 2018-08-01 Tom de Vries <tdevries@suse.de>
4072 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
4073 speculation_safe_value_not_needed.
4075 2018-08-01 Richard Biener <rguenther@suse.de>
4077 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
4080 2018-08-01 Martin Liska <mliska@suse.cz>
4082 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
4083 format how successful transformation is dumped.
4084 (gimple_mod_pow2_value_transform): Likewise.
4085 (gimple_mod_subtract_transform): Likewise.
4086 (gimple_stringops_transform): Likewise.
4088 2018-08-01 Martin Liska <mliska@suse.cz>
4091 * value-prof.c (interesting_stringop_to_profile_p):
4092 Simplify the code and add BUILT_IN_MEMMOVE.
4093 (gimple_stringops_transform): Likewise.
4095 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
4097 * config/aarch64/aarch64-simd.md
4098 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
4099 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
4100 use GPI iterator instead of SI mode.
4102 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
4104 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
4105 rs6000_speculation_barrier.
4106 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
4107 new barrier pattern name.
4109 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
4111 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
4112 (speculation_barrier): New insn.
4114 2018-08-01 Richard Biener <rguenther@suse.de>
4117 * graphite.h: Include isl/id.h and isl/space.h to allow build
4120 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
4123 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4124 mode for COFF targets.
4125 * defaults.h (TARGET_COFF): Define.
4126 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
4127 TARGET_COFF): Define.
4128 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
4129 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
4130 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
4132 2018-07-31 Alexandre Oliva <oliva@adacore.com>
4134 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
4135 * gimple-streamer-out.c (output_bb): Save it.
4136 * lto-streamer-in.c (input_struct_function_base): Restore
4137 instance discriminator if available. Create map on demand.
4138 * lto-streamer-out.c (output_struct_function_base): Save it if
4140 * final.c (decl_to_instance_map): Document LTO strategy.
4142 2018-07-31 Alexandre Oliva <oliva@adacore.com>
4143 Olivier Hainque <hainque@adacore.com>
4145 * debug.h (decl_to_instance_map_t): New type.
4146 (decl_to_instance_map): Declare.
4147 (maybe_create_decl_to_instance_map): New inline function.
4148 * final.c (bb_discriminator, last_bb_discriminator): New statics,
4149 to track basic block discriminators.
4150 (final_start_function_1): Initialize them.
4151 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
4153 (decl_to_instance_map): New variable.
4154 (map_decl_to_instance, maybe_set_discriminator): New functions.
4155 (notice_source_line): Set discriminator.
4157 2018-07-31 Ian Lance Taylor <iant@golang.org>
4159 * targhooks.c (default_have_speculation_safe_value): Add
4162 2018-07-31 David Malcolm <dmalcolm@redhat.com>
4164 * dump-context.h: Include "pretty-print.h".
4165 (dump_context::refresh_dumps_are_enabled): New decl.
4166 (dump_context::emit_item): New decl.
4167 (class dump_context): Add fields "m_test_pp" and
4169 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
4170 (temp_dump_context::get_dumped_text): New decl.
4171 (class temp_dump_context): Add field "m_pp".
4172 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
4173 (dump_context::refresh_dumps_are_enabled): ...and add a test for
4175 (set_dump_file): Update for above change.
4176 (set_alt_dump_file): Likewise.
4177 (dump_loc): New overload, taking a pretty_printer *.
4178 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
4179 to any test pretty-printer.
4180 (make_item_for_dump_gimple_stmt): New function, adapted from
4181 optinfo::add_gimple_stmt.
4182 (dump_context::dump_gimple_stmt): Call it, and use the result,
4183 eliminating the direct usage of dump_file and alt_dump_file in
4184 favor of indirectly using them via emit_item.
4185 (make_item_for_dump_gimple_expr): New function, adapted from
4186 optinfo::add_gimple_expr.
4187 (dump_context::dump_gimple_expr): Call it, and use the result,
4188 eliminating the direct usage of dump_file and alt_dump_file in
4189 favor of indirectly using them via emit_item.
4190 (make_item_for_dump_generic_expr): New function, adapted from
4192 (dump_context::dump_generic_expr): Call it, and use the result,
4193 eliminating the direct usage of dump_file and alt_dump_file in
4194 favor of indirectly using them via emit_item.
4195 (make_item_for_dump_printf_va): New function, adapted from
4196 optinfo::add_printf_va.
4197 (make_item_for_dump_printf): New function.
4198 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
4199 and use the result, eliminating the direct usage of dump_file and
4200 alt_dump_file in favor of indirectly using them via emit_item.
4201 (make_item_for_dump_dec): New function.
4202 (dump_context::dump_dec): Call it, and use the result,
4203 eliminating the direct usage of dump_file and alt_dump_file in
4204 favor of indirectly using them via emit_item.
4205 (make_item_for_dump_symtab_node): New function, adapted from
4206 optinfo::add_symtab_node.
4207 (dump_context::dump_symtab_node): Call it, and use the result,
4208 eliminating the direct usage of dump_file and alt_dump_file in
4209 favor of indirectly using them via emit_item.
4210 (dump_context::begin_scope): Reimplement, avoiding direct usage
4211 of dump_file and alt_dump_file in favor of indirectly using them
4213 (dump_context::emit_item): New member function.
4214 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
4215 Set up test pretty-printer on the underlying context. Call
4216 refresh_dumps_are_enabled.
4217 (temp_dump_context::~temp_dump_context): Call
4218 refresh_dumps_are_enabled.
4219 (temp_dump_context::get_dumped_text): New member function.
4220 (selftest::verify_dumped_text): New function.
4221 (ASSERT_DUMPED_TEXT_EQ): New macro.
4222 (selftest::test_capture_of_dump_calls): Run all tests twice, with
4223 and then without optinfo enabled. Add uses of
4224 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
4225 * dumpfile.h: Update comment for the dump_* API.
4226 * optinfo-emit-json.cc
4227 (selftest::test_building_json_from_dump_calls): Update for new
4228 param for temp_dump_context ctor.
4229 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
4230 and "m_owned" field.
4231 (optinfo_item::~optinfo_item): Likewise.
4232 (optinfo::add_item): New member function.
4233 (optinfo::emit): Update comment.
4234 (optinfo::add_string): Delete.
4235 (optinfo::add_printf): Delete.
4236 (optinfo::add_printf_va): Delete.
4237 (optinfo::add_gimple_stmt): Delete.
4238 (optinfo::add_gimple_expr): Delete.
4239 (optinfo::add_tree): Delete.
4240 (optinfo::add_symtab_node): Delete.
4241 (optinfo::add_dec): Delete.
4242 * optinfo.h (class dump_context): New forward decl.
4243 (optinfo::add_item): New decl.
4244 (optinfo::add_string): Delete.
4245 (optinfo::add_printf): Delete.
4246 (optinfo::add_printf_va): Delete.
4247 (optinfo::add_gimple_stmt): Delete.
4248 (optinfo::add_gimple_expr): Delete.
4249 (optinfo::add_tree): Delete.
4250 (optinfo::add_symtab_node): Delete.
4251 (optinfo::add_dec): Delete.
4252 (optinfo::add_poly_int): Delete.
4253 (optinfo_item::optinfo_item): Remove "owned" param.
4254 (class optinfo_item): Remove field "m_owned".
4256 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4259 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
4260 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
4261 requested variable alignment.
4262 (expand_one_ssa_partition): Likewise.
4263 (expand_one_var): Likewise.
4265 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4267 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
4268 to speculation_safe_value_not_needed.
4270 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4272 * targhooks.h (speculation_safe_value_not_needed): New prototype.
4273 * targhooks.c (speculation_safe_value_not_needed): New function.
4274 * target.def (have_speculation_safe_value): Update documentation.
4275 * doc/tm.texi: Regenerated.
4277 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4279 * config/aarch64/iterators.md (ALLI_TI): New iterator.
4280 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
4282 (despeculate_copy<ALLI:mode>_insn): New insn.
4283 (despeculate_copyti_insn): New insn.
4284 (despeculate_simple<ALLI:mode>): New insn
4285 (despeculate_simpleti): New insn.
4286 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
4288 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
4289 aarch64_speculation_safe_value.
4290 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
4292 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4294 * config/aarch64/aarch64-speculation.cc: New file.
4295 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
4296 before pass_reorder_blocks.
4297 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
4299 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
4300 X14 and X15 when tracking speculation.
4301 * config/aarch64/aarch64.md (register name constants): Add
4302 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
4303 (unspec): Add UNSPEC_SPECULATION_TRACKER.
4304 (speculation_barrier): New insn attribute.
4305 (cmp<mode>): Allow SP in comparisons.
4306 (speculation_tracker): New insn.
4307 (speculation_barrier): Add speculation_barrier attribute.
4308 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
4309 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
4310 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
4312 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4314 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
4315 aarch64_track_speculation is true.
4316 (tb<optab><mode>1): Likewise.
4317 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
4318 generate CB[N]Z when tracking speculation.
4319 (aarch64_split_compare_and_swap): Likewise.
4320 (aarch64_split_atomic_op): Likewise.
4322 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4324 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
4326 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4328 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
4329 (speculation_barrier): New insn.
4331 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4333 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
4334 * config/arm/arm.md (speculation_barrier): New expand.
4335 (speculation_barrier_insn): New pattern.
4337 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4339 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
4340 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
4341 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
4342 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
4344 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
4345 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
4346 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
4347 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
4348 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
4349 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
4350 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
4351 * builtins.c (expand_speculation_safe_value): New function.
4352 (expand_builtin): Call it.
4353 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
4354 * doc/extend.texi: Document __builtin_speculation_safe_value.
4355 * doc/md.texi: Document "speculation_barrier" pattern.
4356 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
4357 TARGET_HAVE_SPECULATION_SAFE_VALUE.
4358 * doc/tm.texi: Regenerated.
4359 * target.def (have_speculation_safe_value, speculation_safe_value): New
4361 * targhooks.c (default_have_speculation_safe_value): New function.
4362 (default_speculation_safe_value): New function.
4363 * targhooks.h (default_have_speculation_safe_value): Add prototype.
4364 (default_speculation_safe_value): Add prototype.
4366 2018-07-31 David Malcolm <dmalcolm@redhat.com>
4368 * dump-context.h (dump_context::dump_loc): New decl.
4369 * dumpfile.c (dump_context::dump_loc): New member function.
4370 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
4371 and dump_gimple_stmt.
4372 (dump_context::dump_gimple_expr_loc): Likewise, using
4374 (dump_context::dump_generic_expr_loc): Likewise, using
4376 (dump_context::dump_printf_loc_va): Likewise, using
4378 (dump_context::begin_scope): Explicitly using the global function
4379 "dump_loc", rather than the member function.
4381 2018-07-31 Martin Sebor <msebor@redhat.com>
4383 PR tree-optimization/86741
4384 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
4386 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
4388 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
4389 depend on whether prefetch instructions will be emitted or not.
4390 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
4391 will be emitted or not.
4392 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
4393 (TARGET_SETMEM_PFD): New macros.
4395 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4397 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
4398 (NULL_STMT_VEC_INFO): Delete.
4399 (stmt_vec_info::operator*): Likewise.
4400 (stmt_vec_info::operator gimple *): Likewise.
4401 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
4402 of NULL_STMT_VEC_INFO.
4403 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4404 (vect_reassociating_reduction_p): Likewise.
4405 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
4406 (vectorizable_store): Likewise.
4407 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
4408 (vec_info::free_stmt_vec_infos): Likewise.
4410 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4412 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
4413 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
4415 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4417 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
4418 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4419 (vec_info::free_stmt_vec_info): New private member functions.
4420 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
4421 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
4422 * tree-parloops.c (gather_scalar_reductions): Remove calls to
4423 set_stmt_vec_info_vec and free_stmt_vec_infos.
4424 * tree-vect-loop.c (_loop_vec_info): Remove call to
4425 set_stmt_vec_info_vec.
4426 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
4427 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
4428 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
4429 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4430 (vec_info::free_stmt_vec_info): ...these new functions. Remove
4431 assignments in {vec_info::,}new_stmt_vec_info that are redundant
4432 with the clearing in the xcalloc.
4433 (stmt_vec_info_vec): Delete.
4434 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
4435 (vectorize_loops): Likewise.
4436 (vec_info::~vec_info): Remove argument from call to
4437 free_stmt_vec_infos.
4438 (vec_info::add_stmt): Remove vinfo argument from call to
4441 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4443 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
4444 rather than a gimple stmt.
4445 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
4446 information for pattern statements when passed the original
4447 statement; instead wait to be passed the pattern statement itself.
4448 Don't call set_vinfo_for_stmt here.
4449 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
4450 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
4451 stmt_vec_infos here.
4452 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
4453 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
4454 stmt_vec_infos entry.
4456 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4458 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
4459 * tree-vectorizer.c (vec_info::replace_stmt): New function.
4460 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
4461 * tree-vect-stmts.c (vectorizable_call): Likewise.
4462 (vectorizable_simd_clone_call): Likewise.
4464 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4466 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
4467 * tree-vectorizer.c (vec_info::remove_stmt): New function.
4468 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
4469 * tree-vect-loop.c (vect_transform_loop): Likewise.
4470 * tree-vect-slp.c (vect_schedule_slp): Likewise.
4471 * tree-vect-stmts.c (vect_remove_stores): Likewise.
4473 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4475 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
4476 (vect_dr_stmt): Delete.
4477 * tree-vectorizer.c (vec_info::lookup_dr): New function.
4478 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
4480 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
4481 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
4482 (vect_verify_datarefs_alignment, vect_peeling_supportable)
4483 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4484 (vect_analyze_data_refs): Likewise.
4485 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
4487 (vect_find_same_alignment_drs): Likewise.
4488 (vect_slp_analyze_node_dependences): Update calls accordingly.
4489 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
4490 instead of DR_VECT_AUX.
4491 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
4492 of a vector data references. Use vec_info::lookup_dr instead of
4494 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
4495 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
4496 instead of DR_VECT_AUX.
4498 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4500 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
4502 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4504 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4505 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4506 (vect_gen_prolog_loop_niters): Likewise.
4508 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4510 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
4511 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
4512 (vect_known_alignment_in_bytes, vect_dr_behavior)
4513 (vect_get_scalar_dr_size): Take references as dr_vec_infos
4514 instead of data_references. Update calls to other routines for
4515 which the same change has been made.
4516 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
4517 dr_vec_infos instead of stmt_vec_infos.
4518 (vect_analyze_data_ref_dependence): Update call accordingly.
4519 (vect_slp_analyze_data_ref_dependence)
4520 (vect_record_base_alignments): Use DR_VECT_AUX.
4521 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
4522 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
4523 (vector_alignment_reachable_p, vect_get_data_access_cost)
4524 (vect_peeling_supportable, vect_analyze_group_access_1)
4525 (vect_analyze_group_access, vect_analyze_data_ref_access)
4526 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
4527 (vect_compile_time_alias, vect_small_gap_p)
4528 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
4529 (vect_supportable_dr_alignment): Take references as dr_vec_infos
4530 instead of data_references. Update calls to other routines for
4531 which the same change has been made.
4532 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
4533 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4534 (vect_slp_analyze_and_verify_node_alignment)
4535 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4536 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4537 (vect_setup_realignment): Use dr_vec_infos. Update calls after
4539 (_vect_peel_info::dr): Replace with...
4540 (_vect_peel_info::dr_info): ...this new field.
4541 (vect_peeling_hash_get_most_frequent)
4542 (vect_peeling_hash_choose_best_peeling): Update accordingly.
4543 (vect_peeling_hash_get_lowest_cost):
4544 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
4545 routines for which the same change has been made.
4546 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
4548 * tree-vect-loop-manip.c (get_misalign_in_elems)
4549 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
4551 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4552 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
4553 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
4554 (get_group_load_store_type, get_negative_load_store_type)
4555 (vect_get_data_ptr_increment, vectorizable_store)
4556 (vectorizable_load): Likewise.
4557 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
4558 Update calls to other routines for which the same change has been made.
4560 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4562 * tree-vectorizer.h (vec_info::move_dr): New member function.
4563 (dataref_aux): Rename to...
4564 (dr_vec_info): ...this and add "dr" and "stmt" fields.
4565 (_stmt_vec_info::dr_aux): Update accordingly.
4566 (_stmt_vec_info::data_ref_info): Delete.
4567 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
4568 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
4569 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
4570 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
4572 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
4573 (STMT_VINFO_DR_INFO): New macro.
4574 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
4575 (set_dr_misalignment): Update after rename of dataref_aux.
4576 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
4577 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
4578 initialization of STMT_VINFO_DATA_REF.
4579 * tree-vectorizer.c (vec_info::move_dr): New function.
4580 * tree-vect-patterns.c (vect_recog_bool_pattern)
4581 (vect_recog_mask_conversion_pattern)
4582 (vect_recog_gather_scatter_pattern): Use it.
4583 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
4584 the "dr" and "stmt" fields of dr_vec_info instead of
4585 STMT_VINFO_DATA_REF.
4587 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4589 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
4590 (is_pattern_stmt_p): Use it.
4591 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
4592 on pattern statements.
4594 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4596 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
4597 original stmt as a stmt_vec_info rather than a gimple stmt.
4598 (vect_pattern_recog_1): Take the statement directly as a
4599 stmt_vec_info, rather than via a gimple_stmt_iterator.
4600 Update call to vect_mark_pattern_stmts.
4601 (vect_pattern_recog): Update calls accordingly.
4603 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4605 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
4606 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
4607 a vect_def_type for the first argument.
4608 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
4609 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
4610 operand if it isn't defined by a vectorized statement.
4611 (vect_build_gather_load_calls): Remove the mask_dt argument and
4612 update calls to vect_get_vec_def_for_stmt_copy.
4613 (vectorizable_bswap): Likewise the dt argument.
4614 (vectorizable_call): Update calls to vectorizable_bswap and
4615 vect_get_vec_def_for_stmt_copy.
4616 (vectorizable_simd_clone_call, vectorizable_assignment)
4617 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
4618 (vectorizable_comparison): Update calls to
4619 vect_get_vec_def_for_stmt_copy.
4620 (vectorizable_store): Likewise. Remove now-unnecessary calls to
4622 (vect_get_loop_based_defs): Remove dt argument and update call
4623 to vect_get_vec_def_for_stmt_copy.
4624 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
4625 and vect_get_vec_def_for_stmt_copy.
4626 (vectorizable_load): Update calls to vect_build_gather_load_calls
4627 and vect_get_vec_def_for_stmt_copy.
4628 * tree-vect-loop.c (vect_create_epilog_for_reduction)
4629 (vectorizable_reduction, vectorizable_live_operation): Update calls
4630 to vect_get_vec_def_for_stmt_copy.
4632 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4634 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
4635 and gimple stmt arguments with a stmt_vec_info.
4636 (vect_record_base_alignments): Update calls accordingly.
4637 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
4638 and gimple stmt arguments with a stmt_vec_info.
4639 (vect_build_slp_tree_1): Remove vinfo argument and update call
4640 to vect_record_max_nunits.
4641 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
4642 and vect_record_max_nunits.
4644 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4646 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
4647 file and take a stmt_vec_info instead of a gimple stmt.
4648 (supportable_widening_operation, vect_finish_replace_stmt)
4649 (vect_finish_stmt_generation, vect_get_store_rhs)
4650 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
4651 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
4652 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
4653 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
4654 (vect_create_data_ref_ptr, bump_vector_ptr)
4655 (vect_permute_store_chain, vect_setup_realignment)
4656 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
4657 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
4658 (vectorizable_reduction, vectorizable_induction)
4659 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
4660 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
4661 than gimple stmts as arguments.
4662 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
4663 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
4664 (can_group_stmts_p, vect_check_gather_scatter)
4665 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4666 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
4667 (vect_permute_load_chain, vect_shift_permute_load_chain)
4668 (vect_transform_grouped_load)
4669 (vect_record_grouped_load_vectors): Likewise.
4670 * tree-vect-loop.c (vect_fixup_reduc_chain)
4671 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
4672 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
4673 (vectorizable_reduction, vectorizable_induction)
4674 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
4675 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
4676 (vect_get_load_store_mask): Likewise.
4677 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
4678 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
4679 * tree-vect-stmts.c (vect_mark_relevant)
4680 (is_simple_and_all_uses_invariant)
4681 (exist_non_indexing_operands_for_use_p, process_use)
4682 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
4683 (vect_get_vec_def_for_operand, vect_get_vec_defs)
4684 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
4685 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
4686 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
4687 (get_negative_load_store_type, get_load_store_type)
4688 (vect_check_load_store_mask, vect_check_store_rhs)
4689 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
4690 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4691 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4692 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4693 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
4694 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
4695 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
4696 (supportable_widening_operation): Likewise.
4698 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4700 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
4701 a stmt_vec_info instead of a gcall.
4702 (vect_check_gather_scatter): Update call accordingly.
4703 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
4705 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
4706 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
4707 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
4708 instead of a gimple stmt.
4709 (vect_transform_loop): Update calls accordingly.
4710 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
4711 stmt_vec_infos instead of gimple stmts.
4712 (vect_analyze_slp_instance): Update use accordingly.
4713 * tree-vect-stmts.c (read_vector_array, write_vector_array)
4714 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
4715 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
4716 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
4717 (vect_gen_widened_results_half, vect_get_loop_based_defs)
4718 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
4719 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
4720 down to subroutines.
4722 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4724 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
4725 of the worklist from a vector of gimple stmts to a vector of
4727 * tree-vect-stmts.c (vect_mark_relevant, process_use)
4728 (vect_mark_stmts_to_be_vectorized): Likewise
4730 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4732 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
4733 statement before passing it to vect_analyze_stmt.
4734 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
4735 the chain of phi vector definitions. Track the exit phi via its
4737 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
4738 STMT_VINFO_REDUC_DEF.
4739 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
4740 stmt_vec_infos to handle the statement chains.
4741 (vect_get_slp_defs): Record the first statement in the node
4742 using a stmt_vec_info.
4743 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
4744 statements here and pass their stmt_vec_info down to subroutines.
4745 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
4746 down to vect_finish_stmt_generation.
4747 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
4748 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
4749 stmt_vec_infos to subroutines.
4750 (vect_remove_stores): Use stmt_vec_infos to handle the statement
4753 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4755 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
4756 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
4757 (vect_permute_store_chain, vect_setup_realignment)
4758 (vect_permute_load_chain, vect_shift_permute_load_chain)
4759 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
4760 stmts internally, and when passing values to other vectorizer routines.
4761 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4762 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
4763 (vect_analyze_loop_operations, get_initial_def_for_reduction)
4764 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
4765 (vectorizable_reduction, vectorizable_induction)
4766 (vectorizable_live_operation, vect_transform_loop_stmt)
4767 (vect_transform_loop): Likewise.
4768 * tree-vect-patterns.c (vect_reassociating_reduction_p)
4769 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
4770 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
4771 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4772 (vect_slp_analyze_node_operations_1): Likewise.
4773 * tree-vect-stmts.c (vect_mark_relevant, process_use)
4774 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
4775 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
4776 (vect_finish_stmt_generation_1, get_group_load_store_type)
4777 (get_load_store_type, vect_build_gather_load_calls)
4778 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4779 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4780 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4781 (vectorizable_store, vectorizable_load, vectorizable_condition)
4782 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
4783 (supportable_widening_operation): Likewise.
4784 (vect_get_vector_types_for_stmt): Likewise.
4785 * tree-vectorizer.h (vect_dr_behavior): Likewise.
4787 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4789 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4790 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
4791 (vect_permute_store_chain, vect_permute_load_chain)
4792 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
4793 repeated stmt_vec_info lookups.
4794 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4795 (vect_update_ivs_after_vectorizer): Likewise.
4796 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
4797 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4798 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
4799 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4800 (vect_bb_slp_scalar_cost): Likewise.
4801 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
4803 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4805 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
4806 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
4807 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
4808 to get gassigns and gcalls, rather than operating on generc gimple
4810 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
4811 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
4812 (vectorizable_load, vect_analyze_stmt): Likewise.
4813 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
4815 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4817 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
4818 return stmt_vec_infos rather than gimple stmts. Do not accept
4820 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
4822 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4823 Update use of get_later_stmt.
4824 (vect_get_constant_vectors): Update call accordingly.
4825 (vect_schedule_slp_instance): Likewise
4826 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
4827 (vect_slp_analyze_instance_dependence): Likewise.
4828 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
4830 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4832 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
4833 (stmt_info_for_cost::stmt_info): ...this new field.
4834 (add_stmt_costs): Update accordingly.
4835 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
4836 (vect_get_known_peeling_cost): Likewise.
4837 (vect_estimate_min_profitable_iters): Likewise.
4838 * tree-vect-stmts.c (record_stmt_cost): Likewise.
4840 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4842 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
4843 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4844 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4846 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
4848 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4850 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
4851 a gimple stmt to a stmt_vec_info.
4852 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
4854 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4856 * tree-vectorizer.h (vec_info::grouped_stores): Change from
4857 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4858 (_loop_vec_info::reduction_chains): Likewise.
4859 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
4861 * tree-vect-slp.c (vect_analyze_slp): Likewise.
4863 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4865 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
4866 a gimple stmt to a stmt_vec_info.
4867 (_stmt_vec_info::next_element): Likewise.
4868 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
4869 (vect_slp_analyze_and_verify_node_alignment)
4870 (vect_analyze_group_access_1, vect_analyze_group_access)
4871 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
4872 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
4873 (vect_supportable_dr_alignment): Update accordingly.
4874 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
4875 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
4876 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
4877 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4878 * tree-vect-slp.c (vect_build_slp_tree_1)
4879 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
4880 (vect_split_slp_store_group, vect_analyze_slp_instance)
4881 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
4882 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
4883 (get_group_load_store_type, get_load_store_type)
4884 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
4885 (vect_transform_stmt, vect_remove_stores): Likewise.
4887 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4889 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
4891 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4892 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
4893 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
4894 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
4895 (vector_alignment_reachable_p, vect_get_data_access_cost)
4896 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
4897 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
4898 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4899 (vect_analyze_group_access_1, vect_analyze_group_access)
4900 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
4901 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
4902 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
4903 result of vect_dr_stmt and use the stmt_vec_info instead of
4904 the associated gimple stmt.
4905 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4906 (vect_gen_prolog_loop_niters): Likewise.
4907 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4909 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4911 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
4912 to a vec<stmt_vec_info>.
4913 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
4914 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
4916 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
4917 to a vec<stmt_vec_info>.
4918 (bst_traits::value_type, bst_traits::value_type): Likewise.
4919 (bst_traits::hash): Update accordingly.
4920 (vect_get_and_check_slp_defs): Change the stmts parameter from
4921 a vec<gimple *> to a vec<stmt_vec_info>.
4922 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
4923 (vect_build_slp_tree): Likewise.
4924 (vect_build_slp_tree_2): Likewise. Update uses of
4925 SLP_TREE_SCALAR_STMTS.
4926 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
4927 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
4928 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
4929 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
4930 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
4931 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
4932 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
4933 (vect_get_constant_vectors, vect_get_slp_defs)
4934 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
4935 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
4936 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
4937 instead of gimple stmts.
4938 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
4939 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
4940 (vect_slp_analyze_instance_dependence): Update uses of
4941 SLP_TREE_SCALAR_STMTS.
4942 (vect_slp_analyze_and_verify_node_alignment): Likewise.
4943 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
4944 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
4945 (get_initial_defs_for_reduction): Likewise.
4946 (vect_create_epilog_for_reduction): Likewise.
4947 (vectorize_fold_left_reduction): Likewise.
4948 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
4949 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
4950 (can_vectorize_live_stmts): Likewise.
4952 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4954 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
4955 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4956 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
4958 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
4960 (gather_scalar_reductions): Update after above interface changes.
4961 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
4962 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
4964 (vect_force_simple_reduction): Likewise.
4965 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
4966 LOOP_VINFO_REDUCTIONS.
4967 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4969 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4971 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
4972 a gimple stmt to a stmt_vec_info.
4973 * tree-vect-loop.c (vect_active_double_reduction_p)
4974 (vect_force_simple_reduction, vectorizable_reduction): Update
4977 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4979 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
4980 vec<gimple *> to a vec<stmt_vec_info>.
4981 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
4982 the reduction_phis argument from a vec<gimple *> to a
4984 (vectorizable_reduction): Likewise the phis local variable that
4985 is passed to vect_create_epilog_for_reduction. Update for new type
4986 of SLP_TREE_VEC_STMTS.
4987 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
4988 (vectorizable_live_operation): Likewise.
4989 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
4990 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
4992 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4994 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
4995 a gimple stmt to a stmt_vec_info.
4996 (vectorizable_condition, vectorizable_live_operation)
4997 (vectorizable_reduction, vectorizable_induction): Pass back the
4998 vectorized statement as a stmt_vec_info.
4999 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
5000 use of STMT_VINFO_VEC_STMT.
5001 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
5002 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
5003 as stmt_vec_infos rather than gimple stmts.
5004 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
5006 (vectorizable_live_operation): Likewise.
5007 (vectorizable_reduction, vectorizable_induction): Likewise,
5008 updating use of STMT_VINFO_VEC_STMT.
5009 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
5010 of STMT_VINFO_VEC_STMT.
5011 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
5012 (vectorizable_simd_clone_call, vectorizable_conversion)
5013 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5014 (vectorizable_store, vectorizable_load, vectorizable_condition)
5015 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
5016 from a gimple stmt to a stmt_vec_info.
5017 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
5018 pointer to a stmt_vec_info to the vectorizable_* routines.
5020 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5022 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
5023 a gimple stmt to a stmt_vec_info.
5024 (is_pattern_stmt_p): Update accordingly.
5025 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
5026 (vect_record_grouped_load_vectors): Likewise.
5027 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
5028 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
5029 (vect_model_reduction_cost): Likewise.
5030 (vect_create_epilog_for_reduction): Likewise.
5031 (vectorizable_reduction, vectorizable_induction): Likewise.
5032 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5033 Return the stmt_vec_info for the pattern statement.
5034 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
5035 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
5036 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
5037 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
5038 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
5039 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
5040 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
5041 (free_stmt_vec_info, vect_is_simple_use): Likewise.
5043 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5045 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
5046 (vect_finish_stmt_generation): Likewise.
5047 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
5048 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
5049 (vect_build_gather_load_calls): Use the return value of the above
5050 functions instead of a separate call to vinfo_for_stmt. Use narrow
5051 scopes for the input gimple stmt and wider scopes for the associated
5052 stmt_vec_info. Use vec_info::lookup_def when setting these
5053 stmt_vec_infos from an SSA_NAME definition.
5054 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
5055 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
5056 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5057 (vectorizable_store, vectorizable_load, vectorizable_condition)
5058 (vectorizable_comparison): Likewise.
5059 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
5060 (vectorizable_reduction): Likewise.
5062 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5064 * tree-vectorizer.h (vect_is_simple_use): Add an optional
5065 stmt_vec_info * parameter before the optional gimple **.
5066 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5067 (process_use, vect_get_vec_def_for_operand_1): Update callers.
5068 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
5069 * tree-vect-loop.c (vectorizable_reduction): Likewise.
5070 (vectorizable_live_operation): Likewise.
5071 * tree-vect-patterns.c (type_conversion_p): Likewise.
5072 (vect_look_through_possible_promotion): Likewise.
5073 (vect_recog_rotate_pattern): Likewise.
5074 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
5076 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5078 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
5079 a typedef to a wrapper class.
5080 (NULL_STMT_VEC_INFO): New macro.
5081 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
5082 (stmt_vec_info::operator*): New function.
5083 (stmt_vec_info::operator gimple *): Likewise.
5084 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
5085 (add_stmt_costs): Likewise.
5086 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
5087 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
5088 (vect_get_known_peeling_cost): Likewise.
5089 (vect_estimate_min_profitable_iters): Likewise.
5090 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5091 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
5092 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
5093 (vectorizable_store, free_stmt_vec_infos): Likewise.
5094 (new_stmt_vec_info): Change return type of xcalloc to
5097 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5099 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
5100 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
5101 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
5102 a single_imm_use-based sequence.
5103 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
5105 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5107 * tree-vectorizer.h (vec_info::lookup_def): Declare.
5108 * tree-vectorizer.c (vec_info::lookup_def): New function.
5109 * tree-vect-patterns.c (vect_get_internal_def): Use it.
5110 (vect_widened_op_tree): Likewise.
5111 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5112 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
5113 (vectorizable_reduction): Likewise.
5114 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
5116 (vect_is_slp_reduction): Update calls accordingly. Use
5117 vec_info::lookup_def.
5118 (vect_is_simple_reduction): Likewise
5119 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
5121 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5123 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
5124 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
5125 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
5127 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
5128 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
5129 (vect_update_vf_for_slp, vect_analyze_loop_operations)
5130 (vect_is_slp_reduction, vectorizable_induction)
5131 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
5132 * tree-vect-patterns.c (vect_init_pattern_stmt):
5133 (vect_determine_min_output_precision_1, vect_determine_precisions)
5134 (vect_pattern_recog): Likewise.
5135 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
5136 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
5137 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
5138 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
5139 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
5140 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
5141 info field from a loop to a loop_vec_info.
5143 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5145 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
5146 (vec_info::add_stmt): Declare.
5147 * tree-vectorizer.c (vec_info::add_stmt): New function.
5148 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
5149 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
5150 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
5151 (vectorizable_induction): Likewise.
5152 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
5153 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
5154 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
5155 (vectorizable_load): Likewise.
5156 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5157 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
5158 (vect_recog_gather_scatter_pattern): Likewise.
5159 (append_pattern_def_seq): Likewise. Remove a check that is
5160 performed by add_stmt itself.
5162 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5164 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
5165 which make_ssa_name was called with new_stmt before new_stmt
5168 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5170 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
5172 (vect_is_slp_reduction): ...here...
5173 (vect_is_simple_reduction): ...and here. Remove repetition of tests
5174 that are already known to be false.
5176 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5178 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
5179 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
5180 STMT_VINFO_NUM_SLP_USES when it's true.
5181 (vect_free_slp_instance): Add a final_p parameter and pass it to
5183 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
5184 (vect_analyze_slp_instance): Likewise.
5185 (vect_slp_analyze_operations): Likewise.
5186 (vect_slp_analyze_bb_1): Likewise.
5187 * tree-vectorizer.c (vec_info): Likewise.
5188 * tree-vect-loop.c (vect_transform_loop): Likewise.
5190 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5192 * tree-vect-loop.c (vectorizable_reduction): Assert that the
5193 function is not called for second and subsequent members of
5196 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5198 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
5199 cases for nested loops from here to ...
5200 (vect_create_epilog_for_reduction): ...here. Only call
5201 vect_is_simple_use for inner-loop reductions.
5203 2018-07-31 Martin Liska <mliska@suse.cz>
5205 PR gcov-profile/85338
5206 PR gcov-profile/85350
5207 PR gcov-profile/85372
5208 * profile.c (struct location_triplet): New.
5209 (struct location_triplet_hash): Likewise.
5210 (output_location): Do not output a BB that
5211 is already recorded for a line.
5212 (branch_prob): Use streamed_locations.
5214 2018-07-31 Martin Liska <mliska@suse.cz>
5216 PR gcov-profile/85370
5217 * coverage.c (coverage_begin_function): Do not mark target
5218 clones as artificial functions.
5220 2018-07-31 Martin Liska <mliska@suse.cz>
5222 PR gcov-profile/83813
5223 PR gcov-profile/84758
5224 PR gcov-profile/85217
5225 PR gcov-profile/85332
5226 * profile.c (branch_prob): Do not record GOTO expressions
5227 for GIMPLE statements which locations are already streamed.
5229 2018-07-31 Olivier Hainque <hainque@adacore.com>
5231 * gcc.c (handle_spec_function): Accept a soft_matched_part
5232 argument, as do_spec_1. Pass it down to ...
5233 (eval_spec_function): Accept a soft_matched_part argument,
5234 and pass it down to ...
5235 (do_spec_2): Accept a soft_matched_part argument, and pass
5236 it down to do_spec_1.
5237 (do_spec_1): Pass soft_matched_part to handle_spec_function.
5238 (handle_braces): Update call to handle_spec_function.
5239 (driver::set_up_specs): Update calls to do_spec_2.
5240 (compare_debug_dump_opt_spec_function): Likewise.
5241 (compare_debug_self_opt_spec_function): Likewise.
5243 2018-07-31 Olivier Hainque <hainque@adacore.com>
5245 * common.opt (nolibc): New option.
5246 * doc/invoke.texi (Link Options): Document it.
5247 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
5248 * config/alpha/linux.h: Likewise.
5249 * config/arc/elf.h: Likewise.
5250 * config/arm/uclinux-elf.h: Likewise.
5251 * config/arm/unknown-elf.h: Likewise.
5252 * config/avr/avrlibc.h: Likewise.
5253 * config/bfin/bfin.h: Likewise.
5254 * config/bfin/linux.h: Likewise.
5255 * config/bfin/uclinux.h: Likewise.
5256 * config/darwin.h: Likewise.
5257 * config/darwin10.h: Likewise.
5258 * config/darwin12.h: Likewise.
5259 * config/gnu-user.h: Likewise.
5260 * config/lm32/uclinux-elf.h: Likewise.
5261 * config/pa/pa-hpux11.h: Likewise.
5262 * config/pa/pa64-hpux.h: Likewise.
5263 * config/sparc/sparc.h: Likewise.
5265 2018-07-31 Olivier Hainque <hainque@adacore.com>
5267 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
5268 undefined variables.
5270 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
5273 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
5276 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
5278 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
5279 terminated string literal.
5281 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
5283 PR rtl-optimization/85160
5284 * combine.c (is_just_move): New function.
5285 (try_combine): Allow combining two instructions into two if neither of
5286 the original instructions was a move.
5288 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
5291 * doc/extend.texi (Global Register Variables): Discourage use of type
5293 (Local Register Variables): Likewise.
5295 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
5297 PR tree-optimization/86506
5298 * hwint.h (ceil_log2): Resync with hwint.c implementation.
5300 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
5303 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
5304 hard_regno, make sure no insn between `from` and `to` clobbers it.
5306 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
5307 Tom de Vries <tdevries@suse.de>
5309 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
5310 (PTX_DEFAULT_RUNTIME_DIM): ... this.
5311 (nvptx_goacc_validate_dims): Set default worker and gang dims to
5312 PTX_DEFAULT_RUNTIME_DIM.
5313 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
5315 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
5317 * config/pa/pa.c (pa_output_addr_vec): Align address table.
5318 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
5319 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
5321 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
5323 * config/rs6000/constraints.md (wG constraint): Delete, no longer
5325 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
5326 predicate to reflect toc fusion has been deleted.
5327 (toc_fusion_mem_raw): Delete, no longer used.
5328 (toc_fusion_mem_wrapped): Likewise.
5329 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
5331 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
5332 Delete, no longer used.
5333 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
5334 meant to be used for toc fusion.
5335 (rs6000_debug_print_mode): Delete toc fusion debugging.
5336 (rs6000_debug_reg_global): Likewise.
5337 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
5338 fusion and secondary reload support that were never used.
5339 (rs6000_option_override_internal): Delete TOC fusion, that was only
5340 partially defined, and it did not work unless you also used the
5342 (rs6000_legitimate_address_p): Delete TOC fusion support.
5343 (rs6000_opt_masks): Likewise.
5344 (fusion_wrap_memory_address): Delete function, no longer used.
5345 (fusion_split_address); Delete TOC fusion support.
5346 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
5347 longer used with toc fusion being deleted.
5348 (TARGET_TOC_FUSION_FP): Likewise.
5349 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
5351 (toc fusion spliter): Delete TOC fusion support.
5352 (toc_fusionload_<mode>): Likewise.
5353 (toc_fusionload_di): Likewise.
5354 (fusion_gpr_load_<mode>): Delete generator function, this insn no
5355 longer needs to be named. Rename predicate to delete TOC fusion.
5356 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5357 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5358 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5359 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5360 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
5362 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
5364 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
5365 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
5366 __int128 in built-in function prototypes.
5367 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
5368 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
5370 2018-07-27 Martin Sebor <msebor@redhat.com>
5372 PR tree-optimization/86696
5373 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
5374 types, including enums.
5375 (handle_char_store): Be prepared for the above function to fail.
5377 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
5379 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
5380 when optimization level is lower than 2 or optimize for size.
5382 2018-07-26 Martin Sebor <msebor@redhat.com>
5384 PR tree-optimization/86043
5385 PR tree-optimization/86042
5386 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
5387 (get_string_cst_length): Rename...
5388 (get_min_string_length): ...to this. Add argument.
5389 (handle_char_store): Extend to handle multi-character stores by
5391 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
5392 * tree.h (initializer_zerop): Add argument.
5394 2018-07-26 Jakub Jelinek <jakub@redhat.com>
5397 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
5398 declare target to variables if they have always,{to,from,tofrom} map
5401 2018-07-26 Martin Liska <mliska@suse.cz>
5404 * lto-wrapper.c: Add linker_output as prefix
5405 for ltrans_output_file.
5407 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
5409 PR rtl-optimization/85805
5410 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
5411 value for hard registers if that was written in the same mode.
5413 2018-07-26 Martin Liska <mliska@suse.cz>
5415 PR gcov-profile/86536
5416 * gcov.c (format_gcov): Use printf format %.*f directly
5417 and do not handle special values.
5419 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5421 * common/config/arc/arc-common.c (arc_option_optimization_table):
5422 Update default optimizations for size.
5424 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5426 * config/arc/arc.md (movsf_insn): Add short instruction selection.
5427 * config/arc/constraints.md (CfZ): New constraint.
5428 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
5429 (subsf3_fpu): Likewise.
5430 (cmpsf_fpu): Likewise.
5431 (cmpsf_fpu_uneq): Likewise.
5433 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5435 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
5437 (arc_is_uncached_mem_p): uncached applies to both the variable and
5440 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5442 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
5445 2018-07-25 David Malcolm <dmalcolm@redhat.com>
5447 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
5448 field "m_scopes" from vec to auto_vec.
5450 2018-07-25 Martin Liska <mliska@suse.cz>
5452 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
5455 2018-07-25 Richard Biener <rguenther@suse.de>
5458 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
5459 special wrt context_die late.
5460 (gen_subprogram_die): Re-use DIEs in local scope.
5462 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
5464 PR tree-optimization/86644
5465 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
5467 2018-07-25 Martin Liska <mliska@suse.cz>
5470 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
5471 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
5473 2018-07-25 Martin Liska <mliska@suse.cz>
5476 * params.def: Explain ASan abbreviation and provide
5477 a documentation link.
5479 2018-07-24 Martin Sebor <msebor@redhat.com>
5481 PR tree-optimization/86622
5482 PR tree-optimization/86532
5483 * builtins.h (string_length): Declare.
5484 * builtins.c (c_strlen): Correct handling of non-constant offsets.
5485 (check_access): Be prepared for non-constant length ranges.
5486 (string_length): Make extern.
5487 * expr.c (string_constant): Only handle the minor non-constant
5488 array index. Use string_constant to compute the length of
5489 a generic string constant.
5491 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
5493 PR tree-optimization/86618
5494 * tree-vect-stmts.c (vectorizable_call): Don't take the address
5495 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
5497 2018-07-24 David Malcolm <dmalcolm@redhat.com>
5499 PR tree-optimization/86636
5500 * json.cc (json::object::set): Fix comment. Add assertions.
5501 (json::array::append): Move here from json.h. Add comment and an
5503 (json::string::string): Likewise.
5504 * json.h (json::array::append): Move to json.cc.
5505 (json::string::string): Likewise.
5506 * optinfo-emit-json.cc
5507 (optrecord_json_writer::impl_location_to_json): Assert that we
5508 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
5509 wrapper around it. Expand the location once, rather than three
5511 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
5512 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
5514 (optrecord_json_writer::optinfo_to_json): Likewise, in four
5515 places. Fix some overlong lines.
5517 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
5519 * config/aarch64/aarch64-simd.md
5520 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
5521 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
5522 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
5523 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
5524 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
5525 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
5526 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
5527 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
5528 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
5530 2018-07-24 Jakub Jelinek <jakub@redhat.com>
5533 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
5534 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
5535 and abs_d == d, do the power of two handling if profitable.
5537 2018-07-24 Richard Biener <rguenther@suse.de>
5539 * match.pd: Add BIT_FIELD_REF canonicalizations.
5541 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5544 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
5546 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5548 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
5549 terminated STRING_CST object.
5551 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5553 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
5555 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
5557 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
5558 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
5559 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
5560 the elements into a register.
5561 (rs6000_split_v4si_init_di_reg): Delete.
5562 (rs6000_split_v4si_init): Delete.
5563 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
5564 (vsx_init_v4si): Rewrite as a define_expand.
5566 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
5568 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
5569 zero_extend argument from memory): New.
5571 2018-07-22 Martin Sebor <msebor@redhat.com>
5574 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
5575 diagnosing calls with unknown arguments unless -Walloca-larger-than
5576 is restricted to less than PTRDIFF_MAX bytes.
5578 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
5580 * doc/gcov.texi (Invoking Gcov): Editorial changes.
5582 2018-07-20 David Malcolm <dmalcolm@redhat.com>
5584 * pretty-print.c (text_info::set_location): Remove redundant
5585 "line_table" parameter from call to rich_location::set_range.
5587 2018-07-20 Martin Sebor <msebor@redhat.com>
5590 * builtins.c (expand_builtin_alloca): Adjust.
5591 * calls.c (alloc_max_size): Simplify.
5592 * cgraphunit.c (cgraph_node::expand): Adjust.
5593 * common.opt (larger_than_size, warn_frame_larger_than): Remove
5595 (frame_larger_than_size): Same.
5596 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
5597 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
5599 * doc/invoke.texi (GCC Command Options): Document option arguments.
5600 Explain byte-size arguments and suffixes.
5601 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
5602 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
5603 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
5604 * doc/options.texi (UInteger): Expand.
5605 (Host_Wide_Int, ByteSize): Document new properties.
5606 * final.c (final_start_function_1): Include sizes in an error message.
5607 * function.c (frame_offset_overflow): Same.
5608 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
5609 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
5610 Diagnose unbounded alloca calls only for limits of less than
5612 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
5613 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
5615 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
5616 only for limits of less than PTRDIFF_MAX.
5617 * langhooks-def.h (lhd_handle_option): Change function argument
5619 * langhooks.c (lhd_handle_option): Same.
5620 * langhooks.h (handle_option): Same.
5621 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
5623 (var_type, var_type_struct): Same.
5624 (var_set): Handle ByteSize flag.
5625 * optc-gen.awk: Add comments to output to ease debugging. Make
5626 use of HOST_WIDE_INT where appropriate.
5627 * opts-gen-save.awk: Use %lx to format unsigned long.
5628 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
5629 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
5630 arguments. Parse bytes-size suffixes.
5631 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
5632 (enum_value_to_arg): Same.
5633 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
5634 (handle_option): Adjust.
5635 (generate_option): Change function argument to HOST_WIDE_INT.
5636 (cmdline_handle_error): Adjust.
5637 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
5638 (set_option): Change function argument to HOST_WIDE_INT.
5639 (option_enabled): Handle cl_host_wide_int.
5640 (get_option_state): Handle CLVC_SIZE.
5641 (control_warning_option): Same.
5642 * opts.c (common_handle_option): Change function argument to
5643 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
5644 OPT_Wvla_larger_than_.
5645 * opts.h (enum cl_var_type): Add an enumerator.
5646 * stor-layout.c (layout_decl): Print a more meaningful warning.
5647 * toplev.c (output_stack_usage): Adjust.
5649 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
5651 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
5652 call to inline_expand_builtin_string_cmp.
5653 (expand_builtin_strcmp): Likewise.
5654 (expand_builtin_strncmp): Likewise.
5655 (inline_string_cmp): Delete the last parameter, change char_type_node
5656 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
5658 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
5659 the inlining expansion on target where the type of the call has same or
5660 narrower precision than unsigned char.
5662 2018-07-20 David Malcolm <dmalcolm@redhat.com>
5664 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
5665 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
5666 * common.opt (fsave-optimization-record): New option.
5667 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
5668 * doc/invoke.texi (-fsave-optimization-record): New option.
5669 * dumpfile.c: Include "optinfo-emit-json.h".
5670 (struct kv_pair): Move to coretypes.h.
5671 (optgroup_options): Make non-static.
5672 (dump_context::end_scope): Call
5673 optimization_records_maybe_pop_dump_scope.
5674 * dumpfile.h (optgroup_options): New decl.
5675 * json.cc: New file.
5677 * optinfo-emit-json.cc: New file.
5678 * optinfo-emit-json.h: New file.
5679 * optinfo.cc: Include "optinfo-emit-json.h".
5680 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
5681 (optinfo_enabled_p): Check optimization_records_enabled_p.
5682 (optinfo_wants_inlining_info_p): Likewise.
5683 * optinfo.h: Update comment.
5684 * profile-count.c (profile_quality_as_string): New function.
5685 * profile-count.h (profile_quality_as_string): New decl.
5686 (profile_count::quality): New accessor.
5687 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
5688 and optinfo_emit_json_cc_tests.
5689 * selftest.h (selftest::json_cc_tests): New decl.
5690 (selftest::optinfo_emit_json_cc_tests): New decl.
5691 * toplev.c: Include "optinfo-emit-json.h".
5692 (compile_file): Call optimization_records_finish.
5693 (do_compile): Call optimization_records_start.
5694 * tree-ssa-live.c: Include optinfo.h.
5695 (remove_unused_scope_block_p): Retain inlining information if
5696 optinfo_wants_inlining_info_p returns true.
5698 2018-07-20 Richard Biener <rguenther@suse.de>
5701 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
5702 to cover -flto-partition=none.
5704 2018-07-20 Martin Liska <mliska@suse.cz>
5706 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
5707 (get_decl_source_range): Remove unused function.
5709 2018-07-20 Richard Biener <rguenther@suse.de>
5711 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
5712 (struct vn_phi_s): Likewise.
5713 (struct vn_reference_s): Likewise.
5714 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
5715 for searching the slot of an entry known to be in the hash itself.
5716 (vn_phi_hasher::equal): Likewise.
5717 (vn_reference_hasher::equal): Likewise.
5718 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
5720 (optimistic_info, current_info): Remove, keeping only valid_info.
5721 (vn_reference_lookup_1): Remove fallback lookup.
5722 (vn_reference_lookup_2): Likewise.
5723 (vn_nary_op_lookup_1): Likewise.
5724 (vn_phi_lookup): Likewise.
5725 (vn_nary_build_or_lookup_1): Make sure to not chain the built
5727 (vn_reference_insert): Adjust, chain the inserted hash element
5728 at last_inserted_ref.
5729 (vn_reference_insert_pieces): Likewise.
5730 (visit_reference_op_call): Likewise.
5731 (vn_nary_op_insert_into): Chain the inserted hash element at
5733 (vn_nary_op_insert_pieces): Adjust.
5734 (vn_nary_op_insert): Likewise.
5735 (vn_nary_op_insert_stmt): Likewise.
5736 (vn_phi_insert): Adjust, chain the inserted hash element at
5738 (process_scc): Remove clearing and copying the optimistic
5739 table. Instead remove elements inserted during an optimistic
5740 iteration from the single table we maintain.
5741 (init_scc_vn): Adjust.
5742 (free_scc_vn): Likewise.
5743 (sccvn_dom_walker::record_cond): Likewise.
5744 (sccvn_dom_walker::after_dom_children): Likewise.
5746 2018-07-19 Martin Sebor <msebor@redhat.com>
5748 PR tree-optimization/84047
5749 PR tree-optimization/83776
5750 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
5751 (check_array_bounds): Call it.
5753 2018-07-19 Martin Sebor <msebor@redhat.com>
5755 * align.h (align_flags): Use member initialization.
5757 2018-07-19 David Malcolm <dmalcolm@redhat.com>
5759 * Makefile.in (OBJS): Add optinfo.o.
5760 * coretypes.h (class symtab_node): New forward decl.
5761 (struct cgraph_node): New forward decl.
5762 (class varpool_node): New forward decl.
5763 * dump-context.h: New file.
5764 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
5766 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
5767 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
5768 (set_alt_dump_file): Likewise.
5769 (dump_context::~dump_context): New dtor.
5770 (dump_gimple_stmt): Move implementation to...
5771 (dump_context::dump_gimple_stmt): ...this new member function.
5772 Add the stmt to any pending optinfo, creating one if need be.
5773 (dump_gimple_stmt_loc): Move implementation to...
5774 (dump_context::dump_gimple_stmt_loc): ...this new member function.
5775 Start a new optinfo and add the stmt to it.
5776 (dump_gimple_expr): Move implementation to...
5777 (dump_context::dump_gimple_expr): ...this new member function.
5778 Add the stmt to any pending optinfo, creating one if need be.
5779 (dump_gimple_expr_loc): Move implementation to...
5780 (dump_context::dump_gimple_expr_loc): ...this new member function.
5781 Start a new optinfo and add the stmt to it.
5782 (dump_generic_expr): Move implementation to...
5783 (dump_context::dump_generic_expr): ...this new member function.
5784 Add the tree to any pending optinfo, creating one if need be.
5785 (dump_generic_expr_loc): Move implementation to...
5786 (dump_context::dump_generic_expr_loc): ...this new member
5787 function. Add the tree to any pending optinfo, creating one if
5789 (dump_printf): Move implementation to...
5790 (dump_context::dump_printf_va): ...this new member function. Add
5791 the text to any pending optinfo, creating one if need be.
5792 (dump_printf_loc): Move implementation to...
5793 (dump_context::dump_printf_loc_va): ...this new member function.
5794 Start a new optinfo and add the stmt to it.
5795 (dump_dec): Move implementation to...
5796 (dump_context::dump_dec): ...this new member function. Add the
5797 value to any pending optinfo, creating one if need be.
5798 (dump_context::dump_symtab_node): New member function.
5799 (dump_context::get_scope_depth): New member function.
5800 (dump_context::begin_scope): New member function.
5801 (dump_context::end_scope): New member function.
5802 (dump_context::ensure_pending_optinfo): New member function.
5803 (dump_context::begin_next_optinfo): New member function.
5804 (dump_context::end_any_optinfo): New member function.
5805 (dump_context::s_current): New global.
5806 (dump_context::s_default): New global.
5807 (dump_scope_depth): Delete global.
5808 (dumpfile_ensure_any_optinfo_are_flushed): New function.
5809 (dump_symtab_node): New function.
5810 (get_dump_scope_depth): Reimplement in terms of dump_context.
5811 (dump_begin_scope): Likewise.
5812 (dump_end_scope): Likewise.
5813 (selftest::temp_dump_context::temp_dump_context): New ctor.
5814 (selftest::temp_dump_context::~temp_dump_context): New dtor.
5815 (selftest::verify_item): New function.
5816 (ASSERT_IS_TEXT): New macro.
5817 (ASSERT_IS_TREE): New macro.
5818 (ASSERT_IS_GIMPLE): New macro.
5819 (selftest::test_capture_of_dump_calls): New test.
5820 (selftest::dumpfile_c_tests): Call it.
5821 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
5822 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
5823 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
5824 descriptive comment.
5825 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
5826 (dump_node, dump_bb): Move these unrelated decls.
5827 (class dump_manager): Add leading comment.
5828 * optinfo.cc: New file.
5829 * optinfo.h: New file.
5831 2018-07-19 Michael Collison <michael.collison@arm.com>
5832 Richard Henderson <rth@redhat.com>
5834 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
5835 (subti): Handle op1 zero.
5836 (subvti4, usub4ti4): New.
5837 (*sub<GPI>3_compare1_imm): New.
5838 (sub<GPI>3_carryinCV): New.
5839 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
5840 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
5842 2018-07-19 Michael Collison <michael.collison@arm.com>
5843 Richard Henderson <rth@redhat.com>
5845 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
5846 (addti3): Create simpler code if low part is already known to be 0.
5847 (addvti4, uaddvti4): New.
5848 (*add<GPI>3_compareC_cconly_imm): New.
5849 (*add<GPI>3_compareC_cconly): New.
5850 (*add<GPI>3_compareC_imm): New.
5851 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
5852 handle constants within this pattern..
5853 (*add<GPI>3_compareV_cconly_imm): New.
5854 (*add<GPI>3_compareV_cconly): New.
5855 (*add<GPI>3_compareV_imm): New.
5856 (add<GPI>3_compareV): New.
5857 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
5858 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
5859 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
5860 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
5861 with 'comparison' operator.
5862 (*add<GPI>3_compareV_cconly_imm): Ditto.
5863 (*add<GPI>3_compareV_cconly): Ditto.
5864 (*add<GPI>3_compareV_imm): Ditto.
5865 (add<GPI>3_compareV): Ditto.
5866 (add<mode>3_carryinC): Ditto.
5867 (*add<mode>3_carryinC_zero): Ditto.
5868 (*add<mode>3_carryinC): Ditto.
5869 (add<mode>3_carryinV): Ditto.
5870 (*add<mode>3_carryinV_zero): Ditto.
5871 (*add<mode>3_carryinV): Ditto.
5873 2018-07-19 Michael Collison <michael.collison@arm.com>
5874 Richard Henderson <rth@redhat.com>
5876 * config/aarch64/aarch64-modes.def (CC_V): New.
5877 * config/aarch64/aarch64-protos.h
5878 (aarch64_addti_scratch_regs): Declare
5879 (aarch64_subvti_scratch_regs): Declare.
5880 (aarch64_expand_subvti): Declare.
5881 (aarch64_gen_unlikely_cbranch): Declare
5882 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
5883 for signed overflow using CC_Vmode.
5884 (aarch64_get_condition_code_1): Handle CC_Vmode.
5885 (aarch64_gen_unlikely_cbranch): New function.
5886 (aarch64_addti_scratch_regs): New function.
5887 (aarch64_subvti_scratch_regs): New function.
5888 (aarch64_expand_subvti): New function.
5890 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5892 * config/aarch64/aarch64-option-extensions.def: New entry for profile
5894 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
5895 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
5898 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5901 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5902 address check not strict.
5904 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5906 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
5908 (store_pair_lanes<mode>): Likewise.
5909 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
5910 enum value 'ADDR_QUERY_LDP_STP_N'.
5911 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
5912 (aarch64_print_address_internal): Add declaration.
5913 (aarch64_print_ldpstp_address): Remove.
5914 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
5915 (aarch64_print_operand): Change printing of 'y'.
5916 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
5917 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
5918 'true' rather than '1'.
5919 * config/aarch64/constraints.md (Uml): Likewise.
5920 (Uml): Rename to Umn.
5923 2018-07-19 Richard Biener <rguenther@suse.de>
5925 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
5927 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
5928 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
5929 (vn_reference_hasher): Likewise.
5930 (struct vn_tables_s): Remove obstack and alloc-pool members.
5931 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
5932 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
5933 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
5934 (vn_reference_insert_pieces): Likewise.
5935 (alloc_vn_nary_op_noinit): Adjust.
5936 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
5937 (vn_phi_eq): Adjust.
5938 (shared_lookup_phiargs): Remove.
5939 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
5940 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
5941 (visit_reference_op_call): Likewise.
5942 (copy_nary, copy_phi, copy_reference): Remove.
5943 (process_scc): Rewind the obstack when iterating. Do not
5944 copy the elements to valid_info but just move them from one
5945 hashtable to the other.
5946 (allocate_vn_table): Adjust.
5947 (free_vn_table): Likewise.
5948 (init_scc_vn): Likewise.
5949 (free_scc_vn): Likewise.
5951 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
5954 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
5955 indirect_return as function type attribute.
5956 (ix86_attribute_table): Change indirect_return to function
5958 * doc/extend.texi: Update indirect_return attribute.
5960 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
5962 * wide-int.h (widest2_int): New.
5963 * gimple-fold.c (arith_overflowed_p): Use it.
5964 * tree.h (widest2_int_cst): New.
5965 * tree-vrp.c (wide_int_binop_overflow): Rename from
5966 vrp_int_const_binop.
5967 Rewrite to work on trees.
5968 (extract_range_from_multiplicative_op_1): Abstract code to...
5969 (wide_int_range_min_max): ...here.
5970 (wide_int_range_cross_product): ...and here.
5971 (extract_range_from_binary_expr_1): Abstract overflow code to...
5972 (wide_int_range_mult_wrapping): ...here.
5973 * tree-vrp.h (wide_int_range_cross_product): New.
5974 (wide_int_range_mult_wrapping): New.
5976 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
5977 Julia Koval <julia.koval@intel.com>
5979 * config/i386/x86-tune-costs.h (skylake_memcpy,
5980 skylake_memset): Replace rep_prefix with unrolling for size 512.
5982 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
5985 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
5986 comparision with EQ_EXPR in last stmt.
5988 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
5990 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
5991 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
5992 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
5993 previously known as "PowerPC AltiVec Built-in Functions". Move
5994 some material to new subsubsections "PowerPC AltiVec Built-in
5995 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
5997 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
5998 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
5999 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
6000 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
6002 2018-07-18 Richard Biener <rguenther@suse.de>
6004 PR tree-optimization/86557
6005 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
6008 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
6010 * config/s390/s390.c (s390_function_profiler): Generate CFI.
6012 2018-07-17 Jeff Law <law@redhat.com>
6014 * config/arm/arm.c (get_label_padding): Update for recent
6015 changes to label_to_alignment.
6017 PR tree-optimization/86010
6018 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
6020 * config/mips/mips.c (vr4130_align_insns): Update for recent
6021 changes to label_to_alignment.
6023 * config/frv/frv.c (frv_label_align): Update for recent changes
6024 to label_to_alignment.
6026 * config/nios2/nios2.c (nios2_label_align): Update for recent
6027 changes which dropped ALIGN_LABELS_LOG.
6029 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
6031 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
6032 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
6034 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
6036 * config/arc/arc.c (arc_label_align): Use align_labels instead of
6037 deprecated align_labels_log.
6039 2018-07-17 Richard Biener <rguenther@suse.de>
6042 * dwarf2out.c (init_sections_and_labels): Always generate
6043 a debug_line_str_section for early LTO debug.
6044 (dwarf2out_finish): Reset debug_line_str_hash output early.
6045 Bump counter for extra dwarf5 .debug_loc labels to not conflict
6047 (dwarf2out_early_finish): Output debug_line_str.
6049 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
6051 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
6052 index register on z196 or later.
6054 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
6056 * config/s390/s390.c (s390_default_align): Set default function
6058 (s390_override_options_after_change): Call s390_default align.
6059 (s390_option_override_internal): Call s390_default_align.
6060 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
6062 2018-07-17 Jakub Jelinek <jakub@redhat.com>
6065 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
6066 to _looptemp_ clauses, other than the first two.
6068 2018-07-17 Martin Liska <mliska@suse.cz>
6070 * opts.c: Do not enable OPT_falign_* for -Os.
6072 2018-07-17 Martin Liska <mliska@suse.cz>
6074 * align.h (MAX_CODE_ALIGN): New.
6075 (MAX_CODE_ALIGN_VALUE): New.
6076 * common/config/i386/i386-common.c (ix86_handle_option):
6077 (MAX_CODE_ALIGN): Moved to align.h.
6078 * final.c (MAX_CODE_ALIGN): Likewise.
6079 * opts.c (parse_and_check_align_values):
6080 (MAX_CODE_ALIGN): Likewise.
6081 (MAX_CODE_ALIGN_VALUE): Likewise.
6083 2018-07-17 Martin Liska <mliska@suse.cz>
6085 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
6086 in order to fulfil coding style.
6087 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
6088 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6089 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6090 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
6091 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
6092 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
6093 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
6094 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
6095 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6097 2018-07-17 Martin Liska <mliska@suse.cz>
6099 * align.h: New file.
6100 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
6102 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
6103 align_flags of label_to_alignment.
6104 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
6106 * config/m68k/m68k.c: Do not use removed align_labels_value and
6108 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
6109 (LOOP_ALIGN): Likewise.
6110 (LABEL_ALIGN): Likewise.
6111 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
6112 Remove not used macro.
6113 (rs6000_loop_align): Change return type to align_flags.
6114 (rs6000_loop_align_max_skip): Remove.
6115 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
6116 Change return type to align_flags.
6117 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
6118 Remove not used macro.
6119 (rs6000_loop_align): Change return type to align_flags.
6120 (rs6000_loop_align_max_skip): Remove.
6121 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
6122 * config/rx/rx-protos.h (rx_align_for_label): Make it
6124 * config/rx/rx.c (rx_align_for_label): Change return type
6126 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
6128 into align_flags class.
6129 (LABEL_ALIGN): Likewise.
6130 (LOOP_ALIGN): Likewise.
6131 * config/s390/s390.c (s390_label_align): Use align_flags
6133 (s390_asm_output_function_label): Likewise.
6134 * config/sh/sh.c (sh_override_options_after_change):
6135 Use align_flags class directly without macros.
6136 (find_barrier): Likewise.
6137 (barrier_align): Likewise.
6138 (sh_loop_align): Likewise.
6139 * config/spu/spu.c (spu_option_override):
6140 Use align_flags_tuple::get_value instead of removed macros.
6141 (spu_sched_init): Likewise.
6142 * config/spu/spu.h (GTY): Likewise.
6143 * config/visium/visium.c (visium_option_override):
6144 Set "8" as default secondary alignment.
6145 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
6146 in order to guarantee secondary alignment of 8.
6147 * coretypes.h: Include align.h header file.
6148 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
6149 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
6150 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
6151 * doc/tm.texi.in: Likewise.
6152 * final.c (struct label_alignment): Remove not used structure.
6153 (LABEL_ALIGN): Change type to align_flags.
6154 (LOOP_ALIGN): Likewise.
6155 (JUMP_ALIGN): Likewise.
6156 (default_loop_align_max_skip): Remove.
6157 (default_label_align_max_skip): Likewise.
6158 (default_jump_align_max_skip): Likewise.
6159 (default_label_align_after_barrier_max_skip):
6160 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
6161 (LABEL_TO_MAX_SKIP): Remove.
6162 (label_to_alignment): Return align_flags type instead of integer.
6163 (label_to_max_skip): Remove.
6164 (align_fuzz): Use align_flags type.
6165 (compute_alignments): Use align_flags type and use align_flags::max
6166 to combine multiple alignments.
6167 (grow_label_align): Grow vec instead of C array.
6168 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
6169 (shorten_branches): Use align_flags type and use align_flags::max
6170 to combine multiple alignments.
6171 (final_scan_insn_1): Remove usage of secondary alignment that comes
6172 from label alignment, but instead use proper secondary alignment
6173 which is computed in grow_label_align.
6174 * flags.h (struct align_flags_tuple): Move to align.h.
6175 (struct align_flags): Likewise.
6176 (state_align_loops): Rename to align_loops.
6177 (state_align_jumps): Rename to align_jumps.
6178 (state_align_labels): Rename to align_labels.
6179 (state_align_functions): Rename to align_functions.
6180 (align_loops_log): Remove.
6181 (align_jumps_log): Remove.
6182 (align_labels_log): Remove.
6183 (align_functions_log): Remove.
6184 (align_loops_max_skip): Remove.
6185 (align_jumps_max_skip): Remove.
6186 (align_labels_max_skip): Remove.
6187 (align_functions_max_skip): Remove.
6188 (align_loops_value): Remove.
6189 (align_jumps_value): Remove.
6190 (align_labels_value): Remove.
6191 (align_functions_value): Remove.
6192 * output.h (label_to_alignment): Change return type to align_flags.
6193 (label_to_max_skip): Remove.
6194 * target.def: Remove loop_align_max_skip, label_align_max_skip,
6195 jump_align_max_skip macros.
6196 * targhooks.h (default_loop_align_max_skip): Remove.
6197 (default_label_align_max_skip): Likewise.
6198 (default_jump_align_max_skip): Likewise.
6199 (default_label_align_after_barrier_max_skip): Remove.
6200 * toplev.c (read_log_maxskip): Use ::normalize function.
6201 (parse_N_M): Remove not used argument and also call ::normalize.
6202 (parse_alignment_opts): Do not pass unused arguments.
6203 * varasm.c (assemble_start_function): Use directly align_functions
6204 instead of removed macros.
6205 * system.h: Do not poison removed macros.
6207 2018-07-17 Jakub Jelinek <jakub@redhat.com>
6210 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
6211 and cond temporaries don't have reference type if iterator has
6212 pointer type. For init use &for_pre_body instead of pre_p if
6213 for_pre_body is non-empty.
6215 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
6217 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
6218 double-double modes to SFmode directly directly.
6219 (trunc<mode>sf2_fprs): Delete.
6221 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
6223 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
6224 for conversions between IFmode and the decimal floating point modes.
6225 (init_float128_ieee): Use the correct names for conversions between
6226 KFmode and the decimal floating point modes.
6228 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
6230 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
6231 for the conversions between TDmode and IFmode.
6232 (init_float128_ieee): Use more correct names for the conversions
6233 between TDmode and KFmode.
6235 2018-07-16 Jakub Jelinek <jakub@redhat.com>
6237 PR tree-optimization/86526
6238 * builtins.c (expand_builtin_memcmp): Formatting fixes.
6239 (inline_expand_builtin_string_cmp): Likewise.
6240 (inline_string_cmp): Likewise. Use c_readstr instead of
6241 builtin_memcpy_read_str. Add unit_mode temporary.
6243 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
6246 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
6247 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
6249 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6251 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
6252 Alphabetize prototypes of built-in functions, separating out
6253 built-in functions that are listed in this section but should be
6254 described elsewhere.
6256 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
6259 * expmed.c (emit_store_flag): Do not emit setcc followed by a
6260 conditional move when trapping comparison was split to a
6261 non-trapping one (and vice versa).
6263 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
6265 * config/s390/s390.c (s390_function_profiler): Generate nops
6266 instead of profiler call sequences.
6267 * config/s390/s390.opt: Add the new option.
6269 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
6271 * config/s390/s390.c (s390_function_profiler): Generate
6272 __mcount_loc section.
6273 * config/s390/s390.opt: Add the new option.
6275 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
6277 * common.opt: Add the new warning.
6278 * config/s390/s390.c (s390_function_profiler): Emit "brasl
6279 %r0,__fentry__" when -mfentry is specified.
6280 (s390_option_override_internal): Disallow -mfentry for 31-bit
6282 * config/s390/s390.opt: Add the new option.
6284 2018-07-16 Richard Biener <rguenther@suse.de>
6287 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
6288 for function-local FUNCTION_DECL and RESULT_DECL immediately.
6290 2018-07-16 Martin Liska <mliska@suse.cz>
6293 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
6296 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
6298 * config/arc/arcHS.md: Update ARCHS scheduling rules.
6300 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
6302 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
6304 * config/arc/arc-cpus.def (hs4x): New cpu.
6306 * config/arc/arc-tables.opt: Regenerate.
6307 * config/arc/arc.c (arc_sched_issue_rate): New function.
6308 (TARGET_SCHED_ISSUE_RATE): Define.
6309 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
6310 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
6312 (attr tune): Add ARCHS4x tune values.
6313 (attr tune_dspmpy): Define.
6314 (*tst): Correct instruction type.
6315 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
6316 * config/arc/arcHS4x.md: New file.
6317 * config/arc/fpu.md: Update instruction type attributes.
6318 * config/arc/t-multilib: Regenerate.
6320 2018-07-16 Tom de Vries <tdevries@suse.de>
6323 * var-tracking.c (vt_initialize): Fix pre_dec handling.
6325 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6327 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
6330 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
6332 PR tree-optimization/86514
6333 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
6334 conversion to a boolean type from a type with greater precision.
6336 2018-07-16 Tom de Vries <tdevries@suse.de>
6338 * var-tracking.c (vt_initialize): Print adjusted insn slim if
6339 dump_flags request TDF_SLIM.
6341 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
6343 * fold-const.c (int_const_binop_1): Abstract...
6344 (wide_int_binop): ...wide int code here.
6345 (poly_int_binop): ...poly int code here.
6346 Abstract the rest of int_const_binop_1 into int_const_binop.
6347 * fold-const.h (wide_int_binop): New.
6348 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
6349 Remove useless PLUS/MINUS_EXPR case.
6350 (zero_nonzero_bits_from_vr): Move wide int code...
6351 (zero_nonzero_bits_from_bounds): ...here.
6352 (extract_range_from_binary_expr_1): Move mask optimization code...
6353 (range_easy_mask_min_max): ...here.
6354 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
6355 (range_easy_mask_min_max): New.
6357 2018-07-15 Jeff Law <law@redhat.com>
6360 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
6363 2018-07-14 Jim Wilson <jimw@sifive.com>
6365 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
6367 2018-07-14 Paul Koning <ni1d@arrl.net>
6369 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
6371 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
6373 * lto-streamer-out.c (copy_function_or_variable): Dump info about
6376 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
6377 Steve Munroe <munroesj52@gmail.com>
6379 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
6380 (_mm_andnot_si128): Likewise.
6381 (_mm_or_si128): Likewise.
6382 (_mm_xor_si128): Likewise.
6384 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
6387 * builtins.c (expand_builtin_memcmp): Inline the calls first
6388 when result_eq is false.
6389 (expand_builtin_strcmp): Inline the calls first.
6390 (expand_builtin_strncmp): Likewise.
6391 (inline_string_cmp): New routine. Expand a string compare
6392 call by using a sequence of char comparison.
6393 (inline_expand_builtin_string_cmp): New routine. Inline expansion
6394 a call to str(n)cmp/memcmp.
6395 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
6397 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
6399 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
6401 * config/arm/driver-arm.c: Include arm-native.h.
6402 (host_detect_local_cpu): Use auto-generated data tables.
6403 (vendors, arm_cpu_table): Delete. Move part information to ...
6404 * config/arm/arm-cpus.in: ... here.
6405 * config/arm/parsecpu.awk (gen_native): New function.
6406 (vendor, part): New CPU fields.
6407 (END): Add support for building the native CPU detection tables.
6408 * config/arm/t-arm (arm-native.h): Add build rule.
6409 (driver-arm.o): Add dependency on arm-native.h.
6411 2018-07-13 Richard Biener <rguenther@suse.de>
6414 * match.pd (addr1 - addr2): Allow either of the operand to
6417 2018-07-13 Tom de Vries <tdevries@suse.de>
6419 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
6422 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
6424 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
6425 arrays instead of numbered variables.
6427 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
6429 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
6430 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
6432 2018-07-13 Richard Biener <rguenther@suse.de>
6435 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
6436 instead of get_context_die.
6438 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
6439 Richard Biener <rguenther@suse.de>
6442 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
6443 that the loop latch destination where phi is defined.
6445 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
6447 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
6448 (riscv_expand_epilogue): Add assertion to check interrupt mode.
6449 (riscv_set_current_function): Extract getting interrupt type to new
6451 (riscv_get_interrupt_type): New function.
6452 (riscv_merge_decl_attributes): New function, checking interrupt type is
6454 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
6456 2018-07-12 Paul Koning <ni1d@arrl.net>
6458 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
6461 2018-07-12 Paul Koning <ni1d@arrl.net>
6463 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
6464 zero reference, add doloop_end instead.
6465 * doc/md.texi (decrement_and_branch_until_zero): Remove.
6466 (Looping patterns): Remove decrement_and_branch_until_zero. Add
6467 detail for doloop_end.
6469 2018-07-12 Martin Sebor <msebor@redhat.com>
6472 * attribs.c (decl_attributes): Reject conflicting attributes before
6473 calling attribute handlers.
6475 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
6477 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
6479 (gcc::dump_manager::get_dump_file_name): likewise.
6480 (dump_begin): Likewise.
6481 * dumpfile.h (dump_begin): Update prototype.
6482 (gcc::dump_manager::get_dump_file_name,
6483 gcc::dump_manager::get_dump_file_name): Update prototype.
6485 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6487 * internal-fn.h (vectorizable_internal_fn_p): New function.
6488 * tree-vect-slp.c (compatible_calls_p): Likewise.
6489 (vect_build_slp_tree_1): Remove nops argument. Handle calls
6490 to internal functions.
6491 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
6493 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6495 * fold-const.h (inverse_conditions_p): Declare.
6496 * fold-const.c (inverse_conditions_p): New function.
6497 * match.pd: Use inverse_conditions_p. Add folds of view_converts
6498 that test the inverse condition of a conditional internal function.
6499 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
6500 * internal-fn.c (internal_fn_mask_index): Handle conditional
6502 (vectorized_internal_fn_supported_p): New function.
6503 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
6504 (any_pred_load_store): Replace with...
6505 (need_to_predicate): ...this new variable.
6506 (redundant_ssa_names): New variable.
6507 (ifcvt_can_use_mask_load_store): Move initial checks to...
6508 (ifcvt_can_predicate): ...this new function. Handle tree codes
6509 for which a conditional internal function exists.
6510 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
6511 instead of ifcvt_can_use_mask_load_store. Update after variable
6513 (predicate_load_or_store): New function, split out from
6514 predicate_mem_writes.
6515 (check_redundant_cond_expr): New function.
6516 (value_available_p): Likewise.
6517 (predicate_rhs_code): Likewise.
6518 (predicate_mem_writes): Rename to...
6519 (predicate_statements): ...this. Use predicate_load_or_store
6520 and predicate_rhs_code.
6521 (combine_blocks, tree_if_conversion): Update after above name changes.
6522 (ifcvt_local_dce): Handle redundant_ssa_names.
6523 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
6524 general conditional functions.
6525 * tree-vect-stmts.c (vectorizable_call): Likewise.
6527 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6528 Alan Hayward <alan.hayward@arm.com>
6529 David Sherwood <david.sherwood@arm.com>
6531 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
6532 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
6533 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
6534 plus and minus and convert them into IFN_COND_FMA-based sequences.
6535 (convert_mult_to_fma): Handle conditional plus and minus.
6537 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6539 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
6540 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
6541 (cond_fnms_optab): New optabs.
6542 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
6544 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
6545 * internal-fn.h (get_conditional_internal_fn): Declare.
6546 (get_unconditional_internal_fn): Likewise.
6547 * internal-fn.c (cond_ternary_direct): New macro.
6548 (expand_cond_ternary_optab_fn): Likewise.
6549 (direct_cond_ternary_optab_supported_p): Likewise.
6550 (FOR_EACH_COND_FN_PAIR): Likewise.
6551 (get_conditional_internal_fn): New function.
6552 (get_unconditional_internal_fn): Likewise.
6553 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
6554 (gimple_match_op::gimple_match_op): Add a new overload for 5
6556 (gimple_match_op::set_op): Likewise.
6557 (gimple_resimplify5): Declare.
6558 * genmatch.c (decision_tree::gen): Generate simplifications for
6560 * gimple-match-head.c (gimple_simplify): Define an overload for
6561 5 operands. Handle calls with 5 arguments in the top-level overload.
6562 (convert_conditional_op): Handle conversions from unconditional
6563 internal functions to conditional ones.
6564 (gimple_resimplify5): New function.
6565 (build_call_internal): Pass a fifth operand.
6566 (maybe_push_res_to_seq): Likewise.
6567 (try_conditional_simplification): Try converting conditional
6568 internal functions to unconditional internal functions.
6569 Handle 3-operand unconditional forms.
6570 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
6571 Define ternary equivalents of the current rules for binary conditional
6573 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
6575 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
6576 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
6577 (optab): Handle them.
6578 (SVE_COND_FP_TERNARY): New int iterator.
6579 (sve_fmla_op, sve_fmad_op): New int attributes.
6580 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
6581 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
6582 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
6584 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6586 * target.def (preferred_else_value): New target hook.
6587 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
6588 * doc/tm.texi: Regenerate.
6589 * targhooks.h (default_preferred_else_value): Declare.
6590 * targhooks.c (default_preferred_else_value): New function.
6591 * internal-fn.h (conditional_internal_fn_code): Declare.
6592 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
6593 (get_conditional_internal_fn): Use it.
6594 (conditional_internal_fn_code): New function.
6595 * gimple-match.h (gimple_match_cond): New struct.
6596 (gimple_match_op): Add a cond member function.
6597 (gimple_match_op::gimple_match_op): Update all forms to take a
6599 * genmatch.c (expr::gen_transform): Use the same condition as res_op
6600 for the suboperation, but don't specify a particular else_value.
6601 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
6602 (visit_nary_op, visit_reference_op_load): Pass
6603 gimple_match_cond::UNCOND to the gimple_match_op constructor.
6604 * gimple-match-head.c: Include tree-eh.h
6605 (convert_conditional_op): New function.
6606 (maybe_resimplify_conditional_op): Likewise.
6607 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
6608 (gimple_resimplify2): Likewise.
6609 (gimple_resimplify3): Likewise.
6610 (gimple_resimplify4): Likewise.
6611 (maybe_push_res_to_seq): Return null for conditional operations.
6612 (try_conditional_simplification): New function.
6613 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
6615 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
6617 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
6619 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
6621 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
6623 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
6625 (hash_tree): Do not hash DECL_FCONTEXT
6626 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
6627 Do not stream DECL_FCONTEXT.
6628 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
6629 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
6631 2018-07-12 Richard Biener <rguenther@suse.de>
6634 * dwarf2out.c (gen_block_die): Only output blocks when they have
6635 at least one !DECL_IGNORED_P variable.
6637 2018-07-12 Richard Biener <rguenther@suse.de>
6640 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
6641 Remove -mieee-fp handling.
6643 2018-07-12 Richard Biener <rguenther@suse.de>
6645 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
6646 left-over from last patch.
6648 2018-07-12 Jakub Jelinek <jakub@redhat.com>
6650 PR tree-optimization/86492
6651 * gimple-ssa-store-merging.c
6652 (imm_store_chain_info::coalesce_immediate_stores): Call
6653 check_no_overlap even for the merge_overlapping case. Formatting fix.
6655 2018-07-12 Richard Biener <rguenther@suse.de>
6658 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
6659 move possibly trapping operations into the conditional.
6661 2018-07-12 Richard Biener <rguenther@suse.de>
6663 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
6664 (vn_lookup_simplify_result): Remove recursion limit applied
6666 (vn_nary_build_or_lookup_1): Adjust.
6667 (try_to_simplify): Likewise.
6668 * gimple-match-head.c (gimple_resimplify1): Instead apply one
6670 (gimple_resimplify2): Likewise.
6671 (gimple_resimplify3): Likewise.
6672 (gimple_resimplify4): Likewise.
6674 2018-07-11 Jakub Jelinek <jakub@redhat.com>
6676 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
6677 Use __mmask64 type instead of __mmask8 for __M argument.
6678 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
6679 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
6681 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
6682 __mmask16 for __M argument.
6683 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
6684 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
6685 to __mmask16 instead of __mmask8.
6686 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
6687 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
6688 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
6689 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
6690 instead of __mmask16 for __U argument.
6691 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
6692 __mmask16 instead of __mmask8 for __U argument.
6693 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
6695 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
6697 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
6699 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
6701 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
6703 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
6705 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
6707 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
6709 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
6711 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
6713 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
6714 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
6715 return type as well as __M argument type and all casts from __mmask8
6717 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
6718 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
6719 return type as well as __M argument type and all casts from __mmask8
6721 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
6722 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
6723 return type as well as __M argument type and all casts from __mmask8
6725 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
6726 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
6727 return type as well as __M argument type and all casts from __mmask8
6729 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
6730 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
6733 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
6735 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
6736 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
6739 2018-07-11 Paul Koning <ni1d@arrl.net>
6741 * doc/md.texi (define_subst): Document how multiple occurrences of
6742 the same argument in the replacement pattern are handled.
6744 2018-07-11 Paul Koning <ni1d@arrl.net>
6746 * doc/extend.texi (Common Variable Attributes): Move "mode" into
6748 (Common Type Attributes): Add "mode" attribute.
6750 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
6752 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
6753 stream DECL_ORIGINAL_TYPE.
6754 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
6755 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
6756 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
6757 Do not walk original type.
6758 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
6760 (write_ts_decl_non_common_tree_pointers): Do not stream
6762 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
6763 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
6765 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
6767 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
6768 thread twice from the same starting edge.
6770 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
6772 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
6773 * gimple.c (gimple_call_nonnull_result_p): ...here...
6774 (gimple_call_nonnull_arg): ...and here.
6775 * gimple.h (gimple_call_nonnull_result_p): New.
6776 (gimple_call_nonnull_arg): New.
6778 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
6780 * config/arm/arm-cpus.in: Move information from fpu field of each
6781 cpu definition to the isa field.
6782 * config/arm/parsecpu.awk (fpu): Delete match rule.
6783 (gen_comm_data): Don't add bits from the CPU's FPU entry.
6785 2018-07-11 Richard Biener <rguenther@suse.de>
6788 * dwarf2out.c (init_sections_and_labels): Use
6789 output_asm_line_debug_info consistently.
6790 (dwarf2out_early_finish): Likewise.
6791 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
6794 2018-07-11 Richard Biener <rguenther@suse.de>
6796 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
6797 Rework father_bb setting in a way to avoid propagating constants
6798 multiple times on a loop body.
6800 2018-07-10 Mark Wielaard <mark@klomp.org>
6803 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
6805 2018-07-10 Richard Biener <rguenther@suse.de>
6807 * hash-map.h (hash_map::iterator::operator*): Return
6808 references to key and value.
6810 2018-07-10 Jakub Jelinek <jakub@redhat.com>
6813 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
6814 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
6815 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
6816 (gimplify_omp_for): For composite loops, move outer
6817 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
6818 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
6819 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
6820 TREE_LIST for both the original class iterator and the "last" helper
6821 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
6822 loop, remember has_decl_expr from outer composite loops for the
6823 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
6825 2018-07-09 Martin Sebor <msebor@redhat.com>
6829 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
6830 accessing implicitly initialized array elements.
6831 * expr.c (string_constant): Handle string initializers of
6832 character arrays within aggregates.
6833 * gimple-fold.c (fold_array_ctor_reference): Add argument.
6834 Store element offset. As a special case, handle zero size.
6835 (fold_nonarray_ctor_reference): Same.
6836 (fold_ctor_reference): Add argument. Store subobject offset.
6837 * gimple-fold.h (fold_ctor_reference): Add argument.
6839 2018-07-09 Paul Koning <ni1d@arrl.net>
6841 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
6842 (pdp11_insn_cost): New function.
6843 (pdp11_md_asm_adjust): New function.
6844 (TARGET_INVALID_WITHIN_DOLOOP): Define.
6845 (pdp11_rtx_costs): Update to match machine better.
6846 (output_addr_const_pdp11): Correct format mismatch warnings.
6847 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
6848 * config/pdp11/pdp11.md: General change to add base_cost and/or
6849 length attributes for use by new pdp11_insn_cost function.
6850 (MIN_BRANCH): Correct definition.
6852 (doloop_end): Use standard pattern name for looping pattern.
6853 (doloop_end_nocc): New.
6854 (movsf): Add another constraint alternative.
6855 (zero_extendqihi2): Add constraint alternatives for not in place
6857 (zero_extendhisi2): Remove.
6858 (shift patterns): Add CC handling variants.
6862 (define_peephole2): New peephole to recognize mov that sets CC for
6865 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6867 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
6868 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
6871 2018-07-09 Jakub Jelinek <jakub@redhat.com>
6874 * real.c (real_nextafter): Return true if result is denormal.
6876 2018-07-09 Martin Liska <mliska@suse.cz>
6878 * common.opt: Add back wrongly removed attribute.
6880 2018-07-09 Richard Biener <rguenther@suse.de>
6883 * dwarf2out.c (gen_block_die): For an early generated DIE
6884 always output high/low PC attributes.
6886 2018-07-09 Tom de Vries <tdevries@suse.de>
6888 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
6889 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
6890 onto VAR_DECL with abstract origin.
6892 2018-07-07 Jim Wilson <jimw@sifive.com>
6894 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
6896 2018-07-07 Tom de Vries <tdevries@suse.de>
6898 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
6900 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
6902 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
6904 (combine_bound): Use wide-int overflow calculation instead of
6906 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
6908 * fold-const.c (int_const_binop_2): Same.
6909 (extract_muldiv_1): Same.
6910 (fold_div_compare): Same.
6911 (fold_abs_const): Same.
6913 * poly-int.h (add): Same.
6917 * predict.c (predict_iv_comparison): Same.
6918 * profile-count.c (slow_safe_scale_64bit): Same.
6919 * simplify-rtx.c (simplify_const_binary_operation): Same.
6920 * tree-chrec.c (tree_fold_binomial): Same.
6921 * tree-data-ref.c (split_constant_offset_1): Same.
6922 * tree-if-conv.c (idx_within_array_bound): Same.
6923 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
6924 * tree-ssa-phiopt.c (minmax_replacement): Same.
6925 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
6926 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
6927 * vr-values.c (vr_values::adjust_range_with_scev): Same.
6928 * wide-int.cc (wi::add_large): Same.
6929 (wi::mul_internal): Same.
6930 (wi::sub_large): Same.
6931 (wi::divmod_internal): Same.
6932 * wide-int.h: Change overflow type to overflow_type for neg, add,
6933 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
6934 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
6935 mul_internal, divmod_internal.
6936 (overflow_type): New enum.
6937 (accumulate_overflow): New.
6939 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
6941 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
6942 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
6944 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
6946 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
6947 argument is checked for zero before entering loop, avoid checking again.
6949 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
6951 * gimplify.h (generic_expr_could_trap_p): Set as global function.
6952 * gimplify.h (generic_expr_could_trap_p): Likwise.
6953 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
6955 2018-07-06 Jakub Jelinek <jakub@redhat.com>
6957 PR tree-optimization/86401
6958 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
6959 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
6960 (fold_bit_and_mask): ... here. New helper function for match.pd.
6961 * fold-const.h (fold_bit_and_mask): Declare.
6962 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
6964 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
6967 * target.def (translate_mode_attribute): New hook.
6968 * targhooks.h (default_translate_mode_attribute): Declare.
6969 * targhooks.c (default_translate_mode_attribute): New function.
6970 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
6971 * doc/tm.texi: Regenerate.
6972 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
6973 (rs6000_translate_mode_attribute): New function.
6975 2018-07-06 Paul Koning <ni1d@arrl.net>
6977 * doc/md.texi (define_split): Document DONE and FAIL.
6978 (define_peephole2): Ditto.
6980 2018-07-05 Jeff Law <law@redhat.com>
6982 PR tree-optimization/86010
6983 * tree-ssa-dse.c (compute_trims): More aggressively trim at
6984 both the head and tail of mem* and str* calls.
6986 2018-07-05 Jim Wilson <jimw@sifive.com>
6988 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
6990 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
6992 * config/aarch64/aarch64-simd.md: correct flags text for
6993 MIN_EXPR replacement.
6995 2018-07-05 James Clarke <jrtc27@jrtc27.com>
6997 * configure: Regenerated.
6999 2018-07-05 Carl Love <cel@us.ibm.com>
7001 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
7002 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
7003 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
7004 VSX_BUILTIN_DOUBLEL_V4SF.
7006 2018-07-05 Martin Sebor <msebor@redhat.com>
7009 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
7010 than its domain to compute its the upper bound of a char array.
7012 2018-07-05 Nathan Sidwell <nathan@acm.org>
7014 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
7015 * doc/cpp.texi: Update comment.
7016 * doc/tm.texi: Rebuilt.
7017 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
7018 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
7019 * doc/extend.texi (Backwards Compatibility): Clarify it is system
7020 headers affected by extern "C".
7021 * system.h: Poison NO_IMPLICIT_EXTERN_C.
7022 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
7023 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
7024 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
7025 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
7026 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
7027 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
7028 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
7029 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
7030 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
7031 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
7032 config/sparc/sp64-elf.h, config/spu/spu.h,
7033 config/stormy16/stormy16.h, config/v850/v850.h,
7034 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
7035 define NO_IMPLICIT_EXTERN_C.
7036 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
7038 2018-07-05 Tamar Christina <tamar.christina@arm.com>
7041 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7042 instead of GET_MODE_SIZE when comparing Units.
7044 2018-07-05 Tamar Christina <tamar.christina@arm.com>
7047 * rtlanal.c (set_noop_p): Constrain on mode change,
7048 include hard-reg-set.h
7050 2018-07-05 Tamar Christina <tamar.christina@arm.com>
7052 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
7054 2018-07-05 Jakub Jelinek <jakub@redhat.com>
7057 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
7060 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
7062 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
7064 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
7067 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
7069 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
7071 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
7074 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
7076 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
7078 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
7079 neg_*_op* variables.
7081 2018-07-04 Martin Liska <mliska@suse.cz>
7083 * tree-switch-conversion.c: Define
7084 max_ratio_for_speed and max_ratio_for_size constants.
7086 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
7087 Martin Liska <mliska@suse.cz>
7092 * common.opt: Rename align options with 'str_' prefix.
7093 * common/config/i386/i386-common.c (set_malign_value): New
7095 (ix86_handle_option): Use it to set -falign-* options/
7096 * config/aarch64/aarch64-protos.h (struct tune_params): Change
7097 type from int to string.
7098 * config/aarch64/aarch64.c: Update default values from int
7100 * config/alpha/alpha.c (alpha_override_options_after_change):
7102 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
7103 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7104 max skip conditionally.
7105 * config/i386/freebsd.h (SUBALIGN_LOG): New.
7106 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7107 max skip conditionally.
7108 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7109 max skip conditionally.
7110 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
7111 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7112 max skip conditionally.
7113 * config/i386/i386.c (struct ptt): Change type from int to
7115 (ix86_default_align): Set default values.
7116 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
7117 max skip conditionally.
7118 * config/i386/iamcu.h (SUBALIGN_LOG): New.
7119 (ASM_OUTPUT_MAX_SKIP_ALIGN):
7120 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
7121 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7122 max skip conditionally.
7123 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
7124 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
7125 * config/i386/x86-64.h (SUBALIGN_LOG): New.
7126 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7127 max skip conditionally.
7128 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7129 * config/ia64/ia64.c (ia64_option_override): Set default values
7130 for alignment options.
7131 * config/m68k/m68k.c: Handle new str_align_* options.
7132 * config/mips/mips.c (mips_set_compression_mode): Change
7134 (mips_option_override): Set default values for options.
7135 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
7137 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7139 * config/rx/rx.c (rx_option_override): Likewise.
7140 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
7141 (LABEL_ALIGN): Use align_labels_log.
7142 (LOOP_ALIGN): Use align_loops_align.
7143 * config/s390/s390.c (s390_asm_output_function_label): Use new
7145 * config/sh/sh.c (sh_override_options_after_change):
7146 Change type of constants.
7147 * config/spu/spu.c (spu_sched_init): Likewise.
7148 * config/sparc/sparc.c (sparc_option_override): Set default
7150 * config/visium/visium.c (visium_option_override): Likewise.
7151 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
7152 emit p2align format with last argument if it's not needed.
7153 * doc/invoke.texi: Document extended format of -falign-*.
7154 * final.c: Use align_labels alignment.
7155 * flags.h (struct target_flag_state): Change type to use
7157 (struct align_flags_tuple): New.
7158 (struct align_flags): Likewise.
7159 (align_loops_log): Redefine macro to use new types.
7160 (align_loops_max_skip): Redefine macro to use new types.
7161 (align_jumps_log): Redefine macro to use new types.
7162 (align_jumps_max_skip): Redefine macro to use new types.
7163 (align_labels_log): Redefine macro to use new types.
7164 (align_labels_max_skip): Redefine macro to use new types.
7165 (align_functions_log): Redefine macro to use new types.
7166 (align_loops): Redefine macro to use new types.
7167 (align_jumps): Redefine macro to use new types.
7168 (align_labels): Redefine macro to use new types.
7169 (align_functions): Redefine macro to use new types.
7170 (align_functions_max_skip): Redefine macro to use new types.
7171 (align_loops_value): New macro.
7172 (align_jumps_value): New macro.
7173 (align_labels_value): New macro.
7174 (align_functions_value): New macro.
7175 * function.c (invoke_set_current_function_hook): Propagate
7176 alignment values from flags to global variables default in
7178 * ipa-icf.c (sem_function::equals_wpa): Use
7179 cl_optimization_option_eq instead of memcmp.
7180 * lto-streamer.h (cl_optimization_stream_out): Support streaming
7182 (cl_optimization_stream_in): Likewise.
7183 * optc-save-gen.awk: Support strings in cl_optimization.
7184 * opth-gen.awk: Likewise.
7185 * opts.c (finish_options): Remove error checking of invalid
7187 (MAX_CODE_ALIGN): Remove.
7188 (MAX_CODE_ALIGN_VALUE): Likewise.
7189 (parse_and_check_align_values): New function.
7190 (check_alignment_argument): Likewise.
7191 (common_handle_option): Use check_alignment_argument.
7192 * opts.h (parse_and_check_align_values): Declare.
7193 * toplev.c (init_alignments): Remove.
7194 (read_log_maxskip): New.
7195 (parse_N_M): Likewise.
7196 (parse_alignment_opts): Likewise.
7197 (backend_init_target): Remove usage of init_alignments.
7198 * toplev.h (parse_alignment_opts): Declare.
7199 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
7201 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
7202 * tree.c (cl_option_hasher::equal): New.
7203 * varasm.c: Use new global macros.
7205 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
7207 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
7208 Use a simpler align directive also if MAXSKIP = ALIGN-1.
7209 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7210 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7211 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7212 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7213 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
7214 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
7215 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
7216 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7217 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7218 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7219 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7221 2018-07-04 Martin Liska <mliska@suse.cz>
7222 Jonathan Wakely <jwakely@redhat.com>
7224 * coverage.c: Use correct type.
7225 * doc/invoke.texi: Language correction.
7227 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
7230 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
7231 ENDBRANCH for non-tail call which may return via indirect branch.
7232 * doc/extend.texi: Document indirect_return attribute.
7234 2018-07-03 Martin Sebor <msebor@redhat.com>
7236 PR tree-optimization/86274
7237 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
7239 (format_floating): Correct handling of infinities and NaNs.
7241 2018-07-03 Martin Sebor <msebor@redhat.com>
7243 * print-tree.c (print_real_cst): New function.
7244 (print_node_brief): Call it.
7245 (print_node): Ditto.
7247 2018-07-03 Jeff Law <law@redhat.com>
7249 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
7250 into a single pattern.
7252 * config/h8300/h8300.md (ors code_iterator): New.
7253 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
7254 a single pattern and single splitter.
7255 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
7256 (iorqi3_1, xorqi3_1): Likewise.
7257 (iorqi3, xorqi3 expanders): Similarly.
7259 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
7260 (movmd_internal) into a single pattern using the P mode iterator.
7261 (movmd splitters): Similarly.
7262 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
7263 (movsd splitters): Similarly.
7265 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
7266 ADDB, ADDW and ADDL into a single ADD attribute which selects the
7267 right table based on the size of the operand.
7268 * config/h8300/h8300.md (length_table): Corresponding changes. All
7269 references to "addb", "addw" and "addl" changed to "add".
7270 (btst patterns): Merge two variants into a single pattern.
7271 (tstqi, tsthi): Likewise.
7272 (addhi3_incdec, addsi3_incdec): Likewise.
7273 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
7274 (mulhi3, mulsi3): Likewise.
7275 (udivhi3, udivsi3): Likewise.
7276 (divhi3, divsi3): Likewise.
7277 (andorqi3, andorhi3, andorsi3): Likewise.
7279 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
7282 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
7283 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
7285 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7287 PR tree-optimization/85694
7288 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
7289 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
7291 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
7292 (<u>avg<mode>3_ceil): New patterns.
7294 2018-07-03 David Malcolm <dmalcolm@redhat.com>
7296 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
7297 scan-tree-dump directive.
7298 * gcc.dg/vect/slp-perm-2.c: Likewise.
7299 * gcc.dg/vect/slp-perm-3.c: Likewise.
7300 * gcc.dg/vect/slp-perm-5.c: Likewise.
7301 * gcc.dg/vect/slp-perm-6.c: Likewise.
7302 * gcc.dg/vect/slp-perm-7.c: Likewise.
7303 * gcc.dg/vect/slp-perm-8.c: Likewise.
7305 2018-07-03 Marek Polacek <polacek@redhat.com>
7308 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
7310 2018-07-03 Richard Biener <rguenther@suse.de>
7313 * tree-ssa-structalias.c (find_func_clobbers): Properly
7314 handle indirect calls.
7316 2018-07-03 Jeff Law <law@redhat.com>
7318 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
7319 (shifts): New code iterator.
7320 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
7321 expander. Fix HImode handling on H8/SX.
7322 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
7323 (subqi3, subhi3, subsi3 expanders): Likewise.
7324 (andqi3, andhi3, andsi3 expanders): Likewise.
7325 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
7326 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
7327 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
7328 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
7329 (zero_extendqihi2, zero_extendqisi2): Likewise.
7330 (extendqihi2, extendqisi2): Likewise.
7331 (rotlqi3, rotlhi3, rotlsi3): Likewise.
7332 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
7333 (rotlqi3_1, rotlhi3_1): Likewise.
7334 (logicalhi3_sn, logicalsi3_sn): Likewise.
7335 (logicalhi3, logicalsi3): Likewise.
7337 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7339 * tree-vect-patterns.c (vect_recog_rotate_pattern)
7340 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7341 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
7342 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
7343 type to append_pattern_def_seq instead of creating a stmt_vec_info
7345 (build_mask_conversion): Likewise. Remove vinfo argument.
7346 (vect_add_conversion_to_patterm): Likewise, renaming to...
7347 (vect_add_conversion_to_pattern): ...this.
7348 (vect_recog_mask_conversion_pattern): Update call to
7349 build_mask_conversion. Pass the vector type to
7350 append_pattern_def_seq here too.
7351 (vect_recog_gather_scatter_pattern): Update call to
7352 vect_add_conversion_to_pattern.
7354 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7356 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
7357 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7358 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
7359 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
7360 STMT_VINFO_PATTERN_DEF_SEQ to null here.
7361 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
7362 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
7363 append_pattern_def_seq instead of new_pattern_def_seq.
7364 (vect_recog_divmod_pattern): Do both of the above.
7365 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
7368 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7370 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
7371 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
7372 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
7373 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
7374 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
7375 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7376 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
7377 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
7378 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
7379 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
7380 parameter with a single stmt_vec_info.
7381 (vect_recog_func_ptr): Likewise.
7382 (vect_recog_gather_scatter_pattern): Likewise, folding in...
7383 (vect_try_gather_scatter_pattern): ...this.
7384 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
7385 the stmt_vec_info of the statement to be matched. Don't clear
7386 STMT_VINFO_RELATED_STMT.
7387 (vect_pattern_recog): Update call accordingly.
7389 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7391 PR tree-optimization/85694
7392 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
7393 (uavgM3_ceil): Document new optabs.
7394 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
7395 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
7397 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
7398 (savg_ceil_optab): New optabs.
7399 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
7400 (vect_vect_recog_func_ptrs): Add it.
7401 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
7402 constant directly from the associated lhs.
7404 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7406 * tree-vect-patterns.c (vect_split_statement): New function.
7407 (vect_convert_input): Use it to try to split an existing cast.
7409 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7411 * poly-int.h (print_hex): New function.
7412 * dumpfile.h (dump_dec, dump_hex): Declare.
7413 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
7414 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
7415 min_input_precision, operation_precision and operation_sign.
7416 * tree-vect-patterns.c (vect_get_range_info): New function.
7417 (vect_same_loop_or_bb_p, vect_single_imm_use)
7418 (vect_operation_fits_smaller_type): Delete.
7419 (vect_look_through_possible_promotion): Add an optional
7420 single_use_p parameter.
7421 (vect_recog_over_widening_pattern): Rewrite to use new
7422 stmt_vec_info infomration. Handle one operation at a time.
7423 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
7424 (vect_truncatable_operation_p, vect_set_operation_type)
7425 (vect_set_min_input_precision): New functions.
7426 (vect_determine_min_output_precision_1): Likewise.
7427 (vect_determine_min_output_precision): Likewise.
7428 (vect_determine_precisions_from_range): Likewise.
7429 (vect_determine_precisions_from_users): Likewise.
7430 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
7431 (vect_vect_recog_func_ptrs): Put over_widening first.
7433 (vect_pattern_recog): Call vect_determine_precisions.
7435 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7437 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
7438 statements that have been replaced by further pattern statements.
7439 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
7441 2018-07-03 Richard Biener <rguenther@suse.de>
7443 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
7444 always set *dt. Dump vectype in vectype overload.
7445 * dumpfile.h (dump_gimple_expr): New function.
7446 (dump_gimple_expr_loc): Likewise.
7447 * dumpfile.c (dump_gimple_expr): New function.
7448 (dump_gimple_expr_loc): Likewise.
7450 2018-07-02 Jeff Law <law@redhat.com>
7452 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
7453 the H8/300, H8/300H and H8/S variants into a single pattern.
7454 (movhi_h8300, movqi_h8300hs): Similarly.
7455 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
7456 (QHI mode iterator): New.
7458 * config/h8300/h8300.md: Remove trailing whitespace.
7460 2018-07-02 Jim Wilson <jimw@sifive.com>
7462 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
7463 instead of emit_insn for interrupt returns.
7464 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
7465 (riscv_sret, riscv_uret): Likewise.
7467 2018-07-02 David Malcolm <dmalcolm@redhat.com>
7469 * pretty-print.c (selftest::test_pp_format): Move save and restore
7470 of quotes to class auto_fix_quotes, and add an instance.
7471 * selftest.c: Include "intl.h".
7472 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
7473 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
7474 * selftest.h (selftest::auto_fix_quotes): New class.
7476 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
7478 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
7479 (aarch64_sve_prepare_conditional_op): Remove.
7480 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
7481 Allow aarch64_simd_reg_or_zero as select operand; remove
7482 the aarch64_sve_prepare_conditional_op call.
7483 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
7484 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7485 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
7486 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
7487 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
7488 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
7489 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
7490 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
7491 and a splitters to match all of the *_any patterns.
7492 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
7494 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
7495 (SVE_COND_FP_BINARY_REV): Remove.
7496 (sve_int_op_rev, sve_fp_op_rev): New.
7497 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
7498 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
7499 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
7500 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
7501 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
7502 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
7503 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
7504 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
7505 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
7507 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
7508 Remove match_dup 1 from the inner unspec.
7509 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7511 * config/aarch64/aarch64.md (movprfx): New attr.
7512 (length): Default movprfx to 8.
7513 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
7514 (*madd<SVE_I>, *msub<SVE_I): Likewise.
7515 (*<su>mul<SVE_I>3_highpart): Likewise.
7516 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
7517 (*v<ASHIFT><SVE_I>3): Likewise.
7518 (*<su><MAXMIN><SVE_I>3): Likewise.
7519 (*<su><MAXMIN><SVE_F>3): Likewise.
7520 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
7521 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
7522 (*div<SVE_F>4): Likewise.
7524 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
7526 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
7529 2018-07-02 Richard Biener <rguenther@suse.de>
7531 PR tree-optimization/86363
7532 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
7533 memset argument refers to a non-variable address.
7535 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
7537 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
7538 {PLUS,MINUS}_EXPR code to...
7539 (adjust_symbolic_bound): ...here,
7540 (combine_bound): ...here,
7541 (set_value_range_with_overflow): ...and here.
7543 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
7545 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
7547 (extract_range_from_abs_expr): ...here.
7549 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
7551 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
7552 -fno-omit-frame-pointer when not optimizing.
7554 2018-07-02 Martin Liska <mliska@suse.cz>
7557 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
7558 (propagate_nothrow): Likewise.
7560 2018-07-02 Martin Liska <mliska@suse.cz>
7563 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
7565 2018-07-02 David Malcolm <dmalcolm@redhat.com>
7567 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
7568 function in r262149, changing "loc" param from source_location to
7569 const dump_location_t &.
7570 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
7571 declaration, as above.
7573 2018-07-01 Paul Koning <ni1d@arrl.net>
7575 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
7576 -munit-asm, -mgnu-asm, -mdec-asm.
7577 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
7578 (pdp11_output_labelref): New.
7579 (pdp11_output_def): New.
7580 (pdp11_output_addr_vec_elt): New.
7581 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
7582 %# and %@ format codes.
7583 (pdp11_option_override): New.
7584 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
7585 (pdp11_output_ident): New.
7586 (pdp11_asm_named_section): New.
7587 (pdp11_asm_init_sections): New.
7588 (pdp11_file_start): New.
7589 (pdp11_file_end): New.
7590 (output_ascii): Use .ascii/.asciz for -mdec-asm.
7591 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
7592 %o, like %c but octal.
7593 (pdp11_option_override): New.
7594 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
7596 (DATA_SECTION_ASM_OP): Ditto.
7597 (READONLY_DATA_SECTION_ASM_OP): New.
7598 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
7599 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
7600 (ASM_OUTPUT_LABELREF): Ditto.
7601 (ASM_OUTPUT_DEF): Ditto.
7602 (ASM_OUTPUT_EXTERNAL): New.
7603 (ASM_OUTPUT_SOURCE_FILENAME): New.
7604 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
7605 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
7606 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
7607 %# and %@ format codes.
7608 * config/pdp11/pdp11.opt (mgnu-asm): New.
7609 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
7610 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
7611 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
7613 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
7615 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
7616 dereferencing path[] beyond its length.
7618 (debug_all_paths): New.
7619 (rewire_first_differing_edge): New.
7620 (adjust_paths_after_duplication): New.
7621 (duplicate_thread_path): Call adjust_paths_after_duplication.
7623 (thread_through_all_blocks): Add new argument to
7624 duplicate_thread_path.
7626 2018-06-30 Jim Wilson <jimw@sifive.com>
7628 * config/riscv/predicates.md (p2m1_shift_operand): New.
7629 (high_mask_shift_operand): New.
7630 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
7631 pattern using p2m1_shift_operand.
7632 (lshsi3_zero_extend_3+2): New combiner pattern using
7633 high_mask_shift_operand.
7635 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7637 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
7639 (vect_recog_rotate_pattern): ...here.
7640 (vect_convert_input): Try to insert casts of invariants in the
7642 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
7643 preheader to be empty.
7645 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7647 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
7648 vector type. If given, install it in the new statement's
7650 (vect_element_precision): New function.
7651 (vect_unpromoted_value): New struct.
7652 (vect_unpromoted_value::vect_unpromoted_value): New function.
7653 (vect_unpromoted_value::set_op): Likewise.
7654 (vect_look_through_possible_promotion): Likewise.
7655 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
7656 (vect_widened_op_tree, vect_convert_input): Likewise.
7657 (vect_convert_inputs, vect_convert_output): Likewise.
7658 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
7659 to handle the optional cast of the multiplication result and
7660 vect_widened_op_tree to detect the widened multiplication itself.
7661 Do not require the input and output of promotion casts to have
7662 the same sign, but base the signedness of the operation on the
7663 input rather than the result. If the pattern includes two
7664 promotions, check that those promotions have the same sign.
7665 Do not restrict the MULT_EXPR handling to a double-width result;
7666 handle quadruple-width results and wider. Use vect_convert_inputs
7667 to convert the inputs to the common type.
7668 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
7669 to handle the optional cast of the ABS result. Also allow a sign
7670 change or a sign extension between the ABS and MINUS.
7671 Use vect_widened_op_tree to detect the widened subtraction and use
7672 vect_convert_inputs to convert the inputs to the common type.
7673 (vect_handle_widen_op_by_const): Delete.
7674 (vect_recog_widen_op_pattern): New function.
7675 (vect_recog_widen_mult_pattern): Use it.
7676 (vect_recog_widen_shift_pattern): Likewise.
7677 (vect_recog_widen_sum_pattern): Use
7678 vect_look_through_possible_promotion to handle the promoted
7681 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7683 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
7684 the containing gimple_seq *.
7685 * gimple-iterator.h (gsi_for_stmt): Declare it.
7686 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
7687 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
7688 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7689 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7690 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
7692 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
7694 (vect_mark_pattern_stmts): ...here. Handle cases in which the
7695 statement being replaced is part of an existing pattern
7696 definition sequence, inserting the new pattern statements before
7698 (vect_pattern_recog_1): Don't return a bool. If the statement
7699 is already part of a pattern, instead apply pattern matching
7700 to the pattern definition statements. Don't clear the
7701 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
7702 (vect_pattern_recog): Don't break after the first match;
7703 continue processing the pattern definition statements instead.
7704 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
7706 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7708 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7709 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7710 (vect_recog_widen_sum_pattern): Use it.
7712 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7714 * tree-vect-loop.c (vectorizable_reduction): Assert that the
7715 phi is not a pattern statement and has not been replaced by
7716 a pattern statement.
7717 * tree-vect-patterns.c (type_conversion_p): Don't check
7718 STMT_VINFO_IN_PATTERN_P.
7719 (vect_recog_vector_vector_shift_pattern): Likewise.
7720 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
7721 the pattern statement rather than the original statement; check
7722 directly for a WIDEN_MULT_EXPR here.
7723 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
7724 vect_is_simple_use to return the pattern statement rather
7725 than the original statement; use is_pattern_stmt_p to check
7726 for such a pattern statement.
7727 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
7728 to return the pattern statement rather than the original statement;
7729 don't do the same transformation here.
7730 (vect_is_simple_use): If the defining statement has been replaced
7731 by a pattern statement, return the pattern statement instead.
7732 Remove the corresponding (local) transformation from the vectype
7735 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7737 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
7738 end and default to null.
7739 * tree-vect-loop.c (vect_create_epilog_for_reduction)
7740 (vectorizable_reduction): Update calls accordingly, dropping the
7741 gimple ** argument if the passed-back statement isn't needed.
7742 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
7743 (vect_recog_rotate_pattern): Likewise.
7744 (vect_recog_mask_conversion_pattern): Likewise.
7745 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7746 (vect_mask_constant_operand_p): Likewise.
7747 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
7748 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
7749 (get_group_load_store_type, get_load_store_type): Likewise.
7750 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
7751 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
7752 (vectorizable_conversion, vectorizable_assignment): Likewise.
7753 (vectorizable_shift, vectorizable_operation): Likewise.
7754 (vectorizable_store, vect_is_simple_cond): Likewise.
7755 (vectorizable_condition, vectorizable_comparison): Likewise.
7756 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
7757 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
7758 and move it to the end. Cope with null def_stmt_outs.
7760 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
7762 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
7764 2018-06-29 Jeff Law <law@redhat.com>
7766 * config/v850/v850.c (v850_legitimate_address_p): Handle large
7767 displacements for TARGET_V850E2V3 and newer.
7768 (TARGET_LRA_P): Remove. Defaults to LRA now.
7769 * config/v850/v850.md (sign23byte_load): Remove.
7770 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
7771 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
7773 2018-06-29 Martin Liska <mliska@suse.cz>
7776 * coverage.c (coverage_init): Mangle full path name.
7777 * doc/invoke.texi: Document the change.
7778 * gcov-io.c (mangle_path): New.
7779 * gcov-io.h (mangle_path): Likewise.
7780 * gcov.c (mangle_name): Use mangle_path for path mangling.
7782 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7784 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
7785 if starting source register is not even.
7787 2018-06-29 Martin Liska <mliska@suse.cz>
7789 PR tree-optimization/86263
7790 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
7791 Make edge redirection.
7793 2018-06-29 David Malcolm <dmalcolm@redhat.com>
7795 * dumpfile.c (dump_loc): Add indentation based on scope depth.
7796 (dump_scope_depth): New variable.
7797 (get_dump_scope_depth): New function.
7798 (dump_begin_scope): New function.
7799 (dump_end_scope): New function.
7800 * dumpfile.h (get_dump_scope_depth): New declaration.
7801 (dump_begin_scope): New declaration.
7802 (dump_end_scope): New declaration.
7803 (class auto_dump_scope): New class.
7804 (AUTO_DUMP_SCOPE): New macro.
7805 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
7808 2018-06-29 Richard Biener <rguenther@suse.de>
7810 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
7811 compute_all_dependences succeeds.
7812 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
7813 exceed --param loop-max-datarefs-for-datadeps.
7815 2018-06-29 Jakub Jelinek <jakub@redhat.com>
7817 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
7819 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
7822 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
7823 alternative 0 in preferred_for_speed attribute.
7825 2018-06-28 Paul Koning <ni1d@arrl.net>
7827 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
7828 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
7829 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
7830 * config/pdp11/pdp11.md: Correct "length" attribute calculation
7831 for shift insn patterns.
7833 2018-06-28 David Malcolm <dmalcolm@redhat.com>
7835 * cgraph.c (cgraph_node::get_body): Replace assignments to
7836 "dump_file" with calls to set_dump_file.
7837 * dumpfile.c (alt_dump_file): Make static, and group with...
7838 (alt_flags): ...this definition.
7839 (dumps_are_enabled): New variable.
7840 (refresh_dumps_are_enabled): New function.
7841 (set_dump_file): New function.
7842 (set_alt_dump_file): New function.
7843 (gcc::dump_manager::dump_start): Replace assignments to
7844 "dump_file" and "alt_dump_file" with calls to set_dump_file and
7846 (gcc::dump_manager::dump_finish): Likewise.
7847 * dumpfile.h (alt_dump_file): Delete decl.
7848 (dumps_are_enabled): New variable decl.
7849 (set_dump_file): New function decl.
7850 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
7852 * tree-nested.c (lower_nested_functions): Replace assignments to
7853 "dump_file" with calls to set_dump_file.
7855 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
7857 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
7858 goto_locus of each outgoing edge of each basic block.
7860 2018-06-28 Richard Biener <rguenther@suse.de>
7862 * dwarf2out.c (decl_scope_table): Remove.
7863 (push_decl_scope): Likewise.
7864 (pop_decl_scope): Likewise.
7865 (gen_type_die_for_member): Do not call push/pop_decl_scope.
7866 (gen_struct_or_union_type_die): Likewise.
7867 (gen_tagged_type_die): Likewise.
7868 (dwarf2out_init): Do not initialize decl_scope_table.
7869 (dwarf2out_c_finalize): Do not free it.
7871 2018-06-28 Richard Biener <rguenther@suse.de>
7873 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
7874 deciding whether to not re-use a DIE.
7876 2018-06-28 Richard Biener <rguenther@suse.de>
7878 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
7879 DW_AT_abstract_origin attribute.
7881 2018-06-28 Martin Liska <mliska@suse.cz>
7883 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
7884 Use newly introduced constants.
7885 * tree-switch-conversion.h (struct jump_table_cluster):
7886 Define max_ratio_for_size and max_ratio_for_speed.
7888 2018-06-28 Martin Liska <mliska@suse.cz>
7890 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7891 Add new checking assert to catch invalid state.
7892 (jump_table_cluster::can_be_handled): Handle single case
7894 (jump_table_cluster::is_beneficial): Bail out for such case.
7895 (bit_test_cluster::find_bit_tests):
7896 Add new checking assert to catch invalid state.
7897 (bit_test_cluster::can_be_handled): Handle single case
7899 (bit_test_cluster::is_beneficial): Bail out for such case.
7900 (switch_decision_tree::analyze_switch_statement):
7903 2018-06-28 Martin Liska <mliska@suse.cz>
7905 * common.opt: Introduce -completion option.
7906 * gcc.c (driver_handle_option): Handle it.
7907 (driver::main): Print completions if completion
7909 * opt-suggestions.c (option_proposer::get_completions):
7911 (option_proposer::suggest_completion): Likewise.
7912 (option_proposer::find_param_completions): Likewise.
7913 (verify_autocompletions): Likewise.
7914 (test_completion_valid_options): Likewise.
7915 (test_completion_valid_params): Likewise.
7916 (in_completion_p): Likewise.
7917 (empty_completion_p): Likewise.
7918 (test_completion_partial_match): Likewise.
7919 (test_completion_garbage): Likewise.
7920 (opt_proposer_c_tests): Likewise.
7921 * opt-suggestions.h: Declare new functions.
7922 * opts.c (common_handle_option): Handle OPT__completion_.
7923 * selftest-run-tests.c (selftest::run_tests): Add
7924 opt_proposer_c_tests.
7925 * selftest.c (assert_str_startswith): New.
7926 * selftest.h (assert_str_startswith): Likewise.
7927 (opt_proposer_c_tests): New.
7928 (ASSERT_STR_STARTSWITH): Likewise.
7930 2018-06-28 Martin Liska <mliska@suse.cz>
7932 * Makefile.in: Add opt-suggestions.o.
7933 * gcc-main.c: Include opt-suggestions.h.
7934 * gcc.c (driver::driver): Likewise.
7935 (driver::~driver): Remove m_option_suggestions.
7936 (driver::build_option_suggestions): Moved to option_proposer.
7937 (driver::suggest_option): Likewise.
7938 (driver::handle_unrecognized_options): Use option_proposer.
7939 * gcc.h (class driver): Add new memver m_option_proposer.
7940 * opt-suggestions.c: New file.
7941 * opt-suggestions.h: New file.
7943 2018-06-28 Martin Liska <mliska@suse.cz>
7945 * vec.h (class auto_string_vec): New (moved from auto_argvec).
7946 (auto_string_vec::~auto_string_vec): Likewise.
7948 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
7950 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
7951 prevent_decl_creation_for_types fields up and add reset_location field.
7952 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
7953 statement if id->reset_location is true.
7954 (copy_edges_for_bb): Do not set goto_locus on the new edges if
7955 id->reset_location is true.
7956 (copy_phis_for_bb): Force input_location on the arguments if
7957 id->reset_location is true.
7958 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
7959 is set on the function to be inlined.
7961 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
7963 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
7965 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
7967 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
7968 registers for Pmode.
7969 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
7970 hard registers for the clobbered pseudo.
7972 2018-06-27 Paul Koning <ni1d@arrl.net>
7974 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
7975 mutually exclusive options.
7976 * config/pdp11/constraints.md (h): New constraint.
7977 (O): Update definition to match shift code generation.
7978 (D): New constraint.
7979 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
7981 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
7983 (output_jump): Change arguments.
7984 (pdp11_fixed_cc_regs): New function.
7985 (pdp11_cc_mode): Ditto.
7986 (pdp11_expand_shift): Ditto.
7987 (pdp11_assemble_shift): Ditto.
7988 (pdp11_small_shift): Ditto.
7989 (pdp11_branch_cost): Remove.
7990 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
7992 (pdp11_register_move_cost): Update for CC registers.
7993 (pdp11_rtx_costs): Add case for LSHIFTRT.
7994 (pdp11_output_jump): Add CCNZ mode conditional branches.
7995 (notice_update_cc_on_set): Remove.
7996 (pdp11_cc_mode): New function.
7997 (simple_memory_operand): Correct pre/post decrement case.
7998 (no_side_effect_operand): New function.
7999 (pdp11_regno_reg_class): Add CC_REGS class.
8000 (pdp11_fixed_cc_regs): New function.
8001 (pdp11_small_shift): New function.
8002 (pdp11_expand_shift): New function to expand shift insns.
8003 (pdp11_assemble_shift): New function to output shifts.
8004 (pdp11_branch_cost): Remove.
8005 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
8006 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
8007 (WCHAR_TYPE): Ditto.
8008 (PTRDIFF_TYPE): Ditto.
8009 (ADJUST_INSN_LENGTH): New macro.
8010 (FIXED_REGISTERS): Add CC registers.
8011 (CALL_USED_REGISTERS): Ditto.
8013 (REG_CLASS_NAMES): Ditto.
8014 (REG_CLASS_CONTENTS): Ditto.
8015 (SELECT_CC_MODE): Use new function.
8016 (TARGET_FLAGS_REGNUM): New macro.
8017 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
8018 (cc0_reg_rtx): Remove.
8019 (CC_STATUS_MDEP): Remove.
8020 (CC_STATUS_MDEFP_INIT): Remove.
8021 (CC_IN_FPU): Remove.
8022 (NOTICE_UPDATE_CC): Remove.
8023 (REGISTER_NAMES): Add CC registers.
8024 (BRANCH_COST): Change to constant 1.
8025 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
8027 * config/pdp11/pdp11.opt (mbcopy): Remove.
8028 (mbcopy-builtin): Remove.
8029 (mbranch-cheap): Remove.
8030 (mbranch-expensive): Remove.
8031 * config/pdp11/predicates.md (expand_shift_operand): Update to
8032 match shift code generation.
8033 (ccnz_operator): New predicate.
8034 * doc/invoke.texi (PDP-11 Options): Remove deleted options
8035 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
8036 Remove non-existent option -mabshi, -mno-abshi. Document mutually
8038 * doc/md.texi (PDP-11): Document new D and h constraints. Update
8039 description of O constraint.
8041 2018-06-27 Jeff Law <law@redhat.com>
8042 Austin Law <austinklaw@gmail.com>
8044 * config/v850/v850.md (addsi3_set_flags): New pattern.
8045 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
8046 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
8047 (zero_extendhisi2_v850_set_flags): Likewise.
8048 (zero_extendqisi2_v850_set_flags): Likewise.
8049 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
8050 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
8051 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
8053 * config/v850/v850-protos.h (notice_update_cc): Remove.
8054 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
8055 (v850_print_operand): Handle 'D' and "d".
8056 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
8057 Add handling of arithmetic/logical operations compared against zero.
8058 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
8059 Do not look at v850_compare_op, instead get mode from last argument.
8060 (v850_gen_compare): Remove
8061 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
8062 after reload for prologue insns.
8063 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
8065 (construct_save_jarl): Likewise.
8066 (TARGET_FLAGS_REGNUM): Define.
8067 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
8068 (NOTICE_UPDATE_CC): Remove.
8069 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
8070 than cc0. Conditionalize on reload_completed.
8071 (cmpsi_insn, setfcc_insn): Likewise.
8072 (tst1 splitter): Turn into define_and_split which sets the flags
8074 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
8075 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
8076 (cstoresf4, cstoredf4): Clobber the flags.
8077 (cmpsi, cmpsf, cmpdf): Remove expanders.
8078 (setf_insn): Remove pattern.
8079 (addsi3): Turn into define_and_split which clobbers the flags after
8080 reload and a suitable pattern (addsi3_clobber_flags) for use after
8082 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
8083 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
8084 (ashrsi3, ashrsi3_v850e2): Likewise.
8085 (bins): Clobber the flags.
8086 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
8087 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
8088 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
8089 (call_value_internal_short, call_value_internal_long): Likewise.
8090 (callt_save_interrupt, callt_return_interrupt): Likewise.
8091 (save_interrupt, return_interrupt): Likewise.
8092 (callt_save_all_interrupt, save_all_interrupt): Likewise.
8093 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
8094 (restore_all_interrupt, _restore_all_interrupt): Likewise.
8095 (All FP comparisons): Only allow after reload has completed.
8097 (divh, divhu): Tweak output template.
8098 (branch_z_normal, branch_z_invert): Remove
8099 (branch_nz_normal, branch_nz_invert): Likewise.
8100 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
8102 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
8103 * config/v850/v850.c (notice_update_cc): Remove.
8104 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
8105 (CC_NO_CARRY): Likewise.
8106 (NOTICE_UPDATE_CC): Define to nothing.
8107 * config/v850/v850.md: Remove block comment on cc0 handling
8108 Remove "cc" attribute from all patterns. Remove cc_status handling
8109 from all patterns. Minor formatting fixes.
8111 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8113 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
8114 (cortex-a76.cortex-a55): Likewise.
8115 * config/aarch64/aarch64-tune.md: Regenerate.
8116 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
8117 cortex-a76.cortex-a55.
8119 2018-06-27 Jeff Law <law@redhat.com>
8121 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
8122 (MULTILIB_DIRNAMES): Similarly.
8124 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
8126 * gimple.h (gimple_return_retbnd): Delete.
8127 (gimple_return_set_retbnd): Likewise.
8128 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
8129 gimple_return_set_retbnd.
8130 * gimple-pretty-print.c (dump_gimple_return): Remove call to
8131 gimple_return_retbnd and adjust.
8132 * tree-inline.h (struct copy_body_data): Remove retbnd field.
8133 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
8134 Explicitly return NULL in a couple more cases. Move assertion
8135 on debug statements and remove unreachable code.
8136 (reset_debug_binding): Do not test id->retbnd.
8137 (expand_call_inline): Do not set it.
8139 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
8141 * configure.ac: Add --disable-gcov option.
8142 * configure: Regenerate.
8143 * Makefile.in: Honour @enable_gcov@.
8144 * doc/install.texi: Document --disable-gcov.
8146 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8148 * config/arm/arm-cpus.in (cortex-a76): New entry.
8149 (cortex-a76.cortex-a55): Likewise.
8150 * config/arm/arm-tables.opt: Regenerate.
8151 * config/arm/arm-tune.md: Likewise.
8152 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
8153 * doc/invoke.texi (ARM Options): Document cortex-a76 and
8154 cortex-a76.cortex-a55.
8156 2018-06-27 Tamar Christina <tamar.christina@arm.com>
8159 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
8161 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
8163 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
8165 (EPILOGUE_USES): Likewise.
8167 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
8169 * tree-inline.c (remap_location): New function extracted from...
8170 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
8171 (copy_phis_for_bb): ...here. Call remap_location.
8172 (copy_cfg_body): Adjust call to copy_edges_for_bb.
8174 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
8176 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
8177 unaligned vsx for 16B memset.
8179 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
8182 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
8183 ieee128_float_type_node to long_double_type_node unless
8184 TARGET_LONG_DOUBLE_128 is set.
8186 2018-06-26 David Malcolm <dmalcolm@redhat.com>
8188 * cfgloop.c (get_loop_location): Convert return type from
8189 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
8190 by implicit construction from rtx_insn *, and using
8191 dump_user_location_t::from_function_decl for the fallback case.
8192 * cfgloop.h (get_loop_location): Convert return type from
8193 location_t to dump_user_location_t.
8194 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
8195 dump_printf_loc to pass in a dump_location_t rather than a
8196 location_t, via the gimple stmt.
8197 * coverage.c (get_coverage_counts): Update calls to
8198 dump_printf_loc to pass in dump_location_t rather than a
8200 * doc/optinfo.texi (Dump types): Convert example of
8201 dump_printf_loc from taking "locus" to taking "insn". Update
8202 description of the "_loc" calls to cover dump_location_t.
8203 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
8205 (dump_user_location_t::dump_user_location_t): New constructors,
8206 from gimple *stmt and rtx_insn *.
8207 (dump_user_location_t::from_function_decl): New function.
8208 (dump_loc): Make static.
8209 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
8210 const dump_location_t &.
8211 (dump_generic_expr_loc): Delete.
8212 (dump_printf_loc): Convert param "loc" from location_t to
8213 const dump_location_t &.
8214 (selftest::test_impl_location): New function.
8215 (selftest::dumpfile_c_tests): New function.
8216 * dumpfile.h: Include "profile-count.h".
8217 (class dump_user_location_t): New class.
8218 (struct dump_impl_location_t): New struct.
8219 (class dump_location_t): New class.
8220 (dump_printf_loc): Convert 2nd param from source_location to
8221 const dump_location_t &.
8222 (dump_generic_expr_loc): Delete.
8223 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
8224 const dump_location_t &.
8225 * gimple-fold.c (fold_gimple_assign): Update call to
8226 dump_printf_loc to pass in a dump_location_t rather than a
8227 location_t, via the gimple stmt.
8228 (gimple_fold_call): Likewise.
8229 * gimple-loop-interchange.cc
8230 (loop_cand::analyze_iloop_reduction_var): Update for change to
8231 check_reduction_path.
8232 (tree_loop_interchange::interchange): Update for change to
8234 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
8235 change in return-type of find_loop_location.
8236 (graphite_regenerate_ast_isl): Likewise.
8237 * graphite-optimize-isl.c (optimize_isl): Likewise.
8238 * graphite.c (graphite_transform_loops): Likewise.
8239 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
8240 pass in a dump_location_t rather than a location_t, via the
8242 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
8243 * ipa.c (walk_polymorphic_call_targets): Likewise.
8244 * loop-unroll.c (report_unroll): Convert "locus" param from
8245 location_t to dump_location_t.
8246 (decide_unrolling): Update for change to get_loop_location's
8248 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
8249 location_t to dump_user_location_t.
8250 (grid_find_single_omp_among_assignments_1): Updates calls to
8251 dump_printf_loc to pass in a dump_location_t rather than a
8252 location_t, via the gimple stmt.
8253 (grid_parallel_clauses_gridifiable): Convert "tloc" from
8254 location_t to dump_location_t. Updates calls to dump_printf_loc
8255 to pass in a dump_location_t rather than a location_t, via the
8257 (grid_inner_loop_gridifiable_p): Likewise.
8258 (grid_dist_follows_simple_pattern): Likewise.
8259 (grid_gfor_follows_tiling_pattern): Likewise.
8260 (grid_target_follows_gridifiable_pattern): Likewise.
8261 (grid_attempt_target_gridification): Convert initialization
8262 of local "grid" from memset to zero-initialization; FIXME: does
8263 this require C++11? Update call to dump_printf_loc to pass in a
8264 optinfo_location rather than a location_t, via the gimple stmt.
8265 * profile.c (read_profile_edge_counts): Updates call to
8266 dump_printf_loc to pass in a dump_location_t rather than a
8268 (compute_branch_probabilities): Likewise.
8269 * selftest-run-tests.c (selftest::run_tests): Call
8271 * selftest.h (dumpfile_c_tests): New decl.
8272 * tree-loop-distribution.c (pass_loop_distribution::execute):
8273 Update for change in return type of find_loop_location.
8274 * tree-parloops.c (parallelize_loops): Likewise.
8275 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
8276 "locus" from location_t to dump_user_location_t.
8277 (canonicalize_loop_induction_variables): Likewise.
8278 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
8279 for change in return type of find_loop_location.
8280 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
8281 to dump_printf_loc to pass in a dump_location_t rather than a
8282 location_t, via the stmt.
8283 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
8285 * tree-vect-loop-manip.c (find_loop_location): Convert return
8286 type from source_location to dump_user_location_t.
8287 (vect_do_peeling): Update for above change.
8288 (vect_loop_versioning): Update for change in type of
8290 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
8291 from location_t to dump_user_location_t.
8292 (vect_estimate_min_profitable_iters): Update for change in type
8294 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
8295 location_t to dump_location_t.
8296 (vect_slp_bb): Update for change in type of vect_location.
8297 * tree-vectorizer.c (vect_location): Convert from source_location
8298 to dump_user_location_t.
8299 (try_vectorize_loop_1): Update for change in vect_location's type.
8300 (vectorize_loops): Likewise.
8301 (increase_alignment): Likewise.
8302 * tree-vectorizer.h (vect_location): Convert from source_location
8303 to dump_user_location_t.
8304 (find_loop_location): Convert return type from source_location to
8305 dump_user_location_t.
8306 (check_reduction_path): Convert 1st param from location_t to
8307 dump_user_location_t.
8308 * value-prof.c (check_counter): Update call to dump_printf_loc to
8309 pass in a dump_user_location_t rather than a location_t; update
8310 call to error_at for change in type of "locus".
8311 (check_ic_target): Update call to dump_printf_loc to
8312 pass in a dump_user_location_t rather than a location_t, via the
8315 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
8317 * config/s390/s390.h (enum processor_flags): Do not use
8318 default tune parameter when -march was specified.
8320 2018-06-26 Jakub Jelinek <jakub@redhat.com>
8323 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
8324 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
8327 2018-06-26 Richard Biener <rguenther@suse.de>
8329 PR tree-optimization/86287
8331 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
8332 (vect_analyze_loop): Initialize n_stmts.
8334 2018-06-26 Richard Biener <rguenther@suse.de>
8337 * fold-const.c (fold_convertible_p): Pointer extension
8340 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
8343 * dwarf2out.c (loc_list_has_views): Adjust comments.
8344 (dw_loc_list): Split single cross-partition range with
8347 2018-06-25 Jeff Law <law@redhat.com>
8349 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
8350 on -mbig-switch by default.
8352 * config/v850/predicates.md (const_float_1_operand): Fix match_code
8354 (const_float_0_operand): Remove unused predicate.
8355 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
8356 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
8357 (recipsf2): New expander. Original pattern now called
8359 (recipdf2, recipdf2_insn): Similarly.
8360 (rsqrtsf2, rsqrtsf2_insn): Similarly
8361 (rsqrtdf2, rsqrtdf2_insn): Similarly
8363 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
8365 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
8366 Simplify logic for FreeBSD (twice).
8368 2018-06-25 Martin Sebor <msebor@redhat.com>
8370 PR tree-optimization/86204
8371 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
8372 a strnlen result if it's less than the length of the string.
8374 2018-06-25 Martin Sebor <msebor@redhat.com>
8376 PR tree-optimization/85700
8377 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
8378 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
8379 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
8381 2018-06-25 Martin Sebor <msebor@redhat.com>
8383 * doc/extend.texi (Zero-length arrays): Update and clarify.
8385 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
8387 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
8388 added IEEE/IBM long double multilib support on PowerPC little
8389 endian Linux systems.
8390 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
8391 (MULTILIB_DEFAULTS): Likewise.
8392 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8394 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
8395 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
8396 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
8398 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
8401 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
8402 (REORDER_45): Likewise.
8404 2018-06-25 Jeff Law <law@redhat.com>
8406 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
8407 dividend to 32 bits. Adjust length.
8408 (udivmodhi4): Cleanup output template. Fix length.
8410 2018-06-25 Carl Love <cel@us.ibm.com>
8412 * config/rs6000/vsx.md: Change word selector to prefered location.
8414 2018-06-25 Richard Biener <rguenther@suse.de>
8416 PR tree-optimization/86304
8417 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
8418 epilogue-if-converted loops as well.
8420 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
8422 * lto-section-out.c (lto_begin_section): Do not print section
8423 name for noaddr and unnumbered dumps.
8425 2018-06-25 Richard Biener <rguenther@suse.de>
8427 * tree-vectorizer.h (struct vec_info_shared): New structure
8428 with parts split out from struct vec_info and loop_nest from
8429 struct _loop_vec_info.
8430 (struct vec_info): Adjust accordingly.
8431 (struct _loop_vec_info): Likewise.
8432 (LOOP_VINFO_LOOP_NEST): Adjust.
8433 (LOOP_VINFO_DATAREFS): Likewise.
8434 (LOOP_VINFO_DDRS): Likewise.
8435 (struct _bb_vec_info): Likewise.
8436 (BB_VINFO_DATAREFS): Likewise.
8437 (BB_VINFO_DDRS): Likewise.
8438 (struct _stmt_vec_info): Add dr_aux member.
8439 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
8440 (DR_MISALIGNMENT_UNINITIALIZED): New.
8441 (set_dr_misalignment): Adjust.
8442 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
8443 (vect_analyze_loop): Adjust prototype.
8444 (vect_analyze_loop_form): Likewise.
8445 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
8446 Compute dependences lazily.
8447 (vect_record_base_alignments): Use shared datarefs/ddrs.
8448 (vect_verify_datarefs_alignment): Likewise.
8449 (vect_analyze_data_refs_alignment): Likewise.
8450 (vect_analyze_data_ref_accesses): Likewise.
8451 (vect_analyze_data_refs): Likewise.
8452 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
8453 constructor parameter for shared part.
8454 (vect_analyze_loop_form): Pass in shared part and adjust.
8455 (vect_analyze_loop_2): Pass in storage for the number of
8456 stmts. Move loop nest finding to the caller. Compute
8458 (vect_analyze_loop): Pass in shared part.
8459 (vect_transform_loop): Verify shared datarefs are unchanged.
8460 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
8461 constructor parameter for shared part.
8462 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
8463 (vect_slp_bb): Verify shared datarefs are unchanged before
8465 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
8467 (new_stmt_vec_info): Initialize DR_AUX misalignment to
8468 DR_MISALIGNMENT_UNINITIALIZED.
8469 * tree-vectorizer.c (vec_info::vec_info): Add constructor
8470 parameter for shared part.
8471 (vec_info::~vec_info): Adjust.
8472 (vec_info_shared::vec_info_shared): New.
8473 (vec_info_shared::~vec_info_shared): Likewise.
8474 (vec_info_shared::save_datarefs): Likewise.
8475 (vec_info_shared::check_datarefs): Likewise.
8476 (try_vectorize_loop_1): Construct shared part live for analyses
8477 of a single loop for multiple vector sizes.
8478 * tree-parloops.c (gather_scalar_reductions): Adjust.
8480 2018-06-25 Richard Biener <rguenther@suse.de>
8482 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
8483 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
8484 (vect_analyze_data_refs): Remove similar code from here and
8485 simplify accordingly.
8487 2018-06-25 Richard Biener <rguenther@suse.de>
8489 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
8490 for reverse storage order accesses rather than asserting
8491 they cannot happen here.
8493 2018-06-25 Tom de Vries <tdevries@suse.de>
8496 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
8497 Use data16 instead of .byte for insn prefix.
8499 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
8502 * parser.c (make_char_string_pack): Pass this literal chars
8503 through cpp_interpret_string.
8504 (cp_parser_userdef_numeric_literal): Check the result of
8505 make_char_string_pack.
8507 2018-06-24 Maya Rashish <coypu@sdf.org>
8509 * ginclude/stddef.h: Simplify conditions around avoiding
8510 re-definition of __size_t.
8512 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8514 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
8515 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
8517 2018-06-22 Maya Rashish <coypu@sdf.org>
8519 * doc/invoke.texi (mno-fancy-math-387): Update for changes
8520 made to OpenBSD and NetBSD through the years.
8522 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8524 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8525 behavior of vec_pack (vector double, vector double) to match
8526 behavior of vec_float2 (vector double, vector double).
8528 2018-06-22 Olivier Hainque <hainque@adacore.com>
8530 * gimplify.c (gimplify_function_tree): Prevent creation
8531 of a trampoline for the address of the current function
8532 passed to entry/exit instrumentation hooks.
8534 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
8537 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
8540 2018-06-22 Martin Liska <mliska@suse.cz>
8542 PR tree-optimization/86263
8543 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8544 Bail out if is_enabled is false.
8545 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
8547 (jump_table_cluster::is_enabled): New function.
8549 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8551 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8552 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
8553 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
8554 (lto_input_ts_binfo_tree_pointers): Likewise.
8555 * tree-streamer-out.c (streamer_write_tree_bitfields,
8556 write_ts_binfo_tree_pointers): Likewise.
8557 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
8559 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8561 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
8563 2018-06-22 Martin Liska <mliska@suse.cz>
8565 * symbol-summary.h (get): Make it pure and inline move
8566 functionality from ::get function.
8567 (get): Remove and inline into ::get and ::get_create.
8568 (get_create): Move code from ::get function.
8570 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8573 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
8574 -x assembler-with-cpp.
8576 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8578 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
8579 _FILE_OFFSET_BITS=64 for C++.
8581 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
8583 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
8584 conversion insn that shows up when pr85657-3.c is compiled using
8585 IEEE 128-bit long double.
8586 (neg<mode>2_internal): Use the correct mode to check whether the
8587 mode is IBM extended.
8588 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
8589 multiply and divide external functions from being created more
8592 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8594 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
8596 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
8597 the edge can be forwarded.
8598 (cfg_layout_merge_blocks): Likewise.
8600 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8602 * except.c (finish_eh_generation): Commit edge insertions only after
8603 the EH edges have been redirected from post-landing to landing pads.
8605 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8607 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
8608 create_tmp_var_for to create the FRAME decl.
8609 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
8611 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8613 * tree-inline.c (copy_edges_for_bb): Minor tweak.
8614 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
8615 debug statement when resetting its value.
8616 (expand_call_inline): Copy the locus of the call onto the assignment
8617 of the return value, if any. Use local variable in more cases.
8619 2018-06-21 Martin Liska <mliska@suse.cz>
8621 * ipa-pure-const.c (propagate_nothrow): Use
8622 funct_state_summaries->get.
8623 (dump_malloc_lattice): Likewise.
8624 (propagate_malloc): Likewise.
8626 2018-06-21 Richard Biener <rguenther@suse.de>
8628 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
8629 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
8630 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
8632 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
8633 BLOCK_ABSTRACT_ORIGIN unconditionally.
8635 2018-06-21 David Malcolm <dmalcolm@redhat.com>
8637 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
8639 * ipa-reference.c (ipa_reference_c_finalize): Delete
8640 ipa_ref_opt_sum_summaries and set it to NULL.
8642 2018-06-21 Tom de Vries <tdevries@suse.de>
8644 PR tree-optimization/85859
8645 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
8646 test with comment from bb_no_side_effects_p.
8648 2018-06-21 Richard Biener <rguenther@suse.de>
8650 PR tree-optimization/86232
8651 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
8652 max for constant niter.
8654 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
8656 * config/aarch64/aarch64-simd.md
8657 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
8659 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
8661 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
8662 Make opernads of the unspec commutative.
8664 2018-06-21 Richard Biener <rguenther@suse.de>
8666 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
8667 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
8668 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
8669 (vect_analyze_data_ref_dependence): Re-order checks to deal with
8671 (vect_record_base_alignments): Do not record base alignment
8672 for gathers or scatters.
8673 (vect_compute_data_ref_alignment): Drop return value that is always
8674 true. Bail out early for gathers or scatters.
8675 (vect_enhance_data_refs_alignment): Bail out early for gathers
8677 (vect_find_same_alignment_drs): Likewise.
8678 (vect_analyze_data_refs_alignment): Remove dead code.
8679 (vect_slp_analyze_and_verify_node_alignment): Likewise.
8680 (vect_analyze_data_refs): For possible gathers or scatters do
8681 not create an alternate DR, just check their possible validity
8682 and mark them. Adjust DECL_NONALIASED handling to not rely
8684 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
8685 update inits of gathers or scatters.
8686 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
8687 Also copy gather/scatter flag to pattern vinfo.
8689 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
8691 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8692 behavior of vec_packsu (vector unsigned long long, vector unsigned
8693 long long) to match behavior of vec_packs with same signature.
8695 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
8696 Thomas Schwinge <thomas@codesourcery.com>
8697 Cesar Philippidis <cesar@codesourcery.com>
8699 * gimplify.c (gimplify_scan_omp_clauses): Add support for
8700 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8701 (gimplify_adjust_omp_clauses): Likewise.
8702 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
8703 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
8704 (gimplify_omp_target_update): Update handling of acc update and
8706 * omp-low.c (install_var_field): Remove unused parameter
8707 base_pointers_restrict.
8708 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
8709 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
8711 (omp_target_base_pointers_restrict_p): Delete.
8712 (scan_omp_target): Update call to scan_sharing_clauses.
8713 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
8715 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
8716 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8717 (convert_local_omp_clauses): Likewise.
8718 * tree-pretty-print.c (dump_omp_clause): Likewise.
8719 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
8721 (omp_clause_code_name): Likewise.
8723 2018-06-20 Jakub Jelinek <jakub@redhat.com>
8726 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
8729 PR tree-optimization/86231
8730 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
8731 anti-range don't overwrite *vr0min before using it to compute *vr0max.
8733 2018-06-20 Tom de Vries <tdevries@suse.de>
8735 PR tree-optimization/86097
8736 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
8737 iv type if signedness of iv type is not the same as that of *nit.
8739 2018-06-20 Jakub Jelinek <jakub@redhat.com>
8741 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
8742 EDGE_EH edges, verify they are all EDGE_EH.
8744 2018-06-20 Maya Rashish <coypu@sdf.org>
8746 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
8748 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8750 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
8751 * config/aarch64/aarch64.c (xgene1_tunings): Add
8752 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
8753 (aarch64_mode_valid_for_sched_fusion_p):
8754 Allow 16-byte modes.
8755 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
8756 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
8758 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
8760 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
8761 * config/aarch64/iterators.md (VQ2): New mode iterator.
8763 2018-06-20 Martin Liska <mliska@suse.cz>
8765 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
8766 Change default ratio from 10 to 8.
8768 2018-06-20 Martin Liska <mliska@suse.cz>
8770 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8772 (bit_test_cluster::find_bit_tests): Likewise.
8773 (switch_decision_tree::analyze_switch_statement): Find clusters.
8774 * tree-switch-conversion.h (struct jump_table_cluster): Document
8777 2018-06-20 Martin Liska <mliska@suse.cz>
8779 * tree-switch-conversion.c (switch_conversion::collect):
8780 Record m_uniq property.
8781 (switch_conversion::expand): Bail out for special conditions.
8782 (group_cluster::~group_cluster): New.
8783 (group_cluster::group_cluster): Likewise.
8784 (group_cluster::dump): Likewise.
8785 (jump_table_cluster::emit): New.
8786 (switch_decision_tree::fix_phi_operands_for_edges): New.
8787 (struct case_node): Remove struct.
8788 (jump_table_cluster::can_be_handled): New.
8789 (case_values_threshold): Moved to header.
8790 (reset_out_edges_aux): Likewise.
8791 (jump_table_cluster::is_beneficial): New.
8792 (bit_test_cluster::can_be_handled): Likewise.
8793 (add_case_node): Remove.
8794 (bit_test_cluster::is_beneficial): New.
8795 (case_bit_test::cmp): New.
8796 (bit_test_cluster::emit): New.
8797 (expand_switch_as_decision_tree_p): Remove.
8798 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
8799 (fix_phi_operands_for_edge): Likewise.
8800 (switch_decision_tree::analyze_switch_statement): New.
8801 (compute_cases_per_edge): Move ...
8802 (switch_decision_tree::compute_cases_per_edge): ... here.
8803 (try_switch_expansion): Likewise.
8804 (switch_decision_tree::try_switch_expansion): Likewise.
8805 (record_phi_operand_mapping): Likewise.
8806 (switch_decision_tree::record_phi_operand_mapping): Likewise.
8807 (emit_case_decision_tree): Likewise.
8808 (switch_decision_tree::emit): Likewise.
8809 (balance_case_nodes): Likewise.
8810 (switch_decision_tree::balance_case_nodes): Likewise.
8811 (dump_case_nodes): Likewise.
8812 (switch_decision_tree::dump_case_nodes): Likewise.
8813 (emit_jump): Likewise.
8814 (switch_decision_tree::emit_jump): Likewise.
8815 (emit_cmp_and_jump_insns): Likewise.
8816 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
8817 (emit_case_nodes): Likewise.
8818 (switch_decision_tree::emit_case_nodes): Likewise.
8819 (conditional_probability): Remove.
8820 * tree-switch-conversion.h (enum cluster_type): New.
8822 (struct cluster): Likewise.
8823 (cluster::cluster): Likewise.
8824 (struct simple_cluster): Likewise.
8825 (simple_cluster::simple_cluster): Likewise.
8826 (struct group_cluster): Likewise.
8827 (struct jump_table_cluster): Likewise.
8828 (struct bit_test_cluster): Likewise.
8829 (struct min_cluster_item): Likewise.
8830 (struct case_tree_node): Likewise.
8831 (case_tree_node::case_tree_node): Likewise.
8832 (jump_table_cluster::case_values_threshold): Likewise.
8833 (struct case_bit_test): Likewise.
8834 (struct switch_decision_tree): Likewise.
8835 (struct switch_conversion): Likewise.
8836 (switch_decision_tree::reset_out_edges_aux): Likewise.
8838 2018-06-20 Martin Liska <mliska@suse.cz>
8840 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
8841 (hoist_edge_and_branch_if_true): Likewise.
8842 (expand_switch_using_bit_tests_p): Likewise.
8843 (struct case_bit_test): Likewise.
8844 (case_bit_test_cmp): Likewise.
8845 (emit_case_bit_tests): Likewise.
8846 (switch_conversion::switch_conversion): New class.
8847 (struct switch_conv_info): Remove old struct.
8848 (collect_switch_conv_info): More to ...
8849 (switch_conversion::collect): ... this.
8850 (check_range): Likewise.
8851 (switch_conversion::check_range): Likewise.
8852 (check_all_empty_except_final): Likewise.
8853 (switch_conversion::check_all_empty_except_final): Likewise.
8854 (check_final_bb): Likewise.
8855 (switch_conversion::check_final_bb): Likewise.
8856 (create_temp_arrays): Likewise.
8857 (switch_conversion::create_temp_arrays): Likewise.
8858 (free_temp_arrays): Likewise.
8859 (gather_default_values): Likewise.
8860 (switch_conversion::gather_default_values): Likewise.
8861 (build_constructors): Likewise.
8862 (switch_conversion::build_constructors): Likewise.
8863 (constructor_contains_same_values_p): Likewise.
8864 (switch_conversion::contains_same_values_p): Likewise.
8865 (array_value_type): Likewise.
8866 (switch_conversion::array_value_type): Likewise.
8867 (build_one_array): Likewise.
8868 (switch_conversion::build_one_array): Likewise.
8869 (build_arrays): Likewise.
8870 (switch_conversion::build_arrays): Likewise.
8871 (gen_def_assigns): Likewise.
8872 (switch_conversion::gen_def_assigns): Likewise.
8873 (prune_bbs): Likewise.
8874 (switch_conversion::prune_bbs): Likewise.
8875 (fix_phi_nodes): Likewise.
8876 (switch_conversion::fix_phi_nodes): Likewise.
8877 (gen_inbound_check): Likewise.
8878 (switch_conversion::gen_inbound_check): Likewise.
8879 (process_switch): Use the newly created class.
8880 (switch_conversion::expand): New.
8881 (switch_conversion::~switch_conversion): New.
8882 * tree-switch-conversion.h: New file.
8884 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8886 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
8887 tree-vect-patterns.c.
8888 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
8889 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
8890 (vect_recog_sad_pattern): Likewise.
8891 (vect_recog_widen_sum_pattern): Likewise.
8892 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
8893 (vect_recog_widen_shift_pattern): Remove the type_in argument.
8894 (vect_recog_rotate_pattern): Likewise.
8895 (vect_recog_mult_pattern): Likewise.
8896 (vect_recog_vector_vector_shift_pattern): Likewise.
8897 (vect_recog_divmod_pattern): Likewise.
8898 (vect_recog_mixed_size_cond_pattern): Likewise.
8899 (vect_recog_bool_pattern): Likewise.
8900 (vect_recog_mask_conversion_pattern): Likewise.
8901 (vect_try_gather_scatter_pattern): Likewise.
8902 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
8903 (vect_recog_over_widening_pattern): Likewise.
8904 (vect_recog_gather_scatter_pattern): Likewise.
8905 (vect_recog_func_ptr): Move from tree-vectorizer.h
8906 (vect_vect_recog_func_ptrs): Move further down the file.
8907 (vect_recog_func): Likewise. Remove the third argument.
8908 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
8909 (vect_pattern_recog_1): Expect the pattern function to do any
8910 necessary target tests. Also expect it to provide a vector type.
8911 Remove the type_in handling.
8913 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8915 * tree-vect-patterns.c (vect_pattern_detected): New function.
8916 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
8917 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
8918 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
8919 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
8920 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
8921 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
8922 (vect_recog_mask_conversion_pattern)
8923 (vect_try_gather_scatter_pattern): Likewise.
8925 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8927 * tree-vect-patterns.c (vect_get_internal_def): New function.
8928 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8929 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
8930 (search_type_for_mask_1): Use it.
8932 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8934 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8935 redundant WIDEN_SUM_EXPR handling.
8936 (vect_recog_sad_pattern): Likewise.
8938 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8940 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8941 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
8942 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
8943 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
8944 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
8946 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8948 * tree-vect-stmts.c (vectorizable_call): Make sure that we
8949 use the stmt_vec_info of the original bb statement for the
8950 new zero assignment, even if the call is part of a pattern.
8952 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8954 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
8955 that the sequence is attached to the original statement rather
8956 than the pattern statement.
8957 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
8958 PATTERN_DEF_SEQ from the original statement rather than
8959 the main pattern statement.
8960 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
8961 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
8962 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
8964 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8966 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
8967 definition statements before the early exit for statements that aren't
8969 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
8971 (vect_transform_loop): ...here. Process pattern definition
8972 statements without first checking whether the main pattern
8973 statement is live or relevant.
8975 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
8977 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
8978 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
8980 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
8982 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
8983 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
8984 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
8985 (expand_block_compare): Change select_block_compare_mode call.
8986 (expand_strncmp_align_check): Use new functions, fix comment.
8987 (emit_final_str_compare_gpr): New function.
8988 (expand_strn_compare): Refactor and clean up code.
8989 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
8991 2018-06-19 Tony Reix <tony.reix@atos.com>
8992 Damien Bergamini <damien.bergamini@atos.com>
8993 David Edelsohn <dje.gcc@gmail.com>
8995 * collect2.c (static_obj): New variable.
8996 (static_libs): New variable.
8997 (is_in_list): Uncomment declaration.
8998 (main): Track AIX libraries linked statically.
8999 (is_in_list): Uncomment definition.
9000 (scan_prog_file): Don't add AIX shared libraries initializer
9001 to constructor list if linking statically.
9003 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
9005 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
9007 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
9009 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
9011 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
9014 2018-06-19 Martin Liska <mliska@suse.cz>
9016 * config/i386/i386.c (ix86_can_inline_p): Do not use
9017 ipa_fn_summaries::get_create.
9018 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
9020 (devirtualization_time_bonus): Likewise.
9021 (ipcp_propagate_stage): Likewise.
9022 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
9023 (edge_set_predicate): Likewise.
9024 (evaluate_conditions_for_known_args): Likewise.
9025 (evaluate_properties_for_edge): Likewise.
9026 (ipa_call_summary::reset): Tranform to ...
9027 (ipa_call_summary::~ipa_call_summary): ... this.
9028 (ipa_fn_summary::reset): Transform to ...
9029 (ipa_fn_summary::~ipa_fn_summary): ... this.
9030 (ipa_fn_summary_t::remove): Rename to ...
9031 (ipa_fn_summary_t::remove_callees): ... this.
9032 (ipa_fn_summary_t::duplicate): Use placement new
9033 instead of memory copy.
9034 (ipa_call_summary_t::duplicate): Likewise.
9035 (ipa_call_summary_t::remove): Remove.
9036 (dump_ipa_call_summary): Change get_create to get.
9037 (ipa_dump_fn_summary): Dump only when summary exists.
9038 (analyze_function_body): Use symbol_summary::get instead
9040 (compute_fn_summary): Likewise.
9041 (estimate_edge_devirt_benefit): Likewise.
9042 (estimate_edge_size_and_time): Likewise.
9043 (inline_update_callee_summaries): Likewise.
9044 (remap_edge_change_prob): Likewise.
9045 (remap_edge_summaries): Likewise.
9046 (ipa_merge_fn_summary_after_inlining): Likewise.
9047 (write_ipa_call_summary): Likewise.
9048 (ipa_fn_summary_write): Likewise.
9049 (ipa_free_fn_summary): Likewise.
9050 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
9051 (struct ipa_call_summary): Likewise.
9052 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
9054 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
9055 (estimate_size_after_inlining): Likewise.
9056 (estimate_growth): Likewise.
9057 (growth_likely_positive): Likewise.
9058 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
9059 (inline_call): Likewise.
9060 * ipa-inline.c (caller_growth_limits): Likewise.
9061 (can_inline_edge_p): Likewise.
9062 (can_inline_edge_by_limits_p): Likewise.
9063 (compute_uninlined_call_time): Likewise.
9064 (compute_inlined_call_time): Likewise.
9065 (want_inline_small_function_p): Likewise.
9066 (edge_badness): Likewise.
9067 (update_caller_keys): Likewise.
9068 (update_callee_keys): Likewise.
9069 (inline_small_functions): Likewise.
9070 (inline_to_all_callers_1): Likewise.
9071 (dump_overall_stats): Likewise.
9072 (early_inline_small_functions): Likewise.
9073 (early_inliner): Likewise.
9074 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
9075 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9076 * ipa-pure-const.c (malloc_candidate_p): Likewise.
9077 * ipa-split.c (execute_split_functions): Likewise.
9078 * symbol-summary.h: Likewise.
9079 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
9081 2018-06-19 Richard Biener <rguenther@suse.de>
9083 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
9084 (vectorize_loops): ... here. Fix dbgcnt handling.
9085 (try_vectorize_loop): Wrap try_vectorize_loop_1.
9087 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
9090 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
9091 ieee128 argument takes up only one (vector) register, not two (floating
9094 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
9096 * gimplify.c (gimplify_init_constructor): Really never clear for an
9097 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
9099 2018-06-19 Richard Biener <rguenther@suse.de>
9101 PR tree-optimization/86179
9102 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
9103 after failed recognition.
9105 2018-06-18 Martin Sebor <msebor@redhat.com>
9108 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
9109 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
9110 Handle integer subtraction.
9111 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
9112 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
9114 2018-06-18 David Malcolm <dmalcolm@redhat.com>
9116 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
9117 param from rtx to rtx_insn *.
9118 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
9120 (frv_ifcvt_modify_insn): Likwise.
9121 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
9122 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
9123 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
9124 as_a <rtx_insn *> cast to local "unprotected_region" once
9125 it's been established that it's not NULL or pc_rtx.
9126 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
9127 param "sethi" from rtx to rtx_insn *.
9128 (nds32_group_float_insns): Likewise for param "insn".
9129 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
9131 (vax_output_int_subtract): Likewise.
9132 * config/vax/vax.c (vax_output_int_add): Likewise for param
9134 (vax_output_int_subtract): Likewise.
9135 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
9136 (emit_pattern_after): Likewise for param "after".
9137 (emit_insn_after): Likewise.
9138 (emit_jump_insn_after): Likewise.
9139 (emit_call_insn_after): Likewise.
9140 (emit_debug_insn_after): Likewise.
9141 (emit_pattern_before): Likewise for param "before".
9142 (emit_insn_before): Likewise.
9143 (emit_jump_insn_before): Likewise.
9144 * final.c (get_insn_template): Likewise for param "insn", removing
9146 * output.h (get_insn_template): Likewise for 2nd param.
9147 * rtl.h (emit_insn_before): Likewise.
9148 (emit_jump_insn_before): Likewise.
9149 (emit_debug_insn_before_noloc): Likewise.
9150 (emit_insn_after): Likewise.
9151 (emit_jump_insn_after): Likewise.
9152 (emit_call_insn_after): Likewise.
9153 (emit_debug_insn_after): Likewise.
9154 (set_insn_deleted): Likewise for param.
9156 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
9159 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
9160 floating point modes, so that IFmode is numerically greater than
9161 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
9162 to declare the ordering. This prevents IFmode from being
9163 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
9164 machine. Include rs6000-modes.h to share the fractional values
9165 between genmodes* and the rest of the compiler.
9169 * config/rs6000/rs6000-modes.h: New file.
9170 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
9171 meaning of rs6000_long_double_size so that 126..128 selects an
9172 appropriate 128-bit floating point type.
9173 (rs6000_option_override_internal): Likewise.
9174 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
9175 (TARGET_LONG_DOUBLE_128): Change the meaning of
9176 rs6000_long_double_size so that 126..128 selects an appropriate
9177 128-bit floating point type.
9178 (LONG_DOUBLE_TYPE_SIZE): Update comment.
9179 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
9180 source and destination to match the standard usage.
9181 (truncifkf2): Likewise.
9182 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
9183 ISA 2.07 to use an explicit clobber, instead of passing in a
9185 (copysign<mode>3_soft): Likewise.
9187 2018-06-18 David Malcolm <dmalcolm@redhat.com>
9189 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
9190 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
9191 (vect_slp_analyze_instance_dependence): Likewise.
9192 (vect_enhance_data_refs_alignment): Likewise.
9193 (vect_analyze_data_refs_alignment): Likewise.
9194 (vect_slp_analyze_and_verify_instance_alignment
9195 (vect_analyze_data_ref_accesses): Likewise.
9196 (vect_prune_runtime_alias_test_list): Likewise.
9197 (vect_analyze_data_refs): Likewise.
9198 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
9199 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9200 (vect_analyze_scalar_cycles_1): Likewise.
9201 (vect_get_loop_niters): Likewise.
9202 (vect_analyze_loop_form_1): Likewise.
9203 (vect_update_vf_for_slp): Likewise.
9204 (vect_analyze_loop_operations): Likewise.
9205 (vect_analyze_loop): Likewise.
9206 (vectorizable_induction): Likewise.
9207 (vect_transform_loop): Likewise.
9208 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
9209 * tree-vect-slp.c (vect_analyze_slp): Likewise.
9210 (vect_make_slp_decision): Likewise.
9211 (vect_detect_hybrid_slp): Likewise.
9212 (vect_slp_analyze_operations): Likewise.
9213 (vect_slp_bb): Likewise.
9214 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
9215 (vectorizable_bswap): Likewise.
9216 (vectorizable_call): Likewise.
9217 (vectorizable_simd_clone_call): Likewise.
9218 (vectorizable_conversion): Likewise.
9219 (vectorizable_assignment): Likewise.
9220 (vectorizable_shift): Likewise.
9221 (vectorizable_operation): Likewise.
9222 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
9224 2018-06-18 Martin Sebor <msebor@redhat.com>
9226 PR tree-optimization/81384
9227 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
9228 * builtins.c (expand_builtin_strnlen): New function.
9229 (expand_builtin): Call it.
9230 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
9231 * builtins.def (BUILT_IN_STRNLEN): New.
9232 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
9233 Warn for bounds in excess of maximum object size.
9234 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
9235 single-value ranges. Handle strnlen.
9236 (handle_builtin_strlen): Handle strnlen.
9237 (strlen_check_and_optimize_stmt): Same.
9238 * doc/extend.texi (Other Builtins): Document strnlen.
9240 2018-06-18 Maya Rashish <coypu@sdf.org>
9242 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
9243 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
9244 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
9246 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
9248 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
9250 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9252 * tree.c (escaped_string::escape): Replace cast to char * by
9253 const_cast<char *> (unescaped).
9255 2018-06-18 Nick Clifton <nickc@redhat.com>
9258 * tree.c (escaped_string): New class. Converts an unescaped
9259 string into its escaped equivalent.
9260 (warn_deprecated_use): Use the new class to convert the
9261 deprecation message, if present.
9262 (test_escaped_strings): New self test.
9263 (test_c_tests): Add test_escaped_strings.
9264 * doc/extend.texi (deprecated): Add a note that the
9265 deprecation message is affected by the -fmessage-length
9266 option, and that control characters will be escaped.
9267 (#pragma GCC error): Document this pragma.
9268 (#pragma GCC warning): Likewise.
9269 * doc/invoke.texi (-fmessage-length): Document this option's
9270 effect on the #warning and #error preprocessor directives and
9271 the deprecated attribute.
9273 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
9275 * tree.c (decl_value_expr_lookup): Revert latest change.
9276 (decl_value_expr_insert): Likewise.
9278 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
9280 * gimplify.c (nonlocal_vlas): Delete.
9281 (nonlocal_vla_vars): Likewise.
9282 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
9284 (gimplify_body): Do not create and destroy nonlocal_vlas.
9285 * tree-nested.c: Include diagnostic.h.
9286 (use_pointer_in_frame): Tweak.
9287 (lookup_field_for_decl): Add assertion and declare the transformation.
9288 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
9289 internal error when the reference is in a wrong context. Do not
9290 create a debug decl by default.
9291 (note_nonlocal_block_vlas): Delete.
9292 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
9293 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
9294 create a debug decl by default.
9295 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
9296 call is in a wrong context.
9297 (fixup_vla_decls): New function.
9298 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
9299 debug variables were created.
9300 * tree.c (decl_value_expr_lookup): Add checking assertion.
9301 (decl_value_expr_insert): Likewise.
9303 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
9306 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
9307 * tree-scalar-evolution.c (interpret_expr): Likewise.
9308 (expression_expensive_p): Likewise.
9309 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
9310 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
9311 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
9312 (ssa_defined_by_minus_one_stmt_p): New.
9314 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
9317 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
9318 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
9319 * dojump.c (do_jump): Likewise.
9320 * expr.c (expand_expr_real_2): Check operand type's sign.
9321 * fold-const.c (const_unop): Handle ABSU_EXPR.
9322 (fold_abs_const): Likewise.
9323 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
9324 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
9325 (strip_sign_op_1): Likesise.
9326 * match.pd: Add new pattern to generate ABSU_EXPR.
9327 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
9328 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
9329 * tree-eh.c (operation_could_trap_helper_p): Likewise.
9330 * tree-inline.c (estimate_operator_cost): Likewise.
9331 * tree-pretty-print.c (dump_generic_node): Likewise.
9332 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
9333 * tree.def (ABSU_EXPR): New.
9335 2018-06-16 Jakub Jelinek <jakub@redhat.com>
9338 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
9339 documented as preserved for backward compatibility only.
9340 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
9342 PR rtl-optimization/86108
9343 * bb-reorder.c (create_forwarder_block): Renamed to ...
9344 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
9345 jump from new landing pad to the second part.
9346 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
9349 2018-06-15 Jakub Jelinek <jakub@redhat.com>
9352 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
9353 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
9354 Only call store_expr for halves if the mode is the same.
9357 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
9358 Fix up comment formatting.
9360 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
9362 * typed-splay-tree.h (typed_splay_tree::remove): New function.
9363 (typed_splay_tree::closure,
9364 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
9365 (typed_splay_tree::typed_splay_tree,
9366 typed_splay_tree::operator =): Declared private.
9367 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
9368 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
9369 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
9370 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
9371 typed_splay_tree::splay_tree_splay,
9372 typed_splay_tree::splay_tree_foreach_helper,
9373 typed_splay_tree::splay_tree_insert,
9374 typed_splay_tree::splay_tree_remove,
9375 typed_splay_tree::splay_tree_lookup,
9376 typed_splay_tree::splay_tree_predecessor,
9377 typed_splay_tree::splay_tree_successor,
9378 typed_splay_tree::splay_tree_min,
9379 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
9380 (typed_splay_tree::root, typed_splay_tree::comp,
9381 typed_splay_tree::delete_key,
9382 typed_splay_tree::delete_value): New data members.
9383 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
9384 typed_splay_tree::remove.
9386 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
9388 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
9389 -mginv and -mno-ginv to the assembler.
9390 * config/mips/mips.opt (-mcrc): New option.
9392 * doc/invoke.text (-mcrc): Document.
9395 2018-06-15 Nick Clifton <nickc@redhat.com>
9398 * tree.c (escaped_string): New class. Converts an unescaped
9399 string into its escaped equivalent.
9400 (warn_deprecated_use): Use the new class to convert the
9401 deprecation message, if present.
9402 (test_escaped_strings): New self test.
9403 (test_c_tests): Add test_escaped_strings.
9404 * doc/extend.texi (deprecated): Add a note that the
9405 deprecation message is affected by the -fmessage-length
9406 option, and that control characters will be escaped.
9407 (#pragma GCC error): Document this pragma.
9408 (#pragma GCC warning): Likewise.
9409 * doc/invoke.texi (-fmessage-length): Document this option's
9410 effect on the #warning and #error preprocessor directives and
9411 the deprecated attribute.
9413 2018-06-15 Richard Biener <rguenther@suse.de>
9415 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
9416 here, also noting vector size used.
9417 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
9418 size used in MSG_OPTIMIZED_LOCATIONS dump.
9419 (pass_slp_vectorize::execute): Adjust.
9421 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
9424 * config/arc/arc.c (arc_return_address_register): Fix
9427 2018-06-15 Richard Biener <rguenther@suse.de>
9430 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
9431 leave useless conversion stripping to force_gimple_operand_gsi.
9432 (gimplify_build2): Likewise.
9433 (gimplify_build1): Likewise.
9435 2018-06-15 Richard Biener <rguenther@suse.de>
9438 * tree-cfg.c (move_stmt_op): unshare invariant addresses
9439 before adjusting their block.
9441 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
9443 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
9444 multilibs for *-*-rtems*.
9445 * config/riscv/t-rtems: New file.
9447 2018-06-14 Jakub Jelinek <jakub@redhat.com>
9450 * match.pd ((A +- CST1) +- CST2): Punt if last resort
9451 unsigned_type_for returns NULL.
9454 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
9455 subregs of multi-word pseudos unless the float mode has word size.
9457 2018-06-14 Richard Biener <rguenther@suse.de>
9460 * tree-vect-generic.c (build_word_mode_vector_type): Remove
9461 duplicate and harmful type_hash_canon.
9462 * tree.c (type_hash_canon): Assert we didn't find ourselves.
9464 2018-06-14 Richard Biener <rguenther@suse.de>
9467 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
9470 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
9472 * config/rtems.h (STDINT_LONG32): Define.
9474 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
9475 Prachi Godbole <prachi.godbole@imgtec.com>
9477 * config/mips/mips-cpus.def: Define P6600.
9478 * config/mips/mips-tables.opt: Regenerate.
9479 * config/mips/mips.c (mips_ucbranch_type): New enum.
9480 (mips_rtx_cost_data): Add support for P6600.
9481 (mips_issue_rate): Likewise.
9482 (mips_multipass_dfa_lookahead): Likewise.
9483 (mips_avoid_hazard): Likewise.
9484 (mips_reorg_process_insns): Likewise.
9485 (mips_classify_branch_p6600): New function.
9486 * config/mips/mips.h (TUNE_P6600): New define.
9487 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
9488 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
9489 * config/mips/mips.md: Include p6600.md.
9490 (processor): Add p6600.
9491 * config/mips/p6600.md: New file.
9492 * doc/invoke.texi: Add p6600 to supported architectures.
9494 2018-06-13 Martin Sebor <msebor@redhat.com>
9496 PR tree-optimization/86114
9497 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
9499 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
9501 2018-06-13 Richard Biener <rguenther@suse.de>
9503 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
9504 Properly set vector type of the intermediate stmt.
9505 * tree-vect-stmts.c (vectorizable_operation): The destination
9506 var always has vectype_out type.
9508 2018-06-13 Jeff Law <law@redhat.com>
9510 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
9511 integer 0 for argument to print_rtl_with_bb.
9512 (rl78_reorg): Likewise.
9514 2018-06-13 David Malcolm <dmalcolm@redhat.com>
9516 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
9517 from rtx to rtx_insn *.
9518 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
9520 (add_sched_insns_for_speculation): Likewise for local "target",
9521 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9522 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
9523 from rtx_insn ** to rtx_code_label **.
9524 (reorg_emit_nops): Likewise.
9525 (c6x_reorg): Likewise for local "call_labels".
9526 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
9528 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
9529 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
9530 the loops over LABEL_REFS.
9531 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
9533 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9534 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
9535 (split_branches): Strengthen local "olabel" from rtx to
9536 rtx_insn *, adding a safe_as_a cast.
9537 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
9539 (add_insn_after): Likewise for first two params.
9540 (add_insn_before): Likewise.
9541 (remove_insn): Likewise for param.
9542 (emit_pattern_before_noloc): Likewise for second and third params.
9543 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
9544 (emit_call_insn_before_noloc): Likewise.
9545 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
9547 (emit_barrier_before): Likewise.
9548 (emit_label_before): Strengthen "label" param from "rtx" to
9549 "rtx_code_label *". Strengthen "before" param from "rtx" to
9551 (emit_insn_after_1): Strengthen "after" param from "rtx" to
9553 (emit_pattern_after_noloc): Likewise.
9554 (emit_insn_after_noloc): Likewise.
9555 (emit_jump_insn_after_noloc): Likewise.
9556 (emit_call_insn_after_noloc): Likewise.
9557 (emit_debug_insn_after_noloc): Likewise.
9558 (emit_barrier_after): Likewise.
9559 (emit_label_after): Likewise for both params.
9560 (emit_pattern_after_setloc): Likewise for "after" param. Convert
9561 "loc" param from "int" to "location_t".
9562 (emit_insn_after_setloc): Likewise.
9563 (emit_jump_insn_after_setloc): Likewise.
9564 (emit_call_insn_after_setloc): Likewise.
9565 (emit_debug_insn_after_setloc): Likewise.
9566 (emit_pattern_before_setloc): Likewise for "before" param. Convert
9567 "loc" param from "int" to "location_t".
9568 (emit_pattern_before): Convert NULL_RTX to NULL.
9569 (emit_insn_before_setloc): Convert "loc" param from "int" to
9571 (emit_jump_insn_before_setloc): Likewise.
9572 (emit_call_insn_before_setloc): Likewise.
9573 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
9574 rtx_insn *. Convert "loc" param from "int" to "location_t".
9575 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
9576 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
9577 Convert 3rd param from "int" to "location_t".
9578 (emit_barrier_before, emit_barrier_after, next_real_insn):
9579 Strengthen param from rtx to rtx_insn *.
9580 (emit_label_before): Strengthen 1st param from "rtx" to
9581 "rtx_code_label *". Strengthen 2nd param from "rtx" to
9583 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
9584 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
9585 Strengthen 2nd param from "rtx" to "rtx_insn *".
9586 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
9587 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
9588 Likewise. Convert 3rd param from "int" to "location_t".
9589 (emit_label_after): Strengthen 1st param from "rtx" to
9591 (next_real_insn, remove_insn): Strengthen param from "rtx" to
9593 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
9594 from "rtx" to "rtx_insn *".
9596 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
9598 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
9599 bodies streamed in with -Q.
9600 * dumpfile.c (dump_files): Add lto-stream-out dump file.
9601 * dumpfile.h (tree_dump_index): Add lto_stream_out.
9602 * gimple-streamer-out.c: Include gimple-pretty-print.h
9603 (output_bb): Dump stmts streamed.
9604 * lto-section-out.c: Include print-tree.h
9605 (lto_begin_section): Dump sections created.
9606 (lto_output_decl_index): Dump decl encoded.
9607 * lto-streamer-out.c: Include print-tree.h
9608 (create_output_block): Dump output block created.
9609 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
9610 (output_function): Dump function output.
9611 (output_constructor): Dump constructor streamed.
9612 (write_global_stream): Output indexes encoded.
9613 (produce_asm_for_decls): Dump streams encoded.
9614 * lto-streamer.c (streamer_dump_file): New global var.
9615 * lto-streamer.h (streamer_dump_file): Declare.
9616 * passes.c (ipa_write_summaries): Initialize streamer dump.
9617 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
9620 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
9623 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
9624 offsets for register save directives. Emit a second batch of save
9625 directives, if need be, when the function accesses prior frames.
9627 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9629 * config/arc/fpu.md (fmasf4): Force operand to register.
9630 (fnmasf4): Likewise.
9632 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9634 * config/arc/arc-protos.h (arc_pad_return): Remove.
9635 * config/arc/arc.c (machine_function): Remove force_short_suffix
9637 (arc_print_operand): Adjust printing of '&'.
9638 (arc_verify_short): Remove conditional printing of short suffix.
9639 (arc_final_prescan_insn): Remove reference to size_reason.
9640 (pad_return): New function.
9641 (arc_reorg): Call pad_return.
9642 (arc_pad_return): Remove.
9643 (arc_init_machine_status): Remove reference to force_short_suffix.
9644 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
9645 (attr length): When attribute iscompact is true force to 2
9646 regardless; in the case of maybe check if we want to force the
9647 instruction to have 4 bytes length.
9648 (nopv): Change it to generate 4 byte long nop as well.
9649 (blockage): New pattern.
9650 (simple_return): Remove call to arc_pad_return.
9651 (p_return_i): Likewise.
9653 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9655 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
9657 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9659 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
9662 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9664 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
9665 for ARC700 and ARCv2.
9667 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
9670 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
9671 operands[2] instead of operands[1].
9674 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
9676 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
9677 case, check whether the outer register overlaps an unallocatable
9678 register, not just whether it fits the required class.
9680 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
9682 * poly-int.h (can_div_trunc_p): Add new overload in which all values
9684 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
9685 (memrefs_conflict_p): Likewise.
9686 (init_alias_analysis): Likewise.
9687 * cfgexpand.c (expand_debug_expr): Likewise.
9688 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
9689 * cse.c (fold_rtx): Likewise.
9690 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
9691 * expr.c (emit_block_move_hints): Likewise.
9692 (clear_storage_hints, push_block, emit_push_insn): Likewise.
9693 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
9694 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
9695 (emit_group_store): Likewise.
9696 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
9697 to read the PRE/POST_MODIFY increment.
9698 * calls.c (store_one_arg): Use strip_offset.
9699 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
9701 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
9703 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
9704 (simplify_binary_operation_1): Extend CONST_INT handling to
9706 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
9707 than a HOST_WIDE_INT.
9708 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
9710 (adjust_mems, add_stores): Update accodingly.
9711 (vt_canonicalize_addr): Track polynomial offsets.
9712 (emit_note_insn_var_location): Likewise.
9713 (vt_add_function_parameter): Likewise.
9714 (vt_initialize): Likewise.
9716 2018-06-12 Jeff Law <law@redhat.com>
9718 * config.gcc (alpha*-*-freebsd*): Remove.
9719 * config/alpha/freebsd.h: Remove.
9721 2018-06-12 David Malcolm <dmalcolm@redhat.com>
9724 * spellcheck-tree.c (levenshtein_distance): Rename to...
9725 (get_edit_distance): ...this, and update for underlying renaming.
9726 * spellcheck-tree.h (levenshtein_distance): Rename to...
9727 (get_edit_distance): ...this.
9728 * spellcheck.c (levenshtein_distance): Rename to...
9729 (get_edit_distance): ...this. Convert from Levenshtein distance
9730 to Damerau-Levenshtein distance by supporting transpositions of
9731 adjacent characters. Rename "v1" to "v_next" and "v0" to
9733 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
9734 (selftest::test_edit_distance_unit_test_oneway): ...this, and
9735 update for underlying renaming.
9736 (selftest::levenshtein_distance_unit_test): Rename to...
9737 (selftest::test_get_edit_distance_unit): ...this, and update for
9738 underlying renaming.
9739 (selftest::test_find_closest_string): Add example from PR 69968
9740 where transposition helps
9741 (selftest::test_metric_conditions): Update for renaming.
9742 (selftest::test_metric_conditions): Likewise.
9743 (selftest::spellcheck_c_tests): Likewise.
9744 * spellcheck.h (levenshtein_distance): Rename both overloads to...
9745 (get_edit_distance): ...this.
9746 (best_match::consider): Update for renaming.
9748 2018-06-12 Martin Sebor <msebor@redhat.com>
9750 PR tree-optimization/85259
9751 * builtins.c (compute_objsize): Handle constant offsets.
9752 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
9753 true iff a warning has been issued.
9754 * gimple.h (gimple_nonartificial_location): New function.
9755 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
9756 gimple_nonartificial_location and handle -Wno-system-headers.
9757 (handle_builtin_stxncpy): Same.
9759 2018-06-12 Martin Sebor <msebor@redhat.com>
9762 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
9764 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
9766 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
9767 BUILTIN_VEC_XST entries for pointer to double and long long.
9769 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
9772 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9774 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9777 2018-06-12 Martin Liska <mliska@suse.cz>
9779 * doc/options.texi: Document IntegerRange.
9781 2018-06-12 Martin Liska <mliska@suse.cz>
9783 * config/i386/i386.opt: Make MPX-related options as Deprecated.
9784 * opt-functions.awk: Handle Deprecated flag.
9785 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
9787 (read_cmdline_option): Report warning for a deprecated option.
9788 * opts.h (struct cl_option): Add new field cl_deprecated.
9789 (CL_ERR_DEPRECATED): New.
9791 2018-06-12 Martin Liska <mliska@suse.cz>
9793 * doc/options.texi: Document Deprecated option flag.
9795 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9797 * config/arc/arc-arch.h (arc_extras): New enum.
9798 (arc_cpu_t):Add field extra.
9799 (arc_cpu_types): Consider the extras.
9800 * config/arc/arc-cpus.def: Add extras info.
9801 * config/arc/arc-opts.h (processor_type): Consider extra field.
9802 * config/arc/arc.c (arc_override_options): Handle extra field.
9804 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9806 * config/arc/arc-arch.h: Update ARC_OPTX macro.
9807 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
9809 * config/arc/arc.c (arc_init): Update pic warning.
9810 (irq_range): Update irq range parsing warnings.
9811 (arc_override_options): Update various warning messages.
9812 (arc_handle_aux_attribute): Likewise.
9814 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
9816 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
9818 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9820 * doc/sourcebuild.texi: Document usage of line number 0 in verify
9821 compiler messages directives.
9823 2018-06-12 Matthew Fortune <mfortune@gmail.com>
9825 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
9826 * config/mips/mips-tables.opt: Regenerate.
9827 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
9829 * doc/invoke.texi: Document -march=i6500.
9831 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
9833 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
9834 (i6400_gpmul): Add cpu_unit.
9835 (i6400_gpdiv): Likewise.
9836 (i6400_msa_add_d): Update reservations.
9837 (i6400_msa_int_add) Likewise.
9838 (i6400_msa_short_logic3) Likewise.
9839 (i6400_msa_short_logic2) Likewise.
9840 (i6400_msa_short_logic) Likewise.
9841 (i6400_msa_move) Likewise.
9842 (i6400_msa_cmp) Likewise.
9843 (i6400_msa_short_float2) Likewise.
9844 (i6400_msa_div_d) Likewise.
9845 (i6400_msa_long_logic1) Likewise.
9846 (i6400_msa_long_logic2) Likewise.
9847 (i6400_msa_mult) Likewise.
9848 (i6400_msa_long_float2) Likewise.
9849 (i6400_msa_long_float4) Likewise.
9850 (i6400_msa_long_float5) Likewise.
9851 (i6400_msa_long_float8) Likewise.
9852 (i6400_fpu_fadd): Include frint type.
9853 (i6400_fpu_store): New define_insn_reservation.
9854 (i6400_fpu_load): Likewise.
9855 (i6400_fpu_move): Likewise.
9856 (i6400_fpu_fcmp): Likewise.
9857 (i6400_fpu_fmadd): Likewise.
9858 (i6400_int_mult): Include imul3nc type and update reservation.
9859 (i6400_int_div): Include idiv3 type and update reservation.
9860 (i6400_int_load): Update to check type not move_type.
9861 (i6400_int_store): Likewise.
9862 (i6400_int_prefetch): Set zero latency.
9864 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
9866 * gcc.c: Document new %@{...} sequence.
9867 (LINK_COMMAND_SPEC): Use it for the -L switches.
9868 (cpp_unique_options): Use it for the -I switches.
9869 (at_file_argbuf): New global variable.
9870 (in_at_file): Likewise.
9871 (alloc_args): Create at_file_argbuf.
9872 (clear_args): Truncate at_file_argbuf.
9873 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
9874 (open_at_file): New function.
9875 (close_at_file): Likewise.
9876 (create_at_file): Delete.
9877 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
9880 (validate_switches_from_spec): Deal with %@{...} sequence.
9881 (validate_switches): Likewise.
9882 (driver::finalize): Call clear_args.
9884 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
9886 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
9888 2018-06-11 Martin Sebor <msebor@redhat.com>
9890 * doc/invoke.texi (-Wall): List -Wc++17-compat.
9891 (Wno-class-memaccess): Add @opindex.
9892 (Wno-templates, Wno-multiple-inheritance): Same.
9893 (Wno-virtual-inheritance, Wno-namespaces): Same.
9894 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
9895 (Wno-format-overflow, Wno-format-truncation): Same.
9896 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
9897 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
9898 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
9899 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
9900 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
9901 (Wno-misspelled-isr): Same.
9903 2018-06-11 Martin Sebor <msebor@redhat.com>
9905 * PR tree-optimization/86083
9906 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
9908 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
9910 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
9912 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
9915 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
9916 on the correct operand.
9917 (*movdi_internal64): Ditto.
9919 2018-06-11 Martin Liska <mliska@suse.cz>
9921 PR tree-optimization/86089
9922 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
9924 2018-06-11 Julia Koval <julia.koval@intel.com>
9926 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
9927 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
9928 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
9930 2018-06-11 Olivier Hainque <hainque@adacore.com>
9932 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
9933 for Ada with strict dwarf2.
9935 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
9938 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
9941 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
9943 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
9945 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
9947 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
9950 2018-06-08 Martin Liska <mliska@suse.cz>
9952 * tree-cfg.h (debug_function): Fix argument type to match
9955 2018-06-08 Martin Liska <mliska@suse.cz>
9957 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
9958 Remove usage of MPX-related (and removed) fields.
9959 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
9961 2018-06-08 David Malcolm <dmalcolm@redhat.com>
9963 * cfg.c (debug): Use TDF_NONE rather than 0.
9964 * cfghooks.c (debug): Likewise.
9965 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
9966 (struct dump_option_value_info): Convert to...
9967 (struct kv_pair): ...this template type.
9968 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
9970 (optinfo_verbosity_options): Likewise.
9971 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
9973 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
9974 than int for "optgroup_flags" param.
9975 (dump_generic_expr_loc): Use dump_flags_t rather than int for
9977 (dump_dec): Likewise.
9978 (dump_finish): Use TDF_NONE rather than 0.
9979 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
9980 rather than int for "optgroup_flags" param. Use TDF_NONE rather
9981 than 0. Update for change to option_ptr.
9982 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
9983 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
9984 0. Update for changes to optinfo_verbosity_options and
9986 (opt_info_switch_p): Convert optgroup_flags from int to
9988 (dump_basic_block): Use dump_flags_t rather than int
9989 for "dump_kind" param.
9990 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
9991 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
9992 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
9993 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
9994 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
9995 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
9996 TDF_NONE): Convert from macros to...
9997 (enum dump_flag): ...this new enum.
9998 (dump_flags_t): Update to use enum.
9999 (operator|, operator&, operator~, operator|=, operator&=):
10000 Implement for dump_flags_t.
10001 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
10002 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
10003 Convert from macros to...
10004 (enum optgroup_flag): ...this new enum.
10005 (optgroup_flags_t): New typedef.
10006 (operator|, operator|=): Implement for optgroup_flags_t.
10007 (struct dump_file_info): Convert field "alt_flags" to
10008 dump_flags_t. Convert field "optgroup_flags" to
10010 (dump_basic_block): Use dump_flags_t rather than int for param.
10011 (dump_generic_expr_loc): Likewise.
10012 (dump_dec): Likewise.
10013 (dump_register): Convert param "optgroup_flags" to
10015 (opt_info_enable_passes): Likewise.
10016 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
10018 * gimple-pretty-print.c (debug): Likewise.
10019 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
10020 (merged_store_group::apply_stores): Likewise.
10021 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
10022 * gimple.c (verify_gimple_pp): Likewise.
10023 * graphite-poly.c (print_pbb_body): Likewise.
10024 * passes.c (pass_manager::register_one_dump_file): Convert
10025 local "optgroup_flags" to optgroup_flags_t.
10026 * print-tree.c (print_node): Use TDF_NONE rather than 0.
10028 (debug_body): Likewise.
10029 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
10030 to optgroup_flags_t.
10031 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
10033 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
10034 (convert_mult_to_fma): Likewise.
10035 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
10036 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
10037 * tree-vect-data-refs.c (dump_lower_bound): Convert param
10038 "dump_kind" to dump_flags_t.
10040 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
10042 * config/rs6000/rs6000.c (min, max): Delete.
10044 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
10046 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
10047 -mabi=spe and -mabi=no-spe.
10049 2018-06-08 Martin Liska <mliska@suse.cz>
10051 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
10052 where we expect an existing summary.
10054 2018-06-08 Martin Liska <mliska@suse.cz>
10056 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
10057 * ipa-inline.h (estimate_edge_growth): Likewise.
10059 2018-06-08 Martin Liska <mliska@suse.cz>
10061 * cgraph.c (function_version_hasher::hash): Use
10062 cgraph_node::get_uid ().
10063 (function_version_hasher::equal):
10064 * cgraph.h (cgraph_node::get_uid): New method.
10065 * ipa-inline.c (update_caller_keys): Use
10066 cgraph_node::get_uid ().
10067 (update_callee_keys): Likewise.
10068 * ipa-utils.c (searchc): Likewise.
10069 (ipa_reduced_postorder): Likewise.
10070 * lto-cgraph.c (input_node): Likewise.
10071 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
10072 * symbol-summary.h (symtab_insertion): Likewise.
10073 (symtab_removal): Likewise.
10074 (symtab_duplication): Likewise.
10075 * tree-pretty-print.c (dump_function_header): Likewise.
10076 * tree-sra.c (convert_callers_for_node): Likewise.
10078 2018-06-08 Martin Liska <mliska@suse.cz>
10080 * cgraph.c (symbol_table::create_edge): Always assign a new
10082 (symbol_table::free_edge): Do not recycle numbers.
10083 * cgraph.h (cgraph_edge::get): New method.
10084 * symbol-summary.h (symtab_removal): Use it.
10085 (symtab_duplication): Likewise.
10086 (call_summary::hashable_uid): Remove.
10088 2018-06-08 Martin Liska <mliska@suse.cz>
10090 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
10091 (initialize_growth_caches): Remove.
10092 (free_growth_caches): Likewise.
10093 (do_estimate_edge_time): Use edge_growth_cache.
10094 (do_estimate_edge_size): Likewise.
10095 (do_estimate_edge_hints): Likewise.
10096 * ipa-inline.c (reset_edge_caches): Likewise.
10097 (recursive_inlining): Likewise.
10098 (inline_small_functions): Likewise.
10099 * ipa-inline.h (initialize_growth_caches): Remove.
10100 (estimate_edge_size): Likewise.
10101 (estimate_edge_time): Likewise.
10102 (estimate_edge_hints): Likewise.
10103 (reset_edge_growth_cache): Likewise.
10104 * symbol-summary.h (call_summary::remove): New method.
10106 2018-06-08 Martin Liska <mliska@suse.cz>
10108 * ipa-cp.c (class edge_clone_summary): New summary.
10109 (grow_edge_clone_vectors): Remove.
10110 (ipcp_edge_duplication_hook): Remove.
10111 (class edge_clone_summary_t): New call_summary class.
10112 (ipcp_edge_removal_hook): Remove.
10113 (edge_clone_summary_t::duplicate): New function.
10114 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
10115 (create_specialized_node): Likewise.
10116 (ipcp_driver): Initialize edge_clone_summaries and do not
10119 2018-06-08 Martin Liska <mliska@suse.cz>
10121 * symbol-summary.h (get): New function.
10122 (call_summary::m_initialize_when_cloning): New class member.
10124 2018-06-08 Martin Liska <mliska@suse.cz>
10126 * cgraph.c (cgraph_node::remove): Do not recycle uid.
10127 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
10128 (symbol_table::allocate_cgraph_symbol): Do not set uid.
10129 * passes.c (uid_hash_t): Record removed_nodes by their uids.
10130 (remove_cgraph_node_from_order): Use the removed_nodes set.
10131 (do_per_function_toporder): Likwise.
10132 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
10133 instead of summary_uid.
10134 (symtab_removal): Likewise.
10135 (symtab_duplication): Likewise.
10137 2018-06-08 Martin Liska <mliska@suse.cz>
10139 * ipa-cp.c (ipcp_store_bits_results): Use
10140 ipcp_transformation_sum.
10141 (ipcp_store_vr_results): Likewise.
10142 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
10144 (ipcp_transformation_initialize): ... this.
10145 (ipa_set_node_agg_value_chain):
10146 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
10147 (write_ipcp_transformation_info): Likewise.
10148 (read_ipcp_transformation_info): Likewise.
10149 (ipcp_update_bits): Likewise.
10150 (ipcp_update_vr): Likewise.
10151 (ipcp_transform_function): Likewise.
10152 * ipa-prop.h: Rename ipcp_transformation_summary to
10153 ipcp_transformation.
10154 (class ipcp_transformation_t): New function summary.
10155 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
10156 (ipa_get_agg_replacements_for_node): Likewise.
10158 2018-06-08 Martin Liska <mliska@suse.cz>
10160 * ipa-pure-const.c (struct funct_state_d): Do it class instead
10162 (class funct_state_summary_t): New function_summary class.
10163 (has_function_state): Remove.
10164 (get_function_state): Likewise.
10165 (set_function_state): Likewise.
10166 (add_new_function): Likewise.
10167 (funct_state_summary_t::insert): New function.
10168 (duplicate_node_data): Remove.
10169 (remove_node_data): Remove.
10170 (funct_state_summary_t::duplicate): New function.
10171 (register_hooks): Create new funct_state_summaries.
10172 (pure_const_generate_summary): Use it.
10173 (pure_const_write_summary): Likewise.
10174 (pure_const_read_summary): Likewise.
10175 (propagate_pure_const): Likewise.
10176 (propagate_nothrow): Likewise.
10177 (dump_malloc_lattice): Likewise.
10178 (propagate_malloc): Likewise.
10179 (execute): Do not register hooks, just remove summary
10181 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
10184 2018-06-08 Martin Liska <mliska@suse.cz>
10186 * ipa-reference.c (remove_node_data): Remove.
10187 (duplicate_node_data): Likewise.
10188 (class ipa_ref_var_info_summary_t): New class.
10189 (class ipa_ref_opt_summary_t): Likewise.
10190 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
10191 (get_reference_optimization_summary): Use
10192 ipa_ref_opt_sum_summaries.
10193 (set_reference_vars_info): Remove.
10194 (set_reference_optimization_summary): Likewise.
10195 (ipa_init): Create summaries.
10196 (init_function_info): Use function summary.
10197 (ipa_ref_opt_summary_t::duplicate): New function.
10198 (ipa_ref_opt_summary_t::remove): New function.
10199 (get_read_write_all_from_node): Fix GNU coding style.
10200 (propagate): Use function summary.
10201 (write_node_summary_p): Fix GNU coding style.
10202 (stream_out_bitmap): Likewise.
10203 (ipa_reference_read_optimization_summary): Use function summary.
10204 (ipa_reference_c_finalize): Do not release hooks.
10206 2018-06-08 Martin Liska <mliska@suse.cz>
10208 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
10209 (analyze_function_body): Extract multiple calls of get_create.
10210 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
10211 * ipa-inline.c (recursive_inlining): Use ::get method.
10212 * ipa-inline.h (estimate_edge_growth): Likewise.
10214 2018-06-08 Martin Liska <mliska@suse.cz>
10216 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
10218 (hsa_function_summary::hsa_function_summary): Use the new enum
10220 (hsa_gpu_implementation_p): Use hsa_summaries::get.
10221 * hsa-gen.c (hsa_get_host_function): Likewise.
10222 (get_brig_function_name): Likewise.
10223 * ipa-hsa.c (process_hsa_functions): Likewise.
10224 (ipa_hsa_write_summary): Likewise.
10225 * symbol-summary.h (symtab_duplication): Use ::get function/
10226 (get): New function.
10228 2018-06-08 Martin Liska <mliska@suse.cz>
10230 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
10232 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
10233 (hsa_register_kernel): Likewise.
10234 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
10235 * hsa-gen.c (hsa_get_host_function): Likewise.
10236 (get_brig_function_name): Likewise.
10237 (generate_hsa): Likewise.
10238 (pass_gen_hsail::execute): Likewise.
10239 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
10240 (devirtualization_time_bonus): Likewise.
10241 (ipcp_propagate_stage): Likewise.
10242 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
10243 (edge_set_predicate): Likewise.
10244 (evaluate_conditions_for_known_args): Likewise.
10245 (evaluate_properties_for_edge): Likewise.
10246 (ipa_fn_summary::reset): Likewise.
10247 (ipa_fn_summary_t::duplicate): Likewise.
10248 (dump_ipa_call_summary): Likewise.
10249 (ipa_dump_fn_summary): Likewise.
10250 (analyze_function_body): Likewise.
10251 (compute_fn_summary): Likewise.
10252 (estimate_edge_devirt_benefit): Likewise.
10253 (estimate_edge_size_and_time): Likewise.
10254 (estimate_calls_size_and_time): Likewise.
10255 (estimate_node_size_and_time): Likewise.
10256 (inline_update_callee_summaries): Likewise.
10257 (remap_edge_change_prob): Likewise.
10258 (remap_edge_summaries): Likewise.
10259 (ipa_merge_fn_summary_after_inlining): Likewise.
10260 (ipa_update_overall_fn_summary): Likewise.
10261 (read_ipa_call_summary): Likewise.
10262 (inline_read_section): Likewise.
10263 (write_ipa_call_summary): Likewise.
10264 (ipa_fn_summary_write): Likewise.
10265 (ipa_free_fn_summary): Likewise.
10266 * ipa-hsa.c (process_hsa_functions): Likewise.
10267 (ipa_hsa_write_summary): Likewise.
10268 (ipa_hsa_read_section): Likewise.
10269 * ipa-icf.c (sem_function::merge): Likewise.
10270 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
10271 (do_estimate_edge_time): Likewise.
10272 (estimate_size_after_inlining): Likewise.
10273 (estimate_growth): Likewise.
10274 (growth_likely_positive): Likewise.
10275 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
10276 (inline_call): Likewise.
10277 * ipa-inline.c (caller_growth_limits): Likewise.
10278 (can_inline_edge_p): Likewise.
10279 (can_inline_edge_by_limits_p): Likewise.
10280 (compute_uninlined_call_time): Likewise.
10281 (compute_inlined_call_time): Likewise.
10282 (want_inline_small_function_p): Likewise.
10283 (edge_badness): Likewise.
10284 (update_caller_keys): Likewise.
10285 (update_callee_keys): Likewise.
10286 (recursive_inlining): Likewise.
10287 (inline_small_functions): Likewise.
10288 (inline_to_all_callers_1): Likewise.
10289 (dump_overall_stats): Likewise.
10290 (early_inline_small_functions): Likewise.
10291 (early_inliner): Likewise.
10292 * ipa-inline.h (estimate_edge_growth): Likewise.
10293 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
10294 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10295 * ipa-prop.h (IPA_NODE_REF): Likewise.
10296 (IPA_EDGE_REF): Likewise.
10297 * ipa-pure-const.c (malloc_candidate_p): Likewise.
10298 (propagate_malloc): Likewise.
10299 * ipa-split.c (execute_split_functions): Likewise.
10300 * symbol-summary.h: Rename get to get_create.
10302 (get_create): Likewise.
10303 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
10305 2018-06-08 Martin Liska <mliska@suse.cz>
10307 * symbol-summary.h (release): Move definition out of class
10309 (symtab_removal): Likewise.
10310 (symtab_duplication): Likewise.
10312 2018-06-08 Martin Liska <mliska@suse.cz>
10314 * symbol-summary.h (function_summary): Move constructor
10315 implementation out of class declaration.
10316 (release): Likewise.
10317 (symtab_insertion): Likewise.
10318 (symtab_removal): Likewise.
10319 (symtab_duplication): Likewise.
10322 2018-06-08 Martin Liska <mliska@suse.cz>
10324 * Makefile.in: Remove support for MPX (macros, related functions,
10325 fields in cgraph_node, ...).
10326 * builtin-types.def (BT_BND): Likewise.
10327 (BT_FN_BND_CONST_PTR): Likewise.
10328 (BT_FN_CONST_PTR_BND): Likewise.
10329 (BT_FN_VOID_PTR_BND): Likewise.
10330 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
10331 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
10332 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
10333 (expand_builtin_mempcpy_with_bounds): Likewise.
10334 (expand_builtin_memset_with_bounds): Likewise.
10335 (expand_builtin_memset_args): Likewise.
10336 (std_expand_builtin_va_start): Likewise.
10337 (expand_builtin): Likewise.
10338 (expand_builtin_with_bounds): Likewise.
10339 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
10340 (DEF_LIB_BUILTIN_CHKP): Likewise.
10341 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
10342 (DEF_CHKP_BUILTIN): Likewise.
10343 (BUILT_IN_MEMCPY): Likewise.
10344 (BUILT_IN_MEMMOVE): Likewise.
10345 (BUILT_IN_MEMPCPY): Likewise.
10346 (BUILT_IN_MEMSET): Likewise.
10347 (BUILT_IN_STPCPY): Likewise.
10348 (BUILT_IN_STRCAT): Likewise.
10349 (BUILT_IN_STRCHR): Likewise.
10350 (BUILT_IN_STRCPY): Likewise.
10351 (BUILT_IN_STRLEN): Likewise.
10352 (BUILT_IN_MEMCPY_CHK): Likewise.
10353 (BUILT_IN_MEMMOVE_CHK): Likewise.
10354 (BUILT_IN_MEMPCPY_CHK): Likewise.
10355 (BUILT_IN_MEMSET_CHK): Likewise.
10356 (BUILT_IN_STPCPY_CHK): Likewise.
10357 (BUILT_IN_STRCAT_CHK): Likewise.
10358 (BUILT_IN_STRCPY_CHK): Likewise.
10359 * calls.c (store_bounds): Likewise.
10360 (emit_call_1): Likewise.
10361 (special_function_p): Likewise.
10362 (maybe_warn_nonstring_arg): Likewise.
10363 (initialize_argument_information): Likewise.
10364 (finalize_must_preallocate): Likewise.
10365 (compute_argument_addresses): Likewise.
10366 (expand_call): Likewise.
10367 * cfgexpand.c (expand_call_stmt): Likewise.
10368 (expand_return): Likewise.
10369 (expand_gimple_stmt_1): Likewise.
10370 (pass_expand::execute): Likewise.
10371 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
10372 (cgraph_node::remove): Likewise.
10373 (cgraph_node::dump): Likewise.
10374 (cgraph_node::verify_node): Likewise.
10375 * cgraph.h (chkp_function_instrumented_p): Likewise.
10376 (symtab_node::get_alias_target): Likewise.
10377 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
10378 (cgraph_local_p): Likewise.
10379 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
10380 (cgraph_edge::rebuild_references): Likewise.
10381 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
10382 (walk_polymorphic_call_targets): Likewise.
10383 (cgraph_node::expand_thunk): Likewise.
10384 (symbol_table::output_weakrefs): Likewise.
10385 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
10386 (ix86_handle_option): Likewise.
10387 * config/i386/constraints.md: Likewise.
10388 * config/i386/i386-builtin-types.def (BND): Likewise.
10392 * config/i386/i386-builtin.def (BDESC_END): Likewise.
10393 (BDESC_FIRST): Likewise.
10395 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
10396 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
10397 * config/i386/i386.c (enum reg_class): Likewise.
10398 (ix86_target_string): Likewise.
10399 (ix86_option_override_internal): Likewise.
10400 (ix86_conditional_register_usage): Likewise.
10401 (ix86_valid_target_attribute_inner_p): Likewise.
10402 (ix86_set_indirect_branch_type): Likewise.
10403 (ix86_set_current_function): Likewise.
10404 (ix86_function_arg_regno_p): Likewise.
10405 (init_cumulative_args): Likewise.
10406 (ix86_function_arg_advance): Likewise.
10407 (ix86_function_arg): Likewise.
10408 (ix86_pass_by_reference): Likewise.
10409 (ix86_function_value_regno_p): Likewise.
10410 (ix86_function_value_1): Likewise.
10411 (ix86_function_value_bounds): Likewise.
10412 (ix86_return_in_memory): Likewise.
10413 (ix86_setup_incoming_vararg_bounds): Likewise.
10414 (ix86_va_start): Likewise.
10415 (indirect_thunk_need_prefix): Likewise.
10416 (print_reg): Likewise.
10417 (ix86_print_operand): Likewise.
10418 (ix86_expand_call): Likewise.
10419 (ix86_output_function_return): Likewise.
10420 (reg_encoded_number): Likewise.
10421 (BDESC_VERIFYS): Likewise.
10422 (ix86_init_mpx_builtins): Likewise.
10423 (ix86_init_builtins): Likewise.
10424 (ix86_emit_cmove): Likewise.
10425 (ix86_emit_move_max): Likewise.
10426 (ix86_expand_builtin): Likewise.
10427 (ix86_builtin_mpx_function): Likewise.
10428 (ix86_get_arg_address_for_bt): Likewise.
10429 (ix86_load_bounds): Likewise.
10430 (ix86_store_bounds): Likewise.
10431 (ix86_load_returned_bounds): Likewise.
10432 (ix86_store_returned_bounds): Likewise.
10433 (ix86_class_likely_spilled_p): Likewise.
10434 (ix86_hard_regno_mode_ok): Likewise.
10435 (x86_order_regs_for_local_alloc): Likewise.
10436 (ix86_mitigate_rop): Likewise.
10437 (ix86_bnd_prefixed_insn_p): Likewise.
10438 (ix86_mpx_bound_mode): Likewise.
10439 (ix86_make_bounds_constant): Likewise.
10440 (ix86_initialize_bounds): Likewise.
10441 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
10442 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
10443 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
10444 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
10445 (TARGET_CHKP_BOUND_MODE): Likewise.
10446 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
10447 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
10448 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
10449 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
10450 * config/i386/i386.h (TARGET_MPX): Likewise.
10451 (TARGET_MPX_P): Likewise.
10452 (VALID_BND_REG_MODE): Likewise.
10453 (FIRST_BND_REG): Likewise.
10454 (LAST_BND_REG): Likewise.
10455 (enum reg_class): Likewise.
10456 (BND_REG_P): Likewise.
10457 (BND_REGNO_P): Likewise.
10458 (BNDmode): Likewise.
10459 (ADJUST_INSN_LENGTH): Likewise.
10460 * config/i386/i386.md: Likewise.
10461 * config/i386/i386.opt: Likewise.
10462 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
10463 (defined): Likewise.
10464 (LINK_MPX): Likewise.
10465 (MPX_SPEC): Likewise.
10466 (LIBMPX_SPEC): Likewise.
10467 (LIBMPXWRAPPERS_SPEC): Likewise.
10468 (CHKP_SPEC): Likewise.
10469 * config/i386/predicates.md: Likewise.
10470 * dbxout.c (dbxout_type): Likewise.
10471 * doc/extend.texi: Likewise.
10472 * doc/invoke.texi: Likewise.
10473 * doc/md.texi: Likewise.
10474 * doc/tm.texi: Likewise.
10475 * doc/tm.texi.in: Likewise.
10476 * dwarf2out.c (is_base_type): Likewise.
10477 (gen_formal_types_die): Likewise.
10478 (gen_subprogram_die): Likewise.
10479 (gen_type_die_with_usage): Likewise.
10480 (gen_decl_die): Likewise.
10481 (dwarf2out_late_global_decl): Likewise.
10482 * expr.c (expand_assignment): Likewise.
10483 (emit_storent_insn): Likewise.
10484 (store_expr_with_bounds): Likewise.
10485 (store_expr): Likewise.
10486 (expand_expr_real_1): Likewise.
10487 * expr.h (store_expr_with_bounds): Likewise.
10488 * function.c (use_register_for_decl): Likewise.
10489 (struct bounds_parm_data): Likewise.
10490 (assign_parms_augmented_arg_list): Likewise.
10491 (assign_parm_find_entry_rtl): Likewise.
10492 (assign_parm_is_stack_parm): Likewise.
10493 (assign_parm_load_bounds): Likewise.
10494 (assign_bounds): Likewise.
10495 (assign_parms): Likewise.
10496 (expand_function_start): Likewise.
10497 * gcc.c (CHKP_SPEC): Likewise.
10498 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
10499 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
10500 (wrestrict_dom_walker::check_call): Likewise.
10501 * gimple.c (gimple_build_call_from_tree): Likewise.
10502 * gimple.h (enum gf_mask): Likewise.
10503 (gimple_call_with_bounds_p): Likewise.
10504 (gimple_call_set_with_bounds): Likewise.
10505 * gimplify.c (gimplify_init_constructor): Likewise.
10506 * ipa-cp.c (initialize_node_lattices): Likewise.
10507 (propagate_constants_across_call): Likewise.
10508 (find_more_scalar_values_for_callers_subset): Likewise.
10509 * ipa-hsa.c (process_hsa_functions): Likewise.
10510 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
10511 * ipa-icf.c (sem_function::merge): Likewise.
10512 * ipa-inline.c (early_inliner): Likewise.
10513 * ipa-pure-const.c (warn_function_noreturn): Likewise.
10514 (warn_function_cold): Likewise.
10515 (propagate_pure_const): Likewise.
10516 * ipa-ref.h (enum GTY): Likewise.
10517 * ipa-split.c (find_retbnd): Likewise.
10518 (consider_split): Likewise.
10519 (split_function): Likewise.
10520 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
10521 * ipa.c (walk_polymorphic_call_targets): Likewise.
10522 (symbol_table::remove_unreachable_nodes): Likewise.
10523 (process_references): Likewise.
10524 (cgraph_build_static_cdtor_1): Likewise.
10525 * lto-cgraph.c (lto_output_node): Likewise.
10526 (output_refs): Likewise.
10527 (compute_ltrans_boundary): Likewise.
10528 (input_overwrite_node): Likewise.
10529 (input_node): Likewise.
10530 (input_cgraph_1): Likewise.
10531 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
10532 * passes.c (pass_manager::execute_early_local_passes): Likewise.
10533 (class pass_chkp_instrumentation_passes): Likewise.
10534 (make_pass_chkp_instrumentation_passes): Likewise.
10535 * passes.def: Likewise.
10536 * rtl.h (struct GTY): Likewise.
10537 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
10538 * stor-layout.c (layout_type): Likewise.
10539 * symtab.c: Likewise.
10540 * target.def: Likewise.
10541 * targhooks.c (default_chkp_bound_type): Likewise.
10542 (default_chkp_bound_mode): Likewise.
10543 (default_builtin_chkp_function): Likewise.
10544 (default_chkp_function_value_bounds): Likewise.
10545 (default_chkp_make_bounds_constant): Likewise.
10546 (default_chkp_initialize_bounds): Likewise.
10547 * targhooks.h (default_chkp_bound_type): Likewise.
10548 (default_chkp_bound_mode): Likewise.
10549 (default_builtin_chkp_function): Likewise.
10550 (default_chkp_function_value_bounds): Likewise.
10551 (default_chkp_make_bounds_constant): Likewise.
10552 (default_chkp_initialize_bounds): Likewise.
10553 * toplev.c (compile_file): Likewise.
10554 (process_options): Likewise.
10555 * tree-core.h (DEF_BUILTIN): Likewise.
10556 (DEF_BUILTIN_CHKP): Likewise.
10557 * tree-inline.c (declare_return_variable): Likewise.
10558 (remap_gimple_stmt): Likewise.
10559 (copy_bb): Likewise.
10560 (initialize_inlined_parameters): Likewise.
10561 (expand_call_inline): Likewise.
10562 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
10563 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
10564 (make_pass_ipa_chkp_produce_thunks): Likewise.
10565 (make_pass_chkp): Likewise.
10566 (make_pass_chkp_opt): Likewise.
10567 (make_pass_chkp_instrumentation_passes): Likewise.
10568 * tree-pretty-print.c (dump_generic_node): Likewise.
10569 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
10570 * tree-ssa-dce.c (propagate_necessity): Likewise.
10571 (eliminate_unnecessary_stmts): Likewise.
10572 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
10573 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
10574 * tree-ssa-sccvn.h: Likewise.
10575 * tree-ssa-strlen.c (get_string_length): Likewise.
10576 (valid_builtin_call): Likewise.
10577 (adjust_last_stmt): Likewise.
10578 (handle_builtin_strchr): Likewise.
10579 (handle_builtin_strcpy): Likewise.
10580 (handle_builtin_stxncpy): Likewise.
10581 (handle_builtin_memcpy): Likewise.
10582 (handle_builtin_strcat): Likewise.
10583 (strlen_check_and_optimize_stmt): Likewise.
10584 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
10585 * tree-streamer-in.c: Likewise.
10586 * tree-streamer.c (record_common_node): Likewise.
10587 * tree.c (tree_code_size): Likewise.
10588 (wide_int_to_tree_1): Likewise.
10589 (type_contains_placeholder_1): Likewise.
10590 (build_common_tree_nodes): Likewise.
10591 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
10592 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
10593 (POINTER_BOUNDS_P): Likewise.
10594 (BOUNDED_TYPE_P): Likewise.
10595 (BOUNDED_P): Likewise.
10596 (CALL_WITH_BOUNDS_P): Likewise.
10597 (pointer_bounds_type_node): Likewise.
10598 * value-prof.c (gimple_ic): Likewise.
10599 * var-tracking.c (vt_add_function_parameters): Likewise.
10600 * varasm.c (make_decl_rtl): Likewise.
10601 (assemble_start_function): Likewise.
10602 (output_constant): Likewise.
10603 (maybe_assemble_visibility): Likewise.
10604 * varpool.c (ctor_for_folding): Likewise.
10605 * chkp-builtins.def: Remove.
10606 * ipa-chkp.c: Remove.
10607 * ipa-chkp.h: Remove.
10608 * rtl-chkp.c: Remove.
10609 * rtl-chkp.h: Remove.
10610 * tree-chkp-opt.c: Remove.
10611 * tree-chkp.c: Remove.
10612 * tree-chkp.h: Remove.
10614 2018-06-07 Carl Love <cel@us.ibm.com>
10616 * config/rs6000/vsx.md (vextract_fp_from_shorth,
10617 vextract_fp_from_shortl): Add BE support.
10619 2018-06-07 Paul Koning <ni1d@arrl.net>
10621 * compare-elim.c (try_merge_compare): Don't merge compare if
10622 address contains a side effect.
10623 (try_eliminate_compare): Likewise.
10625 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
10627 * config.gcc: Support "tremont".
10628 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
10629 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10631 * config/i386/i386.c (m_TREMONT): Define.
10632 (processor_target_table): Add "tremont".
10633 (PTA_TREMONT): Define.
10634 (ix86_lea_outperforms): Add TARGET_TREMONT.
10635 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
10636 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
10637 and M_INTEL_GOLDMONT_PLUS.
10638 (fold_builtin_cpu): Add "tremont".
10639 (ix86_add_stmt_cost): Add TARGET_TREMONT.
10640 (ix86_option_override_internal): Add "tremont".
10641 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
10642 (processor_type): Add PROCESSOR_TREMONT.
10643 * config/i386/x86-tune.def: Add m_TREMONT.
10644 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
10646 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10648 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
10649 symbol defined for msp430i* devices to be lower case.
10651 2018-06-07 Richard Biener <rguenther@suse.de>
10653 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
10654 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
10655 Properly wrap signed arithmetic if overflow wraps.
10657 2018-06-07 Jakub Jelinek <jakub@redhat.com>
10659 PR tree-optimization/69615
10660 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
10661 of a cast from a same precision integral SSA_NAME in a bb dominated
10662 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
10663 cast to utype if rhs2 has already a compatible type.
10665 2018-06-07 Richard Biener <rguenther@suse.de>
10667 PR tree-optimization/85935
10668 * graphite-scop-detection.c (find_params_in_bb): Analyze
10669 condition operands with respect to the correct loop. Assert
10670 the analysis doesn't fail.
10672 2018-06-04 Carl Love <cel@us.ibm.com>
10674 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
10675 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
10676 as it is slightly cheaper.
10677 (first_match_or_eos_index_<mode>):
10678 Calculate index using natural element order.
10679 (first_match_index_<mode>):
10680 Calculate index using natural element order.
10681 (first_match_or_eos_index_<mode>):
10682 Calculate index using natural order.
10683 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
10684 for BE and LE modes.
10685 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
10686 P9V_BUILTIN_VCLZLSBB_V16QI.
10687 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
10690 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
10692 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
10693 indentation and line wrap for many prototypes. Add missing
10694 @smallexample directives around block of prototypes for vec_xl and
10697 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
10699 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
10700 track if we pass or return IEEE 128-bit floating point.
10701 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
10702 C++ mangling that is compatible with GCC 8.1.
10703 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
10704 (init_cumulative_args): Note if we pass or return IEEE 128-bit
10705 floating point types.
10706 (rs6000_function_arg_advance_1): Likewise.
10707 (rs6000_mangle_type): Optionally generate mangled names that match
10708 what GCC 8.1 generated for IEEE 128-bit floating point types.
10709 (rs6000_globalize_decl_name): If we have an external function that
10710 passes or returns IEEE 128-bit types, generate a weak reference
10711 from the mangled name used in GCC 8.1 to the current mangled
10713 (rs6000_init_builtins): Make __ibm128 use the long double type if
10714 long double is IBM extended double. Make __float128 use the long
10715 double type if long double is IEEE 128-bit.
10718 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
10719 macro for __ibm128 built-in functions.
10720 (PACK_IF): Add __ibm128 pack/unpack functions.
10721 (UNPACK_IF): Likewise.
10722 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
10723 enable long double built-in functions if long double is IEEE
10724 128-bit floating point.
10725 (rs6000_invalid_builtin): Update long double built-in function
10727 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
10728 functions, adjust the built-in function to use the long double
10729 built-in function if __ibm128 and long double are the same type.
10730 * doc/extend.texi (PowerPC builtins): Update documention for
10731 __builtin_{,un}pack_longdouble. Add documentation for
10732 __builtin_{,un}pack_ibm128.
10734 2018-06-06 Jim Wilson <jimw@sifive.com>
10736 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
10737 (struct machine_function): New field interrupt_mode.
10738 (riscv_handle_type_attribute): New function. Add forward declaration.
10739 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
10740 (riscv_expand_epilogue): Check interrupt_mode field.
10741 (riscv_set_current_function): Check interrupt attribute args and
10742 set interrupt_mode field.
10743 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
10744 (riscv_sret, riscv_uret): New.
10745 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
10746 new arguments to interrupt attribute.
10748 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
10751 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
10752 Don't handle -mcpu=power8 if -mpower9-vector is also used.
10754 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
10756 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
10757 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
10758 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
10759 several redundant entries.
10761 2018-06-06 David Malcolm <dmalcolm@redhat.com>
10763 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
10764 type from "rtx" to "rtx_insn *".
10765 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
10766 for local "call_insn", removing cast.
10767 (ix86_expand_call): Likewise, introducing a "call_insn" local.
10769 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
10771 PR tree-optimization/86066
10772 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
10773 for BIT_INSERT_EXPR stores.
10775 2018-06-06 Richard Biener <rguenther@suse.de>
10777 PR tree-optimization/86062
10778 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
10779 component refs ontop
10780 of to be offsetted base.
10782 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10784 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
10785 to be static and remove check on interrupt attribute name.
10787 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
10789 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
10790 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
10792 2018-06-05 Steve Ellcey <sellcey@cavium.com>
10795 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
10797 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
10798 Remove second argument, change how error is called.
10799 (aarch64_layout_arg): Remove second argument from
10800 aarch64_err_no_fpadvsimd call.
10801 (aarch64_init_cumulative_args): Ditto.
10802 (aarch64_gimplify_va_arg_expr): Ditto.
10803 * config/aarch64/aarch64.md (mov<mode>): Ditto.
10805 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
10807 * config/i386/i386.md (simple_return_indirect_internal): New expander.
10808 (*simple_return_indirect_internal<mode>): Rename from
10809 simple_return_indirect_internal. Use W mode iterator.
10810 (rstorssp): New expander.
10811 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
10812 (clrssbsy): New expander.
10813 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
10815 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
10817 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
10819 (cmse_check_pointed_object): Likewise.
10821 2018-06-05 Martin Liska <mliska@suse.cz>
10823 PR gcov-profile/47618
10824 * doc/invoke.texi: Document how -fprofile-dir format
10827 2018-06-05 Richard Biener <rguenther@suse.de>
10829 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
10830 removal pretend DOM info isn't available so we do not update
10831 it and only remove edges, not dominated blocks. Actually free
10832 DOM info in case we removed something. Remove unreachable blocks.
10833 (mfb_keep_latches): Work with either DOM info or marked backedges.
10834 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
10835 first. Mark backedges if DOM info isn't available.
10836 (Re-)compute DOM info after cleanup_control_flow_pre.
10838 2018-06-05 Richard Biener <rguenther@suse.de>
10840 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
10841 (locus_discrim_hasher::hash): Adjust.
10842 (locus_discrim_hasher::equal): Likewise.
10843 (next_discriminator_for_locus): Work on line directly.
10844 (same_line_p): Pass in expanded locus1 as well.
10845 (assign_discriminators): Avoid redundant location expansions.
10847 2018-06-05 Richard Biener <rguenther@suse.de>
10849 PR tree-optimization/86046
10850 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
10851 if required after clearing TREE_ADDRESSABLE.
10853 2018-06-05 Richard Biener <rguenther@suse.de>
10855 PR tree-optimization/86047
10856 * tree-ssa-loop.c (for_each_index): Glob handling of all
10857 decls and constants and really handle all of them.
10859 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10862 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
10863 qualifier_void_pointer and qualifier_const_void_pointer.
10864 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
10865 (arm_init_builtins): Handle the above.
10866 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
10867 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
10868 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
10871 2018-06-05 Martin Liska <mliska@suse.cz>
10873 * auto-profile.c (read_autofdo_file): Do not use
10874 gcov_ctr_summary struct.
10875 (afdo_callsite_hot_enough_for_early_inline): Likewise.
10876 * coverage.c (struct counts_entry): Likewise.
10877 (read_counts_file): Read just single summary entry.
10878 (get_coverage_counts): Use gcov_summary struct.
10879 * coverage.h (get_coverage_counts): Likewise.
10880 * gcov-dump.c (dump_working_sets): Likewise.
10881 (tag_summary): Dump just single summary.
10882 * gcov-io.c (gcov_write_summary): Write just histogram
10884 (gcov_read_summary): Read just single summary.
10885 (compute_working_sets): Use gcov_summary struct.
10886 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
10887 of GCOV_COUNTERS_SUMMABLE.
10888 (GCOV_COUNTERS_SUMMABLE): Remove.
10889 (GCOV_FIRST_VALUE_COUNTER): Replace with
10890 GCOV_COUNTER_V_INTERVAL.
10891 (struct gcov_ctr_summary): Remove.
10892 (struct gcov_summary): Directly use fields of former
10894 (compute_working_sets): Use gcov_summary struct.
10895 * gcov.c (read_count_file): Do not use ctrs fields.
10896 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
10898 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
10900 * profile.c: Likewise.
10901 * profile.h: Likewise.
10903 2018-06-05 Martin Liska <mliska@suse.cz>
10905 PR gcov-profile/84846
10906 * gcov.c (output_lines): Print working directory only
10907 in intermediate format.
10909 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
10911 * config/s390/s390-builtin-types.def: Add void function type.
10912 * config/s390/s390-builtins.def: Use the function type for the
10915 2018-06-04 Jim Wilson <jimw@sifive.com>
10917 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
10919 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
10920 and maybe_eh_return. Change regno to unsigned int. Use new args to
10921 handle EH_RETURN_DATA_REGNO registers properly.
10922 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
10923 (riscv_expand_epilogue): Update comment. Change argument name and
10924 type. Update code to use new name and type. Pass new args to
10925 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
10927 * config/riscv/riscv.md (NORMAL_RETURN): New.
10928 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
10929 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
10930 (eh_return): Call gen_eh_return_internal and emit barrier.
10931 (eh_return_internal): Call riscv_expand_epilogue.
10933 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
10935 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
10936 bit_insertion field and declare can_be_merged_into method.
10937 (merged_store_group::can_be_merged_into): New method.
10938 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
10939 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
10940 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
10942 2018-06-04 Richard Biener <rguenther@suse.de>
10944 PR tree-optimization/85955
10945 * builtins.c (fold_builtin_sincos): Convert pointers to
10946 destination to appropriate type before dereferencing.
10948 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
10950 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
10952 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
10954 * expr.c (expand_expr_real_1): Force the operand into memory if
10955 its TYPE_MODE is BLKmode and if there is no integer mode for
10956 the number of bits being extracted.
10958 2018-06-04 Jakub Jelinek <jakub@redhat.com>
10962 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
10963 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
10965 2018-06-04 Richard Biener <rguenther@suse.de>
10967 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
10968 (cleanup_tree_cfg_noloop): ... single caller. Do
10969 start_recording_case_labels later.
10971 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
10973 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
10974 to _IMMINTRIN_H_INCLUDED.
10975 * config/i386/pconfigintrin.h: Ditto.
10976 * config/i386/waitpkgintrin.h: Ditto.
10977 * config/i386/immintrin.h: Add includes for sgxintrin.h,
10978 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
10979 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
10980 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
10981 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
10982 waitpkgintrin.h and cldemoteintrin.h.
10984 2018-06-04 Richard Biener <rguenther@suse.de>
10986 PR tree-optimization/86038
10987 * tracer.c (find_best_successor): Check probability for
10988 being initialized, bail out if not.
10990 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
10993 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
10994 of bits to ignore when comparing architectures.
10996 2018-06-04 Jakub Jelinek <jakub@redhat.com>
10998 PR tree-optimization/69615
10999 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
11000 maximum or minimum of the type, try to merge it also as if
11001 range1 is + [-, x - 1] or + [x + 1, -].
11004 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
11006 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
11008 PR tree-optimization/86034
11009 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
11010 the unsigned bitfield type in a bit insertion sequence if it does not
11011 have a larger precision than the bitfield size.
11012 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
11014 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
11016 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
11018 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
11020 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
11021 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
11022 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
11023 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
11025 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
11027 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
11028 Disable -fdelete-null-pointer-checks for ELF toolchain.
11030 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
11031 Kito Cheng <kito.cheng@gmail.com>
11033 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
11034 (nds32le-*-*, nds32be-*-*): Integrate checking process.
11035 (nds32*-*-*): Add glibc and uclibc conditions.
11036 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
11037 (TARGET_EXCEPT_UNWIND_INFO): Define.
11038 * config/nds32/elf.h: New file.
11039 * config/nds32/linux.h: New file.
11040 * config/nds32/nds32-elf.opt: New file.
11041 * config/nds32/nds32-linux.opt: New file.
11042 * config/nds32/nds32-fp-as-gp.c
11043 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
11044 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
11046 (nds32_asm_file_end): Ditto.
11047 (nds32_print_operand): Ditto.
11048 (nds32_insert_attributes): Ditto.
11049 (nds32_init_libfuncs): New function.
11050 (TARGET_HAVE_TLS): Define.
11051 (TARGET_INIT_LIBFUNCS): Define.
11052 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
11054 (TARGET_ELF): Apply different mcmodel setting.
11055 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
11056 been migrated into elf.h and linux.h files.
11057 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
11058 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
11059 (mcmodel): The content has been migrated into nds32-elf.opt and
11060 nds32-linux.opt files.
11061 * config/nds32/t-elf: New file.
11062 * config/nds32/t-linux: New file.
11064 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
11065 Shiva Chen <shiva0217@gmail.com>
11067 * config/nds32/constants.md (unspec_volatile_element): Add
11068 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
11069 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
11071 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
11072 (make_pass_nds32_fp_as_gp): Declare.
11073 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
11075 (nds32_asm_function_end_prologue): Remove unused asm output.
11076 (nds32_asm_function_begin_epilogue): Remove unused asm output.
11077 (nds32_asm_file_start): Output necessary fp_as_gp information.
11078 (nds32_option_override): Adjust register usage.
11079 (nds32_expand_prologue): Consider fp_as_gp situation.
11080 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
11081 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
11084 (simple_return): Ditto.
11085 (omit_fp_begin): Output special directive for fp_as_gp.
11086 (omit_fp_end): Output special directive for fp_as_gp.
11087 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
11088 mforbid-fp-as-gp): New options.
11090 2018-06-01 Mark Wielaard <mark@klomp.org>
11092 * dwarf2out.c (dwarf2out_finish): Remove generation of
11093 DW_AT_loclists_base.
11095 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
11097 * gimple-ssa-store-merging.c: Include gimple-fold.h.
11098 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
11099 (struct merged_store_group): Add bit_insertion field.
11100 (dump_char_array): Use standard hexadecimal format.
11101 (merged_store_group::merged_store_group): Set bit_insertion to false.
11102 (merged_store_group::apply_stores): Use optimal buffer size. Deal
11103 with BIT_INSERT_EXPR stores. Move up code updating the mask and
11104 also print the mask in the dump file.
11105 (pass_store_merging::gate): Minor tweak.
11106 (imm_store_chain_info::coalesce_immediate): Fix wrong association
11107 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
11108 stores with INTEGER_CST stores.
11109 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
11110 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
11111 and use it throughout. Generate bit insertion sequences if need be.
11112 (pass_store_merging::process_store): Remove redundant condition.
11113 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
11115 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
11117 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
11118 the 128-bit floating point types. Fix function comment.
11120 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11122 * config/aarch64/aarch64-simd.md
11123 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
11125 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
11128 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
11130 PR tree-optimization/85989
11131 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
11133 (backprop::intersect_uses): Check it when deciding whether this
11134 is a backedge reference.
11135 (backprop::process_block): Add each phi to m_visited_phis
11136 after visiting it, then clear it at the end.
11138 2018-06-01 Richard Biener <rguenther@suse.de>
11140 * tree-vectorizer.h (vect_dr_stmt): New function.
11141 (vect_get_load_cost): Adjust.
11142 (vect_get_store_cost): Likewise.
11143 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
11144 Use vect_dr_stmt instead of DR_SMTT.
11145 (vect_record_base_alignments): Likewise.
11146 (vect_calculate_target_alignment): Likewise.
11147 (vect_compute_data_ref_alignment): Likewise and make static.
11148 (vect_update_misalignment_for_peel): Likewise.
11149 (vect_verify_datarefs_alignment): Likewise.
11150 (vector_alignment_reachable_p): Likewise.
11151 (vect_get_data_access_cost): Likewise. Pass down
11152 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
11153 (vect_get_peeling_costs_all_drs): Likewise.
11154 (vect_peeling_hash_get_lowest_cost): Likewise.
11155 (vect_enhance_data_refs_alignment): Likewise.
11156 (vect_find_same_alignment_drs): Likewise.
11157 (vect_analyze_data_refs_alignment): Likewise.
11158 (vect_analyze_group_access_1): Likewise.
11159 (vect_analyze_group_access): Likewise.
11160 (vect_analyze_data_ref_access): Likewise.
11161 (vect_analyze_data_ref_accesses): Likewise.
11162 (vect_vfa_segment_size): Likewise.
11163 (vect_small_gap_p): Likewise.
11164 (vectorizable_with_step_bound_p): Likewise.
11165 (vect_prune_runtime_alias_test_list): Likewise.
11166 (vect_analyze_data_refs): Likewise.
11167 (vect_supportable_dr_alignment): Likewise.
11168 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
11169 (vect_gen_prolog_loop_niters): Likewise.
11170 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11171 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
11173 (vect_recog_mask_conversion_pattern): Likewise.
11174 (vect_try_gather_scatter_pattern): Likewise.
11175 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
11176 to vect_get_store_cost.
11177 (vect_get_store_cost): Get stmt_info instead of DR.
11178 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
11179 (vect_get_load_cost): Get stmt_info instead of DR.
11181 2018-06-01 Richard Biener <rguenther@suse.de>
11183 PR middle-end/86017
11184 * gimple-fold.c (var_decl_component_p): Also allow offsetted
11185 vars wrapped in MEM_REFs.
11187 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
11189 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
11190 Fix subreg tests so that we only return a choice between
11191 GENERAL_REGS and FP_REGS if the original classes included both.
11193 2018-06-01 Richard Biener <rguenther@suse.de>
11196 * tree-ssa-structalias.c (get_function_part_constraint):
11197 Handle NULL fi->decl.
11198 (find_func_aliases_for_call): Properly handle indirect
11199 fi from direct call.
11200 (find_func_clobbers): Likewise.
11201 (ipa_pta_execute): Likewise.
11202 (create_variable_info_for): For functions that are ifunc_resolver
11203 resolve to a varinfo that contains the result of the resolver call.
11204 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
11207 2018-05-31 Michael Collison <michael.collison@arm.com>
11209 * config/aarch64/aarch64.md:
11210 (*fix_to_zero_extenddfdi2): New pattern.
11211 * gcc.target/aarch64/fix_extend1.c: New testcase.
11213 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
11215 PR middle-end/78809
11216 PR middle-end/83026
11217 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
11218 and BUILT_IN_STRNCMP_EQ.
11219 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
11220 BUILT_IN_STRNCMP_EQ.
11221 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
11222 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
11223 (gimple_fold_builtin): Likewise.
11224 * tree-ssa-strlen.c (compute_string_length): New function.
11225 (determine_min_obsize): New function.
11226 (handle_builtin_string_cmp): New function to handle calls to
11227 string compare functions.
11228 (strlen_optimize_stmt): Add handling to builtin string compare
11230 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
11231 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
11232 * tree.c (build_common_builtin_nodes): Add new defines of
11233 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
11235 2018-05-31 Jakub Jelinek <jakub@redhat.com>
11238 * bb-reorder.c (pass_partition_blocks::gate): Return false for
11239 functions with naked attribute.
11241 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
11243 * config/i386/sse.md (avx_vec_concat<mode>):
11244 Substitute concat_tg_mode mode attribute with xtg_mode.
11245 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
11246 (concat_tg_mode): Remove mode attribute.
11248 2018-05-31 Martin Sebor <msebor@redhat.com>
11251 * calls.c (alloc_max_size): Correct a logic error/typo.
11252 Treat excessive arguments as infinite. Warn for invalid arguments.
11253 * doc/invoke.texi (-Walloc-size-larger-than): Update.
11255 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
11258 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
11259 and movx for Haswell.
11261 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
11262 Cesar Philippidis <cesar@codesourcery.com>
11264 PR middle-end/85879
11265 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
11266 when emitting error on private/firstprivate reductions.
11267 * omp-low.c (lower_omp_target): Avoid reference-type processing
11268 on pointers for firstprivate clause.
11270 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
11272 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
11275 * config/aarch64/aarch64-simd.md
11276 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
11277 (aarch64_ld1_x3_<mode>): Likewise
11278 (aarch64_st1x2<VALLDIF:mode>): Likewise
11279 (aarch64_st1_x2_<mode>): Likewise
11280 (aarch64_st1x3<VALLDIF:mode>): Likewise
11281 (aarch64_st1_x3_<mode>): Likewise
11282 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
11283 (vld1_s8_x3): Likewise.
11284 (vld1_u16_x3): Likewise.
11285 (vld1_s16_x3): Likewise.
11286 (vld1_u32_x3): Likewise.
11287 (vld1_s32_x3): Likewise.
11288 (vld1_u64_x3): Likewise.
11289 (vld1_s64_x3): Likewise.
11290 (vld1_f16_x3): Likewise.
11291 (vld1_f32_x3): Likewise.
11292 (vld1_f64_x3): Likewise.
11293 (vld1_p8_x3): Likewise.
11294 (vld1_p16_x3): Likewise.
11295 (vld1_p64_x3): Likewise.
11296 (vld1q_u8_x3): Likewise.
11297 (vld1q_s8_x3): Likewise.
11298 (vld1q_u16_x3): Likewise.
11299 (vld1q_s16_x3): Likewise.
11300 (vld1q_u32_x3): Likewise.
11301 (vld1q_s32_x3): Likewise.
11302 (vld1q_u64_x3): Likewise.
11303 (vld1q_s64_x3): Likewise.
11304 (vld1q_f16_x3): Likewise.
11305 (vld1q_f32_x3): Likewise.
11306 (vld1q_f64_x3): Likewise.
11307 (vld1q_p8_x3): Likewise.
11308 (vld1q_p16_x3): Likewise.
11309 (vld1q_p64_x3): Likewise.
11310 (vst1_s64_x2): Likewise.
11311 (vst1_u64_x2): Likewise.
11312 (vst1_f64_x2): Likewise.
11313 (vst1_s8_x2): Likewise.
11314 (vst1_p8_x2): Likewise.
11315 (vst1_s16_x2): Likewise.
11316 (vst1_p16_x2): Likewise.
11317 (vst1_s32_x2): Likewise.
11318 (vst1_u8_x2): Likewise.
11319 (vst1_u16_x2): Likewise.
11320 (vst1_u32_x2): Likewise.
11321 (vst1_f16_x2): Likewise.
11322 (vst1_f32_x2): Likewise.
11323 (vst1_p64_x2): Likewise.
11324 (vst1q_s8_x2): Likewise.
11325 (vst1q_p8_x2): Likewise.
11326 (vst1q_s16_x2): Likewise.
11327 (vst1q_p16_x2): Likewise.
11328 (vst1q_s32_x2): Likewise.
11329 (vst1q_s64_x2): Likewise.
11330 (vst1q_u8_x2): Likewise.
11331 (vst1q_u16_x2): Likewise.
11332 (vst1q_u32_x2): Likewise.
11333 (vst1q_u64_x2): Likewise.
11334 (vst1q_f16_x2): Likewise.
11335 (vst1q_f32_x2): Likewise.
11336 (vst1q_f64_x2): Likewise.
11337 (vst1q_p64_x2): Likewise.
11338 (vst1_s64_x3): Likewise.
11339 (vst1_u64_x3): Likewise.
11340 (vst1_f64_x3): Likewise.
11341 (vst1_s8_x3): Likewise.
11342 (vst1_p8_x3): Likewise.
11343 (vst1_s16_x3): Likewise.
11344 (vst1_p16_x3): Likewise.
11345 (vst1_s32_x3): Likewise.
11346 (vst1_u8_x3): Likewise.
11347 (vst1_u16_x3): Likewise.
11348 (vst1_u32_x3): Likewise.
11349 (vst1_f16_x3): Likewise.
11350 (vst1_f32_x3): Likewise.
11351 (vst1_p64_x3): Likewise.
11352 (vst1q_s8_x3): Likewise.
11353 (vst1q_p8_x3): Likewise.
11354 (vst1q_s16_x3): Likewise.
11355 (vst1q_p16_x3): Likewise.
11356 (vst1q_s32_x3): Likewise.
11357 (vst1q_s64_x3): Likewise.
11358 (vst1q_u8_x3): Likewise.
11359 (vst1q_u16_x3): Likewise.
11360 (vst1q_u32_x3): Likewise.
11361 (vst1q_u64_x3): Likewise.
11362 (vst1q_f16_x3): Likewise.
11363 (vst1q_f32_x3): Likewise.
11364 (vst1q_f64_x3): Likewise.
11365 (vst1q_p64_x3): Likewise.
11367 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11369 * config/msp430/msp430.c (msp430_output_labelref): Prepend
11370 user_label_prefix to name.
11372 * tree-core.h: Update comment about the format of NAME string
11373 passed to handler in attribute_spec.
11375 * config/msp430/msp430.md: Remove erroneous subreg expression from
11376 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
11377 zero_extend{q,h}isi2.
11379 2018-05-30 Borislav Petkov <bp@suse.de>
11381 * doc/extend.texi: Document some architecture specific
11382 constraints and sort entries.
11384 2018-05-30 Martin Sebor <msebor@redhat.com>
11386 PR middle-end/85369
11387 * builtins.c (expand_builtin_stpcpy_1): New function.
11388 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
11389 only if the former succeeds.
11391 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
11393 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
11396 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11398 * doc/invoke.texi (-flinker-output): Document
11400 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11402 * passes.c (ipa_write_summaries): Only modify statements if body
11404 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
11405 incrementally linking.
11406 (ipa_passes): Likewise.
11407 * lto-cgraph.c (lto_output_node): When incrementally linking do not
11408 pass down resolution info.
11409 * common.opt (flag_incremental_link): Update info.
11410 * gcc.c (plugin specs): Turn flinker-output=* to
11411 -plugin-opt=-linker-output-known
11412 * toplev.c (compile_file): Also cut compilation when doing incremental
11414 * flag-types. (enum lto_partition_model): Add
11415 LTO_LINKER_OUTPUT_NOLTOREL.
11416 (invoke.texi): Add -flinker-output docs.
11417 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
11418 link same way as WPA; do not stream in dead initializers.
11420 * dwarf2out.c (dwarf2out_die_ref_for_decl,
11421 darf2out_register_external_decl): Support incremental link.
11423 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11425 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
11427 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11429 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
11430 it down to simple_object_copy_lto_debug_sections.
11431 (run_gcc): Determine incremental LTO link time and configure
11432 lto1 into non-wpa mode, disable renaming of debug sections.
11434 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
11436 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
11437 descriptions of various incorrectly documented functions.
11439 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
11442 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11443 address check not strict.
11445 2018-05-30 Richard Biener <rguenther@suse.de>
11447 PR tree-optimization/85964
11448 * tracer.c (better_p): Drop initialized count check, we only
11449 call the function with initialized counts now.
11450 (find_best_successor): Do find a best edge if one
11451 has uninitialized count.
11452 (find_best_predecessor): Likewise. Do BB frequency check only
11453 if count is initialized.
11455 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
11457 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
11458 (aarch64_ldrstr_offset_compare): New.
11459 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
11460 load/store orderings.
11461 (aarch64_gen_adjusted_ldpstp): Likewise.
11463 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
11465 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
11466 Check for subset of GENERAL_REGS and FP_REGS.
11467 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
11470 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
11472 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
11473 and wi::to_poly_offset. Add the current offset and then check
11474 whether the sum fits, rather than using an unchecked addition of
11475 a checked term. Check for a shwi rather than a uhwi.
11476 * expr.c (get_bit_range): Use tree_to_poly_uint64.
11477 (store_constructor): Use poly_int_tree_p.
11478 (expand_expr_real_1): Likewise.
11479 * function.c (assign_temp): Likewise.
11480 * fold-const.c (const_binop): Use poly_int_tree_p and
11481 wi::to_poly_offset.
11482 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
11484 * ipa-icf-gimple.c (func_checker::compare_operand): Use
11485 to_poly_offset for MEM offsets.
11486 * ipa-icf.c (sem_variable::equals): Likewise.
11487 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
11488 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
11489 wi::to_poly_offset for BIT_FIELD_REF offsets.
11490 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
11491 wi::to_poly_offset.
11492 * var-tracking.c (emit_note_insn_var_location): Use
11493 tree_to_poly_uint64.
11495 2018-05-29 Jim Wilson <jimw@sifive.com>
11497 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
11499 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
11502 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
11503 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
11505 (sse4_1_round<mode>2): Use nonimmediate_operand
11506 for operand 1 predicate.
11508 2018-05-29 Martin Sebor <msebor@redhat.com>
11509 Richard Biener <rguenther@suse.de>
11512 * calls.c (get_size_range): Call determine_value_range instead
11513 of get_value_range..
11514 * tree-vrp.h (determine_value_range): Declared new function.
11515 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
11517 2018-05-29 Richard Biener <rguenther@suse.de>
11519 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
11520 sure to use non-pattern stmts for get_earlier_stmt arguments.
11521 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
11522 called on pattern stmts.
11523 (get_later_stmt): Likewise.
11525 2018-05-29 Martin Liska <mliska@suse.cz>
11527 PR gcov-profile/85759
11528 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
11531 2018-05-29 Jakub Jelinek <jakub@redhat.com>
11533 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
11535 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
11539 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11540 VEC_PACK_FLOAT_EXPR): New tree codes.
11541 * tree-pretty-print.c (op_code_prio): Handle
11542 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11543 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11544 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11545 * tree-inline.c (estimate_operator_cost): Likewise.
11546 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
11547 * fold-const.c (const_binop): Likewise.
11548 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
11549 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11550 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11551 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
11552 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11553 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11554 * expr.c (expand_expr_real_2): Likewise.
11555 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
11556 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
11557 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
11559 * optabs.c (expand_widen_pattern_expr): For
11560 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
11561 sign from result type rather than operand's type.
11562 (expand_binop_directly): For vec_packu_float_optab and
11563 vec_packs_float_optab allow result type to be different from operand's
11565 * optabs-tree.c (optab_for_tree_code): Handle
11566 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11567 VEC_PACK_FLOAT_EXPR. Formatting fixes.
11568 * tree-vect-generic.c (expand_vector_operations_1): Handle
11569 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11570 VEC_PACK_FLOAT_EXPR.
11571 * tree-vect-stmts.c (supportable_widening_operation): Handle
11573 (supportable_narrowing_operation): Handle FLOAT_EXPR.
11574 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
11575 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
11576 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
11577 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
11579 (vec_pack<floatprefix>_float_<mode>): New expander.
11580 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
11582 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
11583 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
11584 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
11585 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
11586 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
11588 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
11589 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
11590 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11591 VEC_PACK_FLOAT_EXPR): Document.
11593 2018-05-29 Richard Biener <rguenther@suse.de>
11595 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
11597 (stmt_vec_info_vec): Make pointer.
11598 (init_stmt_vec_info_vec): Remove.
11599 (free_stmt_vec_info_vec): Likewise.
11600 (set_stmt_vec_info_vec): New function.
11601 (free_stmt_vec_infos): Likewise.
11602 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
11603 (set_vinfo_for_stmt): Likewise.
11604 (get_earlier_stmt): Likewise.
11605 (get_later_stmt): Likewise.
11606 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
11607 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
11608 (vec_info::~vec_info): Free stmt_vec_infos.
11609 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
11610 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
11611 (pass_slp_vectorize::execute): Likewise.
11612 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
11613 (free_stmt_vec_info_vec): Likewise.
11614 (set_stmt_vec_info_vec): New function.
11615 (free_stmt_vec_infos): Likewise.
11616 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
11617 the global stmt_vec_info_vec.
11618 * tree-parloops.c (gather_scalar_reductions): Use
11619 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
11622 2018-05-29 Richard Biener <rguenther@suse.de>
11624 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
11626 2018-05-29 Martin Liska <mliska@suse.cz>
11627 David Malcolm <dmalcolm@redhat.com>
11629 * vec.c (test_reverse): New.
11630 (vec_c_tests): Add new test.
11631 * vec.h (vl_ptr>::reverse): New function.
11633 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
11635 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
11637 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
11640 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
11642 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
11644 2018-05-28 Richard Biener <rguenther@suse.de>
11646 PR tree-optimization/85933
11647 * tree-vect-data-refs.c (vect_record_base_alignments): Only
11648 look at stmts marked as vectorizable.
11650 2018-05-28 Richard Biener <rguenther@suse.de>
11652 PR tree-optimization/85934
11653 * tree-vect-generic.c (expand_vector_operations_1): Hoist
11654 vector boolean check before scalar optimization.
11656 2018-05-28 Jakub Jelinek <jakub@redhat.com>
11658 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
11661 2018-05-28 Mark Wielaard <mark@klomp.org>
11663 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
11664 if it is an expression containing a minus sign.
11666 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
11668 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
11670 2018-05-27 Paul Koning <ni1d@arrl.net>
11672 * config/pdp11/pdp11.md (truncsihi2): Remove.
11674 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
11675 Chung-Ju Wu <jasonwucj@gmail.com>
11677 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
11679 (unaligned_store_dw): Ditto.
11680 * config/nds32/nds32-memory-manipulation.c
11681 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
11682 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
11683 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
11684 (emit_setmem_word_loop): Rename to ...
11685 (emit_setmem_doubleword_loop): ... this.
11686 (nds32_gen_dup_4_byte_to_word_value): New function.
11687 (nds32_gen_dup_8_byte_to_double_word_value): New function.
11688 (nds32_expand_setmem_loop): Refine implementation.
11689 (nds32_expand_setmem_loop_v3m): Ditto.
11690 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
11693 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
11695 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
11697 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
11699 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
11700 (nds32_init_machine_status): Initialize machine->attr_naked_p and
11701 machine->attr_no_prologue_p.
11702 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
11703 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
11704 (nds32_expand_epilogue): Consider attr_naked_p.
11705 (nds32_expand_epilogue_v3pop): Likewise.
11706 (nds32_can_use_return_insn): Likewise.
11707 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
11708 attr_no_prologue_p fields.
11709 * config/nds32/nds32.opt (mret-in-naked-func): New option.
11711 2018-05-27 Jakub Jelinek <jakub@redhat.com>
11714 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
11716 * config/i386/sse.md
11717 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
11719 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
11721 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
11723 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
11725 (*<floatsuffix>floatv2div2sf2): Rename to ...
11726 (*float<floatunssuffix>v2div2sf2): ... this.
11727 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
11728 (float<floatunssuffix>v2div2sf2_mask): ... this.
11729 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
11730 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
11731 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
11733 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
11735 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11737 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11739 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11741 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11743 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
11744 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
11745 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
11746 gen_ufix_truncv8dfv8si2.
11747 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
11748 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
11749 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
11750 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
11751 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
11752 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
11753 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
11754 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
11756 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
11760 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
11762 2018-05-25 Jim Wilson <jimw@sifive.com>
11764 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
11765 * config/riscv/riscv.c (struct machine_function): Add
11766 interrupt_handler_p and attribute_checked_p fields.
11767 (riscv_attribute_table): Add interrupt.
11768 (riscv_interrupt_type_p): New.
11769 (riscv_save_reg_p): Save extra regs for interrupt handler.
11770 (riscv_use_save_libcall): Return false for interrupt handler.
11771 (riscv_first_stack_step): Add forward declaration.
11772 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
11773 for interrupt handler with large frame. Use it for saved reg list.
11774 (riscv_expand_prologue): Move flag_stack_usage_info support to
11775 eliminate duplication.
11776 (riscv_expand_epilogue): Generate mret for interrupt handler.
11777 (riscv_epilogue_uses): New.
11778 (riscv_can_use_return_insn): Return false for interrupt handler.
11779 (riscv_function_ok_for_sibcall): Likewise.
11780 (riscv_set_current_function): Add interrupt handler support.
11781 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
11782 * config/riscv/riscv.md (UNSPECV_MRET): New.
11784 (riscv_frflags, riscv_fsflags): Use tab after opcode.
11786 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
11788 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
11790 PR tree-optimization/85712
11791 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
11792 this candidate has already been replaced in-situ by a copy.
11794 2018-05-25 Jason Merrill <jason@redhat.com>
11796 PR c++/80485 - inline function non-zero address.
11797 * symtab.c (nonzero_address): Check DECL_COMDAT.
11799 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
11802 * config/alpha/alpha.md (ashlsi3): New insn pattern.
11803 (*ashlsi_se): Rename from *ashldi_se. Define as sign
11804 extension of SImode operation. Use const123_operand predicate.
11805 (*saddsi_1): Remove.
11806 (*saddl_se_1): Ditto.
11807 (*ssubsi_1): Ditto.
11808 (*ssubl_se_1): Ditto.
11809 * config/alpha/predicates.md (const123_operand): New predicate.
11810 * config/alpha/constraints.md (P): Use IN_RANGE.
11812 2018-05-25 Richard Biener <rguenther@suse.de>
11814 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
11816 (ref_maybe_used_by_stmt_p): Likewise.
11817 (stmt_may_clobber_ref_p): Likewise.
11818 (stmt_may_clobber_ref_p_1): Likewise.
11819 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
11821 (ref_maybe_used_by_stmt_p): Likewise.
11822 (stmt_may_clobber_ref_p): Likewise.
11823 (stmt_may_clobber_ref_p_1): Likewise.
11824 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
11825 the alias oracle to disambiguate DRs with stmts DR analysis
11827 (vect_analyze_data_refs): Do not give up on not analyzable
11828 DRs for BB vectorization. Remove code truncating the dataref
11831 2018-05-25 Jakub Jelinek <jakub@redhat.com>
11834 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
11835 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
11836 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
11838 2018-05-25 Richard Biener <rguenther@suse.de>
11840 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
11841 function, combining stmt data ref gathering and fatal analysis
11843 (vect_analyze_data_refs): Remove now redudnant code and simplify.
11844 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
11845 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
11846 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
11847 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
11849 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11851 PR tree-optimization/85720
11852 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
11853 SCC if all partitions are builtins.
11854 (version_loop_by_alias_check): New parameter. Generate cancelable
11855 runtime alias check if all partitions are builtins.
11856 (distribute_loop): Update call to above function.
11858 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11860 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
11861 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
11862 (parm_default_def_partition_arg): Ditto.
11863 (set_parm_default_def_partition): Ditto.
11864 (get_parm_default_def_partitions): Ditto and make it static.
11865 (get_undefined_value_partitions): Ditto and make it static.
11866 (remove_ssa_form): Refactor call to init_var_map here.
11867 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
11868 computation for loop region.
11869 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
11870 (register_default_def): Delete.
11871 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
11872 (parm_default_def_partition_arg): Ditto.
11873 (set_parm_default_def_partition): Ditto.
11874 (get_parm_default_def_partitions): Ditto and make it static.
11875 (get_undefined_value_partitions): Ditto and make it static.
11876 (coalesce_with_default, coalesce_with_default): Update comment.
11877 (create_coalesce_list_for_region): New func factored out from
11878 create_outofssa_var_map.
11879 (populate_coalesce_list_for_outofssa): New func factored out from
11880 create_outofssa_var_map and coalesce_ssa_name.
11881 (create_outofssa_var_map): Delete.
11882 (coalesce_ssa_name): Refactor to support live range computation.
11883 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
11884 (get_parm_default_def_partitions): Delete.
11885 (get_undefined_value_partitions): Ditto.
11886 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
11887 computation for loop region.
11888 (new_tree_live_info, loe_visit_block): Ditto.
11889 (live_worklist, set_var_live_on_entry): Ditto.
11890 (calculate_live_on_exit, verify_live_on_entry): Ditto.
11891 * tree-ssa-live.h (struct _var_map): New fields.
11892 (init_var_map): Change decl.
11893 (region_contains_p): New.
11895 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11897 * tree-ssa-live.h (live_merge_and_clear): Delete.
11899 2018-05-25 Richard Biener <rguenther@suse.de>
11902 * tree-dump.c (dequeue_and_dump): Remove access to removed
11903 operand 2 of a SWITCH_EXPR.
11905 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11907 * doc/sourcebuild.texi (vect_double_cond_arith): Include
11908 multiplication and division.
11909 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
11910 (cond_udiv@var{m}, cond_umod@var{m}): Document.
11911 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
11912 (cond_udiv_optab, cond_umod_optab): New optabs.
11913 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
11914 (IFN_COND_RDIV): New internal functions.
11915 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
11916 TRUNC_MOD_EXPR and RDIV_EXPR.
11917 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
11918 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
11920 (SVE_INT_BINARY): Include mult.
11921 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
11922 (optab, sve_int_op): Handle mult.
11923 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
11925 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
11926 for SVE_INT_BINARY_SD.
11928 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11930 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
11931 (optab, sve_int_op): Handle div and udiv.
11932 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
11933 for SVE_INT_BINARY_SD.
11934 (*<optab><mode>3): New insn for the same.
11936 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11938 * tree-vect-patterns.c: Include predict.h.
11939 (vect_recog_divmod_pattern): Restrict check for division support
11940 to when optimizing for size.
11942 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11944 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
11945 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
11946 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
11947 (gimple_match_op::set_op): Likewise.
11948 (gimple_resimplify4): Declare.
11949 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
11950 (expr::gen_transform): Likewise.
11951 (decision_tree::gen): Generate a simplification routine for 4 operands.
11952 * gimple-match-head.c (gimple_simplify): Add an overload for
11953 4 operands. In the top-level function, handle up to 4 call
11954 arguments and call gimple_resimplify4.
11955 (gimple_resimplify4): New function.
11956 (build_call_internal): Pass a fourth operand.
11957 (maybe_push_to_seq): Likewise.
11958 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
11959 Fold VEC_COND_EXPRs of an operation and a default value into
11960 an IFN_COND_* function if possible.
11961 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
11963 (SVE_COND_FP_BINARY): Include them.
11964 (optab, sve_fp_op): Handle them.
11965 (SVE_INT_BINARY_REV): New code iterator.
11966 (SVE_COND_FP_BINARY_REV): New int iterator.
11967 (commutative): New int attribute.
11968 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
11970 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
11972 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
11973 (*cond_<optab><mode>): New patterns for reversed operands.
11975 2018-05-25 Richard Biener <rguenther@suse.de>
11977 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
11978 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
11979 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
11980 (STMT_VINFO_GROUPED_ACCESS): Adjust.
11981 * tree-vect-data-refs.c (everywhere): Adjust users.
11982 * tree-vect-loop.c (everywhere): Likewise.
11983 * tree-vect-slp.c (everywhere): Likewise.
11984 * tree-vect-stmts.c (everywhere): Likewise.
11985 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
11987 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11989 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
11991 (gcc_cv_as_section_exclude): ... this.
11992 Try Solaris as #exclude syntax.
11993 * configure: Regenerate.
11994 * config.in: Regenerate.
11995 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
11997 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
11998 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
12000 * varasm.c (default_elf_asm_named_section): Don't check if
12001 HAVE_GAS_SECTION_EXCLUDE is defined.
12003 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12005 * doc/md.texi: Update the documentation of the cond_* optabs
12006 to mention the new final operand. Fix GET_MODE_NUNITS call.
12007 Describe the scalar case too.
12008 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
12009 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
12011 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
12012 (get_conditional_internal_fn): Update comment.
12013 * tree-vect-loop.c (vectorizable_reduction): Pass the original
12014 accumulator value as a final argument to conditional functions.
12015 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
12016 a define_expand and add an "else" operand. Assert for now that
12017 the else operand is equal to operand 2. Use SVE_INT_BINARY and
12018 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
12019 (*cond_<optab><mode>): New patterns.
12020 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
12021 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12022 (UNSPEC_COND_EOR): Delete.
12023 (optab): Remove associated mappings.
12024 (SVE_INT_BINARY): New code iterator.
12025 (sve_int_op): Remove int attribute and add "minus" to the code
12027 (SVE_COND_INT_OP): Delete.
12028 (SVE_COND_FP_OP): Rename to...
12029 (SVE_COND_FP_BINARY): ...this.
12031 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12033 * optabs.c (can_reuse_operands_p): New function.
12034 (maybe_legitimize_operands): Try to reuse the results for
12037 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
12039 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
12040 Add {q} suffix to insn mnemonic.
12042 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12044 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
12045 (msp430_warn_func_return): New.
12047 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
12049 * fold-const.c (tree_nonzero_bits): New function.
12050 * fold-const.h (tree_nonzero_bits): Likewise.
12051 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
12052 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
12054 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
12058 * varasm.c (assemble_alias): Check ifunc_resolver only on
12061 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
12064 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
12065 when memory input operand is handled.
12067 2018-05-24 Luis Machado <luis.machado@linaro.org>
12069 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
12071 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
12073 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
12075 * match.pd: Delay FMA folds until after vectorization.
12077 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
12080 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12081 address check not strict.
12083 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
12085 * gimple-match.h (gimple_match_op): New class.
12086 (mprts_hook): Replace parameters with a gimple_match_op *.
12087 (maybe_build_generic_op): Likewise.
12088 (gimple_simplified_result_is_gimple_val): Replace parameters with
12089 a const gimple_match_op *.
12090 (gimple_simplify): Replace code_helper * and tree * parameters with
12091 a gimple_match_op * parameter.
12092 (gimple_resimplify1): Replace code_helper *, tree and tree *
12093 parameters with a gimple_match_op * parameter.
12094 (gimple_resimplify2): Likewise.
12095 (gimple_resimplify3): Likewise.
12096 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
12097 parameters with a gimple_match_op * parameter.
12098 * gimple-match-head.c (gimple_simplify): Change prototypes of
12099 auto-generated functions to take a gimple_match_op * instead of
12100 separate code_helper * and tree * parameters. Make the same
12101 change in the top-level overload and update calls to the
12102 gimple_resimplify routines. Update calls to the auto-generated
12103 functions and to maybe_push_res_to_seq in the publicly-facing
12104 operation-specific gimple_simplify overloads.
12105 (gimple_match_op::MAX_NUM_OPS): Define.
12106 (gimple_resimplify1): Replace rcode and ops with a single res_op
12107 parameter. Update call to gimple_simplify.
12108 (gimple_resimplify2): Likewise.
12109 (gimple_resimplify3): Likewise.
12110 (mprts_hook): Replace parameters with a gimple_match_op *.
12111 (maybe_build_generic_op): Likewise.
12112 (build_call_internal): Replace type, nargs and ops with
12113 a gimple_match_op *.
12114 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
12115 with a single gimple_match_op *. Update calls to mprts_hook,
12116 build_call_internal and gimple_simplified_result_is_gimple_val.
12117 Factor out code that is common to the tree_code and combined_fn cases.
12118 * genmatch.c (expr::gen_transform): Replace tem_code and
12119 tem_ops with a gimple_match_op called tem_op. Update calls
12120 to the gimple_resimplify functions and maybe_push_res_to_seq.
12121 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
12122 res_ops. Update call to the gimple_resimplify functions.
12123 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
12124 (decision_tree::gen): Make the functions take a gimple_match_op *
12125 called res_op instead of separate res_code and res_ops parameters.
12126 Update call accordingly.
12127 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
12128 and ops with a single res_op parameter. Update calls to
12129 maybe_build_generic_op and maybe_push_res_to_seq.
12130 (fold_stmt_1): Update calls to gimple_simplify and
12131 replace_stmt_with_simplification.
12132 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
12133 and gimple_simplified_result_is_gimple_val.
12134 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
12136 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
12137 with a gimple_match_op *.
12138 (vn_nary_build_or_lookup): Likewise. Update call to
12139 vn_nary_build_or_lookup_1.
12140 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
12141 gimple_match_op *. Update calls to the gimple_resimplify routines
12142 and to gimple_simplified_result_is_gimple_val.
12143 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
12144 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
12145 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
12146 (visit_nary_op): Likewise.
12147 (visit_reference_op_load): Likewise.
12149 2018-05-23 Luis Machado <luis.machado@linaro.org>
12151 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
12152 modifier for printing the step amount.
12154 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
12157 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
12160 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
12162 * doc/sourcebuild.texi (Endianness): New subsubsection.
12164 2018-05-23 Luis Machado <luis.machado@linaro.org>
12166 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12167 <prefetch_dynamic_strides>: New const bool field.
12168 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12169 prefetch_dynamic_strides.
12170 (exynosm1_prefetch_tune): Likewise.
12171 (thunderxt88_prefetch_tune): Likewise.
12172 (thunderx_prefetch_tune): Likewise.
12173 (thunderx2t99_prefetch_tune): Likewise.
12174 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
12176 (aarch64_override_options_internal): Update to set
12177 PARAM_PREFETCH_DYNAMIC_STRIDES.
12178 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12179 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12180 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12181 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12182 prefetch-dynamic-strides setting.
12184 2018-05-23 Luis Machado <luis.machado@linaro.org>
12186 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12187 <minimum_stride>: New const int field.
12188 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12189 minimum_stride field defaulting to -1.
12190 (exynosm1_prefetch_tune): Likewise.
12191 (thunderxt88_prefetch_tune): Likewise.
12192 (thunderx_prefetch_tune): Likewise.
12193 (thunderx2t99_prefetch_tune): Likewise.
12194 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
12195 <default_opt_level>: Set to 3.
12196 (aarch64_override_options_internal): Update to set
12197 PARAM_PREFETCH_MINIMUM_STRIDE.
12198 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12199 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12200 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12201 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12202 stride is constant and is below the minimum stride threshold.
12204 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12206 * config/arm/arm-cpus.in (mode26): Delete.
12207 (armv4): Delete mode26 reference.
12208 * config/arm/arm.c (arm_configure_build_target): Delete use of
12211 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
12213 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
12215 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
12216 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
12218 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
12219 (floatunsdidf2): Ditto.
12221 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
12223 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
12224 (fixuns_trunc<mode>si2_avx512f): Ditto.
12225 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
12226 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
12227 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
12229 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
12231 PR rtl-optimization/79985
12232 * df-scan.c (df_insn_refs_collect): Remove special case for
12233 global registers and asm statements.
12235 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
12237 * extend.texi (Global Register Variables): Rewrite the bullet list.
12238 Note that the register is available for allocation. Note that access
12239 via inline asm must use constraints. Add note about async-signal
12240 handlers. Remove paragraph about automagic register selection.
12242 2018-05-23 Richard Biener <rguenther@suse.de>
12244 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
12245 of fixed offset from memset VN.
12247 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
12249 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
12250 first_interp field.
12251 (alloc_cand_and_find_basis): Initialize first_interp field.
12252 (slsr_process_mul): Modify first_interp field.
12253 (slsr_process_add): Likewise.
12254 (slsr_process_cast): Modify first_interp field for each new
12256 (slsr_process_copy): Likewise.
12257 (dump_candidate): Dump first_interp field.
12258 (replace_mult_candidate): Process all interpretations, not just
12260 (replace_rhs_if_not_dup): Likewise.
12261 (replace_one_candidate): Likewise.
12263 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
12265 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
12267 (aarch64_needs_frame_chain): New function.
12268 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
12270 2018-05-23 Sudakshina Das <sudi.das@arm.com>
12273 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
12274 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
12275 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
12276 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
12277 as true for strict-align.
12278 (aarch64_can_inline_p): Perform checks even when callee has no
12279 attributes to check for strict alignment.
12280 * doc/extend.texi (AArch64 Function Attributes): Document
12282 * doc/invoke.texi: (AArch64 Options): Likewise.
12284 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
12286 PR tree-optimization/85853
12287 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
12288 the handling of the root of the node to...
12289 (vect_slp_analyze_node_operations_1): ...this new function,
12290 and run the whole thing with the child nodes' def types
12291 set according to their SLP node's def type.
12293 2018-05-23 Richard Biener <rguenther@suse.de>
12295 PR middle-end/85874
12296 * tree-data-ref.c (create_runtime_alias_checks): Defer
12297 and ignore overflow warnings.
12299 2018-05-23 Yury Gribov <tetra2005@gmail.com>
12301 PR tree-optimization/85822
12302 * tree-vrp.c (is_masked_range_test): Fix handling of negative
12305 2018-05-23 Richard Biener <rguenther@suse.de>
12307 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
12308 memset constants via native_interpret_expr.
12310 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
12313 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
12315 (cgraph_node::create_alias): Likewise.
12316 (cgraph_node::get_availability): Check ifunc_resolver instead
12317 of looking up ifunc attribute.
12318 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
12319 * varasm.c (do_assemble_alias): Likewise.
12320 (assemble_alias): Likewise.
12321 (default_binds_local_p_3): Likewise.
12322 * cgraph.h (cgraph_node): Add ifunc_resolver.
12323 (cgraph_node::only_called_directly_or_aliased_p): Return false
12324 for IFUNC resolver.
12325 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
12327 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
12328 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
12329 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
12330 instead of looking up ifunc attribute.
12332 2018-05-22 Luis Machado <luis.machado@linaro.org>
12334 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
12336 2018-05-22 Martin Sebor <msebor@redhat.com>
12338 PR middle-end/85359
12339 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
12340 only when expasion succeeds.
12341 (expand_builtin_strcmp): Same.
12342 (expand_builtin_strncmp): Same.
12344 2018-05-22 Martin Sebor <msebor@redhat.com>
12346 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
12348 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
12349 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12351 * config/aarch64/aarch64-ldpstp.md: Replace uses of
12352 aarch64_mem_pair_operand with memory_operand and delete operand swapping
12354 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12355 Add check for legitimate_address.
12356 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
12357 (aarch64_swap_ldrstr_operands): New.
12358 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
12361 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
12362 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12364 * config/aarch64/aarch64.md: New patterns to generate stp
12366 (store_pair_sw, store_pair_dw): New patterns to generate stp for
12367 single words and double words.
12368 (load_pair_sw, load_pair_dw): Likewise.
12369 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
12371 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
12373 * config/aarch64/aarch64-ldpstp.md: Modify peephole
12374 for different mode ldpstp and add peephole for merged zero stores.
12375 Likewise for loads.
12376 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12378 (aarch64_gen_store_pair): Rename calls to match new patterns.
12379 (aarch64_gen_load_pair): Rename calls to match new patterns.
12380 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
12381 (load_pair<DREG:mode><DREG2:mode>): ... This.
12382 (store_pair<mode>): Rename to...
12383 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
12384 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
12385 New mode iterators.
12386 (V_INT_EQUIV): Handle SImode.
12387 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
12390 2018-05-22 Martin Sebor <msebor@redhat.com>
12393 * calls.c (maybe_warn_nonstring_arg): Use string length to set
12394 or ajust the presumed bound on an operation to avoid unnecessary
12397 2018-05-22 Martin Sebor <msebor@redhat.com>
12399 PR tree-optimization/85826
12400 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
12401 assuming that a DECL necesarily has a constant size.
12403 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
12405 PR middle-end/85862
12406 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
12408 2018-05-22 Richard Biener <rguenther@suse.de>
12410 PR tree-optimization/85834
12411 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
12412 non-constant and non-zero memset arguments.
12414 2018-05-22 Martin Liska <mliska@suse.cz>
12417 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
12419 2018-05-22 Richard Biener <rguenther@suse.de>
12421 PR tree-optimization/85863
12422 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
12423 comparisons when vectype is specified.
12424 (vectorizable_condition): Do not specify vectype for
12425 vect_is_simple_cond when SLP vectorizing.
12427 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
12430 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
12431 define __ibm128 as long double.
12432 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
12433 as a distinct type when IEEE 128-bit support is enabled.
12434 (init_float128_ieee): Fix up conversions between IFmode and IEEE
12435 128-bit types to use the correct functions.
12436 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
12437 convert between 128-bit floating point types that have different
12438 modes but the same representation, instead of using gen_lowpart to
12440 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
12442 (IFKF_reg): New attributes to give the register constraints for
12444 (extend<mode>tf2_internal): New insns to mark an explicit
12445 conversion between 128-bit floating point types that have a
12446 different mode but share the same representation.
12448 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
12450 PR tree-optimization/85814
12451 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
12452 a null return from get_strinfo when unsharing the next
12453 strinfo in the chain.
12455 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
12458 * varasm.c (weak_finish): Clean up weak_decls.
12460 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12462 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
12463 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
12464 UNSPEC_UADALP values.
12465 * config/aarch64/iterators.md (ABAL): New int iterator.
12468 (sur): Add mappings for the above.
12469 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
12471 (aarch64_<sur>abal<mode>_4): Likewise.
12472 (aarch64_<sur>adalp<mode>_3): Likewise.
12473 (<sur>sadv16qi): New define_expand.
12475 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
12477 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
12478 (*movdf_internal): Ditto.
12479 (*rcpsf2_sse): Ditto.
12480 (*rsqrtsf2_sse): Ditto.
12481 (*sqrt<mode>2_sse): Ditto.
12483 2018-05-21 Tamar Christina <tamar.christina@arm.com>
12485 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
12487 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
12488 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
12489 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
12490 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
12492 * config/aarch64/arm_neon.h: Likewise.
12493 * config/aarch64/iterators.md (VQ_I): New.
12495 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
12497 * config.gcc: Add arc/t-multilib-linux to tmake_file for
12499 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
12502 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
12504 * config/nds32/constraints.md (S): New constraint.
12505 * config/nds32/nds32.md (call_internal): Use constraint S.
12506 (call_value_internal): Likewise.
12507 (sibcall_internal): Likewise.
12508 (sibcall_value_internal): Likewise.
12510 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
12511 Chung-Ju Wu <jasonwucj@gmail.com>
12513 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
12514 into consideration.
12516 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
12517 Chung-Ju Wu <jasonwucj@gmail.com>
12519 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
12520 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
12521 (nds32_rtx_costs_impl): Simplify.
12522 (nds32_address_cost_impl): Simplify.
12523 (nds32_init_rtx_costs): New function.
12524 (nds32_rtx_costs_speed_prefer): Likewise.
12525 (nds32_rtx_costs_size_prefer): Likewise.
12526 (nds32_address_cost_speed_prefer): Likewise.
12527 (nds32_address_cost_speed_fwprop): Likewise.
12528 (nds32_address_cost_size_prefer): Likewise.
12529 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
12530 * config/nds32/nds32.c (nds32_option_override): Use
12531 nds32_init_rtx_costs function.
12533 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
12535 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
12536 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
12537 (TARGET_PIPELINE_N8): Likewise.
12538 (TARGET_PIPELINE_N10): Likewise.
12539 (TARGET_PIPELINE_N13): Likewise.
12540 (TARGET_PIPELINE_GRAYWOLF): Likewise.
12542 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
12544 * config/nds32/nds32-fpu.md: Update copyright year.
12546 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12548 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
12550 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12552 * config/nds32/nds32.c
12553 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
12554 * config/nds32/nds32.opt (minline-asm-r15): New option.
12556 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12558 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
12560 * config/nds32/nds32.md (abssi2): New pattern.
12562 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
12564 * config/i386/i386.md (rex64namesuffix): New mode attribute.
12565 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
12566 Merge insn pattern from sse_cvtsi2ss<round_name> and
12567 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
12568 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
12569 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
12570 using SWI48 mode iterator.
12571 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
12572 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
12573 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
12574 pattern from sse_cvttss2si<round_saeonly_name>
12575 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
12576 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
12577 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
12578 using SWI48 mode iterator.
12579 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
12580 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
12581 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
12582 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
12583 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
12584 using SWI48 mode iterator.
12585 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
12586 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
12587 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
12588 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
12589 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
12590 SWI48 mode iterator.
12591 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
12592 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
12593 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
12594 pattern from sse_cvttsd2si<round_saeonly_name>
12595 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
12597 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12599 * config/nds32/nds32-md-auxiliary.c
12600 (nds32_valid_smw_lwm_base_p): Refine.
12601 (nds32_output_smw_single_word): Refine.
12602 (nds32_output_smw_double_word): New.
12603 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
12605 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12607 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
12608 (nds32_output_stack_pop): Refine.
12609 (nds32_expand_unaligned_load): Refine.
12610 (nds32_expand_unaligned_store): Refine.
12612 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12613 Chung-Ju Wu <jasonwucj@gmail.com>
12615 * config/nds32/constants.md: Add TP_REGNUM constant.
12616 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
12617 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
12619 * config/nds32/nds32-doubleword.md: Consider flag_pic.
12620 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
12621 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
12622 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
12623 and PIC code generation.
12624 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
12626 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
12628 * config/nds32/nds32.md: Support TLS and PIC.
12629 * config/nds32/nds32.c: Support TLS and PIC.
12630 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
12631 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
12634 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12636 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
12637 mode with E_ prefix.
12639 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12640 Chung-Ju Wu <jasonwucj@gmail.com>
12642 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
12643 * config/nds32/nds32-md-auxiliary.c
12644 (symbolic_reference_mentioned_p): New.
12645 (nds32_legitimize_ict_address): New.
12646 (nds32_expand_ict_move): New.
12647 (nds32_indirect_call_referenced_p): New.
12648 (nds32_symbol_binds_local_p): Delete.
12649 (nds32_long_call_p): Modify.
12650 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
12651 * config/nds32/nds32-protos.h
12652 (symbolic_reference_mentioned_p): Declare.
12653 (nds32_legitimize_ict_address): Declare.
12654 (nds32_expand_ict_move): Declare.
12655 (nds32_indirect_call_referenced_p): Declare.
12656 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
12657 (nds32_relax_group): Use nds32_ict_const_p as condition.
12658 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
12659 (nds32_asm_file_start): Output ict_model directive in asm code.
12660 (nds32_legitimate_address_p): Consider indirect call.
12661 (nds32_print_operand): Consider indirect call.
12662 (nds32_print_operand_address): Consider indirect call.
12663 (nds32_insert_attributes): Handle "indirect_call" attribute.
12664 (TARGET_LEGITIMATE_ADDRESS_P): Define.
12665 (TARGET_LEGITIMATE_CONSTANT_P): Define.
12666 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
12667 (TARGET_DELEGITIMIZE_ADDRESS): Define.
12668 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
12669 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
12670 (TARGET_ICT_MODEL_SMALL): Define.
12671 (TARGET_ICT_MODEL_LARGE): Define.
12672 * config/nds32/nds32.md (movsi): Consider ict model.
12673 (call, call_value): Consider ict model.
12674 (sibcall, sibcall_value): Consider ict model.
12675 * config/nds32/nds32.opt (mict-model): New option.
12676 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
12679 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
12680 Monk Chiang <sh.chiang04@gmail.com>
12681 Jim Wilson <jimw@sifive.com>
12683 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
12684 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
12685 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
12686 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
12687 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
12688 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
12689 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
12690 compute save_libcall_adjustment properly.
12691 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
12692 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
12693 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
12694 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
12695 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
12696 (ABI_SPEC): Handle mabi=ilp32e.
12697 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
12698 (RVE): Add RVE mask.
12699 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
12700 <-march>: Add rv32e as an example.
12702 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
12705 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
12706 (intra_create_variable_infos): Handle C++ constructors.
12708 2018-05-18 Martin Liska <mliska@suse.cz>
12710 * passes.def: Remove a redundant pass.
12712 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
12715 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
12717 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12719 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
12721 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
12723 (armv2, armv2a, armv3, armv3m): Delete architectures.
12724 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
12725 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
12726 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
12728 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
12729 (*mulsidi3adddi): Likewise.
12730 (mulsidi3): Likewise.
12731 (*mulsidi3_nov6): Likewise.
12732 (umulsidi3): Likewise.
12733 (umulsidi3_nov6): Likewise.
12734 (umaddsidi4): Likewise.
12735 (*umulsidi3adddi): Likewise.
12736 (smulsi3_highpart): Likewise.
12737 (*smulsi3_highpart_nov6): Likewise.
12738 (umulsi3_highpart): Likewise.
12739 (*umulsi3_highpart_nov6): Likewise.
12740 * config/arm/arm.h (arm_arch3m): Delete.
12741 * config/arm/arm.c (arm_arch3m): Delete.
12742 (arm_option_override_internal): Update armv3-related comment.
12743 (arm_configure_build_target): Delete use of isa_bit_mode32.
12744 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
12745 (arm_rtx_costs_internal): Delete check of arm_arch3m.
12746 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
12747 (mulsa3): Likewise.
12748 (mulusa3): Likewise.
12749 * config/arm/arm-protos.h (arm_arch3m): Delete.
12750 * config/arm/arm-tables.opt: Regenerate.
12751 * config/arm/arm-tune.md: Likewise.
12752 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
12753 deleted architectures.
12755 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12757 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
12758 (armv5t, armv5te): New features.
12759 (ARMv5, ARMv5e): Delete fgroups.
12760 (ARMv5t, ARMv5te): Adjust for above changes.
12761 (ARMv6m): Likewise.
12762 (armv5, armv5e): Delete arches.
12763 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
12765 (*call_reg_arm): Likewise.
12766 (*call_value_reg_armv5): Likewise.
12767 (*call_value_reg_arm): Likewise.
12768 (*call_symbol): Likewise.
12769 (*call_value_symbol): Likewise.
12770 (*sibcall_insn): Likewise.
12771 (*sibcall_value_insn): Likewise.
12772 (clzsi2): Likewise.
12773 (prefetch): Likewise.
12774 (define_split and define_peephole2 dependent on arm_arch5):
12776 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
12778 (TARGET_ARM_QBIT): Likewise.
12779 (TARGET_DSP_MULTIPLY): Likewise.
12780 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
12781 (arm_arch5, arm_arch5e): Delete.
12782 (arm_arch5t, arm_arch5te): Declare.
12783 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
12784 (arm_arch5t): Declare.
12785 (arm_option_reconfigure_globals): Update for the above.
12786 (arm_options_perform_arch_sanity_checks): Update comment, replace
12787 use of arm_arch5 with arm_arch5t.
12788 (use_return_insn): Likewise.
12789 (arm_emit_call_insn): Likewise.
12790 (output_return_instruction): Likewise.
12791 (arm_final_prescan_insn): Likewise.
12792 (arm_coproc_builtin_available): Likewise.
12793 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
12794 arm_arch5e with arm_arch5t and arm_arch5te.
12795 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
12796 (arm_arch5t, arm_arch5te): Declare.
12797 * config/arm/arm-tables.opt: Regenerate.
12798 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
12799 * config/arm/t-multilib: Likewise.
12800 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
12801 instead of arm_arch5.
12802 (*call_reg_thumb1): Likewise.
12803 (*call_value_reg_thumb1_v5): Likewise.
12804 (*call_value_reg_thumb1): Likewise.
12805 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
12807 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
12809 2018-05-18 Martin Liska <mliska@suse.cz>
12811 PR gcov-profile/84846
12812 * doc/gcov.texi: Document -t option of gcov tool.
12814 2018-05-18 Martin Liska <mliska@suse.cz>
12816 PR gcov-profile/84846
12817 * gcov.c (print_usage): Add new -t option.
12818 (process_args): Handle the option.
12819 (generate_results): Use stdout as output when requested by
12822 2018-05-18 Martin Liska <mliska@suse.cz>
12824 PR gcov-profile/84846
12825 * coverage.c (coverage_init): Write PWD to .gcno file.
12826 * doc/gcov.texi: Document how working directory is printed.
12827 * gcov-dump.c (dump_gcov_file): Print PWD.
12828 * gcov.c (output_intermediate_file): Likewise.
12829 (read_graph_file): Read PWD string.
12830 (output_lines): Print PWD.
12832 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12834 PR middle-end/85817
12835 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
12836 for retval and return false if all args to phi are zero.
12838 2018-05-18 Richard Biener <rguenther@suse.de>
12840 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
12842 (evrp_dom_walker::before_dom_children): Call it.
12844 2018-05-18 Richard Biener <rguenther@suse.de>
12846 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
12847 results when processing array refs with variable index.
12849 2018-05-18 Toon Moene <toon@moene.org>
12851 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
12852 directly after that of -floop-interchange. Indicate that both
12853 options are enabled by default when specifying -O3.
12855 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12857 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
12858 iterator. Delete separate integer-mode vec_set<mode> expander.
12859 (aarch64_simd_vec_setv2di): Delete.
12860 (vec_setv2di): Delete.
12861 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
12862 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
12863 the "w, r" alternative.
12865 2018-05-18 Martin Liska <mliska@suse.cz>
12867 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
12868 * tree-pass.h (make_pass_lower_switch_O0): New function.
12869 * tree-switch-conversion.c (node_has_low_bound): Remove.
12870 (node_has_high_bound): Likewise.
12871 (node_is_bounded): Likewise.
12872 (class pass_lower_switch): Make it a template type and create
12874 (pass_lower_switch::execute): Add template argument.
12875 (make_pass_lower_switch): New function.
12876 (make_pass_lower_switch_O0): New function.
12877 (do_jump_if_equal): Remove.
12878 (emit_case_nodes): Simplify to just handle all 3 cases and leave
12879 all the hard work to tree optimization passes.
12881 2018-05-18 Martin Liska <mliska@suse.cz>
12883 * dbgcnt.c (limit_low): Renamed from limit.
12884 (limit_high): New variable.
12885 (dbg_cnt_is_enabled): Check for upper limit.
12886 (dbg_cnt): Adjust dumping.
12887 (dbg_cnt_set_limit_by_index): Add new argument for high
12889 (dbg_cnt_set_limit_by_name): Likewise.
12890 (dbg_cnt_process_single_pair): Parse new format.
12891 (dbg_cnt_process_opt): Use strtok.
12892 (dbg_cnt_list_all_counters): Remove 'value' and add
12894 * doc/invoke.texi: Document changes.
12896 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
12898 * doc/sourcebuild.texi (scalar_all_fma): Document.
12899 * tree.def (FMA_EXPR): Delete.
12900 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
12901 * internal-fn.c (ternary_direct): New macro.
12902 (expand_ternary_optab_fn): Likewise.
12903 (direct_ternary_optab_supported_p): Likewise.
12904 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
12905 * builtins.c (fold_builtin_fma): Delete.
12906 (fold_builtin_3): Don't call it.
12907 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
12908 * expr.c (expand_expr_real_2): Likewise.
12909 * fold-const.c (operand_equal_p): Likewise.
12910 (fold_ternary_loc): Likewise.
12911 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
12912 * gimple.c (DEFTREECODE): Likewise.
12913 * gimplify.c (gimplify_expr): Likewise.
12914 * optabs-tree.c (optab_for_tree_code): Likewise.
12915 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
12916 * tree-eh.c (operation_could_trap_p): Likewise.
12917 (stmt_could_throw_1_p): Likewise.
12918 * tree-inline.c (estimate_operator_cost): Likewise.
12919 * tree-pretty-print.c (dump_generic_node): Likewise.
12920 (op_code_prio): Likewise.
12921 * tree-ssa-loop-im.c (stmt_cost): Likewise.
12922 * tree-ssa-operands.c (get_expr_operands): Likewise.
12923 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
12924 * fold-const-call.h (fold_fma): Delete.
12925 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
12926 CFN_FNMA and CFN_FNMS.
12927 (fold_fma): Delete.
12928 * genmatch.c (combined_fn): New enum.
12929 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
12930 (commutative_op): New function.
12931 (commutate): Use it. Handle more than 2 operands.
12932 (dt_operand::gen_gimple_expr): Use commutative_op.
12933 (parser::parse_expr): Allow :c to be used with non-binary
12934 operators if the commutative operand is known.
12935 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
12936 CFN_FMS, CFN_FNMA and CFN_FNMS.
12937 (backprop::process_assign_use): Remove FMA_EXPR handling.
12938 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
12939 (gen_hsa_fma): New function.
12940 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
12941 IFN_FNMA and IFN_FNMS.
12942 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
12943 * gimple-fold.h (follow_all_ssa_edges): Declare.
12944 * gimple-fold.c (follow_all_ssa_edges): New function.
12945 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
12946 gimple_build interface and use follow_all_ssa_edges to fold the result.
12947 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
12948 instead of checking for optabs directly.
12949 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
12950 rather than FMA_EXPRs.
12951 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
12952 call to IFN_FMA instead of an FMA_EXPR.
12954 2018-05-17 Jim Wilson <jimw@sifive.com>
12956 * expr.c (do_tablejump): When converting index to Pmode, if we have a
12957 sign extended promoted subreg, and the range does not have the sign bit
12958 set, then do a sign extend.
12960 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
12961 test, check for sign extended subreg and/or constant operands, and
12962 do a sign extend in that case.
12964 2018-05-17 Steve Ellcey <sellcey@cavium.com>
12966 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
12967 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
12969 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
12970 Change logics_shift_reg to logics_shift_imm.
12971 (thunderx2t99_fp_loadpair_basic): Delete.
12972 (thunderx2t99_fp_storepair_basic): Delete.
12973 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
12974 (thunderx2t99_asimd_polynomial): Delete.
12975 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
12976 and neon_fp_mul_d_scalar_q.
12977 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
12978 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
12979 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
12980 (thunderx2t99_asimd_lut): Add missing tbl types.
12981 (thunderx2t99_asimd_ext): Delete.
12982 (thunderx2t99_asimd_load1_1_mult): Delete.
12983 (thunderx2t99_asimd_load1_2_mult): Delete.
12984 (thunderx2t99_asimd_load1_ldp): New.
12985 (thunderx2t99_asimd_load1): New.
12986 (thunderx2t99_asimd_load2): Add missing *load2* types.
12987 (thunderx2t99_asimd_load3): New.
12988 (thunderx2t99_asimd_load4): New.
12989 (thunderx2t99_asimd_store1_1_mult): Delete.
12990 (thunderx2t99_asimd_store1_2_mult): Delete.
12991 (thunderx2t99_asimd_store2_mult): Delete.
12992 (thunderx2t99_asimd_store2_onelane): Delete.
12993 (thunderx2t99_asimd_store_stp): New.
12994 (thunderx2t99_asimd_store1): New.
12995 (thunderx2t99_asimd_store2): New.
12996 (thunderx2t99_asimd_store3): New.
12997 (thunderx2t99_asimd_store4): New.
12999 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
13001 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
13002 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
13004 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
13005 Segher Boessenkool <segher@kernel.crashing.org>
13008 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
13011 2018-05-17 Richard Biener <rguenther@suse.de>
13013 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
13014 for pruning loop and prune defs feeding only already visited PHIs.
13016 2018-05-17 Richard Biener <rguenther@suse.de>
13018 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
13020 2018-05-17 Bin Cheng <bin.cheng@arm.com>
13021 Richard Biener <rguenther@suse.de>
13023 PR tree-optimization/85793
13024 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
13025 for VMAT_ELEMENTWISE.
13027 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
13029 * internal-fn.h (lookup_internal_fn): Declare
13030 * internal-fn.c (lookup_internal_fn): New function.
13031 * gimple.c (gimple_build_call_from_tree): Handle calls to
13032 internal functions.
13033 * gimple-pretty-print.c (dump_gimple_call): Print "." before
13034 internal function names.
13035 * tree-pretty-print.c (dump_generic_node): Likewise.
13036 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
13038 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
13040 * gimple-fold.h (gimple_build): Make the function forms take
13041 combined_fn rather than built_in_function.
13042 (gimple_simplify): Likewise.
13043 * gimple-match-head.c (gimple_simplify): Likewise.
13044 * gimple-fold.c (gimple_build): Likewise.
13045 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
13046 rather than gimple_build_call_internal.
13047 (get_initial_defs_for_reduction): Likewise.
13048 (vect_create_epilog_for_reduction): Likewise.
13049 (vectorizable_live_operation): Likewise.
13051 2018-05-17 Martin Liska <mliska@suse.cz>
13053 * gimple-ssa-sprintf.c (format_directive): Do not use
13054 space in between 'G_' and '('.
13056 2018-05-17 Jakub Jelinek <jakub@redhat.com>
13059 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
13060 even if the mask is not all ones.
13063 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
13065 (ix86_gimple_fold_builtin): Likewise.
13068 * config/i386/i386.c: Include tree-vector-builder.h.
13069 (ix86_vector_shift_count): New function.
13070 (ix86_fold_builtin): Fold shift builtins by scalar count.
13071 (ix86_gimple_fold_builtin): Likewise.
13073 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
13074 _mm512_setzero): New intrinsics.
13076 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
13077 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13079 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
13080 code generation for cases where splatting a value is not useful.
13081 * simplify-rtx.c (simplify_ternary_operation): Simplify
13082 vec_merge across a vec_duplicate and a paradoxical subreg forming
13083 a vector mode to a vec_concat.
13085 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
13087 * config.gcc: Support "goldmont-plus".
13088 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
13090 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13091 PROCESSOR_GOLDMONT_PLUS.
13092 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
13093 (processor_target_table): Add "goldmont-plus".
13094 (PTA_GOLDMONT_PLUS): Define.
13095 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
13096 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
13097 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
13098 (fold_builtin_cpu): Add "goldmont-plus".
13099 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
13100 (ix86_option_override_internal): Add "goldmont-plus".
13101 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
13102 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
13103 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
13104 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
13106 2018-05-17 Richard Biener <rguenther@suse.de>
13108 PR tree-optimization/85757
13109 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
13110 remove defs that only feed that PHI from further processing.
13112 2018-05-16 Jim Wilson <jimw@sifive.com>
13114 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
13116 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
13118 2018-05-16 Mark Wielaard <mark@klomp.org>
13120 * dwarf2out.c (count_index_strings): New function.
13121 (output_indirect_strings): Call count_index_strings and generate
13122 header for dwarf_version >= 5.
13124 2018-05-16 Mark Wielaard <mark@klomp.org>
13126 * dwarf2out.c (dwarf_FORM): New function.
13127 (set_indirect_string): Use dwarf_FORM.
13128 (reset_indirect_string): Likewise.
13129 (size_of_die): Likewise.
13130 (value_format): Likewise.
13131 (output_die): Likewise.
13132 (add_skeleton_AT_string): Likewise.
13133 (output_macinfo_op): Likewise.
13134 (index_string): Likewise.
13135 (output_index_string_offset): Likewise.
13136 (output_index_string): Likewise.
13137 (count_index_strings): Likewise.
13139 2018-05-16 Carl Love <cel@us.ibm.com>
13141 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13142 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
13144 2018-05-16 Martin Jambor <mjambor@suse.cz>
13146 * ipa-prop.c (ipa_free_all_edge_args): Remove.
13147 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
13149 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
13151 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
13152 (fnma<mode>4): Likewise.
13153 (fms<mode>4): Likewise.
13154 (fnms<mode>4): Likewise.
13155 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
13156 (aarch64_fnma<mode>4): Likewise.
13157 (aarch64_fms<mode>4): Likewise.
13158 (aarch64_fnms<mode>4): Likewise.
13159 (aarch64_fnmadd<mode>4): Likewise.
13161 2018-05-16 Jason Merrill <jason@redhat.com>
13163 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
13165 2018-05-16 Richard Biener <rguenther@suse.de>
13167 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
13168 (dump_stmt_cost): Declare.
13169 (add_stmt_cost): Dump cost we add.
13170 (add_stmt_costs): New function.
13171 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
13172 No longer exported.
13173 (vect_analyze_stmt): Adjust prototype.
13174 (vectorizable_condition): Likewise.
13175 (vectorizable_live_operation): Likewise.
13176 (vectorizable_reduction): Likewise.
13177 (vectorizable_induction): Likewise.
13178 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
13179 cost vector to pass to vectorizable_ and record afterwards.
13180 (vect_model_reduction_cost): Take cost vector argument and adjust.
13181 (vect_model_induction_cost): Likewise.
13182 (vectorizable_reduction): Likewise.
13183 (vectorizable_induction): Likewise.
13184 (vectorizable_live_operation): Likewise.
13185 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
13186 SLP_TREE_NUMBER_OF_VEC_STMTS.
13187 (vect_analyze_slp_cost_1): Remove.
13188 (vect_analyze_slp_cost): Likewise.
13189 (vect_slp_analyze_node_operations): Take visited args and
13190 a target cost vector. Avoid processing already visited stmt sets.
13191 (vect_slp_analyze_operations): Use a local cost vector to gather
13192 costs and register those of non-discarded instances.
13193 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
13194 (vect_schedule_slp_instance): Remove copying of
13195 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
13197 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
13198 adding cost. Record cost entry location.
13199 (vect_prologue_cost_for_slp_op): Function to compute cost of
13200 a constant or invariant generated for SLP vect in the prologue,
13201 split out from vect_analyze_slp_cost_1.
13202 (vect_model_simple_cost): Make static. Adjust for SLP costing.
13203 (vect_model_promotion_demotion_cost): Likewise.
13204 (vect_model_store_cost): Likewise, make static.
13205 (vect_model_load_cost): Likewise.
13206 (vectorizable_bswap): Add cost vector arg and adjust.
13207 (vectorizable_call): Likewise.
13208 (vectorizable_simd_clone_call): Likewise.
13209 (vectorizable_conversion): Likewise.
13210 (vectorizable_assignment): Likewise.
13211 (vectorizable_shift): Likewise.
13212 (vectorizable_operation): Likewise.
13213 (vectorizable_store): Likewise.
13214 (vectorizable_load): Likewise.
13215 (vectorizable_condition): Likewise.
13216 (vectorizable_comparison): Likewise.
13217 (can_vectorize_live_stmts): Likewise.
13218 (vect_analyze_stmt): Likewise.
13219 (vect_transform_stmt): Adjust calls to vectorizable_*.
13220 * tree-vectorizer.c: Include gimple-pretty-print.h.
13221 (dump_stmt_cost): New function.
13223 2018-05-16 Richard Biener <rguenther@suse.de>
13225 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
13226 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
13227 * tree-ssa-dse.c: Include tree-ssa-loop.h.
13228 (check_name): New callback.
13229 (dse_classify_store): Track cycles via a visited bitmap of PHI
13230 defs and simplify handling of in-loop and across loop dead stores
13231 and properly fail for loop-variant refs. Handle byte-tracking with
13232 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
13235 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
13237 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
13238 (vect_get_mask_type_for_stmt): Likewise.
13239 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
13241 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
13242 to determine the statement's vector type and the vector type that
13243 should be used for calculating nunits. Deal with cases in which
13244 the type has to be deferred.
13245 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
13246 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
13247 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
13248 (vect_determine_vf_for_stmt): New functions, split out from...
13249 (vect_determine_vectorization_factor): ...here.
13250 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
13251 (vect_get_mask_type_for_stmt): New functions, split out from
13252 vect_determine_vectorization_factor.
13254 2018-05-16 Richard Biener <rguenther@suse.de>
13256 * tree-cfg.c (verify_gimple_assign_ternary): Properly
13257 verify the [VEC_]COND_EXPR embedded comparison.
13259 2018-05-15 Martin Sebor <msebor@redhat.com>
13261 PR tree-optimization/85753
13262 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
13263 RECORD_TYPE in addition to ARRAY_TYPE.
13265 2018-05-15 Martin Sebor <msebor@redhat.com>
13267 PR middle-end/85643
13268 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
13270 2018-05-15 Richard Biener <rguenther@suse.de>
13272 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
13273 add by_clobber_p one. Change algorithm to collect all defs
13274 representing uses we need to walk and try reducing them to
13275 a single one before failing.
13276 (dse_dom_walker::dse_optimize_stmt): Adjust.
13278 2018-05-13 Mark Wielaard <mark@klomp.org>
13280 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
13281 (size_of_loc_descr): Likewise.
13282 (output_loc_operands): Likewise.
13283 (output_loc_operands_raw): Likewise.
13284 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
13285 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
13286 (hash_loc_operands): Likewise.
13287 (compare_loc_operands): Likewise.
13289 2018-05-14 Mark Wielaard <mark@klomp.org>
13291 * dwarf2out.c (count_index_addrs): New function.
13292 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
13294 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13296 PR tree-optimization/83648
13297 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
13298 return value as malloc candidate.
13300 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13303 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
13304 param as true in call to suggest_attribute.
13306 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
13308 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
13309 -mreadonly-in-sdata.
13311 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13313 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
13315 (aarch64_crypto_aesd_fused): Likewise.
13317 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
13319 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
13320 (movsi_aarch64): Likewise.
13321 (load_pairsi): Likewise.
13322 (load_pairdi): Likewise.
13323 (store_pairsi): Likewise.
13324 (store_pairdi): Likewise.
13325 (load_pairsf): Likewise.
13326 (load_pairdf): Likewise.
13327 (store_pairsf): Likewise.
13328 (store_pairdf): Likewise.
13329 (zero_extend): Likewise.
13330 (trunc): Swap alternatives.
13331 (fcvt_target): Add '?' to prefer w over r.
13333 2018-05-14 Jakub Jelinek <jakub@redhat.com>
13336 * config/i386/i386.md: Disallow non-commutative arithmetics in
13337 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
13338 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
13339 in the peephole2 before it.
13341 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
13343 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
13344 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
13345 (ix86_handle_option): Handle -mcldemote.
13346 * config.gcc: New header.
13347 * config/i386/cldemoteintrin.h: New file.
13348 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
13349 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
13351 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13352 OPTION_MASK_ISA_CLDEMOTE.
13353 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
13354 (ix86_valid_target_attribute_inner_p): Ditto.
13355 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
13356 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
13357 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
13358 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
13359 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
13361 * config/i386/i386.opt: Add -mcldemote.
13362 * config/i386/x86intrin.h: New header.
13363 * doc/invoke.texi: Add -mcldemote.
13365 2018-05-14 Richard Biener <rguenther@suse.de>
13367 * doc/match-and-simplify.texi: Adjust :s documentation.
13369 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
13371 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
13372 intended memcpy size.
13373 (REORDER_45): Likewise.
13375 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
13377 * sort.cc: New file.
13378 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
13379 * vec.c (qsort_chk): Use gcc_qsort.
13380 * Makefile.in (OBJS-libcommon): Add sort.o.
13381 (build/sort.o): New target. Use it...
13382 (BUILD_RTL): ... here, and...
13383 (build/gencfn-macros): ... here, and...
13384 (build/genmatch): ... here.
13386 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13387 Chung-Ju Wu <jasonwucj@gmail.com>
13389 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
13390 * config/nds32/nds32-graywolf.md: New file.
13391 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
13392 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
13394 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
13395 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
13396 * config/nds32/nds32.md (pipeline_model): Add graywolf.
13397 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
13398 * config/nds32/pipelines.md: Include n15 settings.
13400 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13401 Chung-Ju Wu <jasonwucj@gmail.com>
13403 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
13404 * config/nds32/nds32-n13.md: New file.
13405 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
13406 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
13408 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
13409 * config/nds32/nds32.md (pipeline_model): Add n13.
13410 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
13411 * config/nds32/pipelines.md: Include n13 settings.
13413 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13414 Chung-Ju Wu <jasonwucj@gmail.com>
13416 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
13417 * config/nds32/nds32-n10.md: New file.
13418 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
13419 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
13421 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
13422 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
13423 * config/nds32/nds32.md (pipeline_model): Add n10.
13424 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
13425 * config/nds32/pipelines.md: Include n10 settings.
13427 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
13428 Kito Cheng <kito.cheng@gmail.com>
13429 Chung-Ju Wu <jasonwucj@gmail.com>
13431 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
13432 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13433 Add enum values for DSP extension instructions.
13434 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
13436 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
13437 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
13438 New code iterators.
13439 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
13440 * config/nds32/nds32-dspext.md: New file for DSP implementation.
13441 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
13442 * config/nds32/nds32-intrinsic.md: Likewise.
13443 * config/nds32/nds32_intrinsic.h: Likewise.
13444 * config/nds32/nds32-md-auxiliary.c: Likewise.
13445 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
13446 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
13447 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
13448 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
13449 * config/nds32/nds32-protos.h: New declarations for DSP extension.
13450 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
13451 TYPE_DMAC in switch statement.
13452 * config/nds32/nds32.c: New checking and implementation for DSP
13453 extension instructions.
13454 * config/nds32/nds32.h: Likewise.
13455 * config/nds32/nds32.md: Likewise.
13456 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
13457 * config/nds32/predicates.md: Implement new predicates for DSP
13460 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13462 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
13463 Reformat alternatives and attributes so it is easier to identify
13464 which constraints/attributes go with which instruction.
13465 (mov<mode>_hardfloat32, FMOVE64): Likewise.
13466 (mov<mode>_softfloat32, FMOVE64): Likewise.
13467 (mov<mode>_hardfloat64, FMOVE64): Likewise.
13468 (mov<mode>_softfloat64, FMOVE64): Likewise.
13470 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
13472 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13474 (Basic PowerPC Built-in Functions): The new name of the
13475 subsection previously known as "PowerPC Built-in Functions".
13476 (Basic PowerPC Built-in Functions Available on all Configurations):
13478 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
13479 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
13480 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
13481 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
13483 2018-05-11 Martin Jambor <mjambor@suse.cz>
13486 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
13489 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
13492 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
13494 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
13496 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
13497 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
13498 (ix86_handle_option): Handle -mwaitpkg.
13499 * config.gcc: New header.
13500 * config/i386/cpuid.h (bit_WAITPKG): New bit.
13501 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
13502 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
13504 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13505 OPTION_MASK_ISA_WAITPKG.
13506 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
13507 (ix86_option_override_internal): Add PTA_WAITPKG.
13508 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
13509 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
13510 IX86_BUILTIN_TPAUSE.
13511 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
13512 __builtin_ia32_umwait and __builtin_ia32_tpause.
13513 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
13514 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
13515 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
13516 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
13517 UNSPECV_TPAUSE): New.
13518 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
13519 * config/i386/i386.opt: Add -mwaitpkg.
13520 * config/i386/waitpkgintrin.h: New file.
13521 * config/i386/x86intrin.h: New header.
13522 * doc/invoke.texi: Add -mwaitpkg.
13524 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
13527 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
13529 (cortex-m0): Use armv6s-m isa.
13530 (cortex-m0plus): Likewise.
13531 (cortex-m1): Likewise.
13532 (cortex-m0.small-multiply): Likewise.
13533 (cortex-m0plus.small-multiply): Likewise.
13534 (cortex-m1.small-multiply): Likewise.
13536 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
13537 Jakub Jelinek <jakub@redhat.com>
13539 PR tree-optimization/85692
13540 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
13541 source permute as well.
13543 2018-05-11 Martin Liska <mliska@suse.cz>
13546 * doc/extend.texi: Document LLVM style format for no_sanitize
13549 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13551 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
13552 mode_supports_vsx_dform_quad to mode_supports_dq_form.
13553 (mode_supports_vsx_dform_quad): Likewise.
13554 (mode_supports_vmx_dform): Move these functions to be next to the
13555 other mode_supports functions.
13556 (mode_supports_dq_form): Likewise.
13557 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
13558 mode_supports_dq_form.
13559 (reg_offset_addressing_ok_p): Likewise.
13560 (offsettable_ok_by_alignment): Likewise.
13561 (rs6000_legitimate_offset_address_p): Likewise.
13562 (legitimate_lo_sum_address_p): Likewise.
13563 (rs6000_legitimize_address): Likewise.
13564 (rs6000_legitimize_reload_address): Likewise.
13565 (rs6000_secondary_reload_inner): Likewise.
13566 (rs6000_preferred_reload_class): Likewise.
13567 (rs6000_output_move_128bit): Likewise.
13569 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
13571 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
13572 Generate SImode target register for null target.
13573 <case IX86_BUILTIN_XGETBV>: Ditto.
13574 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
13575 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
13577 2018-05-10 Carl Love <cel@us.ibm.com>
13579 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13580 dcbtt and dcbtstt if operands[2] is 0.
13582 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
13585 * config/i386/sse.md (usadv64qi): New expander.
13587 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
13589 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
13590 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
13591 -maltivec=be support.
13592 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
13593 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
13594 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
13595 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
13596 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
13597 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
13598 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
13599 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
13600 altivec_vsumsws): Adjust.
13601 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
13602 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
13603 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
13605 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
13606 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
13607 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
13608 (altivec_lve<VI_char>x): Delete expand.
13609 (*altivec_lve<VI_char>x_internal): Rename to...
13610 (altivec_lve<VI_char>x): ... this.
13611 (altivec_lvxl_<mode>): Delete expand.
13612 (*altivec_lvxl_<mode>_internal): Rename to ...
13613 (altivec_lvxl_<mode>): ... this.
13614 (altivec_stvxl_<mode>): Delete expand.
13615 (*altivec_stvxl_<mode>_internal): Rename to ...
13616 (altivec_stvxl_<mode>): ... this.
13617 (altivec_stve<VI_char>x): Delete expand.
13618 (*altivec_stve<VI_char>x_internal): Rename to ...
13619 (altivec_stve<VI_char>x): ... this.
13620 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
13621 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
13622 reduc_plus_scal_<mode>): Adjust.
13623 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
13625 (rs6000_cpu_cpp_builtins): Adjust.
13626 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
13627 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
13628 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
13629 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
13630 -maltivec=be support.
13631 (rs6000_split_vec_extract_var): Adjust.
13632 (rs6000_split_v4si_init): Adjust.
13633 (swap_selector_for_mode): Delete.
13634 (altivec_expand_lvx_be, altivec_expand_stvx_be,
13635 altivec_expand_stvex_be): Delete.
13636 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
13637 -maltivec=be support.
13638 (rs6000_gimple_fold_builtin): Ditto.
13639 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
13641 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
13642 (TARGET_DIRECT_MOVE_64BIT): Adjust.
13643 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
13644 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
13645 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
13646 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
13647 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
13648 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
13649 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
13650 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
13651 anonymous split): Adjust.
13652 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
13653 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
13655 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
13657 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
13658 when --with-gxx-include-dir is also specified.
13659 * configure: Regenerate.
13661 2018-05-09 Jim Wilson <jimw@sifive.com>
13664 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
13665 * config/riscv/t-withmultilib: New.
13666 * config/riscv/withmultilib.h: New.
13667 * doc/install.texi: Document RISC-V --with-multilib-list support.
13669 2018-05-09 Richard Biener <rguenther@suse.de>
13671 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
13673 (vect_bb_vectorization_profitable_p): Adjust. Compute
13674 actual scalar cost using the cost vector and the add_stmt_cost
13677 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13679 PR rtl-optimization/85645
13680 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
13681 in the REG_CFA_REGISTER note for LR, don't leave it empty.
13683 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13685 PR rtl-optimization/85645
13686 * shrink-wrap.c (spread_components): Return a boolean saying if
13687 anything was changed.
13688 (try_shrink_wrapping_separate): Iterate spread_components until
13689 nothing changes anymore.
13691 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13693 PR rtl-optimization/85645
13694 * regrename.c (build_def_use): Also kill the chains that include the
13695 destination of a REG_CFA_REGISTER note.
13697 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13699 PR rtl-optimization/85645
13700 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
13701 insn that has a REG_CFA_REGISTER note.
13703 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
13705 * cfgexpand.c (expand_clobber): New function.
13706 (expand_gimple_stmt_1): Use it.
13707 * tree-vect-stmts.c (vect_clobber_variable): New function,
13709 (vectorizable_simd_clone_call): ...here.
13710 (vectorizable_store): Emit a clobber either side of an
13711 IFN_STORE_LANES sequence.
13712 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
13714 2018-05-09 Tom de Vries <tom@codesourcery.com>
13717 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
13718 (define_insn "trap_if_false"): Add exit after trap.
13720 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
13722 PR rtl-optimization/85638
13723 * bb-reorder.c: Include common/common-target.h.
13724 (create_forwarder_block): New function extracted from...
13725 (fix_up_crossing_landing_pad): ...here. Rename into...
13726 (dw2_fix_up_crossing_landing_pad): ...this.
13727 (sjlj_fix_up_crossing_landing_pad): New function.
13728 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
13729 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
13730 from both partitions and exit the loop after one iteration.
13732 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
13735 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13737 (Basic PowerPC Built-in Functions): The new name of the
13738 subsection previously known as "PowerPC Built-in Functions".
13739 (Basic PowerPC Built-in Functions Available on all Configurations):
13741 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13743 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13745 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13747 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13750 2018-05-08 Jim Wilson <jimw@sifive.com>
13752 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
13753 (LD_EMUL_SUFFIX): New.
13754 (LINK_SPEC): Use it.
13756 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
13758 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13760 (Basic PowerPC Built-in Functions): The new name of the
13761 subsection previously known as "PowerPC Built-in Functions".
13762 (Basic PowerPC Built-in Functions Available on all Configurations):
13764 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13766 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13768 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13770 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13773 2018-05-08 Jakub Jelinek <jakub@redhat.com>
13776 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
13777 after cmpelim optimization.
13779 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
13781 * config.gcc: Support "goldmont".
13782 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
13783 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13784 PROCESSOR_GOLDMONT.
13785 * config/i386/i386.c (m_GOLDMONT): Define.
13786 (processor_target_table): Add "goldmont".
13787 (PTA_GOLDMONT): Define.
13788 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
13789 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
13790 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
13791 (fold_builtin_cpu): Add "goldmont".
13792 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
13793 (ix86_option_override_internal): Add "goldmont".
13794 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
13795 (processor_type): Add PROCESSOR_GOLDMONT.
13796 * config/i386/i386.md: Add CPU "glm".
13797 * config/i386/glm.md: New file.
13798 * config/i386/x86-tune.def: Add m_GOLDMONT.
13799 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
13801 2018-05-08 Jakub Jelinek <jakub@redhat.com>
13804 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
13806 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
13807 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
13808 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
13811 * config/i386/i386.c (ix86_fold_builtin): Handle
13812 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
13815 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
13816 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
13818 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
13821 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
13822 (check_arch): Likewise.
13823 (check_fpu): Return the result rather than printing it.
13824 (end arch): Fix operator precedence.
13825 (end cpu): Likewise.
13826 (END): Print the result from check_fpu.
13828 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
13829 Alan Hayward <alan.hayward@arm.com>
13830 David Sherwood <david.sherwood@arm.com>
13832 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
13833 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
13834 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
13835 (*fcmuo<mode>_and): New patterns.
13837 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
13839 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
13840 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
13841 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
13842 (cmp_op, sve_imm_con): New code attributes.
13843 (SVE_COND_INT_CMP, imm_con): Delete.
13844 (cmp_op): Remove above unspecs from int attribute.
13845 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
13847 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
13848 comparison-specific unspecs.
13849 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
13850 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
13851 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
13852 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
13853 (*vec_fcm<cmp_op><mode>): Rename to...
13854 (*fcm<cmp_op><mode>): ...this and adjust likewise.
13855 (*vec_fcmuo<mode>): Rename to...
13856 (*fcmuo<mode>): ...this and adjust likewise.
13857 (*pred_fcm<cmp_op><mode>): New pattern.
13858 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
13859 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
13861 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
13863 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
13864 (aarch64_emit_sve_predicated_cond): New function.
13865 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
13866 (aarch64_emit_unspec_cond_or): Replace with...
13867 (aarch64_emit_sve_or_conds): ...this new function. Use
13868 aarch64_emit_sve_ptrue_op for the individual comparisons and
13869 aarch64_emit_binop to OR them together.
13870 (aarch64_emit_inverted_unspec_cond): Replace with...
13871 (aarch64_emit_sve_inverted_cond): ...this new function. Use
13872 aarch64_emit_sve_ptrue_op for the comparison and
13873 aarch64_emit_unop to invert the result.
13874 (aarch64_expand_sve_vec_cmp_float): Update after the above
13875 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
13877 2018-05-07 Nathan Sidwell <nathan@acm.org>
13879 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
13880 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
13881 (Backwards Compatibility): Likewise.
13883 2018-05-07 Luis Machado <luis.machado@linaro.org>
13887 2018-05-07 Luis Machado <luis.machado@linaro.org>
13889 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13890 <prefetch_dynamic_strides>: New const bool field.
13891 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13892 prefetch_dynamic_strides.
13893 (exynosm1_prefetch_tune): Likewise.
13894 (thunderxt88_prefetch_tune): Likewise.
13895 (thunderx_prefetch_tune): Likewise.
13896 (thunderx2t99_prefetch_tune): Likewise.
13897 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13899 (aarch64_override_options_internal): Update to set
13900 PARAM_PREFETCH_DYNAMIC_STRIDES.
13901 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13902 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13903 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13904 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13905 prefetch-dynamic-strides setting.
13907 2018-05-07 Luis Machado <luis.machado@linaro.org>
13909 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13910 <minimum_stride>: New const int field.
13911 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13912 minimum_stride field.
13913 (exynosm1_prefetch_tune): Likewise.
13914 (thunderxt88_prefetch_tune): Likewise.
13915 (thunderx_prefetch_tune): Likewise.
13916 (thunderx2t99_prefetch_tune): Likewise.
13917 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13918 (aarch64_override_options_internal): Update to set
13919 PARAM_PREFETCH_MINIMUM_STRIDE.
13920 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13921 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13922 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13923 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13924 stride is constant and is below the minimum stride threshold.
13926 2018-05-07 Luis Machado <luis.machado@linaro.org>
13928 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
13931 2018-05-07 Luis Machado <luis.machado@linaro.org>
13933 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13934 <prefetch_dynamic_strides>: New const bool field.
13935 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13936 prefetch_dynamic_strides.
13937 (exynosm1_prefetch_tune): Likewise.
13938 (thunderxt88_prefetch_tune): Likewise.
13939 (thunderx_prefetch_tune): Likewise.
13940 (thunderx2t99_prefetch_tune): Likewise.
13941 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13943 (aarch64_override_options_internal): Update to set
13944 PARAM_PREFETCH_DYNAMIC_STRIDES.
13945 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13946 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13947 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13948 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13949 prefetch-dynamic-strides setting.
13951 2018-05-07 Luis Machado <luis.machado@linaro.org>
13953 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13954 <minimum_stride>: New const int field.
13955 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13956 minimum_stride field.
13957 (exynosm1_prefetch_tune): Likewise.
13958 (thunderxt88_prefetch_tune): Likewise.
13959 (thunderx_prefetch_tune): Likewise.
13960 (thunderx2t99_prefetch_tune): Likewise.
13961 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13962 (aarch64_override_options_internal): Update to set
13963 PARAM_PREFETCH_MINIMUM_STRIDE.
13964 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13965 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13966 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13967 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13968 stride is constant and is below the minimum stride threshold.
13970 2018-05-06 Jakub Jelinek <jakub@redhat.com>
13973 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
13974 the type is addressable. Don't force op into register if it has
13977 2018-05-05 Roland McGrath <mcgrathr@google.com>
13980 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
13981 any section for which we don't know a specific type it should have,
13982 regardless of name. Previously this was done only for the exact
13983 names ".init_array", ".fini_array", and ".preinit_array".
13984 (default_elf_asm_named_section): Add comment about
13985 relationship with default_section_type_flags and SECTION_NOTYPE.
13986 (get_section): Don't consider it a type conflict if one side has
13987 SECTION_NOTYPE and the other doesn't, as long as neither has the
13988 SECTION_BSS et al used in the default_section_type_flags logic.
13990 2018-05-05 Tom de Vries <tom@codesourcery.com>
13993 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
13994 (workaround_barsyncs): New function.
13995 (nvptx_reorg): Use workaround_barsyncs.
13996 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
13997 (define_expand "nvptx_membar_cta"): New define_expand.
13998 (define_insn "*nvptx_membar_cta"): New insn.
14000 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
14002 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
14003 To improve optimization opportunities.
14004 * builtin-types.def: The new needed builtin types for the above.
14006 2018-05-04 Richard Biener <rguenther@suse.de>
14008 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
14009 * gimple-ssa-store-merging.c
14010 (imm_store_chain_info::output_merged_store): Remove redundant create,
14011 release split_store vector contents on failure.
14012 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
14013 scalar stmt vector on cache hit.
14015 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
14017 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
14019 * config.gcc (powerpc-xilinx-eabi*): Remove.
14020 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
14022 (fusion_addis_mem_combo_load): Ditto.
14023 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
14025 (rs6000_cpu_cpp_builtins): Ditto.
14026 * config/rs6000/rs6000-linux.c
14027 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
14028 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
14029 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
14031 (rs6000_setup_reg_addr_masks): Ditto.
14032 (rs6000_init_hard_regno_mode_ok): Ditto.
14033 (rs6000_option_override_internal): Ditto.
14034 (legitimate_lo_sum_address_p): Ditto.
14035 (rs6000_legitimize_address): Ditto.
14036 (rs6000_legitimize_reload_address): Ditto.
14037 (rs6000_legitimate_address_p): Ditto.
14038 (abi_v4_pass_in_fpr): Ditto.
14039 (setup_incoming_varargs): Ditto.
14040 (rs6000_gimplify_va_arg): Ditto.
14041 (rs6000_split_multireg_move): Ditto.
14042 (rs6000_savres_strategy): Ditto.
14043 (rs6000_emit_prologue_components): Ditto.
14044 (rs6000_emit_epilogue_components): Ditto.
14045 (rs6000_emit_prologue): Ditto.
14046 (rs6000_emit_epilogue): Ditto.
14047 (rs6000_elf_file_end): Ditto.
14048 (rs6000_function_value): Ditto.
14049 (rs6000_libcall_value): Ditto.
14050 * config/rs6000/rs6000.h: Ditto.
14051 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
14052 (TARGET_MINMAX): ... this. New.
14053 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
14054 * config/rs6000/rs6000.md: Remove Xilinx FP support.
14055 (*movsi_internal1_single): Delete.
14056 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
14057 mfpu=, mxilinx-fpu): Delete.
14058 * config/rs6000/singlefp.h: Delete.
14059 * config/rs6000/sysv4.h: Remove Xilinx FP support.
14060 * config/rs6000/t-rs6000: Ditto.
14061 * config/rs6000/t-xilinx: Delete.
14062 * config/rs6000/titan.md: Adjust for fp_type removal.
14063 * config/rs6000/vsx.md: Remove Xilinx FP support.
14064 (VStype_simple): Delete.
14065 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
14066 * config/rs6000/xfpu.h: Delete.
14067 * config/rs6000/xfpu.md: Delete.
14068 * config/rs6000/xilinx.h: Delete.
14069 * config/rs6000/xilinx.opt: Delete.
14070 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
14071 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
14073 2018-05-04 Tom de Vries <tom@codesourcery.com>
14076 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
14079 2018-05-04 Richard Biener <rguenther@suse.de>
14081 PR middle-end/85627
14082 * tree-complex.c (update_complex_assignment): We are always in SSA form.
14083 (expand_complex_div_wide): Likewise.
14084 (expand_complex_operations_1): Likewise.
14085 (expand_complex_libcall): Preserve EH info of the original stmt.
14086 (tree_lower_complex): Handle removed blocks.
14087 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
14088 on complex multiplication and division libcall builtins.
14090 2018-05-04 Richard Biener <rguenther@suse.de>
14092 PR middle-end/85574
14093 * fold-const.c (negate_expr_p): Restrict negation of operand
14094 zero of a division to when we know that can happen without
14096 (fold_negate_expr_1): Likewise.
14098 2018-05-04 Jakub Jelinek <jakub@redhat.com>
14101 * real.h (real_nextafter): Declare.
14102 * real.c (real_nextafter): New function.
14103 * fold-const-call.c (fold_const_nextafter): New function.
14104 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
14105 CASE_CFN_NEXTTOWARD.
14106 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
14107 even when arg1_mode is different from arg0_mode.
14109 2018-05-03 Nathan Sidwell <nathan@acm.org>
14111 * doc/extend.texi (Deprecated Features): Remove
14112 -ffriend-injection.
14113 (Backwards Compatibility): Likewise.
14114 * doc/invoke.texi (C++ Language Options): Likewise.
14115 (C++ Dialect Options): Likewise.
14117 2018-05-03 Jakub Jelinek <jakub@redhat.com>
14120 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
14121 _mm512_mask_mullox_epi64): New intrinsics.
14123 2018-05-03 Tom de Vries <tom@codesourcery.com>
14126 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14127 dump files): Add offload-tree.
14129 2018-05-03 Richard Biener <rguenther@suse.de>
14131 PR tree-optimization/85615
14132 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
14133 to loops not nested in BBs loop father to avoid creating multi-entry
14136 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14138 PR tree-optimization/70291
14139 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
14140 arguments. Change return type to tree. Emit libcall as a new
14141 statement rather than replacing existing one when inplace_p is true.
14142 (expand_complex_multiplication_components): New function.
14143 (expand_complex_multiplication): Expand floating-point complex
14144 multiplication using the above.
14145 (expand_complex_division): Rename inner_type parameter to type.
14146 Update expand_complex_libcall call-site.
14147 (expand_complex_operations_1): Update expand_complex_multiplication
14148 and expand_complex_division call-sites.
14150 2018-05-02 Jakub Jelinek <jakub@redhat.com>
14153 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
14154 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
14155 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
14156 the highest significant bit of the shift count mask is clear. In
14157 check whether and[sq]i3 is needed verify that all significant bits
14158 of the shift count other than the highest are set.
14160 2018-05-02 Tom de Vries <tom@codesourcery.com>
14163 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
14164 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
14165 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
14166 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
14167 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
14168 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
14169 __builtin_goacc_parlevel_size.
14171 2018-05-02 Richard Biener <rguenther@suse.de>
14173 PR tree-optimization/85597
14174 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
14175 do not use split vect_get_vec_defs call but call vect_get_slp_defs
14178 2018-05-02 Tom de Vries <tom@codesourcery.com>
14181 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14182 dump files): Add ltrans-tree.
14184 2018-05-02 Tom de Vries <tom@codesourcery.com>
14187 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14188 dump files): Add wpa-ipa.
14190 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
14192 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
14193 powerpc*-*-linux*paired* target.
14194 * config/rs6000/750cl.h: Delete.
14195 * config/rs6000/paired.h: Delete.
14196 * config/rs6000/paired.md: Delete.
14197 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
14199 * config/rs6000/rs6000-builtin.def: Remove paired float support.
14200 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
14201 comment. Remove paired float support.
14202 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
14203 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
14205 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
14206 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
14208 * config/rs6000/rs6000.c: Remove paired float support.
14209 (paired_expand_vector_init, paired_expand_vector_move,
14210 paired_emit_vector_compare, paired_emit_vector_cond_expr,
14211 (paired_expand_lv_builtin, paired_expand_stv_builtin,
14212 paired_expand_builtin, paired_expand_predicate_builtin,
14213 paired_init_builtins): Delete.
14214 * config/rs6000/rs6000.h: Remove paired float support.
14215 * config/rs6000/rs6000.md: Remove paired float support.
14216 (move_from_CR_ov_bit): Delete.
14217 * config/rs6000/rs6000.opt (mpaired): Delete.
14218 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
14219 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
14221 2018-05-02 Richard Biener <rguenther@suse.de>
14223 PR middle-end/85567
14224 * gimplify.c (gimplify_save_expr): When in SSA form allow
14225 SAVE_EXPRs to compute to SSA vars.
14227 2018-05-02 Jakub Jelinek <jakub@redhat.com>
14230 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
14231 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
14232 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
14233 clobber operands[2], instead use a new pseudo. Formatting fixes.
14235 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
14237 PR tree-optimization/85586
14238 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
14239 exit early for statements in the same group if the accesses are
14242 2018-05-02 Tom de Vries <tom@codesourcery.com>
14245 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
14248 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
14250 PR tree-optimization/85143
14251 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
14253 2018-05-01 Tom de Vries <tom@codesourcery.com>
14256 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
14257 not found" error message.
14259 2018-05-01 Tom de Vries <tom@codesourcery.com>
14262 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
14263 * vec.c (test_ordered_remove_if): New function.
14264 (vec_c_tests): Call test_ordered_remove_if.
14265 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
14266 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
14267 * tree-vect-patterns.c (vect_pattern_recog_1): Use
14268 VEC_ORDERED_REMOVE_IF.
14270 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14272 PR tree-optimization/82665
14273 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
14274 pointer subtraction where arguments come from a memchr call.
14276 2018-05-01 Jakub Jelinek <jakub@redhat.com>
14278 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
14279 --push-state --as-needed and --pop-state instead of --as-needed and
14280 --no-as-needed if ld supports it.
14281 * configure: Regenerated.
14284 * doc/install.texi2html: Replace _002d with - and _002a with * in
14285 generated html files using sed.
14287 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14290 * gcc-rich-location.c (blank_line_before_p): New function.
14291 (use_new_line): New function.
14292 (gcc_rich_location::add_fixit_insert_formatted): New function.
14293 * gcc-rich-location.h
14294 (gcc_rich_location::add_fixit_insert_formatted): New function.
14296 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14298 * selftest.c (assert_streq): Rename "expected" and "actual" to
14299 "val1" and "val2". Extend NULL-handling to cover both inputs
14300 symmetrically, while still requiring both to be non-NULL for a pass.
14301 * selftest.h (assert_streq): Rename "expected" and "actual" to
14303 (ASSERT_EQ): Likewise.
14304 (ASSERT_EQ_AT): Likewise.
14305 (ASSERT_KNOWN_EQ): Likewise.
14306 (ASSERT_KNOWN_EQ_AT): Likewise.
14307 (ASSERT_NE): Likewise.
14308 (ASSERT_MAYBE_NE): Likewise.
14309 (ASSERT_MAYBE_NE_AT): Likewise.
14310 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
14311 the assertion to pass.
14312 (ASSERT_STREQ_AT): Likewise.
14314 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
14316 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
14317 interaction with -pie.
14319 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14321 * selftest.h: Fix alphabetization of per-source-file selftest
14324 2018-04-30 Jason Merrill <jason@redhat.com>
14326 PR c++/61982 - dead stores to destroyed objects.
14327 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
14330 2018-04-30 Jason Merrill <jason@redhat.com>
14332 * tree.c (build_clobber): New.
14333 * tree.h: Declare it.
14334 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
14336 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14338 * diagnostic-show-locus.c (layout::layout): Update for
14339 location_get_source_line returning a char_span.
14340 (struct char_span): Move to input.h.
14341 (struct correction): Update for fields in char_span becoming
14343 (struct source_line): Update for location_get_source_line
14344 returning a char_span.
14345 (layout::print_line): Likewise.
14346 * edit-context.c (edited_file::print_content): Likewise.
14347 (edited_file::print_diff_hunk): Likewise.
14348 (edited_file::print_run_of_changed_lines): Likewise.
14349 (edited_file::get_num_lines): Likewise.
14350 (edited_line::edited_line): Likewise.
14351 * final.c (asm_show_source): Likewise.
14352 * input.c (location_get_source_line): Convert return type
14353 from const char * to char_span, losing the final "line_len"
14355 (dump_location_info): Update for the above.
14356 (get_substring_ranges_for_loc): Likewise. Use a char_span
14357 when handling the literal within the line.
14358 (test_reading_source_line): Update for location_get_source_line
14359 returning a char_span.
14360 * input.h (class char_span): Move here from
14361 diagnostic-show-locus.c, converting from a struct to a class.
14362 Make data members private.
14363 (char_span::operator bool): New.
14364 (char_span::length): New.
14365 (char_span::get_buffer): New.
14366 (char_span::operator[]): New.
14367 (char_span::subspan): Make const.
14368 (char_span::xstrdup): New.
14369 (location_get_source_line): Convert return type from const char *
14370 to char_span, losing the final "line_size" param.
14372 2018-04-30 Jan Hubicka <jh@suse.cz>
14374 * lto-wrapper.c (ltrans_priorities): New static var.
14375 (cmp_priority): New.
14376 (run_gcc): Read priorities and if doing parallel build order
14377 the Makefile by them.
14379 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14381 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
14383 2018-04-30 Richard Biener <rguenther@suse.de>
14385 * tree-cfg.c (verify_address): Remove base argument, add
14386 flag whether to check TREE_ADDRESSABLE and do that.
14387 (verify_expr): Remove.
14388 (verify_types_in_gimple_reference): Add pieces from verify_expr.
14389 (verify_gimple_assign_single): Likewise.
14390 (verify_gimple_switch): Likewise.
14391 (verify_expr_location_1): Dereference tp once. Add (disabled)
14392 piece from verify_expr.
14393 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
14395 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14397 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
14399 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14401 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
14402 (small_data_pattern): Likewise.
14403 (arc_rewrite_small_data): Likewise.
14404 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
14405 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
14406 (get_symbol_alignment): New function.
14407 (legitimate_small_data_address_p): Likewise.
14408 (legitimate_scaled_address): Update, call
14409 legitimate_small_data_address_p.
14410 (output_sdata): New static variable.
14411 (arc_print_operand): Update how we handle small data operands.
14412 (arc_print_operand_address): Likewise.
14413 (arc_legitimate_address_p): Update, use
14414 legitimate_small_data_address_p.
14415 (arc_rewrite_small_data_p): Remove.
14416 (arc_rewrite_small_data_1): Likewise.
14417 (arc_rewrite_small_data): Likewise.
14418 (small_data_pattern): Likewise.
14419 (compact_sda_memory_operand): Update to use
14420 legitimate_small_data_address_p and get_symbol_alignment.
14421 (prepare_move_operands): Don't rewite sdata pattern.
14422 (prepare_extend_operands): Remove.
14423 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
14425 (zero_extendqisi2): Likewise.
14426 (zero_extendhisi2): Likewise.
14427 (extendqihi2): Likewise.
14428 (extendqisi2): Likewise.
14429 (extendhisi2): Likewise.
14430 (addsi3): Likewise.
14431 (subsi3): Likewise.
14432 (andsi3): Likewise.
14433 * config/arc/constraints.md (Usd): Change it to memory constraint.
14435 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14437 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
14438 as source of std instructions.
14439 * config/arc/arc.md (movsi_insn): Update pattern predicate to
14440 allow 6-bit constants as source for store instructions.
14441 (movdi_insn): Update instruction pattern to allow 6-bit constants
14442 as source for store instructions.
14444 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
14446 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
14448 2018-04-30 Nathan Sidwell <nathan@acm.org>
14449 Sandra Loosemore <sandra@codesourcery.com>
14451 * dumpfile.c (dump_open): Allow '-' for stdout.
14452 * doc/invoke.texi (Developer Options): Document dump filename
14453 determination early. Document stdin/stdout selection.
14455 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
14457 Microblaze Target: PIC data text relative
14459 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
14460 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
14462 * config/microblaze/microblaze.h (microblaze_constant_address_p):
14463 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
14464 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
14465 New addressing mode for data-text relative position indepenedent code.
14466 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
14467 'ADDRESS_SYMBOLIC_TXT_REL'.
14468 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
14469 (microblaze_legitimate_pic_operand): Exclude function calls from
14470 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
14471 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
14473 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14474 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14475 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
14476 for 'address + offset'.
14477 (microblaze_expand_prologue): Add new function prologue call for
14479 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
14480 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
14481 table in case of TARGET_PIC_DATA_TEXT_REL.
14482 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
14483 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
14484 Add new macros 'UNSPEC_TEXT',
14485 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
14486 + exclude function calls from 'UNSPEC_PLT' in case of data text
14488 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14489 new target hook for generating address diff vector tables in case of
14491 * doc/tm.texi : Regenerate.
14492 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
14493 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
14494 of addr diff vector generation.
14495 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14496 target hook definition.
14497 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
14498 Add default function for generate_pic_addr_diff_vec -> flag_pic.
14499 * doc/invoke.texi (Add new pic option): Add new microblaze pic
14500 option for data text relative.
14502 2018-04-30 Richard Biener <rguenther@suse.de>
14504 * tree-chrec.h (evolution_function_is_constant_p): Remove
14506 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
14508 2018-04-30 Richard Biener <rguenther@suse.de>
14511 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
14513 2018-04-30 Richard Biener <rguenther@suse.de>
14515 PR tree-optimization/28364
14516 PR tree-optimization/85275
14517 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
14518 copying first exit test.
14520 2018-04-28 Mark Wielaard <mark@klomp.org>
14522 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
14523 dwarf_version >= 5.
14524 (dwarf_AT): Handle DW_AT_addr_base.
14525 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
14527 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
14530 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
14531 (*ashl<dwi>3_doubleword_mask_1): Ditto.
14532 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
14533 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
14535 2018-04-28 Richard Biener <rguenther@suse.de>
14537 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
14538 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
14539 to reflect use. Only add interesting stmts.
14541 2018-04-27 Martin Jambor <mjambor@suse.cz>
14544 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
14545 the jump function allows for passing through aggregate values.
14547 2018-04-27 David Malcolm <dmalcolm@redhat.com>
14549 * input.h (in_system_header_at): Convert from macro to inline
14551 (from_macro_expansion_at): Likewise.
14552 (from_macro_definition_at): Likewise.
14554 2018-04-27 Jeff Law <law@redhat.com>
14556 * config.gcc: Mark tile* targets as deprecated/obsolete.
14558 2018-04-27 Richard Biener <rguenther@suse.de>
14560 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
14563 2018-04-27 Richard Biener <rguenther@suse.de>
14565 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
14567 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
14569 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
14570 with Yd constraint. Set "preferred_for_speed" attribute from
14571 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
14572 with Yd constraint.
14573 (*movdi_internal): Ditto.
14574 (movti_interunit splitters): Remove
14575 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
14576 (movdi_interunit splitters): Ditto.
14577 * config/i386/constraints.md (Ye): Remove.
14578 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
14580 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14583 * config/aarch64/constraints.md (Usg): Limit to 31.
14584 (Usj): Limit to 63.
14586 2018-04-27 Jakub Jelinek <jakub@redhat.com>
14588 PR tree-optimization/85529
14589 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
14590 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
14591 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
14592 zero extension or masking of the MSB bit.
14593 (optimize_range_tests): Add FIRST_BB argument, pass it through
14594 to optimize_range_tests_var_bound.
14595 (maybe_optimize_range_tests, reassociate_bb): Adjust
14596 optimize_range_tests callers.
14598 2018-04-26 Richard Biener <rguenther@suse.de>
14599 Jakub Jelinek <jakub@redhat.com>
14601 * cgraph.h (symbol_table): Just declare debug method here.
14602 * symtab.c (symbol_table::debug): Define.
14604 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
14606 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
14608 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
14610 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
14611 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
14612 (*movdi_internal): Substitute Yi and Yj constraint with x
14613 and Ym and Yn constraint with y constraint. Update "isa"
14614 attribute and set "preferred_for_speed" attribute from
14615 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
14616 (*movsi_internal): Ditto.
14617 (*movdf_internal): Ditto.
14618 (*movsf_internal): Ditto.
14619 (*zero_extendsidi2): Ditto.
14620 * config/i386/sse.md (vec_set<mode>_0): Ditto.
14621 (sse2_loadld): Ditto.
14622 (*vec_extract<ssevecmodelower>_0): Ditto.
14623 (*vec_extractv4si_0_zext_sse4): Ditto.
14624 (vec_concatv2di): Ditto.
14625 (*vec_dup<mode>): Ditto.
14626 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
14627 * config/i386/constraints.md (Yi): Remove.
14632 2018-04-26 Nathan Sidwell <nathan@acm.org>
14634 * dumpfile.c (dump_open): New.
14635 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
14636 (dump_finish): Detect stdio/stderr by value not name.
14638 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
14640 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
14642 2018-04-26 Tom de Vries <tom@codesourcery.com>
14645 * config/nvptx/nvptx.c (verify_neutering_jumps)
14646 (verify_neutering_labels): New function
14647 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
14649 2018-04-26 Tom de Vries <tom@codesourcery.com>
14652 * config/nvptx/nvptx.c (needs_neutering_p): New function.
14653 (nvptx_single): Use needs_neutering_p to skip over insns that do not
14656 2018-04-26 Richard Biener <rguenther@suse.de>
14657 Tom de Vries <tom@codesourcery.com>
14660 * lto-streamer-out.c (output_function): Fixup loops if required to match
14661 discovery done in the reader.
14663 2018-04-26 Richard Biener <rguenther@suse.de>
14665 PR tree-optimization/85116
14666 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
14667 have a loop exit from the single latch predecessor. Remove
14668 case of header with just condition.
14669 (ch_base::copy_headers): Exclude infinite loops from any
14671 (pass_ch::execute): Record exits.
14673 2018-04-26 Richard Biener <rguenther@suse.de>
14675 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
14676 prologue cost vector and pass it to vect_get_load_cost.
14677 (vect_get_peeling_costs_all_drs): Likewise.
14678 (vect_peeling_hash_get_lowest_cost): Likewise.
14679 (vect_enhance_data_refs_alignment): Likewise.
14681 2018-04-26 Richard Biener <rguenther@suse.de>
14683 PR middle-end/85450
14684 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
14685 checking of integer<->pointer conversions.
14686 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
14687 sign-/zero-extending pointer types.
14688 (expand_omp_for_static_chunk): Likewise.
14690 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
14691 Jean Lee <xiaoyur347@gmail.com>
14693 * config/mips/mips.c (mips_asan_shadow_offset): New function.
14694 (TARGET_ASAN_SHADOW_OFFSET): Define.
14695 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
14696 true for -fsanitize=address.
14698 2018-04-25 Mark Wielaard <mark@klomp.org>
14700 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
14703 2018-04-25 Jakub Jelinek <jakub@redhat.com>
14705 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
14706 than "alu", remove explicit "memory" and "imm_disp" attributes.
14707 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
14709 PR middle-end/85414
14710 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
14711 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
14712 gen_lowpart_no_emit.
14714 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
14717 * config/i386/i386.c (ix86_expand_builtin): Change memory
14718 operand to XI, extend p0 to Pmode.
14719 * config/i386/i386.md: Change unspec volatile and operand
14720 1 mode to XI, change operand 0 mode to P.
14722 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14724 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
14725 GET_MODE_MASK before any checking.
14726 (nds32_can_use_bset_p): Likewise.
14727 (nds32_can_use_btgl_p): Likewise.
14729 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14731 * config/nds32/nds32-doubleword.md: New define_split pattern for
14732 illegal register number.
14734 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14736 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
14738 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14740 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
14742 2018-04-25 Richard Biener <rguenther@suse.de>
14744 * lto-streamer.h (LTO_major_version): Bump to 8.
14746 2018-04-25 Jakub Jelinek <jakub@redhat.com>
14748 * BASE-VER: Set to 9.0.0.
14750 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
14752 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
14753 in __abskf2 and __powikf2.
14755 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14758 * config/aarch64/constraints.md (Usg, Usj): New constraints.
14759 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
14760 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
14761 Use the above on operand 2. Reindent.
14762 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
14764 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
14767 * common/config/i386/i386-common.c (ix86_handle_option): Don't
14769 * config/i386/i386.opt (mcet): Removed.
14770 * doc/install.texi: Remove -mcet documentation.
14771 * doc/invoke.texi: Likewise.
14773 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
14776 * doc/install.texi: Remove -mcet from bootstrap-cet.
14778 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14781 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
14782 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
14786 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
14787 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
14788 containing a CONST_VECTOR.
14790 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
14792 * doc/install.texi: Update newlib dependency for nvptx.
14794 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14797 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
14798 instead of INTVAL when shifting x left.
14800 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
14802 PR tree-optimization/85478
14803 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
14804 vect_grouped_store_supported for single element vectors.
14806 2018-04-24 Richard Biener <rguenther@suse.de>
14809 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
14810 load cost increase to the case of non-constant step.
14812 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14815 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
14816 destination if any_malformed_asm.
14818 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
14820 PR middle-end/85496
14821 * expr.c (store_field): In the bitfield case, if the value comes from
14822 a function call and is returned in registers by means of a PARALLEL,
14823 do not change the mode of the temporary unless BLKmode and VOIDmode.
14825 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
14827 PR rtl-optimization/85423
14828 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
14829 dependencies to debug insns when the previous insn is non-debug.
14831 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
14833 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
14834 enums into a single definition.
14835 (fls): Fix predicates and printing.
14838 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
14840 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
14841 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
14842 and short u6 immediate.
14843 (check_if_valid_sleep_operand): Remove.
14844 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
14846 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14848 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
14849 flag_always_save_lp condition.
14850 * config/nds32/nds32.opt (malways-save-lp): New option.
14852 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14854 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
14855 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
14856 * config/nds32/nds32.h
14857 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
14858 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
14860 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14862 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
14863 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
14865 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14866 Chung-Ju Wu <jasonwucj@gmail.com>
14868 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
14870 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
14871 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
14873 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14875 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
14877 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14879 * config/nds32/nds32-protos.h (nds32_data_alignment,
14880 nds32_local_alignment): Declare.
14881 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
14882 nds32_local_alignment): New functions.
14883 (TARGET_CONSTANT_ALIGNMENT): Define.
14884 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
14886 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14888 * config/nds32/nds32.c
14889 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
14890 (TARGET_MODES_TIEABLE_P): Likewise.
14892 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14894 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
14895 level Ofast and Og.
14897 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
14898 Chung-Ju Wu <jasonwucj@gmail.com>
14900 * config/nds32/constants.md (unspec_volatile_element): Add enum values
14901 for unaligned access.
14902 * config/nds32/nds32-intrinsic.c: Implementation of expanding
14904 * config/nds32/nds32-intrinsic.md: Likewise.
14905 * config/nds32/nds32_intrinsic.h: Likewise.
14906 * config/nds32/nds32.h (nds32_builtins): Likewise.
14907 * config/nds32/nds32.opt (munaligned-access): New option.
14908 * config/nds32/nds32.c (nds32_asm_file_start): Display
14909 flag_unaligned_access status.
14911 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
14913 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
14914 -mno-relax is present.
14915 * config/riscv/linux.h (LINK_SPEC): Ditto.
14917 2018-04-20 Martin Sebor <msebor@redhat.com>
14920 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
14922 (gimple_fold_builtin_stxcpy_chk): Same.
14923 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
14925 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
14928 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
14929 __powikf2 when long double is IEEE 128-bit.
14931 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
14933 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
14934 step to make sure stack always aligned.
14936 2018-04-20 Carl Love <cel@us.ibm.com>
14939 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
14940 size check for arg0.
14942 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
14943 Tom de Vries <tom@codesourcery.com>
14946 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
14947 Emit insns for calls too.
14948 (nvptx_find_par): Always look for worker-level predecessor insn.
14949 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
14951 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
14952 (nvptx_process_pars): Propagate frames for calls.
14954 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
14957 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
14959 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
14960 (ix86_handle_option): Don't handle OPT_mibt.
14961 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
14963 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
14965 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
14966 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
14967 (ix86_target_macros): Define __CET__ with flag_cf_protection
14968 for -fcf-protection.
14969 * config/i386/i386.c (isa2_opts): Remove -mibt.
14970 * config/i386/i386.h (TARGET_IBT): Removed.
14971 (TARGET_IBT_P): Likewise.
14972 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
14973 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
14974 * config/i386/i386.opt (mcet): Update help message.
14975 (mshstk): Likewise.
14977 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
14978 -mcet as an alias for -mshstk.
14980 2018-04-20 Richard Biener <rguenther@suse.de>
14982 PR middle-end/85475
14983 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
14984 complexity by forcing a single use of the multiply operand.
14986 2018-04-20 Martin Jambor <mjambor@suse.cz>
14989 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
14990 recursion dependency to only apply to non-clones.
14992 2018-04-20 Martin Jambor <mjambor@suse.cz>
14995 * ipa-cp.c (create_specialized_node): Check that clones of
14996 self-recursive edges exist during IPA-CP.
14998 2018-04-19 Toon Moene <toon@moene.org>
15000 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
15003 2018-04-19 Jakub Jelinek <jakub@redhat.com>
15005 PR tree-optimization/85467
15006 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
15007 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
15008 VECTOR_CST element to type.
15010 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
15013 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
15014 * config/i386/i386.md (builtin_setjmp_setup): Removed.
15015 (builtin_longjmp): Likewise.
15016 (save_stack_nonlocal): New pattern.
15017 (restore_stack_nonlocal): Likewise.
15019 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
15022 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
15023 Replace ASM_OUTPUT_LABEL with fprintf.
15025 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
15028 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
15029 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
15030 * config/i386/i386-c.c (ix86_target_macros_internal): Also
15031 define __IBT__ and __SHSTK__ for -fcf-protection.
15032 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
15034 (ix86_trampoline_init): Likewise.
15035 (x86_output_mi_thunk): Likewise.
15036 (ix86_notrack_prefixed_insn_p): Likewise.
15037 (ix86_option_override_internal): Don't disallow -fcf-protection.
15038 * config/i386/i386.md (rdssp<mode>): Also enable for
15040 (incssp<mode>): Likewise.
15041 (nop_endbr): Likewise.
15042 * config/i386/i386.opt (mcet): Change help message to built-in
15045 (mshstk): Likewise.
15046 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
15047 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
15048 enable CET built-in functions.
15050 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
15052 * common/config/i386/i386-common.c
15053 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
15054 OPTION_MASK_ISA_MOVDIRI_UNSET,
15055 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
15056 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
15057 * config.gcc (movdirintrin.h): New header.
15058 * config/i386/cpuid.h (bit_MOVDIRI,
15059 bit_MOVDIR64B): New bits.
15060 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
15062 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
15063 (VOID, PVOID, PCVOID)): New function types.
15064 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
15065 __builtin_ia32_directstoreu_u64,
15066 __builtin_ia32_movdir64b): New builtins.
15067 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
15068 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
15070 (ix86_valid_target_attribute_inner_p): Ditto.
15071 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
15072 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
15073 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
15074 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
15075 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
15076 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
15077 (movdiri<mode>, movdir64b_<mode>): New.
15078 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
15079 * config/i386/immintrin.h: Include movdirintrin.h.
15080 * config/i386/movdirintrin.h: New file.
15081 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
15083 2018-04-19 Richard Biener <rguenther@suse.de>
15085 PR middle-end/85455
15086 * cfg.c (clear_bb_flags): When loop state says we have
15087 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
15089 2018-04-19 Richard Biener <rguenther@suse.de>
15091 PR tree-optimization/84737
15092 * tree-vect-data-refs.c (vect_copy_ref_info): New function
15093 copying restrict info.
15094 (vect_setup_realignment): Use it.
15095 * tree-vectorizer.h (vect_copy_ref_info): Declare.
15096 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
15097 the first DR to all generated stores.
15098 (vectorizable_load): Likewise for loads.
15100 2018-04-19 Jakub Jelinek <jakub@redhat.com>
15102 PR tree-optimization/85446
15103 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
15104 the integral and pointer types to have the same precision.
15106 * doc/install.texi: Document --disable-cet being the default and
15109 2018-04-18 Martin Liska <mliska@suse.cz>
15111 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
15114 2018-04-18 Martin Liska <mliska@suse.cz>
15117 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
15120 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
15121 arguments if they are comparable.
15123 2018-04-18 Martin Liska <mliska@suse.cz>
15126 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
15129 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
15132 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
15135 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
15136 ENDBR after calling __morestack.
15138 2018-04-18 David Malcolm <dmalcolm@redhat.com>
15141 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
15142 by using gcc_base_ver to generate a gcc_driver_version, and use
15143 it when generating GCC_DRIVER_NAME.
15144 * configure: Regenerate.
15146 2018-04-18 Jakub Jelinek <jakub@redhat.com>
15149 * config.gcc: Obsolete powerpc*-*-*spe*.
15151 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15154 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
15155 (stabstr_D): Change type of unum from unsigned int to
15156 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
15159 2018-04-17 Jim Wilson <jimw@sifive.com>
15162 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
15163 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
15164 Set arg_pointer_offset after using pretend_args_size.
15166 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15168 PR rtl-optimization/85431
15169 * dse.c (record_store): Ignore zero width stores.
15172 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
15173 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
15174 __builtin_stack_restore rather than after it.
15175 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
15176 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
15177 argument instead of virtual_dynamic_stack_rtx.
15179 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
15181 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
15183 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15184 Add note to error message to explain internal mapping of overloaded
15185 built-in function name to non-overloaded built-in function name.
15186 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
15189 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
15192 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
15193 where the inputs overlap with the output.
15195 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15197 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
15198 (=v, v) alternative and explicit "memory" attribute.
15199 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
15200 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15202 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
15203 "sselog1" type instead of "sselog".
15204 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
15205 "sselog". Remove explicit "memory" attribute.
15206 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
15207 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15209 (vec_extract_hi_v32hi): Merge all alternatives into one, use
15210 "sselog1" type instead of "sselog". Remove explicit "memory"
15212 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
15213 use "sselog1" type instead of "sselog". Remove explicit "memory"
15215 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
15216 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15218 (vec_extract_hi_v64qi): Merge all alternatives into one, use
15219 "sselog1" type instead of "sselog". Remove explicit "memory"
15221 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
15222 use "sselog1" type instead of "sselog". Remove explicit "memory"
15226 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
15228 PR middle-end/85414
15229 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
15232 2018-04-17 Martin Jambor <mjambor@suse.cz>
15235 * ipa-cp.c (create_specialized_node): Call
15236 expand_all_artificial_thunks if necessary.
15238 2018-04-17 Martin Liska <mliska@suse.cz>
15241 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
15242 in message, remote space in between '_G' and '('.
15244 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15247 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
15248 avx512f_vmcmp<mode>3<round_saeonly_name>,
15249 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
15250 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
15251 avx512f_rndscale<mode><round_saeonly_name>,
15252 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
15253 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
15254 Use %<iptr>2 instead of %2 for -masm=intel.
15255 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
15256 avx512f_vcvttss2usi<round_saeonly_name>,
15257 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
15259 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
15260 avx512f_vcvttsd2usi<round_saeonly_name>,
15261 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
15262 Use %q1 instead of %1 for -masm=intel.
15263 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
15264 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
15265 of %3 for -masm=intel.
15266 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
15268 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
15270 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
15272 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
15273 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
15275 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
15277 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
15278 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
15279 %g1 and one with %0 and %1.
15280 (avx512er_vmrcp28<mode><round_saeonly_name>,
15281 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
15282 %1 for -masm=intel.
15283 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
15284 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
15285 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
15286 of %0 and %{%4%} for -masm=intel.
15287 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
15288 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
15289 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
15290 order of %0 and %{%5%}%{z%} for -masm=intel.
15292 2018-04-17 Jan Hubicka <jh@suse.cz>
15295 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
15297 2018-04-17 Martin Liska <mliska@suse.cz>
15300 * multiple_target.c (create_dispatcher_calls): Set apostrophes
15301 for target_clone error message. Make default implementation
15302 clone to be a local declaration.
15303 (separate_attrs): Add new argument and check for an empty
15305 (expand_target_clones): Handle it.
15306 (ipa_target_clone): Make redirection just for target_clones
15309 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
15310 Tom de Vries <tom@codesourcery.com>
15312 PR middle-end/84955
15313 * omp-expand.c (expand_oacc_for): Add dummy false branch for
15314 tiled basic blocks without omp continue statements.
15316 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
15319 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
15320 vec_extract expression as having side effects to make sure it gets
15323 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
15326 * config/i386/i386.c (get_builtin_code_for_version): Check
15329 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
15332 * config.gcc: Support "skylake".
15333 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15335 * config/i386/i386.c (m_SKYLAKE): Define.
15336 (processor_target_table): Add "skylake".
15337 (ix86_option_override_internal): Add "skylake".
15338 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
15339 PROCESSOR_CANNONLAKE.
15340 (get_builtin_code_for_version): Fix priority for
15341 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
15342 PROCESSOR_SKYLAKE-AVX512.
15343 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
15344 (processor_type): Add PROCESSOR_SKYLAKE.
15346 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
15347 Jason Merrill <jason@redhat.com>
15350 * convert.c (convert_to_integer_1): Use direct recursion for
15351 enumeral types and types with a precision less than the number
15352 of bits in their mode.
15354 2018-04-16 Julia Koval <julia.koval@intel.com>
15357 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
15358 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
15360 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
15363 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
15364 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
15365 and -mno-direct-move.
15367 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
15370 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
15371 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
15372 Ensure negative shifts result in {0}.
15374 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
15376 PR rtl-optimization/79916
15377 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
15378 regs (if any) to define how to gnerate SD moves when LRA is in
15381 2018-04-13 Jakub Jelinek <jakub@redhat.com>
15383 PR rtl-optimization/85393
15384 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
15385 * except.c (expand_dw2_landing_pad_for_region): Make static.
15386 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
15387 a label and unconditional jump to old_bb, rather than
15388 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
15391 PR rtl-optimization/85376
15392 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
15393 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
15394 instead of a specific value.
15396 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
15397 Bin Cheng <bin.cheng@arm.com>
15399 PR tree-optimization/82965
15400 PR tree-optimization/83991
15401 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
15402 by_profile_only parameter.
15403 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
15404 information if the loop was predicted to iterate too many times.
15405 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
15407 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
15410 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
15413 2018-04-13 Martin Liska <mliska@suse.cz>
15414 Jakub Jelinek <jakub@redhat.com>
15416 PR middle-end/81657
15417 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
15418 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
15419 * builtins.c (expand_builtin_memory_copy_args): Use
15420 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
15421 handle dest_addr == pc_rtx.
15423 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
15426 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
15427 asked to not generate direct moves.
15428 (fix_trunc<mode>si2_stfiwx): Similar.
15429 (fix_trunc<mode>si2_internal): Similar.
15431 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15434 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
15435 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
15436 lookup if dest in some wider mode is known to be const0_rtx and
15437 if so, record permanent equivalence for it to be ZERO_EXTEND of
15438 the narrower mode destination.
15440 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
15442 * lto-streamer-out.c (output_function): Revert 259346.
15443 * omp-expand.c (expand_oacc_for): Likewise.
15445 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
15447 PR rtl-optimization/85354
15448 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
15449 * sel-sched.c (sel_global_init): ... here.
15451 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
15454 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
15455 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
15456 mode for PE-COFF targets.
15457 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
15458 (i386_pe_asm_lto_end): Likewise.
15459 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
15460 (TARGET_ASM_LTO_END): Likewise.
15461 * config/i386/winnt.c (saved_debug_info_level): New static variable.
15462 (i386_pe_asm_lto_start): New function.
15463 (i386_pe_asm_lto_end): Likewise.
15465 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
15466 Richard Biener <rguenther@suse.de>
15468 PR middle-end/84955
15469 * lto-streamer-out.c (output_function): Fix CFG loop state before
15471 * omp-expand.c (expand_oacc_for): Handle calls to internal
15472 functions like regular functions.
15474 2018-04-12 Richard Biener <rguenther@suse.de>
15477 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
15478 for the early LTO debug to properly generate references to it
15479 during DIE emission. Do not re-use that for the skeleton for
15481 (dwarf2out_early_finish): Likewise.
15483 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15486 * config/i386/sse.md
15487 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
15488 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
15489 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
15490 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
15491 and output is a reg, avoid creating invalid lowpart subreg, but
15492 instead split into a 512-bit move. Don't split if not AVX512VL,
15493 input is xmm16+ reg and output is a mem.
15494 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
15495 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
15496 xmm16+ reg and output is a mem.
15498 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15500 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
15501 also for flag_dwarf2_cfi_asm.
15503 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15505 PR rtl-optimization/85342
15506 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
15507 a bool scalar var inside of the loop instead. Don't try to update
15508 recog_data.operand after failed apply_change_group.
15510 2018-04-12 Tom de Vries <tom@codesourcery.com>
15513 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
15514 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
15515 array with flexible array member as array without given dimension.
15516 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
15517 argument for undefined param to true.
15519 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
15522 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15523 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
15524 from PowerPC section.
15525 * config/rs6000/sysv4.opt (mcall-): Improve help text.
15526 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
15527 help text that is too long.
15528 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
15529 help text that is too long.
15530 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
15531 help text that is too long.
15533 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
15535 * config/alpha/alpha.md (stack_probe_internal): Rename
15536 from "probe_stack". Update all callers.
15538 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15540 PR rtl-optimization/84566
15541 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
15542 sched_macro_fuse_insns.
15544 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15547 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
15548 (compute_block_dependences): ... from here.
15550 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15552 PR tree-optimization/85331
15553 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
15554 from int to HOST_WIDE_INT.
15556 2018-04-11 Martin Jambor <mjambor@suse.cz>
15559 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
15560 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
15561 not the same as the source val.
15562 (cgraph_edge_brings_value_p): New parameter.
15563 (gather_edges_for_value): Pass destination value to
15564 cgraph_edge_brings_value_p.
15565 (perhaps_add_new_callers): Likewise.
15566 (get_info_about_necessary_edges): Likewise and exclude values brought
15567 only by self-recursive edges.
15568 (create_specialized_node): Redirect only clones of self-calling edges.
15569 (+self_recursive_pass_through_p): New function.
15570 (find_more_scalar_values_for_callers_subset): Use it.
15571 (find_aggregate_values_for_callers_subset): Likewise.
15572 (known_aggs_to_agg_replacement_list): Removed.
15573 (decide_whether_version_node): Re-calculate known constants for all
15574 remaining context clones.
15576 2018-04-11 Richard Biener <rguenther@suse.de>
15579 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
15580 from early DWARF output.
15581 (dwarf2out_early_finish): Output line info unconditionally into
15582 early DWARF and add reference to it.
15584 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15587 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
15588 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
15589 other than V2DFmode using iptr mode attribute.
15590 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
15592 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15594 PR rtl-optimization/84659
15595 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
15597 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15600 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
15602 (output_loc_list): Pass address of a dummy size variable even in the
15603 locview handling loop.
15604 (index_location_lists): Add comment on why skip_loc_list_entry can't
15607 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
15610 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
15613 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
15616 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15617 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
15618 and -mstring-compare-inline-limit.
15620 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15623 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
15624 for stack clash protection in a register whenever we need it to be in
15627 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15629 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
15630 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
15632 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15635 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
15637 (mlong-double-): Ditto.
15638 * config/rs6000/sysv4.opt (msdata=): Ditto.
15639 (mtls-size=): Ditto.
15641 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15643 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
15644 erroneous entries for
15645 "vector int vec_ldl (int, long int *)", and
15646 "vector unsigned int vec_ldl (int, unsigned long int *)".
15647 Add comments and entries for
15648 "vector bool char vec_ldl (int, bool char *)",
15649 "vector bool short vec_ldl (int, bool short *)",
15650 "vector bool int vec_ldl (int, bool int *)",
15651 "vector bool long long vec_ldl (int, bool long long *)",
15652 "vector pixel vec_ldl (int, pixel *)",
15653 "vector long long vec_ldl (int, long long *)",
15654 "vector unsigned long long vec_ldl (int, unsigned long long *)".
15655 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
15656 type tree bool_long_long_type_node and correct definition of
15657 bool_V2DI_type_node to make reference to this new type tree.
15658 (rs6000_mangle_type): Replace erroneous reference to
15659 bool_long_type_node with bool_long_long_type_node.
15660 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
15661 comments to emphasize sign distinctions for char and int types and
15662 replace RS6000_BTI_bool_long constant with
15663 RS6000_BTI_bool_long_long constant. Also add comment to restrict
15664 use of RS6000_BTI_pixel.
15665 (bool_long_type_node): Remove this macro definition.
15666 (bool_long_long_type_node): New macro definition
15668 2018-04-10 Jakub Jelinek <jakub@redhat.com>
15670 PR rtl-optimization/85300
15671 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
15672 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
15673 simplify_unary_operation fails.
15675 2018-04-10 Martin Liska <mliska@suse.cz>
15677 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
15678 cgraph_edge and ipa_ref.
15680 2018-04-10 Jakub Jelinek <jakub@redhat.com>
15684 * config/i386/sse.md
15685 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
15686 computation of the VEC_MERGE selector from mask.
15687 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
15688 Fix decoding of the VEC_MERGE selector into mask.
15690 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
15692 PR tree-optimization/85286
15693 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
15695 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
15697 * final.c (final_1): Set insn_last_address as well as
15698 insn_current_address.
15700 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15703 * explow.c (emit_stack_probe): Call validize_mem on memory location
15704 before passing it to gen_probe_stack. Create address operand and
15705 legitimize it for the probe_stack_address case.
15707 2018-04-09 Jan Hubicka <jh@suse.cz>
15710 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
15711 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
15712 * tree.c (free_lang_data_in_type): Fix handling of binfos;
15714 (free_lang_data): Rebuild type inheritance graph.
15716 2018-04-09 Martin Sebor <msebor@redhat.com>
15718 * invoke.texi (-finline-small-functions): Mention other optimization
15720 (-findirect-inlining, -fpartial-inlining): Same.
15721 (-finline-functions-called-once): Same.
15722 (-freorder-blocks-and-partition): Same.
15724 2018-04-09 Jan Hubicka <jh@suse.cz>
15727 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
15728 jumps; choose last target that matches the criteria (i.e.
15729 no partition changes for non-crossing jumps).
15730 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
15731 support for redirecting crossing jumps to non-crossing.
15733 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
15735 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
15736 also for naked functions.
15738 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
15740 * config/arc/arc.md (add_shift): New pattern.
15741 (add_shift2): Likewise.
15742 (sub_shift): Likewise.
15743 (sub_shift_cmp0_noout): Likewise.
15744 (compare_si_ashiftsi): Likewise.
15745 (xbfu_cmp0_noout): New combine pattern.
15746 (xbfu_cmp0"): Likewise.
15747 (movsi_set_cc_insn): Place the predicable variant first.
15748 (commutative_binary_cmp0_noout): Remove clobber.
15749 (commutative_binary_cmp0): New pattern.
15750 (noncommutative_binary_cmp0): Likewise.
15751 (noncommutative_binary_cmp0_noout): Likewise.
15752 (noncommutative_binary_comparison_result_used): Removed.
15753 (rsub_cmp0): New pattern.
15754 (rsub_cmp0_noout): Likewise.
15755 (extzvsi): Changed, keep only meaningful variants.
15756 (SQH, SEZ): New iterators.
15757 (SQH_postfix): New mode attribute.
15758 (SEZ_prefix): New code attribute.
15759 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
15760 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
15761 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
15762 of numerical value.
15763 (noncommutative_operator): Check the availability of barrel
15766 2018-04-09 Richard Biener <rguenther@suse.de>
15768 PR tree-optimization/85284
15769 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
15770 Only use the niter constraining form of simple_iv when the exit
15771 is always executed.
15773 2018-04-09 Tom de Vries <tom@codesourcery.com>
15776 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15777 (define_expand "*memory_barrier"): New define_expand.
15778 (define_insn "memory_barrier"): New insn.
15780 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15782 PR rtl-optimization/80463
15783 PR rtl-optimization/83972
15784 PR rtl-optimization/83480
15786 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
15787 correct producer for the insn.
15788 (tidy_control_flow): Fixup seqnos in case of debug insns.
15790 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15792 PR rtl-optimization/83913
15794 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
15795 different sched-times when merging exprs.
15797 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15799 PR rtl-optimization/83962
15801 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
15802 tidy_fallthru_edge and tidy_control_flow.
15804 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15806 PR rtl-optimization/83530
15808 * sel-sched.c (force_next_insn): New global variable.
15809 (remove_insn_for_debug): When force_next_insn is true, also leave only
15810 next insn in the ready list.
15811 (sel_sched_region): When the region wasn't scheduled, make another pass
15812 over it with force_next_insn set to 1.
15814 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
15816 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
15818 * config/nds32/constants.md (unspec_volatile_element): Add enum values
15819 for interrupt control.
15820 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15821 functions for interrupt control.
15822 * config/nds32/nds32-intrinsic.md: Likewise.
15823 * config/nds32/nds32_intrinsic.h: Likewise.
15824 * config/nds32/nds32.h (nds32_builtins): Likewise.
15826 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
15828 * config/nds32/nds32.c (nds32_init_machine_status,
15829 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
15830 strict_aligned_p field.
15831 (nds32_expand_to_rtl_hook): New function.
15832 (TARGET_EXPAND_TO_RTL_HOOK): Define.
15833 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
15835 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15836 Chung-Ju Wu <jasonwucj@gmail.com>
15838 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
15839 * config/nds32/nds32-n7.md: New file.
15840 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
15841 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
15843 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
15844 * config/nds32/nds32.md (pipeline_model): Add n7.
15845 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
15846 * config/nds32/pipelines.md: Include n7 settings.
15848 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15849 Chung-Ju Wu <jasonwucj@gmail.com>
15851 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
15852 * config/nds32/nds32-e8.md: New file.
15853 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
15854 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
15856 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
15857 * config/nds32/nds32.md (pipeline_model): Add e8.
15858 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
15859 * config/nds32/pipelines.md: Include e8 settings.
15861 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15862 Chung-Ju Wu <jasonwucj@gmail.com>
15864 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
15865 * config/nds32/nds32-n8.md: New file.
15866 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
15867 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
15869 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
15870 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
15871 * config/nds32/nds32.md (pipeline_model): Add n8.
15872 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
15873 * config/nds32/pipelines.md: Include n8 settings.
15875 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15876 Chung-Ju Wu <jasonwucj@gmail.com>
15878 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
15879 * config/nds32/nds32-n9-2r1w.md: New file.
15880 * config/nds32/nds32-n9-3r2w.md: New file.
15881 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
15882 nds32_register_ports): New or modify for cpu n9.
15883 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
15885 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
15886 * config/nds32/nds32-utils.c: New file.
15887 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
15888 TARGET_MUL_SLOW): Define.
15889 * config/nds32/nds32.md (pipeline_model): New attribute.
15890 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
15891 New options that support cpu n9.
15892 * config/nds32/pipelines.md: Include n9 settings.
15893 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
15895 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
15897 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
15898 information if necessary.
15899 (output_cond_branch_compare_zero): Likewise.
15900 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
15901 (nds32_target_alignment): Refine for alignment.
15902 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
15903 (FUNCTION_BOUNDARY): Modify.
15904 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
15906 * config/nds32/nds32.opt (malways-align, malign-functions): New.
15908 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
15910 * config/nds32/constants.md (unspec_volatile_element): Add values for
15911 TLB operation and data prefetch.
15912 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15913 functions for TLB operation and data prefetch.
15914 * config/nds32/nds32-intrinsic.md: Likewise.
15915 * config/nds32/nds32_intrinsic.h: Likewise.
15916 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
15917 (nds32_print_operand): Likewise.
15918 * config/nds32/nds32.h (nds32_builtins): Likewise.
15920 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
15921 Andrew Pinski <pinsika@gcc.gnu.org>
15923 PR middle-end/82976
15924 * match.pd: Use constant_boolean_node of correct type instead of
15925 boolean_true_node or boolean_false_node for simplifying
15926 pointer comparisons to zero.
15928 2018-04-07 Jakub Jelinek <jakub@redhat.com>
15930 PR tree-optimization/80021
15931 * tree.c (verify_type_variant): Make error call in verify_variant_match
15932 translatable and remove final full stop.
15934 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15936 * config/nds32/constants.md (unspec_volatile_element): Add
15937 UNSPEC_VOLATILE_EH_RETURN.
15938 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
15939 nds32_output_stack_pop): Support dwarf exception handling process.
15940 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
15941 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
15942 exception handling process.
15943 (nds32_compute_stack_frame): Likewise.
15944 (nds32_return_addr_rtx): Likewise.
15945 (nds32_initial_elimination_offset): Likewise.
15946 (nds32_expand_prologue): Likewise.
15947 (nds32_expand_epilogue): Likewise.
15948 (nds32_dynamic_chain_address): New function.
15949 * config/nds32/nds32.h (machine_function): Add fields for dwarf
15950 exception handling.
15951 (DYNAMIC_CHAIN_ADDRESS): Define.
15952 (EH_RETURN_DATA_REGNO): Define.
15953 (EH_RETURN_STACKADJ_RTX): Define.
15954 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
15955 patterns for dwarf exception handling.
15957 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15959 * config/nds32/nds32.h: Clean up obsolete macros.
15961 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15963 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15964 Add enum values for particular instructions.
15965 * config/nds32/nds32-intrinsic.c: Implementation of expanding
15966 particular intrinsic functions.
15967 * config/nds32/nds32-intrinsic.md: Likewise.
15968 * config/nds32/nds32_intrinsic.h: Likewise.
15969 * config/nds32/nds32.h (nds32_builtins): Likewise.
15970 * config/nds32/nds32.md (type): Add pbsad and pbsada.
15971 (btst, ave): New patterns for particular instructions.
15973 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15975 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15976 Add enum values for atomic load/store and memory sync.
15977 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
15979 * config/nds32/nds32-intrinsic.md: Likewise.
15980 * config/nds32/nds32_intrinsic.h: Likewise.
15981 * config/nds32/nds32.h (nds32_builtins): Likewise.
15983 2018-04-07 Jakub Jelinek <jakub@redhat.com>
15985 PR tree-optimization/85257
15986 * fold-const.c (native_encode_vector): If not all elts could fit
15987 and off is -1, return 0 rather than offset.
15988 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
15989 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
15990 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
15991 adjust buffer in native_interpret_expr call.
15993 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15995 * config/nds32/constants.md (unspec_volatile_element): Add cache
15996 control enum values.
15997 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
15998 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
15999 * config/nds32/nds32.c (nds32_cctl_names): New.
16000 (nds32_print_operand): Handle cache control register names.
16001 * config/nds32/nds32.h (nds32_builtins): New enum values.
16002 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
16004 * config/nds32/nds32.md (type): Add mmu.
16005 * config/nds32/pipelines.md (simple_insn): Add mmu.
16007 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
16009 * config/nds32/nds32.md (type): Remove call.
16010 * config/nds32/pipelines.md (simple_insn): Likewise.
16012 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16014 * config/nds32/constants.md (unspec_volatile_element): Add
16015 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
16016 UNSPEC_VOLATILE_FMFCFG.
16017 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
16018 description for fmfcfg and fmfcsr.
16019 (bdesc_1arg): Add fmtcsr.
16020 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
16021 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
16022 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
16023 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
16024 unspec_fmfcfg): New patterns.
16025 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
16026 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
16027 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
16028 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
16029 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
16030 __nds32__fmfcfg): Define.
16032 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16034 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
16035 intrinsic register names.
16036 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
16037 intrinsic register enum values and macros.
16039 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
16041 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
16042 for load/store addressing form.
16043 (nds32_print_operand_address): Likewise.
16045 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
16048 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
16049 based on LABEL_REF. Remove useless assertion.
16050 (pic_address_needs_scratch): Fix formatting.
16051 (sparc_legitimize_pic_address): Minor tweaks.
16052 (sparc_delegitimize_address): Adjust assertion accordingly.
16053 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
16054 into symbolic_operand.
16055 (movsi_high_pic_label_ref): Likewise.
16056 (movsi_lo_sum_pic_label_ref): Likewise.
16057 (movdi_pic_label_ref): Likewise.
16058 (movdi_high_pic_label_ref): Likewise.
16059 (movdi_lo_sum_pic_label_ref): Likewise.
16061 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
16063 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
16064 custom LIB_SPEC setup.
16066 2018-04-06 Ruslan Bukin <br@bsdpad.com>
16067 Kito Cheng <kito.cheng@gmail.com>
16069 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
16070 * config/riscv/freebsd.h: New.
16072 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
16074 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
16075 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
16078 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
16079 Kito Cheng <kito.cheng@gmail.com>
16081 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
16082 nds32_output_call, nds32_symbol_binds_local_p): New functions.
16083 * config/nds32/nds32-protos.h (nds32_output_call,
16084 nds32_output_return): Declare.
16085 * config/nds32/nds32.md: Refine all the call and return patterns.
16087 2018-04-06 Jakub Jelinek <jakub@redhat.com>
16090 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
16091 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
16093 PR rtl-optimization/84872
16094 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
16095 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
16096 EDGE_CROSSING edge.
16098 2018-04-06 Tamar Christina <tamar.christina@arm.com>
16100 * expr.c (copy_blkmode_to_reg): Revert 254862.
16101 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
16103 2018-04-06 Richard Biener <rguenther@suse.de>
16105 PR middle-end/85244
16106 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
16107 after seeing a component reference with an adjacent field. Treat
16108 refs to arrays at struct end of external decls similar to
16109 refs to unconstrained commons.
16111 2018-04-06 Jakub Jelinek <jakub@redhat.com>
16114 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
16115 look through SAVE_EXPRs with non-side-effects argument. Adjust
16117 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
16120 2018-04-06 Richard Biener <rguenther@suse.de>
16122 PR middle-end/85180
16123 * alias.c (find_base_term): New wrapper around find_base_term
16124 unwinding CSELIB_VAL_PTR changes.
16125 (find_base_term): Do not restore CSELIB_VAL_PTR during the
16128 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16130 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
16132 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
16133 constant definitions.
16134 ("nop"): lr 0,0 -> nopr r0
16135 ("nop_lr0", "nop_lr1"): New insn definitions.
16137 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
16139 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
16140 NDS32_V3PUSH_AVAILABLE_P macro.
16142 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
16143 Chung-Ju Wu <jasonwucj@gmail.com>
16145 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
16146 (nds32*-*-*): Add float and fpu_config into supported_defaults.
16147 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
16148 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
16149 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
16150 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
16151 * config/nds32/constraints.md: New constraints and checking for hard
16152 float configuration.
16153 * config/nds32/iterators.md: New mode iterator and attribute for hard
16154 float configuration.
16155 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
16157 * config/nds32/nds32-fpu.md: New file.
16158 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
16159 deal with hard float code generation.
16160 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
16162 (abi_type, float_reg_number): New enum type.
16163 * config/nds32/nds32-predicates.c: New predicates for hard float.
16164 * config/nds32/nds32-protos.h: Declare functions for hard float.
16165 * config/nds32/nds32.c: Implementation for hard float configuration.
16166 * config/nds32/nds32.h: Definitions for hard float configuration.
16167 * config/nds32/nds32.md: Include hard float machine description and
16168 modify patterns for hard float configuration.
16169 * config/nds32/nds32.opt: New options for hard float configuration.
16170 * config/nds32/predicates.md: New predicates for hard float
16173 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
16175 * common/config/nds32/nds32-common.c
16176 (nds32_option_optimization_table): Enable -mreleax-hint by default.
16178 2018-04-05 Jakub Jelinek <jakub@redhat.com>
16180 PR middle-end/85195
16181 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
16182 CONSTRUCTOR_ELT (ctor, ...)->value.
16184 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
16187 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
16189 2018-04-05 Tom de Vries <tom@codesourcery.com>
16192 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
16195 2018-04-05 Shiva Chen <shiva0217@gmail.com>
16196 Kito Cheng <kito.cheng@gmail.com>
16198 * config/nds32/constraints.md (U33): Fine-tune checking condition.
16199 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
16200 * config/nds32/nds32.h (nds32_16bit_address_type): Add
16201 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
16203 2018-04-05 Shiva Chen <shiva0217@gmail.com>
16204 Kito Cheng <kito.cheng@gmail.com>
16206 * config/nds32/constraints.md (Ufe): New memory constraint.
16207 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
16208 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
16209 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
16211 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
16212 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
16214 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16216 * config/nds32/nds32.md: Use optimize_size in the condition for
16217 alu-shift instructions.
16219 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16221 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
16223 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16225 * config/nds32/nds32.md (negsi2): Refine pattern.
16227 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
16228 Chung-Ju Wu <jasonwucj@gmail.com>
16230 * config/nds32/iterators.md (shift_rotate): New code iterator.
16231 (shift): New code attribute.
16232 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
16233 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
16234 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
16235 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
16236 bit-wise operations.
16237 (andsi3, *andsi3): Ditto.
16238 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
16239 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
16240 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
16241 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
16242 nds32_ior_operand, nds32_xor_operand): New predicates.
16244 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16246 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
16247 (addsi3, subsi3): ... this.
16249 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16251 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
16253 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16255 * config/nds32/nds32.md: Adjust indention.
16257 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
16259 * config/nds32/nds32.md (feature): New attribute.
16261 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16263 * config/nds32/nds32.md (subtype): New attribute.
16265 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
16268 * config/arm/arm-builtins.c (arm_expand_builtin): Change
16269 expansion to perform a bitwise AND of the argument followed by a
16270 boolean negation of the result.
16272 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
16274 PR rtl-optimization/84878
16275 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
16276 the basic block. Assert the use reference is not artificial and that
16277 it has an associated insn.
16279 2018-04-04 Michael Matz <matz@suse.de>
16281 * builtins.c (compute_objsize): Pass correct operand
16282 to array_at_struct_end_p.
16284 2018-04-04 Richard Biener <rguenther@suse.de>
16287 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
16288 from contexts for DINFO_LEVEL_TERSE and below.
16290 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
16292 * config/nds32/nds32-doubleword.md (move_<mode>): Require
16293 resiter_operand condition.
16294 * config/nds32/nds32.md (*move<mode>): Ditto.
16296 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
16297 Monk Chiang <sh.chiang04@gmail.com>
16299 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
16301 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16303 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
16305 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16306 Kito Cheng <kito.cheng@gmail.com>
16308 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
16309 nds32_cond_code_str, output_cond_branch,
16310 output_cond_branch_compare_zero, nds32_expand_cbranch,
16311 nds32_expand_cstore, nds32_expand_movcc,
16312 nds32_output_cbranchsi4_equality_zero,
16313 nds32_output_cbranchsi4_equality_reg,
16314 nds32_output_cbranchsi4_equality_reg_or_const_int,
16315 nds32_output_cbranchsi4_greater_less_zero: New functions.
16316 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
16317 nds32_expand_cstore, nds32_expand_movcc,
16318 nds32_output_cbranchsi4_equality_zero,
16319 nds32_output_cbranchsi4_equality_reg,
16320 nds32_output_cbranchsi4_equality_reg_or_const_int,
16321 nds32_output_cbranchsi4_greater_less_zero): Declare.
16322 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
16323 nds32_rimm11s_operand): New predicates.
16324 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
16325 * config/nds32/nds32.md: Rewrite all the branch and conditional move
16328 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
16330 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
16331 * config/nds32/nds32.md: Ditto.
16332 * config/nds32/pipelines.md: Ditto.
16334 2018-04-04 Richard Biener <rguenther@suse.de>
16336 PR tree-optimization/85168
16337 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
16338 propagating abnormals.
16340 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16342 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
16344 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16345 Kito Cheng <kito.cheng@gmail.com>
16347 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
16348 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
16349 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
16350 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
16351 * config/nds32/nds32.md (sibcall_internal): New.
16352 (sibcall_register): Remove.
16353 (sibcall_immediate): Remove.
16354 (sibcall_value_internal): New.
16355 (sibcall_value_register): Remove.
16356 (sibcall_value_immediate): Remove.
16357 * config/nds32/predicates.md (nds32_general_register_operand): New.
16358 (nds32_call_address_operand): New.
16360 2018-04-03 Jakub Jelinek <jakub@redhat.com>
16362 PR rtl-optimization/85167
16363 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
16364 bb_defs if *split_p, instead preinitialize it to NULL.
16366 PR tree-optimization/85156
16367 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
16368 evaluating the argument multiple times.
16370 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
16372 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
16374 (_mm_cvtpd_ps): Likewise.
16375 (_mm_cvttpd_epi32): Likewise.
16376 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
16377 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
16378 vector, pixel, and bool following altivec.h include.
16380 2018-04-03 Martin Sebor <msebor@redhat.com>
16382 * doc/extend.texi (Common Function Attributes): Clarify.
16383 (const attribute): Likewise.
16384 (pure attribute): Likewise.
16386 2018-04-03 Jakub Jelinek <jakub@redhat.com>
16389 * config/i386/i386.c (ix86_expand_vector_set): Use
16390 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
16392 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
16394 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
16395 instructions when changing rounding bits to preserve precision bits
16396 in the x87 control word.
16398 2018-04-03 Martin Liska <mliska@suse.cz>
16400 PR tree-optimization/82491
16401 * rtl.h (strip_offset_and_add): Replace += suboffset with
16402 poly_uint64 () + suboffset.
16404 2018-03-29 Martin Liska <mliska@suse.cz>
16405 Martin Jambor <mjambor@suse.cz>
16408 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
16409 param_type is not an integral or pointer type.
16411 2018-04-03 Richard Biener <rguenther@suse.de>
16413 * sese.h (recompute_all_dominators): Remove.
16415 2018-04-02 Martin Sebor <msebor@redhat.com>
16417 * doc/invoke.texi (-Wrestrict): Fix typos.
16419 2018-04-02 Jim Wilson <jimw@sifive.com>
16421 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
16422 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
16423 (<optab>di3, <optab>si3_extend): Likewise.
16424 (<optab>si3_mask, <optab>si3_mask_1): New.
16425 (<optab>di3_mask, <optab>di3_mask_1): New.
16426 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
16427 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
16428 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
16430 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
16432 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
16435 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
16437 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
16438 (nds32_canonicalize_comparison): New function.
16440 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16441 Kito Cheng <kito.cheng@gmail.com>
16442 Kuan-Lin Chen <kuanlinchentw@gmail.com>
16444 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
16445 * config/nds32/constants.md (unspec_volatile_element): Add
16446 UNSPEC_VOLATILE_RELAX_GROUP.
16447 * config/nds32/nds32-relax-opt.c: New file.
16448 * config/nds32/nds32-predicates.c
16449 (nds32_symbol_load_store_p): New function.
16450 * config/nds32/nds32-protos.h
16451 (nds32_symbol_load_store_p): Declare function.
16452 (make_pass_nds32_relax_opt): Declare new rtl pass function.
16453 * config/nds32/nds32.c
16454 (nds32_register_pass): New function to register pass.
16455 (nds32_register_passes): New function to register passes.
16456 * config/nds32/nds32.md (relax_group): New pattern.
16457 * config/nds32/nds32.opt (mrelax-hint): New option.
16458 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
16460 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
16462 * config/nds32/t-nds32: Modify files dependency.
16464 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16466 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
16467 (PROFILE_HOOK): Define its implementation.
16469 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16471 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
16472 type and 32-bit size.
16474 2018-04-01 Jakub Jelinek <jakub@redhat.com>
16476 PR middle-end/85090
16477 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
16478 (V_128_256): New mode iterator.
16479 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
16480 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
16481 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
16483 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
16484 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
16486 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
16489 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
16490 NaN inputs correctly.
16492 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
16495 * config/rs6000/vsx.md (??r): New mode attribute.
16496 (*vsx_mov<mode>_64bit): Use it.
16497 (*vsx_mov<mode>_32bit): Likewise.
16499 2018-03-30 Martin Sebor <msebor@redhat.com>
16501 PR tree-optimization/84818
16502 * builtins.c (check_access): Use warning_n.
16504 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16507 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
16509 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
16512 2018-03-30 Julia Koval <julia.koval@intel.com>
16515 * x86-tune.def (movx, partial_reg_dependency): Enable for
16518 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
16520 PR inline-asm/84985
16521 * lra-constraints.c (process_alt_operands): Move setting
16522 this_alternative_matches below.
16524 2018-03-29 Martin Liska <mliska@suse.cz>
16527 * doc/invoke.texi: Document how LTO works with debug info.
16528 Describe auto-load support of binutils. Mention 'x86-64'
16529 as valid option value of -march option.
16531 2018-03-29 Jakub Jelinek <jakub@redhat.com>
16533 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
16536 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
16537 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
16538 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
16541 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
16544 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
16545 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
16546 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
16547 for RS6000_BTM_POWERPC64.
16548 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
16549 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
16550 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
16553 (DIVDEU): Likewise.
16555 2018-03-28 Carl Love <cel@us.ibm.com>
16558 2017-09-27 Carl Love <cel@us.ibm.com>
16560 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
16561 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
16562 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
16565 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16567 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
16568 instead of __vector bool.
16569 (_mm_max_pu8): Likewise.
16570 (_mm_min_pi16): Likewise.
16572 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
16575 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
16576 (DIVWEUO): Likewise.
16577 (DIVDEO): Likewise.
16578 (DIVDEUO): Likewise.
16579 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
16580 DIVWEUO and DIVDEUO.
16581 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
16582 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
16583 (div_extend): Likewise.
16584 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
16586 (__builtin_divweuo): Likewise.
16587 (__builtin_divdeo): Likewise.
16588 (__builtin_divdeuo): Likewise.
16590 2018-03-28 Jakub Jelinek <jakub@redhat.com>
16593 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
16594 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
16596 PR tree-optimization/82004
16597 * gimple-match-head.c (optimize_pow_to_exp): New function.
16598 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
16599 Don't fold to exp if optimize_pow_to_exp is false.
16601 2018-03-28 Martin Liska <mliska@suse.cz>
16604 * calls.c (initialize_argument_information): Fix trailing space.
16605 * common.opt: Fix typo and provide better explanation for
16606 -fsanitize-coverage option.
16607 * config/i386/i386.opt: Fix typo.
16609 2018-03-28 Jakub Jelinek <jakub@redhat.com>
16610 Martin Liska <mliska@suse.cz>
16613 * gimplify.c (asan_poison_variable): Don't do the check for
16614 gimplify_omp_ctxp here.
16615 (gimplify_decl_expr): Do it here.
16616 (gimplify_target_expr): Likewise.
16618 2018-03-28 Martin Liska <mliska@suse.cz>
16621 * config/i386/i386.c (ix86_function_arg_advance): Do not call
16622 chkp_type_bounds_count if MPX is not enabled.
16624 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
16626 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
16628 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
16631 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
16632 function to create the function decl for complex long double
16633 multiply and divide for -mabi=ieeelongdouble.
16634 (init_float128_ieee): Call it.
16636 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
16639 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
16640 -fcf-protection=branch -mibt.
16641 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
16643 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16646 * config/arm/arm.c (arm_valid_symbolic_address): Handle
16647 arm_word_relocations.
16649 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
16652 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
16653 extern array declarations.
16655 2018-03-27 Richard Biener <rguenther@suse.de>
16657 PR middle-end/84067
16658 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
16659 explicit single_use checks.
16661 2018-03-27 Richard Biener <rguenther@suse.de>
16663 PR tree-optimization/85082
16664 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
16667 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16669 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
16670 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
16671 Turn on fasynchronous-unwind-tables and funwind-tables.
16673 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
16676 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
16677 (*bmi_blsr_<mode>_ccz): Ditto.
16679 2018-03-26 Tom de Vries <tom@codesourcery.com>
16681 PR tree-optimization/85063
16682 * omp-general.c (offloading_function_p): New function. Factor out
16684 * omp-offload.c (pass_omp_target_link::gate): ... here.
16685 * omp-general.h (offloading_function_p): Declare.
16686 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
16687 with attribute omp declare target for offloading functions.
16689 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
16691 PR tree-optimization/84005
16692 * tree-data-ref.h (get_base_for_alignment): Declare.
16693 * tree-data-ref.c (get_base_for_alignment_1): New function.
16694 (get_base_for_alignment): Likewise.
16695 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
16696 get_base_for_alignment to find a suitable base object, instead
16697 of always using drb->base_address.
16699 2018-03-23 Jakub Jelinek <jakub@redhat.com>
16701 PR inline-asm/85022
16702 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
16703 known size by default.
16705 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
16707 PR inline-asm/85030
16708 * lra-constraints.c (process_alt_operands): Don't match BLKmode
16709 and non BLKmode operands.
16711 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16714 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
16715 Clean up attributes.
16717 2018-03-23 Richard Biener <rguenther@suse.de>
16720 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
16721 we are going to emit early debug for LTO.
16723 2018-03-23 Jakub Jelinek <jakub@redhat.com>
16725 PR inline-asm/85034
16726 * function.c (match_asm_constraints_1): Don't optimize if input
16727 doesn't satisfy general_operand predicate for output's mode.
16729 PR inline-asm/85022
16730 * alias.c (write_dependence_p): Don't require for x_canonicalized
16731 non-VOIDmode if x has VOIDmode.
16734 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
16735 just don't try to optimize it rather than assert it never happens.
16737 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
16739 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
16740 macro expansions for definition of ST_INTERNAL_<mode> and
16741 LD_INTERNAL_<mode> builtins.
16742 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
16744 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
16746 (altivec_expand_st_builtin): Likewise.
16747 (altivec_expand_builtin): Remove calls to deleted functions.
16748 (rs6000_address_for_altivec): Delete this function.
16749 * config/rs6000/vector.md: Remove expands for
16750 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
16752 2018-03-22 Sudakshina Das <sudi.das@arm.com>
16755 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
16756 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
16757 re-computing once computed.
16758 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
16759 (arm_init_machine_status): Initialize
16760 machine->static_chain_stack_bytes.
16762 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
16765 * doc/extend.texi: Add four new prototypes for vec_ld.
16766 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
16767 definitions for more logical presentation.
16768 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
16769 entries for V1TI variants of __builtin_altivec_ld builtin.
16770 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
16771 handling of V1TI variant of LVX icode pattern.
16772 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
16773 (rs6000_gimple_fold_builtin): Likewise.
16774 (altivec_init_builtins): Add code to define
16775 __builtin_altivec_lvx_v1ti function.
16777 2018-03-22 Jakub Jelinek <jakub@redhat.com>
16779 PR inline-asm/84941
16780 * function.c (match_asm_constraints_1): Don't do the optimization
16781 if input isn't a REG, SUBREG, MEM or constant.
16783 2018-03-22 Tom de Vries <tom@codesourcery.com>
16785 PR tree-optimization/84956
16786 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
16787 bb_has_abnormal_pred.
16789 2018-03-22 Jakub Jelinek <jakub@redhat.com>
16792 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
16793 DECL_INITIAL (decl) to decl at the end.
16794 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
16795 adjust the comment.
16797 2018-03-21 Joseph Myers <joseph@codesourcery.com>
16799 * doc/extend.texi (__builtin_tgmath): Document when complex
16800 integer types are treated as _Complex _Float64.
16802 2018-03-21 Tom de Vries <tom@codesourcery.com>
16804 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
16806 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16808 PR tree-optimization/84960
16809 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
16810 if it is ENTRY block, move them into single succ of ENTRY in that case.
16812 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
16814 PR tree-optimization/84811
16815 * poly-int.h (poly_span_traits): Remove the T3 parameter and
16816 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
16817 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
16818 (known_subrange_p): Update accordingly. Cast each value involved
16819 in the size comparison, rather than casting the result of the
16822 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16824 PR tree-optimization/84982
16825 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
16826 by flipping the least significant bit rather than all bits from
16827 bitpos to bitpos + bitsize - 1.
16829 2018-03-21 Nathan Sidwell <nathan@acm.org>
16831 * doc/extend.texi (Deprecated Features): Remove mention of
16832 long-deleted deprecations.
16834 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16837 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
16838 * configure: Regenerate.
16840 2018-03-21 Tom de Vries <tom@codesourcery.com>
16842 PR tree-optimization/83126
16843 * tree-parloops.c (num_phis): New function.
16844 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
16846 2018-03-21 Nathan Sidwell <nathan@acm.org>
16848 * doc/extend.texi (Deprecated Features): Update deprecated flags,
16849 mention anon-struct/union members and trailing attributes.
16851 2018-03-21 Bin Cheng <bin.cheng@arm.com>
16853 PR tree-optimization/84969
16854 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
16855 builtin memset partitions if they set different rhs values.
16857 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16859 PR rtl-optimization/84989
16860 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
16861 VEC_DUPLICATE with scalar result mode.
16863 2018-03-21 Martin Liska <mliska@suse.cz>
16866 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
16867 not intended return statement.
16869 2018-03-21 Martin Liska <mliska@suse.cz>
16872 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
16873 (chkp_find_bound_slots_1): Limit number of iterations.
16875 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
16878 * Minor grammar fixes for x86 options.
16880 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16883 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
16884 holding REG_CFA_RESTORE notes, instead turn them into a USE.
16886 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
16889 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
16890 (altivec_lvx_<mode>_1op): Likewise.
16891 (altivec_stvx_<mode>_2op): Likewise.
16892 (altivec_stvx_<mode>_1op): Likewise.
16893 (altivec_lvx_<VM2:mode>): New define_expand.
16894 (altivec_stvx_<VM2:mode>): Likewise.
16895 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
16896 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16897 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
16898 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16899 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
16900 (rs6000_gen_lvx): Likewise.
16901 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
16902 (altivec_expand_stv_builtin): Likewise.
16903 (altivec_expand_builtin): Likewise.
16904 * config/rs6000/vector.md: Likewise.
16906 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16909 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
16912 2018-03-20 Richard Biener <rguenther@suse.de>
16915 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
16916 sign-conversions as zero, fall back to standard scalar_stmt
16919 2018-03-20 Martin Liska <mliska@suse.cz>
16922 * predict.c (rebuild_frequencies): Handle case when we have
16923 PROFILE_ABSENT, but flag_guess_branch_prob is false.
16925 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16928 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
16929 flag_section_anchors.
16930 * varasm.c (use_blocks_for_decl_p): Remove hack for
16931 dw2_force_const_mem.
16934 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
16936 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
16937 be created, use lowpart_subreg of operands[0] rather than operands[0]
16939 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
16940 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
16941 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
16942 and n constraint instead of aarch64_shift_imm_di and Usd.
16943 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
16944 (*aarch64_<optab>_reg_minus<mode>3): ... this.
16946 2018-03-20 Sudakshina Das <sudi.das@arm.com>
16949 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
16950 to favor GPR over NEON registers.
16951 (<shift>di3_neon): Likewise.
16953 2018-03-20 Tom de Vries <tom@codesourcery.com>
16956 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
16957 (nvptx_process_pars): Emit bar.sync asap and alap.
16959 2018-03-20 Tom de Vries <tom@codesourcery.com>
16962 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
16963 seen_label if seen_label is already set.
16965 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16968 * config/i386/i386.c (fold_builtin_cpu): For features above 31
16969 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
16970 Use 1U instead of 1. Formatting fixes.
16973 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
16974 instead of TREE_TYPE (s1) for the return value.
16976 2018-03-19 Jakub Jelinek <jakub@redhat.com>
16978 PR tree-optimization/84946
16979 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
16980 bitsize + bitsize in poly_uint64 rather than poly_int64.
16983 * dwarf2asm.c: Include fold-const.c.
16984 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
16985 of decl rather than decl itself.
16987 PR rtl-optimization/84643
16988 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
16990 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
16993 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
16994 calling assemble_variable.
16996 2018-03-19 Sudakshina Das <sudi.das@arm.com>
16999 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
17000 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
17002 2018-03-19 Jim Wilson <jimw@sifive.com>
17005 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
17006 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
17007 (riscv_first_stack_step): Likewise.
17008 (riscv_option_override): Use STACK_BOUNDARY instead of
17009 MIN_STACK_BOUNDARY.
17010 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
17011 MIN_STACK_BOUNDARY.
17012 (BIGGEST_ALIGNMENT): Set to 128.
17013 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
17014 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
17017 2018-03-19 Richard Biener <rguenther@suse.de>
17019 PR tree-optimization/84933
17020 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
17021 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
17023 2018-03-19 Richard Biener <rguenther@suse.de>
17025 PR tree-optimization/84859
17026 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
17027 (cond_if_else_store_replacement): Perform sinking operation on
17028 single-store BBs regardless of MAX_STORES_TO_SINK setting.
17029 Generalize what a BB with a single eligible store is.
17031 2018-03-19 Richard Biener <rguenther@suse.de>
17033 PR tree-optimization/84929
17034 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
17035 chrec_is_positive against non-chrec arg.
17037 2018-03-19 Tamar Christina <tamar.christina@arm.com>
17040 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
17042 2018-03-18 Martin Liska <mliska@suse.cz>
17044 PR rtl-optimization/84635
17045 * regrename.c (build_def_use): Use matches_mode only when
17048 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
17050 PR tree-optimization/84913
17051 * tree-vect-loop.c (vectorizable_reduction): Don't try to
17052 vectorize chains of COND_EXPRs.
17054 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
17056 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
17058 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
17060 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
17062 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
17064 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
17066 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
17067 Kito Cheng <kito.cheng@gmail.com>
17069 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
17070 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
17071 (nds32_adjust_reg_alloc_order): New function.
17072 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
17074 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
17076 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
17077 nds32_print_operand, nds32_print_operand_address): Use
17078 HOST_WIDE_INT_PRINT_DEC instead.
17080 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
17082 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
17084 2018-03-17 Jakub Jelinek <jakub@redhat.com>
17087 * config/i386/i386.c (initial_ix86_tune_features,
17088 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
17089 unsigned long long.
17090 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
17091 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
17092 rather than 1u << ix86_tune. Formatting fix.
17093 (ix86_option_override_internal): Change ix86_arch_mask from
17094 unsigned int to unsigned HOST_WIDE_INT, initialize to
17095 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
17096 (ix86_function_specific_restore): Likewise.
17098 2018-03-16 Jakub Jelinek <jakub@redhat.com>
17101 * postreload.c (reload_combine_recognize_pattern): Perform
17102 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
17103 truncate_int_for_mode the result for the destination's mode.
17106 * hsa-gen.c (mem_type_for_type): Fix comment typo.
17107 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
17109 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17112 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
17115 * lra-assigns.c (lra_split_hard_reg_for): Don't use
17116 regno_allocno_class_array and sorted_pseudos.
17117 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
17118 insns where regno is used.
17120 2018-03-16 Martin Liska <mliska@suse.cz>
17123 * multiple_target.c (create_dispatcher_calls): Redirect
17124 reference in the symbol table.
17126 2018-03-16 Martin Liska <mliska@suse.cz>
17129 * multiple_target.c (create_dispatcher_calls): Redirect also
17132 2018-03-16 Jakub Jelinek <jakub@redhat.com>
17136 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
17137 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
17138 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
17140 2018-03-16 Julia Koval <julia.koval@intel.com>
17142 * doc/invoke.texi (Skylake Server): Add CLWB.
17143 Cannonlake): Remove CLWB.
17145 2018-03-16 Jakub Jelinek <jakub@redhat.com>
17147 PR tree-optimization/84841
17148 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
17150 (FLOAT_ONE_CONST_TYPE): Define.
17151 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
17152 (sort_by_operand_rank): Put entries with higher constant_type last
17153 rather than first to match comments.
17155 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
17157 * config/nios2/nios2.md (movsi_internal): Fix thinko in
17160 2018-03-15 Jakub Jelinek <jakub@redhat.com>
17163 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
17164 check and use address of target always.
17166 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
17169 * config/i386/i386.c (indirect_thunk_needed): Update comments.
17170 (indirect_thunk_bnd_needed): Likewise.
17171 (indirect_thunks_used): Likewise.
17172 (indirect_thunks_bnd_used): Likewise.
17173 (indirect_return_needed): New.
17174 (indirect_return_bnd_needed): Likewise.
17175 (output_indirect_thunk_function): Add a bool argument for
17177 (output_indirect_thunk_function): Don't generate alias for
17178 function return thunk.
17179 (ix86_code_end): Call output_indirect_thunk_function to generate
17180 function return thunks.
17181 (ix86_output_function_return): Set indirect_return_bnd_needed
17182 and indirect_return_needed instead of indirect_thunk_bnd_needed
17183 and indirect_thunk_needed.
17185 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
17187 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
17188 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
17189 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
17191 2018-03-15 David Malcolm <dmalcolm@redhat.com>
17192 Paul Hua <paul.hua.gm@gmail.com>
17195 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
17197 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
17199 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
17200 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
17201 resp. SFmode cases.
17203 2018-03-15 Tamar Christina <tamar.christina@arm.com>
17206 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
17207 instead of GET_MODE_SIZE when comparing Units.
17209 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
17212 * varasm.c (hash_section): Return an unchangeble hash value
17213 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
17214 Return !aarch64_can_use_per_function_literal_pools_p ().
17216 2018-03-15 Jakub Jelinek <jakub@redhat.com>
17219 * optabs.c (emit_conditional_move): Pass address of cmode's copy
17220 rather than address of cmode as last argument to prepare_cmp_insn.
17222 2018-03-15 Julia Koval <julia.koval@intel.com>
17224 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
17225 F_AVX512VNNI, F_AVX512BITALG): New.
17227 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
17230 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
17231 insn for floating-point loads and stores.
17233 2018-03-14 Carl Love <cel@us.ibm.com>
17235 * config/rs6000/rs6000-c.c: Add macro definitions for
17236 ALTIVEC_BUILTIN_VEC_PERMXOR.
17237 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
17238 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
17239 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
17240 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
17242 * config/doc/extend.texi: Add prototypes for vec_permxor.
17244 2018-03-14 David Malcolm <dmalcolm@redhat.com>
17247 * diagnostic-show-locus.c (class layout_point): Convert m_line
17248 from int to linenum_type.
17249 (line_span::comparator): Use linenum "compare" function when
17250 comparing line numbers.
17251 (test_line_span): New function.
17252 (layout_range::contains_point): Convert param "row" from int to
17254 (layout_range::intersects_line_p): Likewise.
17255 (layout::will_show_line_p): Likewise.
17256 (layout::print_source_line): Likewise.
17257 (layout::should_print_annotation_line_p): Likewise.
17258 (layout::print_annotation_line): Likewise.
17259 (layout::print_leading_fixits): Likewise.
17260 (layout::annotation_line_showed_range_p): Likewise.
17261 (struct line_corrections): Likewise for field m_row.
17262 (line_corrections::line_corrections): Likewise for param "row".
17263 (layout::print_trailing_fixits): Likewise.
17264 (layout::get_state_at_point): Likewise.
17265 (layout::get_x_bound_for_row): Likewise.
17266 (layout::print_line): Likewise.
17267 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
17268 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
17269 * input.c (selftest::test_linenum_comparisons): New function.
17270 (selftest::input_c_tests): Call it.
17271 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
17272 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
17273 * selftest.h (ASSERT_GT): New macro.
17274 (ASSERT_GT_AT): New macro.
17275 (ASSERT_LT): New macro.
17276 (ASSERT_LT_AT): New macro.
17278 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
17280 PR rtl-optimization/84780
17281 * combine.c (distribute_links): Don't make a link based on pc_rtx.
17283 2018-03-14 Martin Liska <mliska@suse.cz>
17285 * tree.c (record_node_allocation_statistics): Use
17286 get_stats_node_kind.
17287 (get_stats_node_kind): New function extracted from
17288 record_node_allocation_statistics.
17289 (free_node): Use get_stats_node_kind.
17291 2018-03-14 Richard Biener <rguenther@suse.de>
17293 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
17294 that the value-set of ANTIC_IN doesn't grow.
17297 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
17299 (BB_VISITED_WITH_VISITED_SUCCS): New define.
17300 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
17302 2018-03-14 Julia Koval <julia.koval@intel.com>
17304 * config.gcc (icelake-client, icelake-server): New.
17306 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
17307 (initial_ix86_arch_features): Ditto.
17308 (PTA_SKYLAKE): Add SGX.
17309 (PTA_ICELAKE): Remove.
17310 (PTA_ICELAKE_CLIENT): New.
17311 (PTA_ICELAKE_SERVER): New.
17312 (ix86_option_override_internal): Split up icelake on icelake client and
17314 (get_builtin_code_for_version): Ditto.
17315 (fold_builtin_cpu): Ditto.
17316 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
17317 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
17318 * config/i386/i386.h (processor_type): Ditto.
17319 * doc/invoke.texi: Ditto.
17321 2018-03-14 Jakub Jelinek <jakub@redhat.com>
17324 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
17325 INTEGER_CST offset, add it together with bitpos / 8 and
17326 sign extend based on POINTER_SIZE.
17330 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
17333 * config/i386/constraints.md (Yc): New register constraint.
17334 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
17335 Use Yc constraint for alternative 2 of operand 0. Remove
17336 preferred_for_speed attribute.
17338 2018-03-14 Richard Biener <rguenther@suse.de>
17340 PR tree-optimization/84830
17341 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
17342 with the old one to avoid oscillations.
17344 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
17347 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
17349 (assign_by_spills): Return a flag of reload assignment failure.
17350 Do not process the reload assignment failures. Do not spill other
17351 reload pseudos if they has the same reg class. Update n if
17353 (lra_assign): Add a return arg. Set up from the result of
17354 assign_by_spills call.
17355 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17356 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17357 usage_insns if it is not NULL.
17358 (spill_hard_reg_in_range): New function.
17359 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17360 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17361 function prototypes.
17362 (lra_assign): Change prototype.
17363 * lra.c (lra): Add code to deal with fails by splitting hard reg
17366 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
17368 * config/riscv/riscv.opt (mrelax): New option.
17369 * config/riscv/riscv.c (riscv_file_start): Emit ".option
17370 "norelax" when riscv_mrelax is disabled.
17371 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
17373 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17376 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
17377 reassociation for int modes.
17379 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17381 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
17382 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
17384 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
17385 * config/aarch64/aarch64-sve.md
17386 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
17387 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
17388 (*extend<mode><Vwide>2): Rename to...
17389 (aarch64_sve_extend<mode><Vwide>2): ...this.
17390 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
17391 renaming the old pattern to...
17392 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
17394 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
17395 define_expand, renaming the old pattern to...
17396 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
17397 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
17398 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
17399 account when deciding which SVE instruction the optab should use.
17400 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
17402 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17404 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
17405 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
17406 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
17407 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
17408 (tlsdesc_small_<mode>): Turn a define_expand and use
17409 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
17410 (tlsdesc_small_advsimd_<mode>): ...this.
17411 (tlsdesc_small_sve_<mode>): New pattern.
17413 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17415 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
17416 (UNSPEC_UMUL_HIGHPART): New constants.
17417 (MUL_HIGHPART): New int iteraor.
17418 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
17419 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
17421 (*<su>mul<mode>3_highpart): New define_insn.
17423 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
17426 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
17429 2018-03-13 Martin Liska <mliska@suse.cz>
17432 * (sem_item_optimizer::sem_item_optimizer): Initialize new
17434 (sem_item_optimizer::~sem_item_optimizer): Release it.
17435 (sem_item_optimizer::merge_classes): Register variable aliases.
17436 (sem_item_optimizer::fixup_pt_set): New function.
17437 (sem_item_optimizer::fixup_points_to_sets): Likewise.
17438 * ipa-icf.h: Declare new variables and functions.
17440 2018-03-13 Jakub Jelinek <jakub@redhat.com>
17442 PR middle-end/84834
17443 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
17444 integer_pow2p@2 and test integer_pow2p in condition.
17445 (A < 0 ? C : 0): Similarly for @1.
17447 PR middle-end/84831
17448 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
17449 characters starting at p contain '\0' character, don't look beyond
17453 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
17454 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
17457 * reg-stack.c (change_stack): Change update_end var from int to
17458 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
17459 also call set_block_for_insn on the newly added insns and rescan.
17462 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
17463 on the last operand.
17466 * tree.c (stabilize_reference_1): Return save_expr (e) for
17467 STATEMENT_LIST even if it doesn't have side-effects.
17469 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
17471 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
17473 2018-03-12 Renlin Li <renlin.li@arm.com>
17475 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
17476 aarch64_output_scalar_simd_mov_immediate.
17478 2018-03-12 Martin Sebor <msebor@redhat.com>
17480 PR tree-optimization/83456
17481 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
17482 for perfectly overlapping calls to memcpy.
17483 (gimple_fold_builtin_memory_chk): Same.
17484 (gimple_fold_builtin_strcpy): Handle no-warning.
17485 (gimple_fold_builtin_stxcpy_chk): Same.
17486 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
17488 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
17490 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
17491 parameter. Use it for SFmode.
17492 (rs6000_function_arg_advance_1): Adjust.
17493 (rs6000_function_arg): Adjust.
17494 (rs6000_gimplify_va_arg): Pass false for that new parameter.
17496 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
17498 PR rtl-optimization/84169
17499 PR rtl-optimization/84780
17500 * combine.c (can_combine_p): Check for a 2-insn combination whether
17501 the destination register is used between the two insns, too.
17503 2018-03-12 Richard Biener <rguenther@suse.de>
17505 PR tree-optimization/84803
17506 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
17507 for refs DR analysis didn't process.
17509 2018-03-12 Richard Biener <rguenther@suse.de>
17511 PR tree-optimization/84777
17512 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
17513 force-vectorize loops ignore whether we are optimizing for size.
17515 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
17517 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
17518 (TARGET_MD_ASM_ADJUST): Define.
17520 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
17521 Kito Cheng <kito.cheng@gmail.com>
17522 Chung-Ju Wu <jasonwucj@gmail.com>
17524 * config/nds32/nds32.c (nds32_compute_stack_frame,
17525 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
17526 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
17527 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
17528 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
17529 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
17530 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
17531 * config/nds32/nds32.md (prologue, epilogue): Use macro
17532 NDS32_V3PUSH_AVAILABLE_P to do checking.
17534 2018-03-11 Jakub Jelinek <jakub@redhat.com>
17537 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
17538 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
17539 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
17540 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
17541 addition of most attributes on !orig_type_die or the attribute not
17542 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
17544 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17545 Chung-Ju Wu <jasonwucj@gmail.com>
17547 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
17548 __NDS32_VH__ macro.
17549 * config/nds32/nds32.opt (mvh): New option.
17551 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17552 Chung-Ju Wu <jasonwucj@gmail.com>
17554 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
17556 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
17557 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
17560 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17561 Chung-Ju Wu <jasonwucj@gmail.com>
17563 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
17565 * config/nds32/nds32-multiple.md (strlensi): New pattern.
17566 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
17568 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
17569 Kito Cheng <kito.cheng@gmail.com>
17570 Chung-Ju Wu <jasonwucj@gmail.com>
17572 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
17573 UNSPEC_FFMISM and UNSPEC_FLMISM.
17574 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
17575 for ffb, ffmism and flmism.
17576 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
17577 (unspec_ffmism): Ditto.
17578 (unspec_flmism): Ditto.
17579 (nds32_expand_builtin_impl): Check if string extension is available.
17580 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
17581 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
17583 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
17586 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
17589 * lra-assigns.c (assign_by_spills): Return a flag of reload
17590 assignment failure. Do not process the reload assignment
17591 failures. Do not spill other reload pseudos if they has the same
17593 (lra_assign): Add a return arg. Set up from the result of
17594 assign_by_spills call.
17595 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17596 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17597 usage_insns if it is not NULL.
17598 (spill_hard_reg_in_range): New function.
17599 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17600 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17601 function prototypes.
17602 (lra_assign): Change prototype.
17603 * lra.c (lra): Add code to deal with fails by splitting hard reg
17606 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
17609 * config/i386/i386.opt: Replace Enforcment with Enforcement.
17611 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
17614 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
17615 (dw_val_node): Add val_symbolic_view.
17616 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
17617 (symview_upper_bound): New.
17618 (new_line_info_table): Initialize symviews_since_reset.
17619 (dwarf2out_source_line): Count symviews_since_reset and set
17620 symview_upper_bound.
17621 (dw_val_equal_p): Handle symview.
17622 (add_AT_symview): New.
17623 (print_dw_val): Handle symview.
17624 (attr_checksum, attr_checksum_ordered): Likewise.
17625 (same_dw_val_p, size_of_die): Likewise.
17626 (value_format, output_die): Likewise.
17627 (add_high_low_attributes): Use add_AT_symview for entry_view.
17628 (dwarf2out_finish): Reset symview_upper_bound, clear
17631 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
17634 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
17635 Add strict argument and use it.
17636 (rs6000_split_multireg_move): Update for new strict argument.
17637 (mem_operand_gpr): Disallow all non-offsettable addresses.
17638 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
17640 2018-03-09 Jakub Jelinek <jakub@redhat.com>
17643 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
17644 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
17645 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
17648 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
17649 decl, use remap_type if we want to use the type.
17651 2018-03-09 Martin Sebor <msebor@redhat.com>
17653 PR tree-optimization/84526
17654 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17656 (builtin_access::generic_overlap): Be prepared to handle non-array
17659 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
17661 PR rtl-optimization/84682
17662 * lra-constraints.c (process_address_1): Check is_address flag
17663 for address constraints.
17664 (process_alt_operands): Likewise.
17665 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
17666 preprocess_constraints.
17667 * recog.h (preprocess_constraints): Add oploc parameter.
17669 * recog.c (preprocess_constraints): Test address_operand for
17670 CT_ADDRESS constraints.
17672 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
17675 * lra-assigns.c (assign_by_spills): Return a flag of reload
17676 assignment failure. Do not process the reload assignment
17677 failures. Do not spill other reload pseudos if they has the same
17679 (lra_assign): Add a return arg. Set up from the result of
17680 assign_by_spills call.
17681 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17682 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17683 usage_insns if it is not NULL.
17684 (spill_hard_reg_in_range): New function.
17685 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17686 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17687 function prototypes.
17688 (lra_assign): Change prototype.
17689 * lra.c (lra): Add code to deal with fails by splitting hard reg
17692 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17695 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
17696 Accept complain bool parameter. Only emit errors if it is true.
17697 (arm_parse_cpu_option_name): Likewise.
17698 (arm_target_thumb_only): Adjust callers of the above.
17699 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
17700 prototype to take a default true bool parameter.
17701 (arm_parse_arch_option_name): Likewise.
17703 2018-03-09 David Malcolm <dmalcolm@redhat.com>
17704 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
17708 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
17709 * configure: Regenerate.
17710 * configure.ac ("linker --version-script option"): New.
17711 ("linker soname option"): New.
17713 2018-03-09 Richard Biener <rguenther@suse.de>
17715 PR tree-optimization/84775
17716 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
17717 immediate uses of predicate stmts and mark them modified.
17720 PR tree-optimization/84178
17721 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17723 (version_loop_for_if_conversion): Delay update_ssa call.
17724 (tree_if_conversion): Delay update_ssa until after predicate
17727 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
17730 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
17731 when the function accesses prior frames.
17733 2018-03-08 Jakub Jelinek <jakub@redhat.com>
17736 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
17737 gen_llsym, otherwise call maybe_gen_llsym.
17739 PR inline-asm/84742
17740 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
17741 has ',' character inside of it.
17743 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17746 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
17747 as clobbering CC_REGNUM.
17749 2018-03-08 Richard Biener <rguenther@suse.de>
17751 PR middle-end/84552
17752 * tree-scalar-evolution.c: Include tree-into-ssa.h.
17753 (follow_copies_to_constant): Do not follow SSA names registered
17756 2018-03-08 Richard Biener <rguenther@suse.de>
17758 PR tree-optimization/84178
17759 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17761 (version_loop_for_if_conversion): Delay update_ssa call.
17762 (tree_if_conversion): Delay update_ssa until after predicate
17765 2018-03-08 David Malcolm <dmalcolm@redhat.com>
17767 PR tree-optimization/84178
17768 * tree-if-conv.c (release_bb_predicate): Remove the
17769 the assertion that the stmts have NULL use_ops.
17770 Discard the statements, asserting that they haven't
17771 yet been added to a BB.
17773 2018-03-08 Richard Biener <rguenther@suse.de>
17775 PR tree-optimization/84746
17776 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
17777 (phi_translate): Pass in destination ANTIC_OUT set.
17778 (phi_translate_1): Likewise. For a simplified result lookup
17779 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
17780 (phi_translate_set): Adjust.
17781 (do_pre_regular_insertion): Likewise.
17782 (do_pre_partial_partial_insertion): Likewise.
17784 2018-03-08 Martin Liska <mliska@suse.cz>
17786 PR gcov-profile/84735
17787 * doc/gcov.texi: Document usage of profile files.
17788 * gcov-io.h: Document changes in the format.
17790 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
17794 * dwarf2out.c (struct dw_line_info_table): Update comments for
17796 (FORCE_RESET_NEXT_VIEW): New.
17797 (FORCE_RESETTING_VIEW_P): New.
17798 (RESETTING_VIEW_P): Check for -1 too.
17799 (ZERO_VIEW_P): Likewise.
17800 (new_line_info_table): Force-reset next view.
17801 (dwarf2out_begin_function): Likewise.
17802 (dwarf2out_source_line): Simplify zero_view_p initialization.
17803 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
17804 view directly. Omit view when omitting .loc at line 0.
17806 2018-03-08 Jakub Jelinek <jakub@redhat.com>
17808 PR tree-optimization/84740
17809 * tree-switch-conversion.c (process_switch): Call build_constructors
17810 only if info.phi_count is non-zero.
17812 PR tree-optimization/84739
17813 * tree-tailcall.c (find_tail_calls): Check call arguments against
17814 DECL_ARGUMENTS (current_function_decl) rather than
17815 DECL_ARGUMENTS (func) when checking for tail recursion.
17817 2018-03-07 Jakub Jelinek <jakub@redhat.com>
17819 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
17820 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
17821 Volker Reichelt's entry and add entries for people that perform
17822 GCC fuzzy testing and report numerous bugs.
17824 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
17827 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
17828 readonly data in sdata, if that is disabled.
17829 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
17830 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
17831 -mreadonly-in-sdata option.
17833 2018-03-07 Martin Sebor <msebor@redhat.com>
17835 PR tree-optimization/84468
17836 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
17837 basic block when looking for nul assignment.
17839 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
17842 * except.h (output_function_exception_table): Adjust prototype.
17843 * except.c (output_function_exception_table): Remove FNNAME parameter
17844 and add SECTION parameter. Ouput one part of the table at a time.
17845 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
17846 the first part of the exception table and emit unwind directives.
17847 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
17848 (i386_pe_seh_cold_init): Likewise.
17849 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
17850 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
17851 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
17852 (ix86_output_call_insn): Emit a nop in one more case for SEH.
17853 * config/i386/winnt.c: Include except.h.
17854 (struct seh_frame_state): Add reg_offset, after_prologue and
17855 in_cold_section fields.
17856 (i386_pe_seh_end_prologue): Set seh->after_prologue.
17857 (i386_pe_seh_cold_init): New function.
17858 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
17859 to seh->in_cold_section.
17860 (seh_emit_push): Record the offset of the push.
17861 (seh_emit_save): Record the offet of the save.
17862 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
17863 Test seh->after_prologue to disregard the epilogue.
17864 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
17865 (i386_pe_end_cold_function): New function.
17867 2018-03-07 Jakub Jelinek <jakub@redhat.com>
17870 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
17871 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
17874 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
17876 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
17877 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
17879 PR middle-end/84723
17880 * multiple_target.c: Include tree-inline.h and intl.h.
17881 (expand_target_clones): Diagnose and fail if node->definition and
17882 !tree_versionable_function_p (node->decl).
17884 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
17886 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
17888 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
17889 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
17890 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
17892 2018-03-06 Jakub Jelinek <jakub@redhat.com>
17895 * combine.c (try_combine): Use reg_or_subregno instead of handling
17896 just paradoxical SUBREGs and REGs.
17898 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
17900 * config/arc/arc.c (arc_finalize_pic): Remove function.
17901 (arc_must_save_register): We use single base PIC register, remove
17902 checks to save/restore the PIC register.
17903 (arc_expand_prologue): Likewise.
17904 * config/arc/arc-protos.h (arc_set_default_type_attributes):
17906 (arc_verify_short): Likewise.
17907 (arc_attr_type): Likewise.
17908 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
17909 (walk_stores): Likewise.
17910 (arc_address_cost): Make it static.
17911 (arc_verify_short): Likewise.
17912 (branch_dest): Likewise.
17913 (arc_attr_type): Likewise.
17914 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
17915 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
17916 (arc_final_prescan_insn): Remove inserting the nops due to
17917 hardware hazards. It is done in reorg step.
17918 (insn_length_variant_t): Remove.
17919 (insn_length_parameters_t): Likewise.
17920 (arc_insn_length_parameters): Likewise.
17921 (arc_get_insn_variants): Likewise.
17922 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
17924 2018-03-06 Jakub Jelinek <jakub@redhat.com>
17926 PR inline-asm/84683
17927 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
17930 PR tree-optimization/84687
17931 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
17933 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
17935 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17937 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
17938 Rename to ppc_speculation_barrier.
17939 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
17940 __builtin_ppc_speculation_barrier.
17942 2018-03-05 Jakub Jelinek <jakub@redhat.com>
17945 * combine.c (combine_simplify_rtx): Don't try to simplify if
17946 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
17949 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
17951 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
17952 to 32 bytes when compiling for POWER9.
17954 2018-03-05 Jakub Jelinek <jakub@redhat.com>
17957 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
17958 regparm >= 3 with no arg reg available also for calls with
17959 flag_force_indirect_call. Pass decl to ix86_function_regparm.
17962 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
17964 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
17966 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
17969 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
17971 2018-03-05 Richard Biener <rguenther@suse.de>
17973 PR tree-optimization/84486
17974 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
17975 When inserting a __builtin_assume_aligned call set the LHS
17976 SSA name alignment info accordingly.
17978 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
17980 PR tree-optimization/84114
17981 * config/aarch64/aarch64.c (aarch64_reassociation_width)
17982 Avoid reassociation of FLOAT_MODE addition.
17984 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
17986 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
17987 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
17988 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
17989 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
17990 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
17991 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
17992 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
17994 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
17995 __builtin_ia32_wbinvd): New builtins.
17996 (SPECIAL_ARGS2): New.
17997 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
17998 (SPECIAL_ARGS2): New.
17999 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
18000 (ix86_valid_target_attribute_inner_p): Ditto.
18001 (ix86_init_mmx_sse_builtins): Add special_args2.
18002 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
18003 TARGET_WBNOINVD_P): New.
18004 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
18005 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
18006 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
18007 * config/i386/immintrin.h (_wbinvd): New intrinsic.
18008 * config/i386/pconfigintrin.h: New file.
18009 * config/i386/wbnoinvdintrin.h: Ditto.
18010 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
18012 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
18014 2018-03-05 Richard Biener <rguenther@suse.de>
18016 PR tree-optimization/84670
18017 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
18019 (BB_VISITED_WITH_VISITED_SUCCS): New define.
18020 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
18021 (compute_antic_aux): Only assert the number of values in ANTIC_IN
18022 doesn't grow if all successors (recursively) were visited at least
18025 2018-03-05 Richard Biener <rguenther@suse.de>
18027 PR tree-optimization/84650
18028 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
18029 if executed in the loop pipeline.
18031 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
18033 * doc/configfiles.texi (Configuration Files): Move info about
18034 conditionalizing $target-protos.h to...
18035 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
18036 differs from $target-protos.h.
18038 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
18039 Chung-Ju Wu <jasonwucj@gmail.com>
18041 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
18042 * config/nds32/nds32-multiple.md (setmemsi): Define.
18043 * config/nds32/nds32-memory-manipulation.c
18044 (nds32_gen_dup_4_byte_to_word_value): New.
18045 (emit_setmem_word_loop): New.
18046 (emit_setmem_byte_loop): New.
18047 (nds32_expand_setmem_loop): New.
18048 (nds32_expand_setmem_loop_v3m): New.
18049 (nds32_expand_setmem_unroll): New.
18050 (nds32_expand_setmem): New.
18052 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
18053 Chung-Ju Wu <jasonwucj@gmail.com>
18055 * config/nds32/nds32-memory-manipulation.c
18056 (nds32_emit_load_store): New.
18057 (nds32_emit_post_inc_load_store): New.
18058 (nds32_emit_mem_move): New.
18059 (nds32_emit_mem_move_block): New.
18060 (nds32_expand_movmemsi_loop_unknown_size): New.
18061 (nds32_expand_movmemsi_loop_known_size): New.
18062 (nds32_expand_movmemsi_loop): New.
18063 (nds32_expand_movmemsi_unroll): New.
18064 (nds32_expand_movmemqi): Rename ...
18065 (nds32_expand_movmemsi): ... to this.
18066 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
18067 (movmemsi): ... to this.
18068 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
18069 (nds32_expand_movmemsi): ... to this.
18071 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
18072 Monk Chiang <sh.chiang04@gmail.com>
18073 Chung-Ju Wu <jasonwucj@gmail.com>
18075 * config/nds32/nds32-protos.h
18076 (nds32_expand_load_multiple): New arguments.
18077 (nds32_expand_store_multiple): Ditto.
18078 (nds32_valid_multiple_load_store): Rename ...
18079 (nds32_valid_multiple_load_store_p): ... to this.
18080 * config/nds32/nds32-memory-manipulation.c
18081 (nds32_expand_load_multiple): Refine implementation.
18082 (nds32_expand_store_multiple): Ditto.
18083 * config/nds32/nds32-multiple.md
18084 (load_multiple): Update nds32_expand_load_multiple interface.
18085 (store_multiple): Update nds32_expand_store_multiple interface.
18086 * config/nds32/nds32-predicates.c
18087 (nds32_valid_multiple_load_store): Rename ...
18088 (nds32_valid_multiple_load_store_p): ... to this and refine
18090 * config/nds32/predicates.md
18091 (nds32_load_multiple_and_update_address_operation): New predicate.
18092 (nds32_store_multiple_and_update_address_operation): New predicate.
18094 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
18095 Chung-Ju Wu <jasonwucj@gmail.com>
18097 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
18098 (combo): New attribute.
18099 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
18101 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
18103 * config/nds32/nds32.opt: Change -mcmodel= default value.
18105 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
18106 Monk Chiang <sh.chiang04@gmail.com>
18107 Chung-Ju Wu <jasonwucj@gmail.com>
18109 * config/nds32/constants.md (unspec_element): New enum.
18110 * config/nds32/constraints.md (Umw): New constraint.
18111 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
18112 * config/nds32/nds32-intrinsic.md: Likewise.
18113 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
18114 (nds32_valid_smw_lwm_base_p): New.
18115 (nds32_output_smw_single_word): New.
18116 (nds32_output_lmw_single_word): New.
18117 (nds32_expand_unaligned_load): New.
18118 (nds32_expand_unaligned_store): New.
18119 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
18120 (nds32_output_smw_single_word): Declare.
18121 (nds32_output_lmw_single_word): Declare.
18122 (nds32_expand_unaligned_load): Declare.
18123 (nds32_expand_unaligned_store): Declare.
18124 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
18125 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
18126 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
18127 NDS32_BUILTIN_UASTORE_DW.
18128 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
18131 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
18132 Kito Cheng <kito.cheng@gmail.com>
18133 Chung-Ju Wu <jasonwucj@gmail.com>
18135 * config/nds32/nds32-intrinsic.c
18136 (nds32_expand_builtin_null_ftype_reg): Delete.
18137 (nds32_expand_builtin_reg_ftype_imm): Ditto.
18138 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
18139 (nds32_read_argument): New.
18140 (nds32_legitimize_target): Ditto.
18141 (nds32_legitimize_argument): Ditto.
18142 (nds32_check_constant_argument): Ditto.
18143 (nds32_expand_unop_builtin): Ditto.
18144 (nds32_expand_unopimm_builtin): Ditto.
18145 (nds32_expand_binop_builtin): Ditto.
18146 (nds32_builtin_decl_impl): Ditto.
18147 (builtin_description): Ditto.
18148 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
18149 (nds32_init_builtins_impl): Ditto.
18150 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
18151 (nds32_builtin_decl): New.
18152 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
18153 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
18155 2018-03-02 Jeff Law <law@redhat.com>
18157 * reorg.c (stop_search_p): Handle DEBUG_INSN.
18158 (redundant_insn, fill_simple_delay_slots): Likewise.
18159 (fill_slots_from_thread): Likewise.
18160 * resource.c (mark_referenced_resources): Likewise.
18161 (mark_set_resources, find_dead_or_set_registers): Likewise.
18163 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18165 * substring-locations.h (format_warning_va): Formatting fix for
18166 ATTRIBUTE_GCC_DIAG.
18167 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
18169 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
18170 * substring-locations.c: Include intl.h.
18171 (format_warning_va): Turned into small wrapper around
18172 format_warning_n_va, renamed to ...
18173 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
18174 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
18176 (format_warning_at_substring_n): New function.
18177 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
18178 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
18179 format_warning_at_substring with just a shorter name instead of
18180 const function pointer.
18181 (fmtwarn_n): New function.
18182 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
18183 appropriate, get rid of all the fmtstr temporaries, move conditionals
18184 with G_() wrapped string literals directly into fmtwarn arguments,
18185 cast dir.len to (int), formatting fixes.
18187 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
18189 * doc/invoke.texi: Remove "Cilk Plus" references.
18191 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18192 Richard Biener <rguenther@suse.de>
18195 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
18196 for error or warning attributes if CALL_FROM_THUNK_P is set.
18199 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18202 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
18203 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
18206 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
18207 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
18209 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
18210 instead of -1U in last predictors element's probability member.
18212 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
18215 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
18216 arguments if they are comparable.
18218 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
18220 PR tree-optimization/84634
18221 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
18222 masks and masked_loop_p with a single loop_masks, making sure it's
18223 null for bb vectorization.
18225 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
18227 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
18228 (vect_analyze_data_ref_access): Use loop->safe_len rather than
18229 loop->force_vectorize to check whether there is no alias.
18231 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18234 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
18236 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
18238 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
18239 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
18240 instead of a loop around prev_real_insn.
18241 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
18244 PR inline-asm/84625
18245 * config/i386/i386.c (ix86_print_operand): Use conditional
18246 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
18249 2018-03-02 Richard Biener <rguenther@suse.de>
18251 PR tree-optimization/84427
18252 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
18253 (bitmap_set_subtract_values): Rewrite to handle multiple
18256 (prune_clobbered_mems): Likewise.
18257 (phi_translate): Take edge instead of pred/phiblock.
18258 (phi_translate_1): Likewise.
18259 (phi_translate_set): Likewise. Insert all translated
18260 exprs for a value into the set, keeping possibly multiple
18261 expressions per value.
18262 (compute_antic_aux): Adjust for phi_translate changes.
18263 When intersecting union the expressions and prune those
18264 not in the final value set, keeping possibly multiple
18265 expressions per value. Do not use value-insertion
18266 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
18267 all expressions. Add verification that the value-sets
18268 only shrink during iteration.
18269 (compute_partial_antic_aux): Adjust for the phi_translate changes.
18270 (do_pre_regular_insertion): Likewise.
18271 (do_pre_partial_partial_insertion): Likewise.
18273 2018-03-02 Richard Biener <rguenther@suse.de>
18276 * config/darwin.c (saved_debug_info_level): New static global.
18277 (darwin_asm_lto_start): Disable debug info generation for LTO out.
18278 (darwin_asm_lto_end): Restore debug info generation settings.
18280 2018-03-01 Martin Liska <mliska@suse.cz>
18283 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
18284 volatile arguments.
18286 2018-03-01 Richard Biener <rguenther@suse.de>
18289 * dwarf2out.c (gen_variable_die): Properly handle late VLA
18290 type annotation with LTO when debug was disabled at compile-time.
18292 2018-03-01 Matthew Fortune <mfortune@gmail.com>
18294 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
18296 (mips_final_postscan_insn): Likewise.
18298 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
18300 PR rtl-optimization/84528
18301 * alias.c (init_alias_target): Add commentary.
18302 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
18303 a unique base value if the frame pointer is not eliminated
18304 to the stack pointer.
18306 2018-03-01 Tom de Vries <tom@codesourcery.com>
18308 PR rtl-optimization/83327
18309 * lra-int.h (hard_regs_spilled_into): Declare.
18310 * lra.c (hard_regs_spilled_into): Define.
18311 (init_reg_info): Init hard_regs_spilled_into.
18312 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
18313 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
18314 (process_bb_lives): Handle hard_regs_spilled_into.
18315 (lra_create_live_ranges_1): Before doing liveness propagation, clear
18316 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
18318 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
18320 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
18321 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
18322 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
18323 * config/rs6000/aix72.h: New file.
18325 2018-02-28 Jakub Jelinek <jakub@redhat.com>
18327 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
18328 instead of warning_at with conditional singular and plural messages
18332 * stor-layout.c (update_alignment_for_field): For
18333 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
18334 && !DECL_PACKED (field), do the alignment update, just use
18335 only desired_align instead of MAX (type_align, desired_align)
18337 (place_field): Don't do known_align < desired_align handling
18338 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
18339 is non-NULL, instead do it after rli->prev_field handling and
18340 only if not within a bitfield word. For DECL_PACKED (field)
18341 use type_align of BITS_PER_UNIT.
18343 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
18345 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
18346 superfluous parentheses and trailing spaces.
18348 2018-02-28 Richard Biener <rguenther@suse.de>
18350 PR tree-optimization/84584
18351 * graphite-scop-detection.c (scop_detection::add_scop): Discard
18352 SCoPs with fake exit edge.
18354 2018-02-28 Martin Liska <mliska@suse.cz>
18357 * timevar.c (timer::print): Fix format to properly print 100%
18360 2018-02-28 Richard Biener <rguenther@suse.de>
18362 PR middle-end/84607
18363 * genmatch.c (capture_info::walk_match): Do not mark
18364 captured expressions without operands as expr_p given
18365 they act more like predicates and should be subject to
18366 "lost tail" side-effect preserving.
18368 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
18370 PR rtl-optimization/81611
18371 * auto-inc-dec.c (attempt_change): Move dead note from
18372 mem_insn if it's the next use of regno
18373 (find_address): Take address use of reg holding
18374 non-incremented value. Add parm to limit search to the named
18376 (merge_in_block): Attempt to use a mem insn that is the next
18377 use of the original regno.
18379 2018-02-27 Martin Sebor <msebor@redhat.com>
18382 * doc/invoke.texi (-Wmissing-attributes): New option.
18383 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
18385 2018-02-27 Martin Sebor <msebor@redhat.com>
18387 PR translation/84207
18388 * diagnostic-core.h (warning_n, error_n, inform_n): Change
18389 n argument to unsigned HOST_WIDE_INT.
18390 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
18391 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
18392 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
18393 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
18395 2018-02-27 Richard Biener <rguenther@suse.de>
18397 PR tree-optimization/84512
18398 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
18399 Do not use the estimate returned from record_stmt_cost for
18400 the scalar iteration cost but sum properly using add_stmt_cost.
18402 2018-02-27 Richard Biener <rguenther@suse.de>
18404 PR tree-optimization/84466
18405 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
18406 Adjust last change to less strictly validate use operands.
18408 2018-02-27 Martin Liska <mliska@suse.cz>
18410 PR gcov-profile/84548
18411 * gcov.c (process_file): Allow partial overlap and consider it
18412 also as group functions.
18413 (output_lines): Properly calculate range of lines for a group.
18415 2018-02-27 Martin Liska <mliska@suse.cz>
18417 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
18418 'ggc' suffixes. Change first column width.
18419 (timer::print): Fix formatting of the column.
18421 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
18423 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
18424 preserve inline entry blocks for the sake of debug inline
18425 entry point markers alone.
18426 (remove_unused_locals): Suggest in comments a better place to
18427 force the preservation of inline entry blocks that are
18428 otherwise unused, but do not preserve them.
18430 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18432 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
18434 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18437 * config/i386/constraints.md (Bs): Replace
18438 ix86_indirect_branch_register with
18439 TARGET_INDIRECT_BRANCH_REGISTER.
18441 * config/i386/i386.md (indirect_jump): Likewise.
18442 (tablejump): Likewise.
18443 (*sibcall_memory): Likewise.
18444 (*sibcall_value_memory): Likewise.
18445 Peepholes of indirect call and jump via memory: Likewise.
18446 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
18447 (*sibcall_value_GOT_32): Likewise.
18448 * config/i386/predicates.md (indirect_branch_operand): Likewise.
18449 (GOT_memory_operand): Likewise.
18450 (call_insn_operand): Likewise.
18451 (sibcall_insn_operand): Likewise.
18452 (GOT32_symbol_operand): Likewise.
18453 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
18455 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
18457 PR rtl-optimization/83496
18458 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
18459 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
18460 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
18461 redundant insn, if any.
18462 (relax_delay_slots): Likewise.
18463 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
18465 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
18467 PR tree-optimization/83965
18468 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
18469 that grouped statements are part of a reduction chain. Return
18470 true if the statement is not marked as a reduction itself but
18471 is part of a group.
18472 (vect_recog_dot_prod_pattern): Don't check whether the statement
18473 is part of a group here.
18474 (vect_recog_sad_pattern): Likewise.
18475 (vect_recog_widen_sum_pattern): Likewise.
18477 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
18480 * final.c (rest_of_clean_state): Also look for calls inside sequences.
18482 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18485 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
18487 (ix86_output_indirect_function_return): New prototype.
18488 (ix86_split_simple_return_pop_internal): Likewise.
18489 * config/i386/i386.c (indirect_return_via_cx): New.
18490 (indirect_return_via_cx_bnd): Likewise.
18491 (indirect_thunk_name): Handle return va CX_REG.
18492 (output_indirect_thunk_function): Create alias for
18493 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
18494 (ix86_output_indirect_jmp): Remove the bool argument.
18495 (ix86_output_indirect_function_return): New function.
18496 (ix86_split_simple_return_pop_internal): Likewise.
18497 * config/i386/i386.md (*indirect_jump): Don't pass false
18498 to ix86_output_indirect_jmp.
18499 (*tablejump_1): Likewise.
18500 (simple_return_pop_internal): Change it to define_insn_and_split.
18501 Call ix86_split_simple_return_pop_internal to split it for
18502 -mfunction-return=.
18503 (simple_return_indirect_internal): Call
18504 ix86_output_indirect_function_return instead of
18505 ix86_output_indirect_jmp.
18507 2018-02-26 Jakub Jelinek <jakub@redhat.com>
18510 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
18511 memset and value initialization afterwards.
18513 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
18515 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
18517 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
18520 * common/config/aarch64/aarch64-common.c
18521 (aarch_option_optimization_table[]): Switch
18522 off fomit-frame-pointer
18524 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
18525 Chung-Ju Wu <jasonwucj@gmail.com>
18527 * config/nds32/nds32-multiple.md (load_multiple): Disallow
18529 (store_multiple): Ditto.
18531 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
18533 * config.gcc: Add --with-cpu support for nds32 target.
18534 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
18535 * config/nds32/nds32.opt: Add -mcpu= option.
18537 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
18539 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
18540 isel=yes): Warn for these deprecated options.
18542 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
18544 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
18545 ISA_2_5_MASKS_EMBEDDED.
18547 2018-02-23 Jakub Jelinek <jakub@redhat.com>
18549 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
18550 p->max as pointers rather than using iterative_hash_expr.
18552 2018-02-23 Carl Love <cel@us.ibm.com>
18554 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
18555 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
18557 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
18558 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
18559 P8V_BUILTIN_VEC_VUNSIGNED2.
18561 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
18564 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
18565 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
18567 * lra-constraints.c (curr_insn_transform): Set up
18568 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
18570 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
18571 * lra-eliminations.c (spill_pseudos): Ditto.
18572 (process_insn_for_elimination): Ditto.
18573 * lra-lives.c (reg_early_clobber_p): Use the new macros.
18574 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
18575 LRA_NON_CLOBBERED_ALT.
18577 2018-02-22 Martin Sebor <msebor@redhat.com>
18579 PR tree-optimization/84480
18580 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
18581 to maybe_diag_stxncpy_trunc. Call it.
18582 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
18583 from gimple_fold_builtin_strcpy. Print inlining stack.
18584 (handle_builtin_stxncpy): Print inlining stack.
18585 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
18587 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
18590 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
18591 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
18592 and -fcheck-pointer-bounds are used together.
18593 (indirect_thunk_prefix): New enum.
18594 (indirect_thunk_need_prefix): New function.
18595 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
18596 "_nt" instead of "_bnd" for NOTRACK prefix.
18597 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
18598 (output_indirect_thunk_function): Likewise.
18600 (ix86_code_end): Update output_indirect_thunk_function calls.
18601 (ix86_output_indirect_branch_via_reg): Replace
18602 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
18603 (ix86_output_indirect_branch_via_push): Likewise.
18604 (ix86_output_function_return): Likewise.
18605 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
18606 incompatible with -fcf-protection=branch and
18607 -fcheck-pointer-bounds.
18609 2018-02-22 Steve Ellcey <sellcey@cavium.com>
18612 * config/aarch64/aarch64.c (aarch64_print_address_internal):
18613 Change gcc_assert call to output_operand_lossage.
18615 2018-02-22 Steve Ellcey <sellcey@cavium.com>
18617 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
18619 2018-02-22 DJ Delorie <dj@redhat.com>
18620 Sebastian Perta <sebastian.perta@renesas.com>
18621 Oleg Endo <olegendo@gcc.gnu.org>
18623 * config/rx/rx.c (rx_rtx_costs): New function.
18624 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
18626 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
18628 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
18630 2018-02-22 Martin Liska <mliska@suse.cz>
18633 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
18634 Add "native" as a possible value.
18636 2018-02-22 Martin Liska <mliska@suse.cz>
18639 * config/i386/i386.c (ix86_option_override_internal):
18640 Add "native" as a possible value for -march and -mtune.
18642 2018-02-22 Jakub Jelinek <jakub@redhat.com>
18645 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
18646 to all type variants.
18648 PR tree-optimization/84503
18649 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
18650 width as info->bitpos + info->bitsize - start.
18651 (merged_store_group::merge_overlapping): Simplify width computation.
18652 (check_no_overlap): New function.
18653 (imm_store_chain_info::try_coalesce_bswap): Compute expected
18654 start + width and last_order of the group, fail if check_no_overlap
18656 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
18657 to group if check_no_overlap fails.
18659 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
18661 * config/rs6000/altivec.md: Delete contraint arguments to
18662 define_expand, define_split, and define_peephole2, and in
18663 define_insn_and_split if always unused.
18664 * config/rs6000/darwin.md: Ditto.
18665 * config/rs6000/dfp.md: Ditto.
18666 * config/rs6000/rs6000.md: Ditto.
18667 * config/rs6000/sync.md: Ditto.
18668 * config/rs6000/vector.md: Ditto.
18669 * config/rs6000/vsx.md: Ditto.
18671 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
18673 * config/rs6000/altivec.md: Write output control strings as braced
18674 blocks instead of double-quoted strings.
18675 * config/rs6000/darwin.md: Ditto.
18676 * config/rs6000/rs6000.md: Ditto.
18677 * config/rs6000/vector.md: Ditto.
18678 * config/rs6000/vsx.md: Ditto.
18680 2018-02-21 Jason Merrill <jason@redhat.com>
18682 PR c++/84314 - ICE with templates and fastcall attribute.
18683 * attribs.c (build_type_attribute_qual_variant): Remove assert.
18685 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
18687 * ipa-cp.c (determine_versionability): Fix comment typos.
18689 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
18692 * ipa-cp.c (determine_versionability): Do not version functions caling
18695 2018-02-21 Martin Liska <mliska@suse.cz>
18698 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
18699 Add "native" as a possible value.
18700 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
18701 the macro when native cpu detection is available.
18703 2018-02-21 Martin Liska <mliska@suse.cz>
18706 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
18707 Add "native" as a possible value.
18708 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
18709 when native cpu detection is available.
18711 2018-02-21 Jakub Jelinek <jakub@redhat.com>
18712 Martin Sebor <msebor@redhat.com>
18714 PR tree-optimization/84478
18715 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
18717 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
18718 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
18719 support which is conservatively correct, for 2 only stay conservative
18720 for maxlen. Formatting and comment capitalization fixes. Add STRICT
18721 argument to the 2 argument get_range_strlen, adjust 6 arg
18722 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
18724 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
18725 (gimple_fold_builtin_strlen): Pass true as last argument to
18728 2018-02-20 Martin Sebor <msebor@redhat.com>
18730 PR middle-end/84095
18731 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
18732 (builtin_memref::set_base_and_offset): Same. Handle inner references.
18733 (builtin_memref::builtin_memref): Factor out parts into
18734 set_base_and_offset and call it.
18736 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
18738 PR middle-end/84406
18739 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
18740 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
18741 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
18742 search at the associated MODE_INT.
18744 2018-02-20 Jeff Law <law@redhat.com>
18746 PR middle-end/82123
18747 PR tree-optimization/81592
18748 PR middle-end/79257
18749 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
18750 for range data rather than using global data.
18751 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
18752 range data rather than using global data.
18753 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
18754 pass it to children as needed.
18755 (struct directive::fmtresult): Similarly.
18756 (struct directive::set_width): Similarly.
18757 (struct directive::set_precision): Similarly.
18758 (format_integer, format_directive, parse_directive): Similarly.
18759 (format_none): Accept unnamed vr_values parameter.
18760 (format_percent, format_floating, format_character): Similarly.
18761 (format_string, format_plain): Similarly.
18762 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
18763 the EVRP range analyzer for range data rather than using global data.
18764 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
18765 gimple-ssa-evrp-analyze.h
18766 (class sprintf_dom_walker): Add after_dom_children member function.
18767 Add evrp_range_analyzer member.
18768 (sprintf_dom_walker::before_dom_children): Call into the EVRP
18769 range analyzer as needed.
18770 (sprintf_dom_walker::after_dom_children): New member function.
18771 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
18773 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
18774 (evrp_range_analyzer::pop_to_marker): Likewise.
18776 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
18778 PR tree-optimization/84419
18779 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
18780 with the required type if its current type is compatible but
18783 2018-02-20 Jakub Jelinek <jakub@redhat.com>
18785 PR middle-end/82004
18786 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
18787 after vectorization.
18789 2018-02-20 Martin Liska <mliska@suse.cz>
18792 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
18793 possible values if we don't have a hint.
18795 2018-02-20 Martin Liska <mliska@suse.cz>
18799 * final.c (shorten_branches): Build align_tab array with one
18801 * opts.c (finish_options): Add alignment option limit check.
18802 (MAX_CODE_ALIGN): Likewise.
18803 (MAX_CODE_ALIGN_VALUE): Likewise.
18804 * doc/invoke.texi: Document maximum allowed option value for
18805 all -falign-* options.
18807 2018-02-19 Jakub Jelinek <jakub@redhat.com>
18810 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
18811 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
18812 * var-tracking.c (emit_note_insn_var_location): Remove all references
18813 to NOTE_INSN_CALL_ARG_LOCATION.
18814 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
18815 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
18816 Use copy_rtx_if_shared.
18817 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
18818 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
18819 (dwarf2out_var_location): Remove handling of
18820 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
18822 * final.c (final_scan_insn): Remove all references to
18823 NOTE_INSN_CALL_ARG_LOCATION.
18824 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
18825 before dumping final insns.
18826 * except.c (emit_note_eh_region_end): Remove all references to
18827 NOTE_INSN_CALL_ARG_LOCATION.
18828 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
18829 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
18830 * config/arc/arc.c (hwloop_optimize): Likewise.
18831 * config/arm/arm.c (create_fix_barrier): Likewise.
18832 * config/s390/s390.c (s390_chunkify_start): Likewise.
18833 * config/sh/sh.c (find_barrier): Likewise.
18834 * config/i386/i386.c (rest_of_insert_endbranch,
18835 ix86_seh_fixup_eh_fallthru): Likewise.
18836 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
18837 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
18838 * config/frv/frv.c (frv_function_prologue): Likewise.
18839 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
18841 (note_outside_basic_block_p): Remove all references to
18842 NOTE_INSN_CALL_ARG_LOCATION.
18843 * gengtype.c (adjust_field_rtx_def): Likewise.
18844 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
18846 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
18847 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
18850 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
18853 PR tree-optimization/84452
18854 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
18855 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
18858 2018-02-19 Martin Liska <mliska@suse.cz>
18861 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
18863 2018-02-19 Martin Liska <mliska@suse.cz>
18864 Richard Sandiford <richard.sandiford@linaro.org>
18866 PR tree-optimization/82491
18867 * gimple-fold.c (get_base_constructor): Make earlier bail out
18870 2018-02-19 Carl Love <cel@us.ibm.com>
18872 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
18873 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
18875 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
18876 P8V_BUILTIN_VEC_NEG.
18878 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
18880 * config/rl78/rl78.md (movdf): New define expand.
18882 2018-02-19 Martin Liska <mliska@suse.cz>
18885 * doc/invoke.texi: Fix typo.
18886 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
18888 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
18890 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
18891 handle rs6000_single_float and rs6000_double_float specially for
18894 2018-02-16 Jeff Law <law@redhat.com>
18896 * config/rx/rx.c (add_pop_cfi_notes): New function.;
18897 (pop_regs): Use it.
18899 2018-02-16 Jakub Jelinek <jakub@redhat.com>
18902 * ipa-inline.c (inline_small_functions): Fix a typo.
18904 2018-02-16 Nathan Sidwell <nathan@acm.org>
18906 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
18908 2018-02-16 Carl Love <cel@us.ibm.com>
18910 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
18911 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
18912 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
18913 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
18914 expansion to P8V_BUILTIN_VEC_FLOAT2.
18916 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
18918 PR rtl-optimization/70023
18919 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
18920 src_regno into account.
18922 2018-02-16 Carl Love <cel@us.ibm.com>
18924 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
18925 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
18926 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
18927 * config/rs6000/rs6000.c: Remove case statements for
18928 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
18929 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
18930 and P9V_BUILTIN_VEC_VINSERT4B.
18931 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
18932 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
18933 * config/rs6000/vsx.md:
18934 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
18937 2018-02-16 Carl Love <cel@us.ibm.com>
18939 * config/rs6000/altivec.h: Add builtin names vec_extract4b
18941 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
18943 * config/rs6000/rs6000-c.c: Add the definitions for
18944 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
18945 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
18946 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
18947 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
18948 definition for insert4b and define insn *insert3b_internal.
18949 * doc/extend.texi: Add documentation for vec_extract4b.
18951 2018-02-16 Nathan Sidwell <nathan@acm.org>
18953 * doc/extend.texi (Backwards Compatibility): Mention friend
18954 injection. Note for-scope is deprecated.
18955 * doc/invoke.texi (-ffriend-injection): Deprecate.
18957 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
18959 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
18960 that moved to I2, also allow destinations that are a paradoxical
18961 subreg (instead of a normal reg).
18963 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
18966 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
18969 2018-02-16 Richard Biener <rguenther@suse.de>
18971 PR tree-optimization/84037
18972 PR tree-optimization/84016
18974 * config/i386/i386.c (ix86_builtin_vectorization_cost):
18975 Adjust vec_construct for the fact we need additional higher latency
18976 128bit inserts for AVX256 and AVX512 vector builds.
18977 (ix86_add_stmt_cost): Scale vector construction cost for
18980 2018-02-16 Richard Biener <rguenther@suse.de>
18982 PR tree-optimization/84417
18983 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
18984 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
18985 (non_rewritable_lvalue_p): Likewise, use poly-ints.
18987 2018-02-16 Martin Liska <mliska@suse.cz>
18990 * internal-fn.def (ASAN_CHECK): Set proper flags.
18991 (ASAN_MARK): Likewise.
18993 2018-02-16 Julia Koval <julia.koval@intel.com>
18995 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
18996 from PTA_CANNONLAKE.
18998 2018-02-16 Jakub Jelinek <jakub@redhat.com>
19001 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
19002 Use ++iter rather than iter++ for std::list iterators.
19003 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
19004 defer deleting them until all nodes in the forest are processed. Do
19005 free even leaf nodes. Change to_process into auto_vec.
19008 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
19009 * vec.h (vec_default_construct): Use memset instead of placement new
19010 if BROKEN_VALUE_INITIALIZATION is defined.
19011 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
19012 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
19015 PR rtl-optimization/83723
19016 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
19017 * lra.c (lra_substitute_pseudo): Likewise. If true, use
19018 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
19020 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
19022 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
19024 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
19026 PR rtl-optimization/81443
19027 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
19028 from inner REGs to paradoxical SUBREGs.
19030 2018-02-16 Richard Biener <rguenther@suse.de>
19032 PR tree-optimization/84399
19033 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
19034 For operands we can analyze at their definition make sure we can
19035 analyze them at each use as well.
19037 2018-02-16 Richard Biener <rguenther@suse.de>
19039 PR tree-optimization/84190
19040 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
19041 volatile accesses if the decl isn't volatile.
19043 2018-02-15 Jason Merrill <jason@redhat.com>
19045 PR c++/84314 - ICE with templates and fastcall attribute.
19046 * attribs.c (build_type_attribute_qual_variant): Don't clobber
19047 TYPE_CANONICAL on an existing type.
19049 2018-02-15 Jakub Jelinek <jakub@redhat.com>
19051 PR tree-optimization/84383
19052 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
19053 dstoff nor call operand_equal_p if dstbase is NULL.
19055 PR tree-optimization/84334
19056 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
19057 also a CONSTANT_CLASS_P, punt.
19059 2018-02-14 Jim Wilson <jimw@sifive.com>
19061 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
19062 first SMALL_OPERAND check. New local min_second_step. Move assert
19063 to where locals are set. Add TARGET_RVC support.
19064 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
19066 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
19068 * doc/invoke.texi: Correct -Wformat-overflow code sample.
19070 2018-02-14 Martin Sebor <msebor@redhat.com>
19072 PR tree-optimization/83698
19073 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
19074 arrays constrain the offset range to their bounds.
19075 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
19076 (builtin_access::overlap): Avoid setting the size of overlap if it's
19078 (maybe_diag_overlap): Also consider arrays when deciding what values
19079 of offsets to include in diagnostics.
19081 2018-02-14 Martin Sebor <msebor@redhat.com>
19084 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
19085 that correspond to the kind of a declaration.
19087 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
19090 * config/pa/pa.md: Load address of PIC label using the linkage table
19091 if the label is nonlocal.
19093 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
19095 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
19096 warning message if user requests -maltivec=be.
19097 * doc/invoke.texi: Document deprecation of -maltivec=be.
19099 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
19102 * config/rs6000/rs6000-c.c: Update definitions for
19103 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
19104 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
19106 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
19109 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
19110 add _get_ssp intrinsics. Remove argument from
19111 __builtin_ia32_rdssp[d|q].
19112 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
19113 * config/i386/i386-builtin.def: Remove argument from
19114 __builtin_ia32_rdssp[d|q].
19115 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
19116 ix86_expand_special_args_builtin for _rdssp[d|q].
19117 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
19118 Clear register before usage.
19119 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
19120 Add documentation for new _get_ssp and _inc_ssp intrinsics.
19122 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
19124 PR tree-optimization/84357
19125 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
19126 operand 1 of an ARRAY_REF too.
19128 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
19131 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
19132 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
19134 (set_of_reg): New struct.
19135 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
19136 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
19137 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
19139 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
19140 Split into bitclr, bitset, bitinvert patterns if appropriate.
19141 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
19142 use rx_fuse_in_memory_bitop.
19143 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
19144 to named insn, correct maximum insn length.
19146 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
19149 * machmode.def: Define a complex mode for PARTIAL_INT.
19150 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
19152 * doc/rtl.texi: Document CSPImode.
19153 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
19155 (msp430_hard_regno_nregs_with_padding): Likewise.
19157 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
19160 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
19162 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
19164 PR rtl-optimization/84169
19165 * combine.c (try_combine): New variable split_i2i3. Set it to true if
19166 we generated a parallel as new i3 and we split that to new i2 and i3
19167 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
19168 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
19169 those to i2, not i1. Partially rewrite this scan code.
19171 2018-02-13 Jakub Jelinek <jakub@redhat.com>
19174 * stor-layout.c (place_field): For variable length fields, adjust
19175 offset_align afterwards not just based on the field's alignment,
19176 but also on the size.
19178 PR middle-end/84309
19179 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
19180 of exps and logs in the use_exp2 case.
19182 2018-02-13 Jeff Law <law@redhat.com>
19184 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
19185 entry for "vector".
19187 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
19190 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
19194 * common.opt (gas-loc-support, gas-locview-support): New.
19195 (ginline-points, ginternal-reset-location-views): New.
19196 * doc/invoke.texi: Document them. Use @itemx where intended.
19197 (gvariable-location-views): Adjust.
19198 * target.def (reset_location_view): New.
19199 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
19200 (TARGET_RESET_LOCATION_VIEW): New.
19201 * doc/tm.texi: Rebuilt.
19202 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
19203 (dwarf2out_default_as_locview_support): New.
19204 (output_asm_line_debug_info): Use option variables.
19205 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
19206 (output_loc_list): Likewise.
19207 (add_high_low_attributes): Check option variables.
19208 Don't output entry view attribute in strict mode.
19209 (gen_inlined_subroutine_die): Check option variables.
19210 (dwarf2out_inline_entry): Likewise.
19211 (init_sections_and_labels): Likewise.
19212 (dwarf2out_early_finish): Likewise.
19213 (maybe_reset_location_view): New, from...
19214 (dwarf2out_var_location): ... here. Call it.
19215 * debug.h (dwarf2out_default_as_loc_support): Declare.
19216 (dwarf2out_default_as_locview_support): Declare.
19217 * hooks.c (hook_int_rtx_insn_0): New.
19218 * hooks.h (hook_int_rtx_insn_0): Declare.
19219 * toplev.c (process_options): Take -gas-loc-support and
19220 -gas-locview-support from dwarf2out. Enable
19221 -gvariable-location-views by default only with locview
19222 assembler support. Enable -ginternal-reset-location-views by
19223 default only if the target defines the corresponding hook.
19224 Enable -ginline-points by default if location views are
19225 enabled; force it disabled if statement frontiers are
19227 * tree-inline.c (expand_call_inline): Check option variables.
19228 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
19230 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
19232 PR tree-optimization/84321
19233 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
19234 handling. Also check whether the anti-range contains any values
19235 that satisfy the mask; switch to a VR_RANGE if not.
19237 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
19240 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
19242 2018-02-13 Martin Jambor <mjambor@suse.cz>
19245 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
19246 of call statements, also set location of a load to a temporary.
19248 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
19250 * config/rl78/rl78.c (add_vector_labels): New function.
19251 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
19252 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
19253 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
19254 which checks that no arguments are passed.
19255 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
19256 * doc/extend.texi: Documentation for the new attribute.
19258 2018-02-13 Andreas Schwab <schwab@suse.de>
19260 * config/riscv/linux.h (CPP_SPEC): Define.
19262 2018-02-13 Jakub Jelinek <jakub@redhat.com>
19265 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
19266 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
19267 OPTION_MASK_ISA_AES as first argument to def_builtin_const
19268 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
19269 instead of OPTION_MASK_ISA_PCLMUL as first argument to
19270 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
19271 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
19272 temporarily for AES and PCLMUL builtins.
19274 PR tree-optimization/84339
19275 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
19276 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
19279 PR middle-end/84309
19280 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
19281 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
19282 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
19284 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
19286 * omp-simd-clone.h: New file.
19287 * omp-simd-clone.c: Include omp-simd-clone.h.
19288 (expand_simd_clones): No longer static.
19289 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
19290 cgraph.h and omp-simd-clone.h.
19291 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
19292 (vect_recog_widen_shift_pattern): Formatting fix.
19293 (vect_pattern_recog_1): Don't check optab for calls.
19296 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
19297 operands[2] into a REG before using gen_lowpart on it.
19299 2018-02-12 Jeff Law <law@redhat.com>
19302 * config/sh/sh.c (find_barrier): Consider a sibling call
19305 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
19306 successfully back substituting a reg.
19308 2018-02-12 Richard Biener <rguenther@suse.de>
19310 PR tree-optimization/84037
19311 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
19312 parameter, move visited init to caller.
19313 (vect_slp_analyze_operations): Separate cost from validity
19314 check, initialize visited once for all instances.
19315 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
19317 * tree-vect-stmts.c (vect_model_simple_cost): Make early
19319 (vect_model_promotion_demotion_cost): Likewise.
19320 (vectorizable_bswap): Guard cost modeling with !slp_node
19321 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
19323 (vectorizable_call): Likewise.
19324 (vectorizable_conversion): Likewise.
19325 (vectorizable_assignment): Likewise.
19326 (vectorizable_shift): Likewise.
19327 (vectorizable_operation): Likewise.
19328 (vectorizable_store): Likewise.
19329 (vectorizable_load): Likewise.
19330 (vectorizable_condition): Likewise.
19331 (vectorizable_comparison): Likewise.
19333 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
19336 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
19337 (ASAN_MARK): Fix fnspec to account for return value, change pointer
19338 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
19340 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
19342 PR middle-end/83665
19343 * params.def (inline-min-speedup): Increase from 8 to 15.
19344 (max-inline-insns-auto): Decrease from 40 to 30.
19345 * ipa-split.c (consider_split): Add some buffer for function to
19346 be considered inlining candidate.
19347 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
19350 2018-02-12 Richard Biener <rguenther@suse.de>
19352 PR tree-optimization/84037
19353 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
19354 matched stmts if we cannot swap the non-matched ones.
19356 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
19358 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
19359 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
19360 _mm_maskz_scalef_round_ss): New intrinsics.
19361 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
19362 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
19363 __builtin_ia32_scalefss_round): Remove.
19364 (__builtin_ia32_scalefsd_mask_round,
19365 __builtin_ia32_scalefss_mask_round): New intrinsics.
19366 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
19367 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
19368 ((match_operand:VF_128 2 "<round_nimm_predicate>"
19369 "<round_constraint>")): Changed to ...
19370 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
19371 "<round_scalar_constraint>")): ... this.
19372 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
19373 %0, %1, %2<round_op3>}"): Changed to ...
19374 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
19375 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
19376 %2<round_scalar_mask_op3>}"): ... this.
19377 * config/i386/subst.md (round_scalar_nimm_predicate): New.
19379 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
19381 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
19382 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
19383 (_mm_maskz_sqrt_round_ss): New intrinsics.
19384 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
19385 (__builtin_ia32_sqrtsd_mask_round)
19386 (__builtin_ia32_sqrtss_mask_round): New builtins.
19387 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
19388 (__builtin_ia32_sqrtss_round): Remove.
19389 (__builtin_ia32_sqrtsd_mask_round)
19390 (__builtin_ia32_sqrtss_mask_round): New builtins.
19391 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
19392 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
19393 ((match_operand:VF_128 1 "vector_operand"
19394 "xBm,<round_constraint>")): Changed to ...
19395 ((match_operand:VF_128 1 "vector_operand"
19396 "xBm,<round_scalar_constraint>")): ... this.
19397 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
19398 %0, %2, %<iptr>1<round_op3>}): Changed to ...
19399 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
19400 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
19401 %<iptr>1<round_scalar_mask_op3>}): ... this.
19402 ((set_attr "prefix" "<round_prefix>")): Changed to ...
19403 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
19405 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
19408 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
19409 Cast vec_cmpeq result to correct type.
19410 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
19411 Cast vec_cmpgt result to correct type.
19413 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
19415 * final.c (final_scan_insn_1): Renamed from...
19416 (final_scan_insn): ... this. New wrapper, to recover
19417 seen from the outermost call in recursive ones.
19418 * config/sparc/sparc.c (output_return): Drop seen from call.
19419 (output_sibcall): Likewise.
19420 * config/visium/visium.c (output_branch): Likewise.
19422 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
19424 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
19427 2018-02-10 Alan Modra <amodra@gmail.com>
19430 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
19431 Specify LR as an input.
19433 2018-02-10 Jakub Jelinek <jakub@redhat.com>
19436 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
19437 remove_member_access_dummy_vars): New functions.
19438 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
19439 lower_omp_1, execute_lower_omp): Use them.
19441 PR rtl-optimization/84308
19442 * shrink-wrap.c (spread_components): Release todo vector.
19444 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
19446 PR rtl-optimization/57193
19447 * ira-color.c (struct allocno_color_data): Add member
19448 conflict_allocno_hard_prefs.
19449 (update_conflict_allocno_hard_prefs): New.
19450 (bucket_allocno_compare_func): Add a preference based on
19451 conflict_allocno_hard_prefs.
19452 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
19453 (color_allocnos): Remove a dead code. Initiate
19454 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
19456 2018-02-09 Jakub Jelinek <jakub@redhat.com>
19459 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
19460 constraint from =wa to wa. Avoid a subreg on the output operand,
19461 instead use a pseudo and subreg it in a move.
19462 (p9_xxbrd_<mode>): Changed to ...
19463 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
19464 (p9_xxbrd_v2df): New expander.
19465 (p9_xxbrw_<mode>): Changed to ...
19466 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
19467 (p9_xxbrw_v4sf): New expander.
19469 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
19471 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
19473 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
19476 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
19477 multiply in 32-bit mode.
19478 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
19479 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
19482 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
19484 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
19485 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
19486 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
19487 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
19489 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
19492 * dwarf2out.c (is_trivial_indirect_ref): New function.
19493 (dwarf2out_late_global_decl): Do not generate a location
19494 attribute for variables that have a non-trivial DECL_VALUE_EXPR
19495 and that are not defined in the current unit.
19497 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
19499 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
19500 instead of a libcall for UNORDERED.
19502 2018-02-09 Tamar Christina <tamar.christina@arm.com>
19505 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
19506 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
19508 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19511 * config/s390/s390.c (s390_set_current_function): Invoke
19512 s390_indirect_branch_settings also if fndecl didn't change.
19514 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
19516 * config/rs6000/rs6000.md (blockage): Set length to zero.
19518 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
19520 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
19522 2018-02-09 Jakub Jelinek <jakub@redhat.com>
19525 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
19526 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
19530 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
19531 PARALLEL incoming that failed vt_get_decl_and_offset check.
19533 PR middle-end/84237
19534 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
19535 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
19537 (get_variable_section): For decls in named .bss* sections pass true as
19538 second argument to bss_initializer_p.
19540 2018-02-09 Marek Polacek <polacek@redhat.com>
19541 Jakub Jelinek <jakub@redhat.com>
19544 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
19545 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
19546 Sync some changes from cxx_fold_indirect_ref.
19548 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
19550 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
19552 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
19553 (BLOCK_INLINE_ENTRY_LABEL): New.
19554 (dwarf2out_var_location): Disregard inline entry markers.
19555 (inline_entry_data): New struct.
19556 (inline_entry_data_hasher): New hashtable type.
19557 (inline_entry_data_hasher::hash): New.
19558 (inline_entry_data_hasher::equal): New.
19559 (inline_entry_data_table): New variable.
19560 (add_high_low_attributes): Add DW_AT_entry_pc and
19561 DW_AT_GNU_entry_view attributes if a pending entry is found
19562 in inline_entry_data_table. Add old entry_pc attribute only
19563 if debug nonbinding markers are disabled.
19564 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
19565 markers are enabled.
19566 (block_within_block_p, dwarf2out_inline_entry): New.
19567 (dwarf2out_finish): Check that no entries remained in
19568 inline_entry_data_table.
19569 * final.c (reemit_insn_block_notes): Handle inline entry notes.
19570 (final_scan_insn, notice_source_line): Likewise.
19571 (rest_of_clean_state): Skip inline entry markers.
19572 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
19574 * gimple.c (gimple_build_debug_inline_entry): New.
19575 * gimple.h (enum gimple_debug_subcode): Add
19576 GIMPLE_DEBUG_INLINE_ENTRY.
19577 (gimple_build_debug_inline_entry): Declare.
19578 (gimple_debug_inline_entry_p): New.
19579 (gimple_debug_nonbind_marker_p): Adjust.
19580 * insn-notes.def (INLINE_ENTRY): New.
19581 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
19582 inline entry marker notes.
19583 (print_insn): Likewise.
19584 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
19585 (INSN_DEBUG_MARKER_KIND): Likewise.
19586 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
19587 * tree-inline.c (expand_call_inline): Build and insert
19588 debug_inline_entry stmt.
19589 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
19590 inline entry blocks early, if nonbind markers are enabled.
19591 (dump_scope_block): Dump fragment info.
19592 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
19593 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
19594 (gimple_build_debug_inline_entry): New.
19595 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
19596 Enable/disable inline entry points too.
19597 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
19598 (DEBUG_INSN): Describe inline entry markers.
19600 * common.opt (gvariable-location-views): New.
19601 (gvariable-location-views=incompat5): New.
19602 * config.in: Rebuilt.
19603 * configure: Rebuilt.
19604 * configure.ac: Test assembler for view support.
19605 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
19606 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
19607 * dwarf2out.c (var_loc_view): New typedef.
19608 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
19609 (dwarf2out_locviews_in_attribute): New.
19610 (dwarf2out_locviews_in_loclist): New.
19611 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
19612 (enum dw_line_info_opcode): Add LI_adv_address.
19613 (struct dw_line_info_table): Add view.
19614 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
19615 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
19616 (zero_view_p): New variable.
19617 (ZERO_VIEW_P): New macro.
19618 (output_asm_line_debug_info): New.
19619 (struct var_loc_node): Add view.
19620 (add_AT_view_list, AT_loc_list): New.
19621 (add_var_loc_to_decl): Add view param. Test it against last.
19622 (new_loc_list): Add view params. Record them.
19623 (AT_loc_list_ptr): Handle loc and view lists.
19624 (view_list_to_loc_list_val_node): New.
19625 (print_dw_val): Handle dw_val_class_view_list.
19626 (size_of_die): Likewise.
19627 (value_format): Likewise.
19628 (loc_list_has_views): New.
19629 (gen_llsym): Set vl_symbol too.
19630 (maybe_gen_llsym, skip_loc_list_entry): New.
19631 (dwarf2out_maybe_output_loclist_view_pair): New.
19632 (output_loc_list): Output view list or entries too.
19633 (output_view_list_offset): New.
19634 (output_die): Handle dw_val_class_view_list.
19635 (output_dwarf_version): New.
19636 (output_compilation_unit_header): Use it.
19637 (output_skeleton_debug_sections): Likewise.
19638 (output_rnglists, output_line_info): Likewise.
19639 (output_pubnames, output_aranges): Update version comments.
19640 (output_one_line_info_table): Output view numbers in asm comments.
19641 (dw_loc_list): Determine current endview, pass it to new_loc_list.
19642 Call maybe_gen_llsym.
19643 (loc_list_from_tree_1): Adjust.
19644 (add_AT_location_description): Create view list attribute if
19645 needed, check it's absent otherwise.
19646 (convert_cfa_to_fb_loc_list): Adjust.
19647 (maybe_emit_file): Call output_asm_line_debug_info for test.
19648 (dwarf2out_var_location): Reset views as needed. Precompute
19649 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
19650 attribute. Set view.
19651 (new_line_info_table): Reset next view.
19652 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
19653 (dwarf2out_source_line): Likewise. Output view resets and labels to
19654 the assembler, or select appropriate line info opcodes.
19655 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
19656 (optimize_string_length): Catch it. Adjust.
19657 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
19658 dw_val_class_view_list, and remove it if no longer needed.
19659 (hash_loc_list): Hash view numbers.
19660 (loc_list_hasher::equal): Compare them.
19661 (optimize_location_lists): Check whether a view list symbol is
19662 needed, and whether the locview attribute is present, and
19663 whether they match. Remove the locview attribute if no longer
19665 (index_location_lists): Call skip_loc_list_entry for test.
19666 (dwarf2out_finish): Call output_asm_line_debug_info for test.
19667 Use output_dwarf_version.
19668 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
19669 (struct dw_val_node): Add val_view_list.
19670 * final.c (SEEN_NEXT_VIEW): New.
19671 (set_next_view_needed): New.
19672 (clear_next_view_needed): New.
19673 (maybe_output_next_view): New.
19674 (final_start_function): Rename to...
19675 (final_start_function_1): ... this. Take pointer to FIRST,
19676 add SEEN parameter. Emit param bindings in the initial view.
19677 (final_start_function): Reintroduce SEEN-less interface.
19678 (final): Rename to...
19679 (final_1): ... this. Take SEEN parameter. Output final pending
19680 next view at the end.
19681 (final): Reintroduce seen-less interface.
19682 (final_scan_insn): Output pending next view before switching
19683 sections or ending a block. Mark the next view as needed when
19684 outputting variable locations. Notify debug backend of section
19685 changes, and of location view changes.
19686 (rest_of_handle_final): Adjust.
19687 * toplev.c (process_options): Autodetect value for debug variable
19688 location views option. Warn on incompat5 without -gdwarf-5.
19689 * doc/invoke.texi (gvariable-location-views): New.
19690 (gvariable-location-views=incompat5): New.
19691 (gno-variable-location-views): New.
19693 2018-02-08 David Malcolm <dmalcolm@redhat.com>
19695 PR tree-optimization/84136
19696 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
19697 that the result of find_edge is non-NULL.
19699 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
19702 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
19703 storing integer register in SImode. Fix cost of 256 and 512
19704 byte aligned SSE register store.
19706 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
19708 * config/i386/i386.c (ix86_multiplication_cost): Fix
19709 multiplication cost for TARGET_AVX512DQ.
19711 2018-02-08 Marek Polacek <polacek@redhat.com>
19713 PR tree-optimization/84238
19714 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
19717 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19719 PR tree-optimization/84265
19720 * tree-vect-stmts.c (vectorizable_store): Don't treat
19721 VMAT_CONTIGUOUS accesses as grouped.
19722 (vectorizable_load): Likewise.
19724 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19726 PR tree-optimization/81635
19727 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
19728 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
19729 (test_round_for_mask): New functions.
19730 (wide_int_cc_tests): Call test_round_for_mask.
19731 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
19732 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
19733 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
19734 range returned by get_range_info.
19736 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
19739 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
19740 * symtab.c: Include builtins.h
19741 (symtab_node::output_to_lto_symbol_table_p): Move here
19742 from lto-streamer-out.c:output_symbol_p.
19743 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
19744 (output_symbol_p): Move all logic to symtab.c
19745 (produce_symtab): Update.
19747 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19749 * config/s390/s390-opts.h (enum indirect_branch): Define.
19750 * config/s390/s390-protos.h (s390_return_addr_from_memory)
19751 (s390_indirect_branch_via_thunk)
19752 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
19753 (enum s390_indirect_branch_type): Define.
19754 * config/s390/s390.c (struct s390_frame_layout, struct
19755 machine_function): Remove.
19756 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
19757 (indirect_branch_table_label_no, indirect_branch_table_name):
19759 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
19760 (enum s390_indirect_branch_option): Define.
19761 (s390_return_addr_from_memory): New function.
19762 (s390_handle_string_attribute): New function.
19763 (s390_attribute_table): Add new attribute handler.
19764 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
19765 (s390_indirect_branch_via_thunk): New function.
19766 (s390_indirect_branch_via_inline_thunk): New function.
19767 (s390_function_ok_for_sibcall): When jumping via thunk disallow
19768 sibling call optimization for non z10 compiles.
19769 (s390_emit_call): Force indirect branch target to be a single
19770 register. Add r1 clobber for non-z10 compiles.
19771 (s390_emit_epilogue): Emit return jump via return_use expander.
19772 (s390_reorg): Handle JUMP_INSNs as execute targets.
19773 (s390_option_override_internal): Perform validity checks for the
19774 new command line options.
19775 (s390_indirect_branch_attrvalue): New function.
19776 (s390_indirect_branch_settings): New function.
19777 (s390_set_current_function): Invoke s390_indirect_branch_settings.
19778 (s390_output_indirect_thunk_function): New function.
19779 (s390_code_end): Implement target hook.
19780 (s390_case_values_threshold): Implement target hook.
19781 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
19783 * config/s390/s390.h (struct s390_frame_layout)
19784 (struct machine_function): Move here from s390.c.
19785 (TARGET_INDIRECT_BRANCH_NOBP_RET)
19786 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
19787 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
19788 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
19789 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
19790 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
19791 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
19792 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
19793 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
19794 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
19795 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
19796 (mnemonic attribute): Add values which aren't recognized
19798 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
19799 pattern for branch conversion. Fix mnemonic attribute.
19800 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
19801 indirect branch via thunk if requested.
19802 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
19803 ("*indirect_jump"): Disable for branch conversion using out of
19805 ("indirect_jump_via_thunk<mode>_z10")
19806 ("indirect_jump_via_thunk<mode>")
19807 ("indirect_jump_via_inlinethunk<mode>_z10")
19808 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
19809 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
19810 ("casesi_jump_via_inlinethunk<mode>_z10")
19811 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
19812 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
19813 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
19814 ("*indirect2_jump"): Disable for branch conversion.
19815 ("casesi_jump"): Turn into expander and expand patterns for branch
19817 ("return_use"): New expander.
19818 ("*return"): Emit return via thunk and rename it to ...
19819 ("*return<mode>"): ... this one.
19820 * config/s390/s390.opt: Add new options and and enum for the
19823 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19825 * lra-constraints.c (match_reload): Unconditionally use
19826 gen_lowpart_SUBREG, rather than selecting between that
19827 and equivalent gen_rtx_SUBREG code.
19829 2018-02-08 Richard Biener <rguenther@suse.de>
19831 PR tree-optimization/84233
19832 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
19833 changed flag instead of boguously re-using phi_inserted.
19835 2018-02-08 Martin Jambor <mjambor@suse.cz>
19837 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
19838 static local variables.
19840 2018-02-08 Richard Biener <rguenther@suse.de>
19842 PR tree-optimization/84278
19843 * tree-vect-stmts.c (vectorizable_store): When looking for
19844 smaller vector types to perform grouped strided loads/stores
19845 make sure the mode is supported by the target.
19846 (vectorizable_load): Likewise.
19848 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
19850 * config/aarch64/aarch64.c (aarch64_components_for_bb):
19851 Increase LDP/STP opportunities by adding adjacent callee-saves.
19853 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
19855 PR rtl-optimization/84068
19856 PR rtl-optimization/83459
19857 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
19859 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
19861 PR tree-optimization/84224
19862 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
19863 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
19864 non-zero arguments.
19866 2018-02-07 Iain Sandoe <iain@codesourcery.com>
19869 * config/rs6000/altivec.md (*restore_world): Remove LR use.
19870 * config/rs6000/predicates.md (restore_world_operation): Adjust op
19871 count, remove one USE.
19873 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
19875 * doc/install.texi (Configuration): Document the
19876 --with-long-double-format={ibm,ieee} PowerPC configuration
19880 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
19881 Convert from define_expand to be define_insn_and_split. Rework
19882 float/double/_Float128 conversions to QI/HI/SImode to work with
19883 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
19884 conversions to QI/HImode types did a store and then a load to
19885 truncate the value. For conversions to VSX registers, don't split
19886 the insn, instead emit the code directly. Use the code iterator
19887 any_fix to combine signed and unsigned conversions.
19888 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
19889 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
19890 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19891 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
19892 (fix_<mode>di2_hw): Likewise.
19893 (fixuns_<mode>di2_hw): Likewise.
19894 (fix_<mode>si2_hw): Likewise.
19895 (fixuns_<mode>si2_hw): Likewise.
19896 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
19897 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19898 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
19899 fix<uns>_trunc<SFDF:mode>si2_p8.
19900 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
19901 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
19902 (fix<uns>_<mode>_mem): Likewise.
19903 (fctiw<u>z_<mode>_mem): Likewise.
19904 (fix<uns>_<mode>_mem): Likewise.
19905 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
19906 the register allocator from doing a direct move to the GPRs to do
19907 a store, and instead use the ISA 3.0 store byte/half-word from
19908 vector register instruction. For IEEE 128-bit floating point,
19909 also optimize stores of 32-bit ints.
19910 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
19912 2018-02-07 Alan Hayward <alan.hayward@arm.com>
19914 * genextract.c (push_pathstr_operand): New function to support
19916 (walk_rtx): Call push_pathstr_operand.
19917 (print_path): Support [a-zA-Z].
19919 2018-02-07 Richard Biener <rguenther@suse.de>
19921 PR tree-optimization/84037
19922 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
19923 (cse_and_gimplify_to_preheader): Declare.
19924 (vect_get_place_in_interleaving_chain): Likewise.
19925 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19927 (_loop_vec_info::~_loop_vec_info): Delete it.
19928 (cse_and_gimplify_to_preheader): New function.
19929 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
19930 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
19931 (vectorizable_load): Likewise. For grouped stores always base
19932 the IV on the first element.
19933 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
19934 condition before gimplifying.
19936 2018-02-07 Jakub Jelinek <jakub@redhat.com>
19938 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
19939 *DIV_EXPR and *MOD_EXPR.
19941 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
19944 * config/i386/i386.c (ix86_option_override_internal): Mask out
19945 the CF_SET bit when checking -fcf-protection.
19947 2018-02-07 Tom de Vries <tom@codesourcery.com>
19950 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
19953 2018-02-07 Richard Biener <rguenther@suse.de>
19955 PR tree-optimization/84204
19956 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
19959 PR tree-optimization/84205
19960 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
19961 special-case isl_ast_op_zdiv_r.
19963 PR tree-optimization/84223
19964 * graphite-scop-detection.c (gather_bbs::before_dom_children):
19965 Only add conditions from within the region.
19966 (gather_bbs::after_dom_children): Adjust.
19968 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
19971 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
19972 * config/avr/avr.md: Only post-reload split REG-REG moves if
19973 either register is GENERAL_REG_P.
19975 2018-02-07 Jakub Jelinek <jakub@redhat.com>
19977 PR tree-optimization/84235
19978 * tree-ssa-scopedtables.c
19979 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
19980 if the subtraction is performed in floating point type where NaNs are
19981 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
19982 build 1. Formatting fix.
19984 2018-02-06 Jakub Jelinek <jakub@redhat.com>
19987 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
19988 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
19989 and skip it regardless of bb boundaries. Use CALL_P macro,
19990 don't test INSN_P (insn) together with CALL_P or JUMP_P check
19991 unnecessarily, formatting fix.
19993 2018-02-06 Michael Collison <michael.collison@arm.com>
19995 * config/arm/thumb2.md:
19996 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
19997 (*thumb_mov_notscc): Ditto.
19999 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
20002 * config/rs6000/rs6000.md (su code attribute): Use "u" for
20003 unsigned_fix, not "s".
20005 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20007 * configure.ac (gcc_fn_eh_frame_ro): New function.
20008 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
20009 correct .eh_frame permissions.
20010 * configure: Regenerate.
20012 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
20014 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
20015 irrelevant options.
20017 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20019 * config/rs6000/rs6000.c (rs6000_option_override_internal):
20020 Display warning message for -mno-speculate-indirect-jumps.
20022 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
20024 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
20026 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
20028 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
20030 PR tree-optimization/84225
20031 * tree-eh.c (find_trapping_overflow): Only call
20032 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
20034 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
20037 * config/i386/i386.c: Reimplement the check of possible options
20038 -mibt/-mshstk conbination. Change error messages.
20039 * doc/invoke.texi: Fix a typo: remove extra '='.
20041 2018-02-06 Marek Polacek <polacek@redhat.com>
20043 PR tree-optimization/84228
20044 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
20046 2018-02-06 Tamar Christina <tamar.christina@arm.com>
20049 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
20050 emitted arch directives.
20051 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
20052 __ARM_FEATURE_COPROC before changing architectures.
20054 2018-02-06 Richard Biener <rguenther@suse.de>
20056 * config/i386/i386.c (print_reg): Fix typo.
20057 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
20059 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
20061 * configure: Regenerate.
20063 2018-02-05 Martin Sebor <msebor@redhat.com>
20065 PR tree-optimization/83369
20066 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
20069 2018-02-05 Martin Liska <mliska@suse.cz>
20071 * doc/invoke.texi: Cherry-pick upstream r323995.
20073 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
20075 * ira.c (ira_init_register_move_cost): Adjust comment.
20077 2018-02-05 Martin Liska <mliska@suse.cz>
20079 PR gcov-profile/84137
20080 * doc/gcov.texi: Fix typo in documentation.
20082 2018-02-05 Martin Liska <mliska@suse.cz>
20084 PR gcov-profile/83879
20085 * doc/gcov.texi: Document necessity of --dynamic-list-data when
20086 using dlopen functionality.
20088 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
20090 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
20091 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
20092 _mm_maskz_range_ss, _mm_mask_range_round_ss,
20093 _mm_maskz_range_round_ss): New intrinsics.
20094 (__builtin_ia32_rangesd128_round)
20095 (__builtin_ia32_rangess128_round): Remove.
20096 (__builtin_ia32_rangesd128_mask_round,
20097 __builtin_ia32_rangess128_mask_round): New builtins.
20098 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
20099 __builtin_ia32_rangess128_round): Remove.
20100 (__builtin_ia32_rangesd128_mask_round,
20101 __builtin_ia32_rangess128_mask_round): New builtins.
20102 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
20103 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
20104 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
20105 "<round_saeonly_constraint>")): Changed to ...
20106 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
20107 "<round_saeonly_scalar_constraint>")): ... this.
20108 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
20109 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
20110 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
20111 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
20112 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
20114 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
20116 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
20118 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
20119 Remove all values except native, 8540 and 8548.
20121 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
20123 * config/i386/i386.c (ix86_output_function_return): Pass
20124 INVALID_REGNUM, instead of -1, as invalid register number to
20125 indirect_thunk_name and output_indirect_thunk.
20127 2018-02-02 Julia Koval <julia.koval@intel.com>
20129 * config.gcc: Add -march=icelake.
20130 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
20131 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
20132 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
20133 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
20134 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
20135 (processor_target_table): Add icelake.
20136 (ix86_option_override_internal): Handle new PTAs.
20137 (get_builtin_code_for_version): Handle icelake.
20138 (M_INTEL_COREI7_ICELAKE): New.
20139 (fold_builtin_cpu): Handle icelake.
20140 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
20141 * doc/invoke.texi: Add -march=icelake.
20143 2018-02-02 Julia Koval <julia.koval@intel.com>
20145 * config/i386/i386.c (ix86_option_override_internal): Change flags type
20146 to wide_int_bitmask.
20147 * wide-int-bitmask.h: New.
20149 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
20152 * config/i386/i386.md: Replace Pmode with word_mode in
20153 builtin_setjmp_setup and builtin_longjmp to support x32.
20155 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
20159 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
20161 (rs6000_supported_cpu_names): New static variable.
20162 (linux_cpu_translation_table): Likewise.
20163 (elf_platform) <cpu>: Define new static variable and use it.
20164 Translate kernel AT_PLATFORM name to canonical name if needed.
20165 Error if platform name is unknown.
20167 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
20170 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
20172 2018-02-01 Jeff Law <law@redhat.com>
20175 * config/i386/i386.c (release_scratch_register_on_entry): Add new
20176 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
20177 the scratch if RELEASE_VIA_POP is false.
20178 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
20179 If we have to save a temporary register, decrement SIZE appropriately.
20180 Pass new arguments to release_scratch_register_on_entry.
20181 (ix86_adjust_stack_and_probe): Likewise.
20182 (ix86_emit_probe_stack_range): Pass new arguments to
20183 release_scratch_register_on_entry.
20185 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
20187 PR rtl-optimization/84157
20188 * combine.c (change_zero_ext): Use REG_P predicate in
20189 front of HARD_REGISTER_P predicate.
20191 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
20193 * config/avr/avr.c (avr_option_override): Move disabling of
20194 -fdelete-null-pointer-checks to...
20195 * common/config/avr/avr-common.c (avr_option_optimization_table):
20198 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20200 PR tree-optimization/81635
20201 * tree-data-ref.c (split_constant_offset_1): For types that
20202 wrap on overflow, try to use range info to prove that wrapping
20205 2018-02-01 Renlin Li <renlin.li@arm.com>
20208 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
20209 TAILCALL_ADDR_REGS.
20210 (aarch64_register_move_cost): Likewise.
20211 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
20212 TAILCALL_ADDR_REGS.
20213 (REG_CLASS_NAMES): Likewise.
20214 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
20215 TAILCALL_ADDR_REGS. Remove IP registers.
20216 * config/aarch64/aarch64.md (Ucs): Update register constraint.
20218 2018-02-01 Richard Biener <rguenther@suse.de>
20220 * domwalk.h (dom_walker::dom_walker): Add additional constructor
20221 for specifying RPO order and allow NULL for that.
20222 * domwalk.c (dom_walker::dom_walker): Likewise.
20223 (dom_walker::walk): Handle NULL RPO order.
20224 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
20226 (rewrite_update_dom_walker): Likewise.
20227 (mark_def_dom_walker): Likewise.
20229 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20231 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
20232 (aarch64_maybe_expand_sve_subreg_move): Declare.
20233 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
20234 * config/aarch64/predicates.md (aarch64_any_register_operand): New
20236 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
20237 that are semantically a reverse operation.
20238 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
20239 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
20240 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
20242 (aarch64_can_change_mode_class): For big-endian, forbid changes
20243 between two SVE modes if they have different element sizes.
20245 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20247 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
20248 the TImode handling for big-endian targets.
20250 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20252 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
20253 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
20255 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
20256 Remove BSWAP handing for big-endian targets and use the form of
20257 LD1RQ appropariate for the mode.
20259 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20261 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
20262 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
20263 duplicated element.
20265 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20268 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
20269 check for operands that need to go through aarch64_sve_reload_be.
20271 2018-02-01 Jakub Jelinek <jakub@redhat.com>
20273 PR tree-optimization/81661
20274 PR tree-optimization/84117
20275 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
20276 * tree-eh.c: Include gimplify.h.
20277 (find_trapping_overflow, replace_trapping_overflow,
20278 rewrite_to_non_trapping_overflow): New functions.
20279 * tree-vect-loop.c: Include tree-eh.h.
20280 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
20281 * tree-data-ref.c: Include tree-eh.h.
20282 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
20284 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
20286 PR rtl-optimization/84123
20287 * combine.c (change_zero_ext): Check if hard register satisfies
20288 can_change_dest_mode before calling gen_lowpart_SUBREG.
20290 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
20293 * ira.c (ira_init_register_move_cost): Remove assert.
20295 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
20297 PR rtl-optimization/84071
20298 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
20299 * doc/tm.texi: Regenerate.
20301 2018-01-31 Richard Biener <rguenther@suse.de>
20303 PR tree-optimization/84132
20304 * tree-data-ref.c (analyze_miv_subscript): Properly
20305 check whether evolution_function_is_affine_multivariate_p
20306 before calling gcd_of_steps_may_divide_p.
20308 2018-01-31 Julia Koval <julia.koval@intel.com>
20311 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
20312 * config/i386/i386.md (rdpid_rex64) New.
20313 (rdpid): Make 32bit only.
20315 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
20318 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
20319 an IDENTIFIER_NODE for FUNCTION_TYPE's.
20321 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
20324 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20326 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
20328 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
20330 PR rtl-optimization/84071
20331 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
20332 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
20334 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
20336 * config/arc/arc.c (arc_handle_aux_attribute): New function.
20337 (arc_attribute_table): Add 'aux' attribute.
20338 (arc_in_small_data_p): Consider aux like variables.
20339 (arc_is_aux_reg_p): New function.
20340 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
20341 (arc_get_aux_arg): New function.
20342 (prepare_move_operands): Handle aux-register access.
20343 (arc_handle_aux_attribute): New function.
20344 * doc/extend.texi (ARC Variable attributes): Add subsection.
20346 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
20348 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
20349 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
20350 (arc_attribute_table): Add 'uncached' attribute.
20351 (arc_print_operand): Print '.di' flag for uncached memory
20353 (arc_in_small_data_p): Do not consider for small data the uncached
20355 (arc_is_uncached_mem_p): New function.
20356 * config/arc/predicates.md (compact_store_memory_operand): Check
20357 for uncached memory accesses.
20358 (nonvol_nonimm_operand): Likewise.
20359 * doc/extend.texi (ARC Type Attribute): New subsection.
20361 2018-01-31 Jakub Jelinek <jakub@redhat.com>
20364 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
20365 falign-loops=): Add Optimization flag.
20367 2018-01-30 Jeff Law <law@redhat.com>
20370 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
20371 INT_REGISTERS_SAVED. Check it prior to calling
20372 get_scratch_register_on_entry.
20373 (ix86_adjust_stack_and_probe): Similarly.
20374 (ix86_emit_probe_stack_range): Similarly.
20375 (ix86_expand_prologue): Corresponding changes.
20377 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20380 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
20381 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
20383 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
20386 * lra-constraints.c (curr_insn_transform): Process AND in the
20389 2018-01-30 Jakub Jelinek <jakub@redhat.com>
20391 PR rtl-optimization/83986
20392 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
20393 dependence against last_pending_memory_flush in addition to
20394 pending_jump_insns.
20396 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
20398 PR tree-optimization/81611
20399 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
20402 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20405 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
20408 2018-01-30 Richard Biener <rguenther@suse.de>
20409 Jakub Jelinek <jakub@redhat.com>
20411 PR tree-optimization/84111
20412 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
20413 inner loops added during recursion, as they don't have up-to-date
20416 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
20419 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
20420 (can_inline_edge_by_limits_p): ... here.
20421 (can_early_inline_edge_p, check_callers,
20422 update_caller_keys, update_callee_keys, recursive_inlining,
20423 add_new_edges_to_heap, speculation_useful_p,
20424 inline_small_functions,
20425 inline_small_functions, flatten_function,
20426 inline_to_all_callers_1): Update.
20428 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
20430 * profile-count.c (profile_count::combine_with_ipa_count): Handle
20433 2018-01-30 Richard Biener <rguenther@suse.de>
20435 PR tree-optimization/83008
20436 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
20437 invariant and constant vector uses in stmts when they need
20438 more than one stmt.
20440 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20443 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
20444 * configure: Regenerate.
20446 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20448 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
20450 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
20451 Use gen_rtx_REG rather than gen_lowpart.
20453 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20455 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
20456 rather than 0 when creating partial subregs.
20458 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20460 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
20463 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
20466 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
20467 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
20468 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
20469 flags. This restores the settings used before the 2017-07-24.
20470 Turning off pre increment/decrement/modify allows IVOPTS to
20471 optimize DF/SF loops where the index is an int.
20473 2018-01-29 Richard Biener <rguenther@suse.de>
20474 Kelvin Nilsen <kelvin@gcc.gnu.org>
20477 * tree-vect-stmts.c (vectorizable_call): Don't call
20478 targetm.vectorize_builtin_md_vectorized_function if callee is
20481 2018-01-22 Carl Love <cel@us.ibm.com>
20483 * doc/extend.tex: Fix typo in second arg in
20484 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
20486 2018-01-29 Richard Biener <rguenther@suse.de>
20488 PR tree-optimization/84086
20489 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
20490 (flush_ssaname_freelist): When SSA names were released reset
20491 the SCEV hash table.
20493 2018-01-29 Richard Biener <rguenther@suse.de>
20495 PR tree-optimization/84057
20496 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
20497 removed paths when removing edges.
20499 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
20501 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
20502 -mfunction-return=@var{choice}.
20504 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
20506 PR diagnostic/84034
20507 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
20508 Handle CR like TAB.
20509 (layout::print_source_line): Likewise.
20510 (test_get_line_width_without_trailing_whitespace): Add test cases.
20512 2018-01-27 Jakub Jelinek <jakub@redhat.com>
20514 PR middle-end/84040
20515 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
20518 2018-01-26 Jim Wilson <jimw@sifive.com>
20520 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
20522 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
20525 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20527 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
20528 and CMP + SUB-immediate -> SUBS.
20530 2018-01-26 Martin Sebor <msebor@redhat.com>
20532 PR tree-optimization/83896
20533 * tree-ssa-strlen.c (get_string_len): Rename...
20534 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
20535 Avoid assuming length is constant.
20536 (handle_char_store): Use HOST_WIDE_INT for string length.
20538 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
20541 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
20542 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
20544 2018-01-26 Richard Biener <rguenther@suse.de>
20546 PR rtl-optimization/84003
20547 * dse.c (record_store): Only record redundant stores when
20548 the earlier store aliases at least all accesses the later one does.
20550 2018-01-26 Jakub Jelinek <jakub@redhat.com>
20552 PR rtl-optimization/83985
20553 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
20554 REG_CFA_RESTORE insns.
20555 (delete_unmarked_insns): Don't ignore separate shrink wrapping
20556 REG_CFA_RESTORE insns here.
20559 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
20560 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
20562 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20564 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
20565 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
20566 (arc_init): Likewise.
20567 (arc_override_options): Likewise.
20568 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
20570 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
20572 * config/arc/arc.h (TARGET_DBNZ): Define.
20573 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
20574 properly set the tune attribute.
20575 (dbnz): Use TARGET_DBNZ guard.
20576 * config/arc/arc.opt (mtune): Add core3 option.
20578 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20580 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
20581 recognize new pic like addresses.
20582 (arc_delegitimize_address): Clean up.
20584 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20586 * config/arc/arc-arches.def: Option mrf16 valid for all
20588 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
20589 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
20590 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
20591 * config/arc/arc-tables.opt: Regenerate.
20592 * config/arc/arc.c (arc_conditional_register_usage): Handle
20593 reduced register file case.
20594 (arc_file_start): Set must have build attributes.
20595 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
20596 mrf16 option value.
20597 * config/arc/arc.opt (mrf16): Add new option.
20598 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
20599 * config/arc/genmultilib.awk: Handle new mrf16 option.
20600 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
20601 * config/arc/t-multilib: Regenerate.
20602 * doc/invoke.texi (ARC Options): Document mrf16 option.
20604 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20606 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
20607 * config/arc/arc.c (arc_handle_secure_attribute): New function.
20608 (arc_attribute_table): Add 'secure_call' attribute.
20609 (arc_print_operand): Print secure call operand.
20610 (arc_function_ok_for_sibcall): Don't optimize tail calls when
20612 (arc_is_secure_call_p): New function. * config/arc/arc.md
20613 (call_i): Add support for sjli instruction.
20614 (call_value_i): Likewise.
20615 * config/arc/constraints.md (Csc): New constraint.
20617 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20618 John Eric Martin <John.Martin@emmicro-us.com>
20620 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
20621 * config/arc/arc.c (_arc_jli_section): New struct.
20622 (arc_jli_section): New type.
20623 (rc_jli_sections): New static variable.
20624 (arc_handle_jli_attribute): New function.
20625 (arc_attribute_table): Add jli_always and jli_fixed attribute.
20626 (arc_file_end): New function.
20627 (TARGET_ASM_FILE_END): Define.
20628 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
20629 (arc_add_jli_section): New function.
20630 (jli_call_scan): Likewise.
20631 (arc_reorg): Call jli_call_scan.
20632 (arc_output_addsi): Remove 'S' from printing asm operand.
20633 (arc_is_jli_call_p): New function.
20634 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
20636 (movhi_insn): Likewise.
20637 (movsi_insn): Likewise.
20638 (movsi_set_cc_insn): Likewise.
20639 (loadqi_update): Likewise.
20640 (load_zeroextendqisi_update): Likewise.
20641 (load_signextendqisi_update): Likewise.
20642 (loadhi_update): Likewise.
20643 (load_zeroextendhisi_update): Likewise.
20644 (load_signextendhisi_update): Likewise.
20645 (loadsi_update): Likewise.
20646 (loadsf_update): Likewise.
20647 (movsicc_insn): Likewise.
20648 (bset_insn): Likewise.
20649 (bxor_insn): Likewise.
20650 (bclr_insn): Likewise.
20651 (bmsk_insn): Likewise.
20652 (bicsi3_insn): Likewise.
20653 (cmpsi_cc_c_insn): Likewise.
20654 (movsi_ne): Likewise.
20655 (movsi_cond_exec): Likewise.
20656 (clrsbsi2): Likewise.
20657 (norm_f): Likewise.
20667 (call_i): Remove 'S' asm letter, add jli instruction.
20668 (call_value_i): Likewise.
20669 * config/arc/arc.op (mjli-always): New option.
20670 * config/arc/constraints.md (Cji): New constraint.
20671 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
20673 (subsf3_fpx): Likewise.
20674 (mulsf3_fpx): Likewise.
20675 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
20677 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
20678 function attrbutes.
20679 * doc/invoke.texi (ARC): Document mjli-always option.
20681 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
20683 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
20684 avoid addition with 0 and use incw and decw where possible.
20686 2018-01-26 Richard Biener <rguenther@suse.de>
20688 PR tree-optimization/81082
20689 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
20690 association if it requires casting to unsigned.
20691 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
20692 from fold_plusminus_mult_expr to catch important cases late when
20693 range info is available.
20695 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20697 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
20698 * configure.ac (hidden_linkonce): New test.
20699 * configure: Regenerate.
20700 * config.in: Regenerate.
20702 2018-01-26 Julia Koval <julia.koval@intel.com>
20704 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
20705 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
20706 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
20707 _mm_mask_bitshuffle_epi64_mask): Fix type.
20708 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
20709 USI_FTYPE_V4DI_V4DI_USI): Remove.
20710 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
20711 __builtin_ia32_vpshufbitqmb256_mask,
20712 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
20713 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
20714 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
20716 2018-01-26 Alan Modra <amodra@gmail.com>
20719 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
20720 UNSPEC_VBPERMQ. Sort other unspecs.
20722 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
20724 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
20726 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
20728 PR middle-end/83055
20729 * predict.c (drop_profile): Do not push/pop cfun; update also
20731 (handle_missing_profiles): Fix logic looking for zero profiles.
20733 2018-01-25 Jakub Jelinek <jakub@redhat.com>
20735 PR middle-end/83977
20736 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
20737 on functions with #pragma omp declare simd or functions with simd
20739 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
20740 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
20741 Remove trailing \n from warning_at calls.
20743 2018-01-25 Tom de Vries <tom@codesourcery.com>
20746 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20747 for neutered workers.
20749 2018-01-24 Joseph Myers <joseph@codesourcery.com>
20752 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
20753 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
20755 2018-01-24 Jeff Law <law@redhat.com>
20758 * i386.c (get_probe_interval): Move to earlier point.
20759 (ix86_compute_frame_layout): If -fstack-clash-protection and
20760 the frame is larger than the probe interval, then use pushes
20761 to save registers rather than reg->mem moves.
20762 (ix86_expand_prologue): Remove conditional for int_registers_saved
20765 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
20768 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
20769 min/max for never referenced object.
20771 2018-01-24 Jakub Jelinek <jakub@redhat.com>
20773 PR middle-end/83977
20774 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
20776 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
20777 attributes from DECL_ATTRIBUTES (decl) without affecting
20778 DECL_ATTRIBUTES (current_function_decl).
20779 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
20780 functions with non-NULL DECL_ABSTRACT_ORIGIN.
20782 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
20784 PR tree-optimization/83979
20785 * fold-const.c (fold_comparison): Use constant_boolean_node
20786 instead of boolean_{true,false}_node.
20788 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
20790 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
20793 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20795 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20796 Simplify the clause that sets the length attribute.
20797 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20798 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
20799 clause that sets the length attribute.
20800 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20802 2018-01-24 Tom de Vries <tom@codesourcery.com>
20805 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
20806 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
20807 Add strict parameter.
20808 (prevent_branch_around_nothing): Insert dummy insn between branch to
20809 label and label with no ptx insn inbetween.
20810 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
20812 2018-01-24 Tom de Vries <tom@codesourcery.com>
20815 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20816 for neutered threads in warp.
20817 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
20819 2018-01-24 Richard Biener <rguenther@suse.de>
20821 PR tree-optimization/83176
20822 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
20825 2018-01-24 Richard Biener <rguenther@suse.de>
20827 PR tree-optimization/82819
20828 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
20829 code generating pluses that are no-ops in the target precision.
20831 2018-01-24 Richard Biener <rguenther@suse.de>
20833 PR middle-end/84000
20834 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
20836 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20838 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
20839 to merge probabilities.
20840 * predict.c (probably_never_executed): Also mark as cold functions
20841 with global 0 profile and guessed local profile.
20842 * profile-count.c (profile_probability::combine_with_count): New
20844 * profile-count.h (profile_probability::operator*,
20845 profile_probability::operator*=, profile_probability::operator/,
20846 profile_probability::operator/=): Reduce precision to adjusted
20847 and set value to guessed on contradictory divisions.
20848 (profile_probability::combine_with_freq): Remove.
20849 (profile_probability::combine_wiht_count): Declare.
20850 (profile_count::force_nonzero):: Set to adjusted.
20851 (profile_count::probability_in):: Set quality to adjusted.
20852 * tree-ssa-tail-merge.c (replace_block_by): Use
20853 combine_with_count.
20855 2018-01-23 Andrew Waterman <andrew@sifive.com>
20856 Jim Wilson <jimw@sifive.com>
20858 * config/riscv/riscv.c (riscv_stack_boundary): New.
20859 (riscv_option_override): Set riscv_stack_boundary. Handle
20860 riscv_preferred_stack_boundary_arg.
20861 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
20862 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
20863 (STACK_BOUNDARY): Set to riscv_stack_boundary.
20864 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
20865 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
20866 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
20868 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
20871 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
20872 of struct ix86_frame.
20873 (ix86_expand_epilogue): Likewise. Add a local variable for
20874 the reg_save_offset field in struct ix86_frame.
20876 2018-01-23 Bin Cheng <bin.cheng@arm.com>
20878 PR tree-optimization/82604
20879 * tree-loop-distribution.c (enum partition_kind): New enum item
20880 PKIND_PARTIAL_MEMSET.
20881 (partition_builtin_p): Support above new enum item.
20882 (generate_code_for_partition): Ditto.
20883 (compute_access_range): Differentiate cases that equality can be
20884 proven at all loops, the innermost loops or no loops.
20885 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
20886 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
20887 (finalize_partitions, distribute_loop): Don't fuse partition of
20888 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
20889 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
20890 parloop is enabled.
20892 2018-01-23 Martin Liska <mliska@suse.cz>
20894 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
20895 order to ignore the predictor.
20896 (PRED_POLYMORPHIC_CALL): Likewise.
20897 (PRED_RECURSIVE_CALL): Likewise.
20899 2018-01-23 Martin Liska <mliska@suse.cz>
20901 * tree-profile.c (tree_profiling): Print function header to
20902 aware reader which function we are working on.
20903 * value-prof.c (gimple_find_values_to_profile): Do not print
20904 not interesting value histograms.
20906 2018-01-23 Martin Liska <mliska@suse.cz>
20908 * profile-count.h (enum profile_quality): Add
20909 profile_uninitialized as the first value. Do not number values
20910 as they are zero based.
20911 (profile_count::verify): Update sanity check.
20912 (profile_probability::verify): Likewise.
20914 2018-01-23 Nathan Sidwell <nathan@acm.org>
20916 * doc/invoke.texi (ffor-scope): Deprecate.
20918 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20920 PR tree-optimization/83510
20921 * domwalk.c (set_all_edges_as_executable): New function.
20922 (dom_walker::dom_walker): Convert bool param
20923 "skip_unreachable_blocks" to enum reachability. Move setup of
20924 edge flags to set_all_edges_as_executable and only do it when
20925 reachability is REACHABLE_BLOCKS.
20926 * domwalk.h (enum dom_walker::reachability): New enum.
20927 (dom_walker::dom_walker): Convert bool param
20928 "skip_unreachable_blocks" to enum reachability.
20929 (set_all_edges_as_executable): New decl.
20930 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
20931 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
20933 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
20934 but converting true to REACHABLE_BLOCKS.
20935 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
20937 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
20938 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
20939 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
20941 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
20942 if check_all_array_refs will be called.
20944 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20946 * tree.c (selftest::test_location_wrappers): Add more test
20949 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20951 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
20952 (selftest::test_bit_in_range): Likewise.
20954 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
20957 * doc/sourcebuild.texi (vect_float): Say that the selector
20958 only describes the situation when -funsafe-math-optimizations is on.
20959 (vect_float_strict): Document.
20961 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
20963 PR tree-optimization/83965
20964 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
20965 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
20966 instead of checking only for a reduction.
20967 (vect_recog_widen_sum_pattern): Likewise.
20969 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20971 * predict.c (probably_never_executed): Only use precise profile info.
20972 (compute_function_frequency): Skip after inlining hack since we now
20973 have quality checking.
20975 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20977 * profile-count.h (profile_probability::very_unlikely,
20978 profile_probability::unlikely, profile_probability::even): Set
20979 precision to guessed.
20981 2018-01-23 Richard Biener <rguenther@suse.de>
20983 PR tree-optimization/83963
20984 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
20985 Properly terminate dominator walk when crossing the exit edge not
20986 when visiting its source block.
20988 2018-01-23 Jakub Jelinek <jakub@redhat.com>
20991 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
20992 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
20994 2018-01-22 Jakub Jelinek <jakub@redhat.com>
20996 PR tree-optimization/83957
20997 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
20998 semicolon after for body surrounded by braces.
21000 PR tree-optimization/83081
21001 * profile-count.h (profile_probability::split): New method.
21002 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
21003 Use profile_probability::split.
21004 (do_compare_rtx_and_jump): Fix adjustment of probabilities
21005 when splitting a single conditional jump into 2.
21007 2018-01-22 David Malcolm <dmalcolm@redhat.com>
21009 PR tree-optimization/69452
21010 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
21013 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21015 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
21016 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
21017 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
21019 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21021 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
21023 * config/rl78/rl78.md (movdi): New define_expand.
21024 * config/rl78/rl78.c (rl78_split_movdi): New function.
21026 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
21029 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
21031 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
21032 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
21033 128-bit to produce an UNSPEC move to get the double word with the
21034 signbit and then a shift directly to do signbit.
21035 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
21036 implementation with a new version that just does either a direct
21037 move or a regular move. Move memory interface to separate insns.
21038 Move insns so they are next to the expander.
21039 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
21040 with signbit move. Split big and little endian case.
21041 (signbit<mode>2_dm_mem_le): Likewise.
21042 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
21043 (signbit<mode>2_dm2): Likewise.
21045 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21047 * config/rl78/rl78.md (anddi3): New define_expand.
21049 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21051 * config/rl78/rl78.md (umindi3): New define_expand.
21053 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21055 * config/rl78/rl78.md (smindi3): New define_expand.
21057 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21059 * config/rl78/rl78.md (smaxdi3): New define_expand.
21061 2018-01-22 Carl Love <cel@us.ibm.com>
21063 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
21064 LVX_V1TI): Add macro expansion.
21065 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
21066 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
21067 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
21068 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
21069 Change check to determine if the instruction is a byte reversing
21070 entry. Fix typo in comment.
21071 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
21072 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
21073 Add def_builtin calls for new builtins.
21074 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
21075 Add define_insn expansion.
21077 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21079 * config/rl78/rl78.md (umaxdi3): New define_expand.
21081 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21083 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
21084 for non-QImode registers.
21086 2018-01-22 Richard Biener <rguenther@suse.de>
21088 PR tree-optimization/83963
21089 * graphite-scop-detection.c (scop_detection::get_sese): Delay
21090 including the loop exit block.
21091 (scop_detection::merge_sese): Likewise.
21092 (scop_detection::add_scop): Do it here instead.
21094 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21096 * doc/sourcebuild.texi (arm_softfloat): Document.
21098 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
21101 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
21102 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
21103 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
21105 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21106 David Edelsohn <dje.gcc@gmail.com>
21109 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21110 Change "crset eq" to "crset 2".
21111 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21112 (*call_indirect_aix<mode>_nospec): Likewise.
21113 (*call_value_indirect_aix<mode>_nospec): Likewise.
21114 (*call_indirect_elfv2<mode>_nospec): Likewise.
21115 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
21116 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
21117 change assembly output from . to $.
21118 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21119 (indirect_jump<mode>_nospec): Change assembly output from . to $.
21120 (*tablejump<mode>_internal1_nospec): Likewise.
21122 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
21125 * config/sh/sh_optimize_sett_clrt.cc:
21126 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
21128 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
21130 PR tree-optimization/83940
21131 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
21132 offset_dt to vect_constant_def rather than vect_unknown_def_type.
21133 (vect_check_load_store_mask): Add a mask_dt_out parameter and
21134 use it to pass back the definition type.
21135 (vect_check_store_rhs): Likewise rhs_dt_out.
21136 (vect_build_gather_load_calls): Add a mask_dt argument and use
21137 it instead of a call to vect_is_simple_use.
21138 (vectorizable_store): Update calls to vect_check_load_store_mask
21139 and vect_check_store_rhs. Use the dt returned by the latter instead
21140 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
21141 instead of calls to vect_is_simple_use. Pass the scalar rather
21142 than the vector operand to vect_is_simple_use when handling
21143 second and subsequent copies of an rhs value.
21144 (vectorizable_load): Update calls to vect_check_load_store_mask
21145 and vect_build_gather_load_calls. Use the cached mask_dt and
21146 gs_info.offset_dt instead of calls to vect_is_simple_use.
21148 2018-01-20 Jakub Jelinek <jakub@redhat.com>
21150 PR middle-end/83945
21151 * tree-emutls.c: Include gimplify.h.
21152 (lower_emutls_2): New function.
21153 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
21154 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
21155 it before further processing.
21158 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
21159 UINTVAL (trueop1) instead of INTVAL (op1).
21161 2018-01-19 Jakub Jelinek <jakub@redhat.com>
21165 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
21166 INCOMING_FRAME_SP_OFFSET if not defined.
21167 (scan_trace): Add ENTRY argument. If true and
21168 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
21169 emit a note to adjust the CFA offset.
21170 (create_cfi_notes): Adjust scan_trace callers.
21171 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
21172 INCOMING_FRAME_SP_OFFSET in the CIE.
21173 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
21174 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
21176 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
21177 * doc/tm.texi: Regenerated.
21179 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
21181 PR rtl-optimization/83147
21182 * lra-constraints.c (remove_inheritance_pseudos): Use
21183 lra_substitute_pseudo_within_insn.
21185 2018-01-19 Tom de Vries <tom@codesourcery.com>
21186 Cesar Philippidis <cesar@codesourcery.com>
21189 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
21191 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
21194 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
21195 spaces for function labels.
21197 2018-01-19 Martin Liska <mliska@suse.cz>
21199 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
21200 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
21201 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
21202 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
21203 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
21204 (PRED_CONST_RETURN): Change from 69 to 65.
21205 (PRED_NULL_RETURN): Change from 91 to 71.
21206 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
21207 (PRED_LOOP_GUARD): Change from 66 to 73.
21209 2018-01-19 Martin Liska <mliska@suse.cz>
21211 * predict.c (predict_insn_def): Add new assert.
21212 (struct branch_predictor): Change type to signed integer.
21213 (test_prediction_value_range): Amend test to cover
21214 PROB_UNINITIALIZED.
21215 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
21216 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
21217 (PRED_LOOP_ITERATIONS_MAX): Likewise.
21218 (PRED_LOOP_IV_COMPARE): Likewise.
21219 * predict.h (PROB_UNINITIALIZED): Define new constant.
21221 2018-01-19 Martin Liska <mliska@suse.cz>
21223 * predict.c (dump_prediction): Add new format for
21224 analyze_brprob.py script which is enabled with -details
21226 * profile-count.h (precise_p): New function.
21228 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
21230 PR tree-optimization/83922
21231 * tree-vect-loop.c (vect_verify_full_masking): Return false if
21232 there are no statements that need masking.
21233 (vect_active_double_reduction_p): New function.
21234 (vect_analyze_loop_operations): Use it when handling phis that
21235 are not in the loop header.
21237 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
21239 PR tree-optimization/83914
21240 * tree-vect-loop.c (vectorizable_induction): Don't convert
21241 init_expr or apply the peeling adjustment for inductions
21242 that are nested within the vectorized loop.
21244 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21246 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
21249 2018-01-18 Jakub Jelinek <jakub@redhat.com>
21253 * function.h (gimplify_parameters): Add gimple_seq * argument.
21254 * function.c: Include gimple.h and options.h.
21255 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
21256 for the added local temporaries if needed.
21257 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
21258 if there are any parameter cleanups, wrap whole body into a
21259 try/finally with the cleanups.
21261 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
21264 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
21265 Use GET_MODE_CLASS for scalar floating point.
21267 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
21271 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
21272 Fix call of call_cgraph_insertion_hooks.
21274 2018-01-18 Martin Sebor <msebor@redhat.com>
21276 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
21278 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
21281 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
21284 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
21287 * common.opt: (-ffile-prefix-map): New option.
21288 * opts.c (common_handle_option): Defer it.
21289 * opts-global.c (handle_common_deferred_options): Handle it.
21290 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
21291 * file-prefix-map.h: New file.
21292 (remap_debug_filename, add_debug_prefix_map): ...here.
21293 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
21294 * final.c (debug_prefix_map, add_debug_prefix_map
21295 remap_debug_filename): Move to...
21296 * file-prefix-map.c: New file.
21297 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
21298 generalize, get rid of alloca(), use strrchr() instead of strchr().
21299 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
21300 Implement in terms of add_prefix_map().
21301 (remap_macro_filename, remap_debug_filename): Implement in term of
21303 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
21304 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
21305 * dbxout.c: Include file-prefix-map.h.
21306 * varasm.c: Likewise.
21307 * vmsdbgout.c: Likewise.
21308 * xcoffout.c: Likewise.
21309 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
21310 * doc/cppopts.texi (-fmacro-prefix-map): Document.
21311 * doc/invoke.texi (-ffile-prefix-map): Document.
21312 (-fdebug-prefix-map): Update description.
21314 2018-01-18 Martin Liska <mliska@suse.cz>
21316 * config/i386/i386.c (indirect_thunk_name): Document that also
21318 (output_indirect_thunk): Document why both instructions
21319 (pause and lfence) are generated.
21321 2018-01-18 Richard Biener <rguenther@suse.de>
21323 PR tree-optimization/83887
21324 * graphite-scop-detection.c
21325 (scop_detection::get_nearest_dom_with_single_entry): Remove.
21326 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
21327 (scop_detection::merge_sese): Re-implement with a flood-fill
21328 algorithm that properly finds a SESE region if it exists.
21330 2018-01-18 Jakub Jelinek <jakub@redhat.com>
21333 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
21334 pointer_diff optimizations use view_convert instead of convert.
21336 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21338 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21339 Generate different code for -mno-speculate-indirect-jumps.
21340 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21341 (*call_indirect_aix<mode>): Disable for
21342 -mno-speculate-indirect-jumps.
21343 (*call_indirect_aix<mode>_nospec): New define_insn.
21344 (*call_value_indirect_aix<mode>): Disable for
21345 -mno-speculate-indirect-jumps.
21346 (*call_value_indirect_aix<mode>_nospec): New define_insn.
21347 (*sibcall_nonlocal_sysv<mode>): Generate different code for
21348 -mno-speculate-indirect-jumps.
21349 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21351 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
21353 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
21354 long double type, set the flags for noting the default long double
21355 type, even if we don't pass or return a long double type.
21357 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
21360 * ipa-inline.c (flatten_function): Do not overwrite final inlining
21363 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
21365 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
21366 support for merge[hl].
21367 (fold_mergehl_helper): New helper function.
21368 (tree-vector-builder.h): New #include for tree_vector_builder usage.
21369 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
21370 (altivec_vmrglw_direct): Add xxmrglw insn.
21372 2018-01-17 Andrew Waterman <andrew@sifive.com>
21374 * config/riscv/riscv.c (riscv_conditional_register_usage): If
21375 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
21377 2018-01-17 David Malcolm <dmalcolm@redhat.com>
21380 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
21381 call the lto_location_cache before reading the
21382 DECL_SOURCE_LOCATION of the types.
21384 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
21385 Richard Sandiford <richard.sandiford@linaro.org>
21387 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
21388 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
21389 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
21390 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
21391 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
21393 * config/aarch64/constraints.md (aarch64_movti_operand):
21395 * config/aarch64/predicates.md (Uti): Add new constraint.
21397 2018-01-17 Carl Love <cel@us.ibm.com>
21399 * config/rs6000/vsx.md (define_expand xl_len_r,
21400 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
21401 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
21403 (define_expand, define_insn): Move the shift left from the
21404 define_insn to the define_expand for lxvl and stxvl instructions.
21405 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
21406 and XL_LEN_R definitions to PURE.
21408 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
21410 * config/i386/i386.c (indirect_thunk_name): Declare regno
21411 as unsigned int. Compare regno with INVALID_REGNUM.
21412 (output_indirect_thunk): Ditto.
21413 (output_indirect_thunk_function): Ditto.
21414 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
21415 in the call to output_indirect_thunk_function.
21417 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
21419 PR middle-end/83884
21420 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
21421 rather than the size of inner_type to determine the stack slot size
21422 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
21424 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
21427 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
21430 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
21432 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
21433 endian Linux systems to optionally enable multilibs for selecting
21434 the long double type if the user configured an explicit type.
21435 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
21436 have no long double multilibs if not defined.
21437 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
21438 warn if the user used -mabi={ieee,ibm}longdouble and we built
21439 multilibs for long double.
21440 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
21441 appropriate multilib option.
21442 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
21444 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
21445 for building long double multilibs.
21446 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
21448 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
21450 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
21453 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
21455 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
21458 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
21461 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
21464 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
21466 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
21467 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
21469 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
21471 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
21472 different rtl trees depending on TARGET_64BIT.
21473 (rs6000_gen_lvx): Likewise.
21475 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
21477 * config/visium/visium.md (nop): Tweak comment.
21478 (hazard_nop): Likewise.
21480 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21482 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
21483 -mspeculate-indirect-jumps.
21484 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
21485 for -mno-speculate-indirect-jumps.
21486 (*call_indirect_elfv2<mode>_nospec): New define_insn.
21487 (*call_value_indirect_elfv2<mode>): Disable for
21488 -mno-speculate-indirect-jumps.
21489 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
21490 (indirect_jump): Emit different RTL for
21491 -mno-speculate-indirect-jumps.
21492 (*indirect_jump<mode>): Disable for
21493 -mno-speculate-indirect-jumps.
21494 (*indirect_jump<mode>_nospec): New define_insn.
21495 (tablejump): Emit different RTL for
21496 -mno-speculate-indirect-jumps.
21497 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
21498 (tablejumpsi_nospec): New define_expand.
21499 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
21500 (tablejumpdi_nospec): New define_expand.
21501 (*tablejump<mode>_internal1): Disable for
21502 -mno-speculate-indirect-jumps.
21503 (*tablejump<mode>_internal1_nospec): New define_insn.
21504 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
21507 2018-01-16 Artyom Skrobov tyomitch@gmail.com
21509 * caller-save.c (insert_save): Drop unnecessary parameter. All
21512 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21513 Richard Biener <rguenth@suse.de>
21516 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
21517 return early, inline manually is_gimple_sizepos. Make sure if we
21518 call gimplify_expr we don't end up with a gimple constant.
21519 * tree.c (variably_modified_type_p): Don't return true for
21520 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
21521 * gimplify.h (is_gimple_sizepos): Remove.
21523 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21525 PR tree-optimization/83857
21526 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
21527 vectorizable_live_operation for pure SLP statements.
21528 (vectorizable_live_operation): Handle PHIs.
21530 2018-01-16 Richard Biener <rguenther@suse.de>
21532 PR tree-optimization/83867
21533 * tree-vect-stmts.c (vect_transform_stmt): Precompute
21534 nested_in_vect_loop_p since the scalar stmt may get invalidated.
21536 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21539 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
21540 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
21541 If off is not INTEGER_CST, issue a may not be aligned warning
21542 rather than isn't aligned. Use isn%'t rather than isn't.
21543 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
21545 <case MULT_EXPR>: Improve the case when bottom and one of the
21546 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
21547 operand, in that case check if the other operand is multiple of
21548 bottom divided by the INTEGER_CST operand.
21550 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21553 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
21554 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
21555 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
21556 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
21557 * config/pa/pa.c (pa_function_arg_advance): Likewise.
21558 (pa_function_arg, pa_arg_partial_bytes): Likewise.
21559 (pa_function_arg_size): New function.
21561 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21563 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
21564 in a separate statement.
21566 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21568 PR tree-optimization/83847
21569 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
21570 group gathers and scatters.
21572 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21574 PR rtl-optimization/86620
21575 * params.def (max-sched-ready-insns): Bump minimum value to 1.
21577 PR rtl-optimization/83213
21578 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
21579 to last if both are JUMP_INSNs.
21581 PR tree-optimization/83843
21582 * gimple-ssa-store-merging.c
21583 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
21584 store_immediate_info for bswap/nop orig_stores.
21586 2018-01-15 Andrew Waterman <andrew@sifive.com>
21588 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
21590 <UDIV>: Increase cost if !TARGET_DIV.
21592 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
21594 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
21595 (define_attr "cr_logical_3op"): New.
21596 (cceq_ior_compare): Adjust.
21597 (cceq_ior_compare_complement): Adjust.
21598 (*cceq_rev_compare): Adjust.
21599 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
21600 (is_cracked_insn): Adjust.
21601 (insn_must_be_first_in_group): Adjust.
21602 * config/rs6000/40x.md: Adjust.
21603 * config/rs6000/440.md: Adjust.
21604 * config/rs6000/476.md: Adjust.
21605 * config/rs6000/601.md: Adjust.
21606 * config/rs6000/603.md: Adjust.
21607 * config/rs6000/6xx.md: Adjust.
21608 * config/rs6000/7450.md: Adjust.
21609 * config/rs6000/7xx.md: Adjust.
21610 * config/rs6000/8540.md: Adjust.
21611 * config/rs6000/cell.md: Adjust.
21612 * config/rs6000/e300c2c3.md: Adjust.
21613 * config/rs6000/e500mc.md: Adjust.
21614 * config/rs6000/e500mc64.md: Adjust.
21615 * config/rs6000/e5500.md: Adjust.
21616 * config/rs6000/e6500.md: Adjust.
21617 * config/rs6000/mpc.md: Adjust.
21618 * config/rs6000/power4.md: Adjust.
21619 * config/rs6000/power5.md: Adjust.
21620 * config/rs6000/power6.md: Adjust.
21621 * config/rs6000/power7.md: Adjust.
21622 * config/rs6000/power8.md: Adjust.
21623 * config/rs6000/power9.md: Adjust.
21624 * config/rs6000/rs64.md: Adjust.
21625 * config/rs6000/titan.md: Adjust.
21627 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21629 * config/i386/predicates.md (indirect_branch_operand): Rewrite
21630 ix86_indirect_branch_register logic.
21632 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21634 * config/i386/constraints.md (Bs): Update
21635 ix86_indirect_branch_register check. Don't check
21636 ix86_indirect_branch_register with GOT_memory_operand.
21638 * config/i386/predicates.md (GOT_memory_operand): Don't check
21639 ix86_indirect_branch_register here.
21640 (GOT32_symbol_operand): Likewise.
21642 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21644 * config/i386/predicates.md (constant_call_address_operand):
21645 Rewrite ix86_indirect_branch_register logic.
21646 (sibcall_insn_operand): Likewise.
21648 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21650 * config/i386/constraints.md (Bs): Replace
21651 ix86_indirect_branch_thunk_register with
21652 ix86_indirect_branch_register.
21654 * config/i386/i386.md (indirect_jump): Likewise.
21655 (tablejump): Likewise.
21656 (*sibcall_memory): Likewise.
21657 (*sibcall_value_memory): Likewise.
21658 Peepholes of indirect call and jump via memory: Likewise.
21659 * config/i386/i386.opt: Likewise.
21660 * config/i386/predicates.md (indirect_branch_operand): Likewise.
21661 (GOT_memory_operand): Likewise.
21662 (call_insn_operand): Likewise.
21663 (sibcall_insn_operand): Likewise.
21664 (GOT32_symbol_operand): Likewise.
21666 2018-01-15 Jakub Jelinek <jakub@redhat.com>
21668 PR middle-end/83837
21669 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
21670 type rather than type addr's type points to.
21671 (expand_omp_atomic_mutex): Likewise.
21672 (expand_omp_atomic): Likewise.
21674 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21677 * config/i386/i386.c (output_indirect_thunk_function): Use
21678 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
21679 for __x86_return_thunk.
21681 2018-01-15 Richard Biener <rguenther@suse.de>
21683 PR middle-end/83850
21684 * expmed.c (extract_bit_field_1): Fix typo.
21686 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21689 * config/arm/iterators.md (VF): New mode iterator.
21690 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
21691 Remove integer-related logic from pattern.
21692 (neon_vabd<mode>_3): Likewise.
21694 2018-01-15 Jakub Jelinek <jakub@redhat.com>
21696 PR middle-end/82694
21697 * common.opt (fstrict-overflow): No longer an alias.
21698 (fwrapv-pointer): New option.
21699 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
21700 also for pointer types based on flag_wrapv_pointer.
21701 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
21702 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
21703 opts->x_flag_wrapv got set.
21704 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
21705 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
21706 POINTER_TYPE_OVERFLOW_UNDEFINED.
21707 * match.pd: Likewise in address comparison pattern.
21708 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
21710 2018-01-15 Richard Biener <rguenther@suse.de>
21713 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
21714 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
21715 Reset type names to their identifier if their TYPE_DECL doesn't
21716 have linkage (and thus is used for ODR and devirt).
21717 (save_debug_info_for_decl): Remove.
21718 (save_debug_info_for_type): Likewise.
21719 (add_tree_to_fld_list): Adjust.
21720 * tree-pretty-print.c (dump_generic_node): Make dumping of
21721 type names more robust.
21723 2018-01-15 Richard Biener <rguenther@suse.de>
21725 * BASE-VER: Bump to 8.0.1.
21727 2018-01-14 Martin Sebor <msebor@redhat.com>
21730 * builtins.c (check_access): Avoid warning when the no-warning bit
21733 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
21735 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
21736 * ira-color (allocno_hard_regs_compare): Likewise.
21738 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
21741 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
21742 Use .pushsection/.popsection.
21744 2018-01-14 Martin Sebor <msebor@redhat.com>
21747 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
21749 2018-01-14 Jakub Jelinek <jakub@redhat.com>
21751 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
21752 entry from extra_headers.
21753 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
21754 extra_headers, make the list bitwise identical to the i?86-*-* one.
21756 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21758 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
21759 -mcmodel=large with -mindirect-branch=thunk,
21760 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
21761 -mfunction-return=thunk-extern.
21762 * doc/invoke.texi: Document -mcmodel=large is incompatible with
21763 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
21764 -mfunction-return=thunk and -mfunction-return=thunk-extern.
21766 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21768 * config/i386/i386.c (print_reg): Print the name of the full
21769 integer register without '%'.
21770 (ix86_print_operand): Handle 'V'.
21771 * doc/extend.texi: Document 'V' modifier.
21773 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21775 * config/i386/constraints.md (Bs): Disallow memory operand for
21776 -mindirect-branch-register.
21778 * config/i386/predicates.md (indirect_branch_operand): Likewise.
21779 (GOT_memory_operand): Likewise.
21780 (call_insn_operand): Likewise.
21781 (sibcall_insn_operand): Likewise.
21782 (GOT32_symbol_operand): Likewise.
21783 * config/i386/i386.md (indirect_jump): Call convert_memory_address
21784 for -mindirect-branch-register.
21785 (tablejump): Likewise.
21786 (*sibcall_memory): Likewise.
21787 (*sibcall_value_memory): Likewise.
21788 Disallow peepholes of indirect call and jump via memory for
21789 -mindirect-branch-register.
21790 (*call_pop): Replace m with Bw.
21791 (*call_value_pop): Likewise.
21792 (*sibcall_pop_memory): Replace m with Bs.
21793 * config/i386/i386.opt (mindirect-branch-register): New option.
21794 * doc/invoke.texi: Document -mindirect-branch-register option.
21796 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21798 * config/i386/i386-protos.h (ix86_output_function_return): New.
21799 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
21800 set function_return_type.
21801 (indirect_thunk_name): Add ret_p to indicate thunk for function
21803 (output_indirect_thunk_function): Pass false to
21804 indirect_thunk_name.
21805 (ix86_output_indirect_branch_via_reg): Likewise.
21806 (ix86_output_indirect_branch_via_push): Likewise.
21807 (output_indirect_thunk_function): Create alias for function
21808 return thunk if regno < 0.
21809 (ix86_output_function_return): New function.
21810 (ix86_handle_fndecl_attribute): Handle function_return.
21811 (ix86_attribute_table): Add function_return.
21812 * config/i386/i386.h (machine_function): Add
21813 function_return_type.
21814 * config/i386/i386.md (simple_return_internal): Use
21815 ix86_output_function_return.
21816 (simple_return_internal_long): Likewise.
21817 * config/i386/i386.opt (mfunction-return=): New option.
21818 (indirect_branch): Mention -mfunction-return=.
21819 * doc/extend.texi: Document function_return function attribute.
21820 * doc/invoke.texi: Document -mfunction-return= option.
21822 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21824 * config/i386/i386-opts.h (indirect_branch): New.
21825 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
21826 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
21827 with local indirect jump when converting indirect call and jump.
21828 (ix86_set_indirect_branch_type): New.
21829 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
21830 (indirectlabelno): New.
21831 (indirect_thunk_needed): Likewise.
21832 (indirect_thunk_bnd_needed): Likewise.
21833 (indirect_thunks_used): Likewise.
21834 (indirect_thunks_bnd_used): Likewise.
21835 (INDIRECT_LABEL): Likewise.
21836 (indirect_thunk_name): Likewise.
21837 (output_indirect_thunk): Likewise.
21838 (output_indirect_thunk_function): Likewise.
21839 (ix86_output_indirect_branch_via_reg): Likewise.
21840 (ix86_output_indirect_branch_via_push): Likewise.
21841 (ix86_output_indirect_branch): Likewise.
21842 (ix86_output_indirect_jmp): Likewise.
21843 (ix86_code_end): Call output_indirect_thunk_function if needed.
21844 (ix86_output_call_insn): Call ix86_output_indirect_branch if
21846 (ix86_handle_fndecl_attribute): Handle indirect_branch.
21847 (ix86_attribute_table): Add indirect_branch.
21848 * config/i386/i386.h (machine_function): Add indirect_branch_type
21849 and has_local_indirect_jump.
21850 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
21852 (tablejump): Likewise.
21853 (*indirect_jump): Use ix86_output_indirect_jmp.
21854 (*tablejump_1): Likewise.
21855 (simple_return_indirect_internal): Likewise.
21856 * config/i386/i386.opt (mindirect-branch=): New option.
21857 (indirect_branch): New.
21860 (thunk-inline): Likewise.
21861 (thunk-extern): Likewise.
21862 * doc/extend.texi: Document indirect_branch function attribute.
21863 * doc/invoke.texi: Document -mindirect-branch= option.
21865 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
21868 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
21870 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
21872 * ipa-inline.c (want_inline_small_function_p): Return false if
21873 inlining has already failed with CIF_FINAL_ERROR.
21874 (update_caller_keys): Call want_inline_small_function_p before
21876 (update_callee_keys): Likewise.
21878 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
21880 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
21882 (rs6000_quadword_masked_address_p): Likewise.
21883 (quad_aligned_load_p): Likewise.
21884 (quad_aligned_store_p): Likewise.
21885 (const_load_sequence_p): Add comment to describe the outer-most loop.
21886 (mimic_memory_attributes_and_flags): New function.
21887 (rs6000_gen_stvx): Likewise.
21888 (replace_swapped_aligned_store): Likewise.
21889 (rs6000_gen_lvx): Likewise.
21890 (replace_swapped_aligned_load): Likewise.
21891 (replace_swapped_load_constant): Capitalize argument name in
21892 comment describing this function.
21893 (rs6000_analyze_swaps): Add a third pass to search for vector loads
21894 and stores that access quad-word aligned addresses and replace
21895 with stvx or lvx instructions when appropriate.
21896 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
21897 New function prototype.
21898 (rs6000_quadword_masked_address_p): Likewise.
21899 (rs6000_gen_lvx): Likewise.
21900 (rs6000_gen_stvx): Likewise.
21901 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
21902 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
21903 when memory address is aligned.
21904 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
21905 this split to select lvx instruction when memory address is aligned.
21906 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
21907 instruction when memory address is aligned.
21908 (*vsx_le_perm_load_v16qi): Likewise.
21909 (four unnamed splitters): Modify to select the stvx instruction
21910 when memory is aligned.
21912 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
21914 * predict.c (determine_unlikely_bbs): Handle correctly BBs
21915 which appears in the queue multiple times.
21917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21918 Alan Hayward <alan.hayward@arm.com>
21919 David Sherwood <david.sherwood@arm.com>
21921 * tree-vectorizer.h (vec_lower_bound): New structure.
21922 (_loop_vec_info): Add check_nonzero and lower_bounds.
21923 (LOOP_VINFO_CHECK_NONZERO): New macro.
21924 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
21925 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
21926 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
21927 fields. Make seg_len the distance travelled, not including the
21929 (dr_direction_indicator): Declare.
21930 (dr_zero_step_indicator): Likewise.
21931 (dr_known_forward_stride_p): Likewise.
21932 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
21934 (runtime_alias_check_p): Allow runtime alias checks with
21936 (operator ==): Compare access_size and align.
21937 (prune_runtime_alias_test_list): Rework for new distinction between
21938 the access_size and seg_len.
21939 (create_intersect_range_checks_index): Likewise. Cope with polynomial
21941 (get_segment_min_max): New function.
21942 (create_intersect_range_checks): Use it.
21943 (dr_step_indicator): New function.
21944 (dr_direction_indicator): Likewise.
21945 (dr_zero_step_indicator): Likewise.
21946 (dr_known_forward_stride_p): Likewise.
21947 * tree-loop-distribution.c (data_ref_segment_size): Return
21948 DR_STEP * (niters - 1).
21949 (compute_alias_check_pairs): Update call to the dr_with_seg_len
21951 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
21952 (vect_preserves_scalar_order_p): New function, split out from...
21953 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
21954 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
21955 (vect_vfa_access_size): New function.
21956 (vect_vfa_align): Likewise.
21957 (vect_compile_time_alias): Take access_size_a and access_b arguments.
21958 (dump_lower_bound): New function.
21959 (vect_check_lower_bound): Likewise.
21960 (vect_small_gap_p): Likewise.
21961 (vectorizable_with_step_bound_p): Likewise.
21962 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
21963 depencies if the vectorization factor is 1. Convert the checks
21964 for nonzero steps into checks on the bounds of DR_STEP. Try using
21965 a bunds check for variable steps if the minimum required step is
21966 relatively small. Update calls to the dr_with_seg_len
21967 constructor and to vect_compile_time_alias.
21968 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
21970 (vect_loop_versioning): Call it.
21971 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
21973 (vect_estimate_min_profitable_iters): Account for any bounds checks.
21975 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21976 Alan Hayward <alan.hayward@arm.com>
21977 David Sherwood <david.sherwood@arm.com>
21979 * doc/sourcebuild.texi (vect_scatter_store): Document.
21980 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
21982 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
21984 * genopinit.c (main): Add supports_vec_scatter_store and
21985 supports_vec_scatter_store_cached to target_optabs.
21986 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
21987 IFN_MASK_SCATTER_STORE.
21988 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
21990 * internal-fn.h (internal_store_fn_p): Declare.
21991 (internal_fn_stored_value_index): Likewise.
21992 * internal-fn.c (scatter_store_direct): New macro.
21993 (expand_scatter_store_optab_fn): New function.
21994 (direct_scatter_store_optab_supported_p): New macro.
21995 (internal_store_fn_p): New function.
21996 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
21997 IFN_MASK_SCATTER_STORE.
21998 (internal_fn_mask_index): Likewise.
21999 (internal_fn_stored_value_index): New function.
22000 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
22001 for scatter stores.
22002 * optabs-query.h (supports_vec_scatter_store_p): Declare.
22003 * optabs-query.c (supports_vec_scatter_store_p): New function.
22004 * tree-vectorizer.h (vect_get_store_rhs): Declare.
22005 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
22006 true for scatter stores.
22007 (vect_gather_scatter_fn_p): Handle scatter stores too.
22008 (vect_check_gather_scatter): Consider using scatter stores if
22009 supports_vec_scatter_store_p.
22010 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
22011 scatter stores too.
22012 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
22013 internal_fn_stored_value_index.
22014 (check_load_store_masking): Handle scatter stores too.
22015 (vect_get_store_rhs): Make public.
22016 (vectorizable_call): Use internal_store_fn_p.
22017 (vectorizable_store): Handle scatter store internal functions.
22018 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
22019 when deciding whether the end of the group has been reached.
22020 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
22021 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
22022 (mask_scatter_store<mode>): New insns.
22024 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22025 Alan Hayward <alan.hayward@arm.com>
22026 David Sherwood <david.sherwood@arm.com>
22028 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
22029 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
22030 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
22032 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
22033 Use vect_truncate_gather_scatter_offset if we can't treat the
22034 operation as a normal gather load or scatter store.
22035 (get_group_load_store_type): Take the gather_scatter_info
22036 as argument. Try using a gather load or scatter store for
22037 single-element groups.
22038 (get_load_store_type): Update calls to get_group_load_store_type
22039 and vect_use_strided_gather_scatters_p.
22041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22042 Alan Hayward <alan.hayward@arm.com>
22043 David Sherwood <david.sherwood@arm.com>
22045 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
22046 optional tree argument.
22047 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
22049 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
22050 but continue to use the current value as a fallback.
22051 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
22052 to compare the updates.
22053 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
22054 (get_load_store_type): Use it when handling a strided access.
22055 (vect_get_strided_load_store_ops): New function.
22056 (vect_get_data_ptr_increment): Likewise.
22057 (vectorizable_load): Handle strided gather loads. Always pass
22058 a step to vect_create_data_ref_ptr and bump_vector_ptr.
22060 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22061 Alan Hayward <alan.hayward@arm.com>
22062 David Sherwood <david.sherwood@arm.com>
22064 * doc/md.texi (gather_load@var{m}): Document.
22065 (mask_gather_load@var{m}): Likewise.
22066 * genopinit.c (main): Add supports_vec_gather_load and
22067 supports_vec_gather_load_cached to target_optabs.
22068 * optabs-tree.c (init_tree_optimization_optabs): Use
22069 ggc_cleared_alloc to allocate target_optabs.
22070 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
22071 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
22073 * internal-fn.h (internal_load_fn_p): Declare.
22074 (internal_gather_scatter_fn_p): Likewise.
22075 (internal_fn_mask_index): Likewise.
22076 (internal_gather_scatter_fn_supported_p): Likewise.
22077 * internal-fn.c (gather_load_direct): New macro.
22078 (expand_gather_load_optab_fn): New function.
22079 (direct_gather_load_optab_supported_p): New macro.
22080 (direct_internal_fn_optab): New function.
22081 (internal_load_fn_p): Likewise.
22082 (internal_gather_scatter_fn_p): Likewise.
22083 (internal_fn_mask_index): Likewise.
22084 (internal_gather_scatter_fn_supported_p): Likewise.
22085 * optabs-query.c (supports_at_least_one_mode_p): New function.
22086 (supports_vec_gather_load_p): Likewise.
22087 * optabs-query.h (supports_vec_gather_load_p): Declare.
22088 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
22089 and memory_type field.
22090 (NUM_PATTERNS): Bump to 15.
22091 * tree-vect-data-refs.c: Include internal-fn.h.
22092 (vect_gather_scatter_fn_p): New function.
22093 (vect_describe_gather_scatter_call): Likewise.
22094 (vect_check_gather_scatter): Try using internal functions for
22095 gather loads. Recognize existing calls to a gather load function.
22096 (vect_analyze_data_refs): Consider using gather loads if
22097 supports_vec_gather_load_p.
22098 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
22099 (vect_get_gather_scatter_offset_type): Likewise.
22100 (vect_convert_mask_for_vectype): Likewise.
22101 (vect_add_conversion_to_patterm): Likewise.
22102 (vect_try_gather_scatter_pattern): Likewise.
22103 (vect_recog_gather_scatter_pattern): New pattern recognizer.
22104 (vect_vect_recog_func_ptrs): Add it.
22105 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
22106 internal_fn_mask_index and internal_gather_scatter_fn_p.
22107 (check_load_store_masking): Take the gather_scatter_info as an
22108 argument and handle gather loads.
22109 (vect_get_gather_scatter_ops): New function.
22110 (vectorizable_call): Check internal_load_fn_p.
22111 (vectorizable_load): Likewise. Handle gather load internal
22113 (vectorizable_store): Update call to check_load_store_masking.
22114 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
22115 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
22116 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
22117 (aarch64_gather_scale_operand_d): New predicates.
22118 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
22119 (mask_gather_load<mode>): New insns.
22121 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22122 Alan Hayward <alan.hayward@arm.com>
22123 David Sherwood <david.sherwood@arm.com>
22125 * optabs.def (fold_left_plus_optab): New optab.
22126 * doc/md.texi (fold_left_plus_@var{m}): Document.
22127 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
22128 * internal-fn.c (fold_left_direct): Define.
22129 (expand_fold_left_optab_fn): Likewise.
22130 (direct_fold_left_optab_supported_p): Likewise.
22131 * fold-const-call.c (fold_const_fold_left): New function.
22132 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
22133 * tree-parloops.c (valid_reduction_p): New function.
22134 (gather_scalar_reductions): Use it.
22135 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
22136 (vect_finish_replace_stmt): Declare.
22137 * tree-vect-loop.c (fold_left_reduction_fn): New function.
22138 (needs_fold_left_reduction_p): New function, split out from...
22139 (vect_is_simple_reduction): ...here. Accept reductions that
22140 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
22141 (vect_force_simple_reduction): Also store the reduction type in
22142 the assignment's STMT_VINFO_REDUC_TYPE.
22143 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
22144 (merge_with_identity): New function.
22145 (vect_expand_fold_left): Likewise.
22146 (vectorize_fold_left_reduction): Likewise.
22147 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
22148 scalar phi in place for it. Check for target support and reject
22149 cases that would reassociate the operation. Defer the transform
22150 phase to vectorize_fold_left_reduction.
22151 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
22152 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
22153 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
22155 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22157 * tree-if-conv.c (predicate_mem_writes): Remove redundant
22158 call to ifc_temp_var.
22160 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22161 Alan Hayward <alan.hayward@arm.com>
22162 David Sherwood <david.sherwood@arm.com>
22164 * target.def (legitimize_address_displacement): Take the original
22165 offset as a poly_int.
22166 * targhooks.h (default_legitimize_address_displacement): Update
22168 * targhooks.c (default_legitimize_address_displacement): Likewise.
22169 * doc/tm.texi: Regenerate.
22170 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
22171 as an argument, moving assert of ad->disp == ad->disp_term to...
22172 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
22173 Try calling targetm.legitimize_address_displacement before expanding
22174 the address rather than afterwards, and adjust for the new interface.
22175 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
22176 Match the new hook interface. Handle SVE addresses.
22177 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
22178 new hook interface.
22180 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22182 * Makefile.in (OBJS): Add early-remat.o.
22183 * target.def (select_early_remat_modes): New hook.
22184 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
22185 * doc/tm.texi: Regenerate.
22186 * targhooks.h (default_select_early_remat_modes): Declare.
22187 * targhooks.c (default_select_early_remat_modes): New function.
22188 * timevar.def (TV_EARLY_REMAT): New timevar.
22189 * passes.def (pass_early_remat): New pass.
22190 * tree-pass.h (make_pass_early_remat): Declare.
22191 * early-remat.c: New file.
22192 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
22194 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
22196 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22197 Alan Hayward <alan.hayward@arm.com>
22198 David Sherwood <david.sherwood@arm.com>
22200 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
22201 vfm1 with a bound_epilog parameter.
22202 (vect_do_peeling): Update calls accordingly, and move the prologue
22203 call earlier in the function. Treat the base bound_epilog as 0 for
22204 fully-masked loops and retain vf - 1 for other loops. Add 1 to
22205 this base when peeling for gaps.
22206 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
22207 with fully-masked loops.
22208 (vect_estimate_min_profitable_iters): Handle the single peeled
22209 iteration in that case.
22211 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22212 Alan Hayward <alan.hayward@arm.com>
22213 David Sherwood <david.sherwood@arm.com>
22215 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
22216 single-element interleaving even if the size is not a power of 2.
22217 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
22218 accesses for single-element interleaving if the group size is
22221 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22222 Alan Hayward <alan.hayward@arm.com>
22223 David Sherwood <david.sherwood@arm.com>
22225 * doc/md.texi (fold_extract_last_@var{m}): Document.
22226 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
22227 * optabs.def (fold_extract_last_optab): New optab.
22228 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
22229 * internal-fn.c (fold_extract_direct): New macro.
22230 (expand_fold_extract_optab_fn): Likewise.
22231 (direct_fold_extract_optab_supported_p): Likewise.
22232 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
22233 * tree-vect-loop.c (vect_model_reduction_cost): Handle
22234 EXTRACT_LAST_REDUCTION.
22235 (get_initial_def_for_reduction): Do not create an initial vector
22236 for EXTRACT_LAST_REDUCTION reductions.
22237 (vectorizable_reduction): Leave the scalar phi in place for
22238 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
22239 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
22240 epilogue code for EXTRACT_LAST_REDUCTION and defer the
22241 transform phase to vectorizable_condition.
22242 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
22244 (vect_finish_stmt_generation): ...here.
22245 (vect_finish_replace_stmt): New function.
22246 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
22247 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
22249 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
22251 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22252 Alan Hayward <alan.hayward@arm.com>
22253 David Sherwood <david.sherwood@arm.com>
22255 * doc/md.texi (extract_last_@var{m}): Document.
22256 * optabs.def (extract_last_optab): New optab.
22257 * internal-fn.def (EXTRACT_LAST): New internal function.
22258 * internal-fn.c (cond_unary_direct): New macro.
22259 (expand_cond_unary_optab_fn): Likewise.
22260 (direct_cond_unary_optab_supported_p): Likewise.
22261 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
22262 loops using EXTRACT_LAST.
22263 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
22264 (extract_last_<mode>): ...this optab.
22265 (vec_extract<mode><Vel>): Update accordingly.
22267 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22268 Alan Hayward <alan.hayward@arm.com>
22269 David Sherwood <david.sherwood@arm.com>
22271 * target.def (empty_mask_is_expensive): New hook.
22272 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
22273 * doc/tm.texi: Regenerate.
22274 * targhooks.h (default_empty_mask_is_expensive): Declare.
22275 * targhooks.c (default_empty_mask_is_expensive): New function.
22276 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
22277 if the target says that empty masks are expensive.
22278 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
22280 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
22282 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22283 Alan Hayward <alan.hayward@arm.com>
22284 David Sherwood <david.sherwood@arm.com>
22286 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
22287 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
22288 (vect_use_loop_mask_for_alignment_p): New function.
22289 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
22290 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
22291 niters_skip argument. Make sure that the first niters_skip elements
22292 of the first iteration are inactive.
22293 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
22294 Update call to vect_set_loop_masks_directly.
22295 (get_misalign_in_elems): New function, split out from...
22296 (vect_gen_prolog_loop_niters): ...here.
22297 (vect_update_init_of_dr): Take a code argument that specifies whether
22298 the adjustment should be added or subtracted.
22299 (vect_update_init_of_drs): Likewise.
22300 (vect_prepare_for_masked_peels): New function.
22301 (vect_do_peeling): Skip prologue peeling if we're using a mask
22302 instead. Update call to vect_update_inits_of_drs.
22303 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22305 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
22306 alignment. Do not include the number of peeled iterations in
22307 the minimum threshold in that case.
22308 (vectorizable_induction): Adjust the start value down by
22309 LOOP_VINFO_MASK_SKIP_NITERS iterations.
22310 (vect_transform_loop): Call vect_prepare_for_masked_peels.
22311 Take the number of skipped iterations into account when calculating
22313 * tree-vect-stmts.c (vect_gen_while_not): New function.
22315 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22316 Alan Hayward <alan.hayward@arm.com>
22317 David Sherwood <david.sherwood@arm.com>
22319 * doc/sourcebuild.texi (vect_fully_masked): Document.
22320 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
22321 default value to 0.
22322 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
22324 (vect_analyze_loop_2): ...here. Don't check the vectorization
22325 factor against the number of loop iterations if the loop is
22328 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22329 Alan Hayward <alan.hayward@arm.com>
22330 David Sherwood <david.sherwood@arm.com>
22332 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
22333 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
22334 (dump_groups): Update accordingly.
22335 (iv_use::mem_type): New member variable.
22336 (address_p): New function.
22337 (record_use): Add a mem_type argument and initialize the new
22339 (record_group_use): Add a mem_type argument. Use address_p.
22340 Remove obsolete null checks of base_object. Update call to record_use.
22341 (find_interesting_uses_op): Update call to record_group_use.
22342 (find_interesting_uses_cond): Likewise.
22343 (find_interesting_uses_address): Likewise.
22344 (get_mem_type_for_internal_fn): New function.
22345 (find_address_like_use): Likewise.
22346 (find_interesting_uses_stmt): Try find_address_like_use before
22347 calling find_interesting_uses_op.
22348 (addr_offset_valid_p): Use the iv mem_type field as the type
22349 of the addressed memory.
22350 (add_autoinc_candidates): Likewise.
22351 (get_address_cost): Likewise.
22352 (split_small_address_groups_p): Use address_p.
22353 (split_address_groups): Likewise.
22354 (add_iv_candidate_for_use): Likewise.
22355 (autoinc_possible_for_pair): Likewise.
22356 (rewrite_groups): Likewise.
22357 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
22358 (determine_group_iv_cost): Update after split of USE_ADDRESS.
22359 (get_alias_ptr_type_for_ptr_address): New function.
22360 (rewrite_use_address): Rewrite address uses in calls that were
22361 identified by find_address_like_use.
22363 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22364 Alan Hayward <alan.hayward@arm.com>
22365 David Sherwood <david.sherwood@arm.com>
22367 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
22369 * gimple-expr.h (is_gimple_addressable: Likewise.
22370 * gimple-expr.c (is_gimple_address): Likewise.
22371 * internal-fn.c (expand_call_mem_ref): New function.
22372 (expand_mask_load_optab_fn): Use it.
22373 (expand_mask_store_optab_fn): Likewise.
22375 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22376 Alan Hayward <alan.hayward@arm.com>
22377 David Sherwood <david.sherwood@arm.com>
22379 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
22380 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
22381 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
22382 (cond_umax@var{mode}): Document.
22383 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
22384 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
22385 (cond_umin_optab, cond_umax_optab): New optabs.
22386 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
22387 (COND_IOR, COND_XOR): New internal functions.
22388 * internal-fn.h (get_conditional_internal_fn): Declare.
22389 * internal-fn.c (cond_binary_direct): New macro.
22390 (expand_cond_binary_optab_fn): Likewise.
22391 (direct_cond_binary_optab_supported_p): Likewise.
22392 (get_conditional_internal_fn): New function.
22393 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
22394 Cope with reduction statements that are vectorized as calls rather
22396 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
22397 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
22398 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
22399 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
22400 (UNSPEC_COND_EOR): New unspecs.
22401 (optab): Add mappings for them.
22402 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
22403 (sve_int_op, sve_fp_op): New int attributes.
22405 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22406 Alan Hayward <alan.hayward@arm.com>
22407 David Sherwood <david.sherwood@arm.com>
22409 * optabs.def (while_ult_optab): New optab.
22410 * doc/md.texi (while_ult@var{m}@var{n}): Document.
22411 * internal-fn.def (WHILE_ULT): New internal function.
22412 * internal-fn.h (direct_internal_fn_supported_p): New override
22413 that takes two types as argument.
22414 * internal-fn.c (while_direct): New macro.
22415 (expand_while_optab_fn): New function.
22416 (convert_optab_supported_p): Likewise.
22417 (direct_while_optab_supported_p): New macro.
22418 * wide-int.h (wi::udiv_ceil): New function.
22419 * tree-vectorizer.h (rgroup_masks): New structure.
22420 (vec_loop_masks): New typedef.
22421 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
22422 and fully_masked_p.
22423 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
22424 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
22425 (vect_max_vf): New function.
22426 (slpeel_make_loop_iterate_ntimes): Delete.
22427 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
22428 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
22429 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
22430 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
22431 internal-fn.h, stor-layout.h and optabs-query.h.
22432 (vect_set_loop_mask): New function.
22433 (add_preheader_seq): Likewise.
22434 (add_header_seq): Likewise.
22435 (interleave_supported_p): Likewise.
22436 (vect_maybe_permute_loop_masks): Likewise.
22437 (vect_set_loop_masks_directly): Likewise.
22438 (vect_set_loop_condition_masked): Likewise.
22439 (vect_set_loop_condition_unmasked): New function, split out from
22440 slpeel_make_loop_iterate_ntimes.
22441 (slpeel_make_loop_iterate_ntimes): Rename to..
22442 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
22443 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
22444 (vect_do_peeling): Update call accordingly.
22445 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
22447 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22448 mask_compare_type, can_fully_mask_p and fully_masked_p.
22449 (release_vec_loop_masks): New function.
22450 (_loop_vec_info): Use it to free the loop masks.
22451 (can_produce_all_loop_masks_p): New function.
22452 (vect_get_max_nscalars_per_iter): Likewise.
22453 (vect_verify_full_masking): Likewise.
22454 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
22455 retries, and free the mask rgroups before retrying. Check loop-wide
22456 reasons for disallowing fully-masked loops. Make the final decision
22457 about whether use a fully-masked loop or not.
22458 (vect_estimate_min_profitable_iters): Do not assume that peeling
22459 for the number of iterations will be needed for fully-masked loops.
22460 (vectorizable_reduction): Disable fully-masked loops.
22461 (vectorizable_live_operation): Likewise.
22462 (vect_halve_mask_nunits): New function.
22463 (vect_double_mask_nunits): Likewise.
22464 (vect_record_loop_mask): Likewise.
22465 (vect_get_loop_mask): Likewise.
22466 (vect_transform_loop): Handle the case in which the final loop
22467 iteration might handle a partial vector. Call vect_set_loop_condition
22468 instead of slpeel_make_loop_iterate_ntimes.
22469 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
22470 (check_load_store_masking): New function.
22471 (prepare_load_store_mask): Likewise.
22472 (vectorizable_store): Handle fully-masked loops.
22473 (vectorizable_load): Likewise.
22474 (supportable_widening_operation): Use vect_halve_mask_nunits for
22476 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
22477 (vect_gen_while): New function.
22478 * config/aarch64/aarch64.md (umax<mode>3): New expander.
22479 (aarch64_uqdec<mode>): New insn.
22481 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22482 Alan Hayward <alan.hayward@arm.com>
22483 David Sherwood <david.sherwood@arm.com>
22485 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
22486 (reduc_xor_scal_optab): New optabs.
22487 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
22488 (reduc_xor_scal_@var{m}): Document.
22489 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
22490 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
22491 internal functions.
22492 * fold-const-call.c (fold_const_call): Handle them.
22493 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
22494 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
22495 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
22496 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
22497 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
22498 (UNSPEC_XORV): New unspecs.
22499 (optab): Add entries for them.
22500 (BITWISEV): New int iterator.
22501 (bit_reduc_op): New int attributes.
22503 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22504 Alan Hayward <alan.hayward@arm.com>
22505 David Sherwood <david.sherwood@arm.com>
22507 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
22508 * internal-fn.def (VEC_SHL_INSERT): New internal function.
22509 * optabs.def (vec_shl_insert_optab): New optab.
22510 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
22511 (duplicate_and_interleave): Likewise.
22512 * tree-vect-loop.c: Include internal-fn.h.
22513 (neutral_op_for_slp_reduction): New function, split out from
22514 get_initial_defs_for_reduction.
22515 (get_initial_def_for_reduction): Handle option 2 for variable-length
22516 vectors by loading the neutral value into a vector and then shifting
22517 the initial value into element 0.
22518 (get_initial_defs_for_reduction): Replace the code argument with
22519 the neutral value calculated by neutral_op_for_slp_reduction.
22520 Use gimple_build_vector for constant-length vectors.
22521 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
22522 but the first group_size elements have a neutral value.
22523 Use duplicate_and_interleave otherwise.
22524 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
22525 Update call to get_initial_defs_for_reduction. Handle SLP
22526 reductions for variable-length vectors by creating one vector
22527 result for each scalar result, with the elements associated
22528 with other scalar results stubbed out with the neutral value.
22529 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
22530 Require IFN_VEC_SHL_INSERT for double reductions on
22531 variable-length vectors, or SLP reductions that have
22532 a neutral value. Require can_duplicate_and_interleave_p
22533 support for variable-length unchained SLP reductions if there
22534 is no neutral value, such as for MIN/MAX reductions. Also require
22535 the number of vector elements to be a multiple of the number of
22536 SLP statements when doing variable-length unchained SLP reductions.
22537 Update call to vect_create_epilog_for_reduction.
22538 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
22539 and remove initial values.
22540 (duplicate_and_interleave): Make public.
22541 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
22542 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
22544 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22545 Alan Hayward <alan.hayward@arm.com>
22546 David Sherwood <david.sherwood@arm.com>
22548 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
22549 (can_duplicate_and_interleave_p): New function.
22550 (vect_get_and_check_slp_defs): Take the vector of statements
22551 rather than just the current one. Remove excess parentheses.
22552 Restriction rejectinon of vect_constant_def and vect_external_def
22553 for variable-length vectors to boolean types, or types for which
22554 can_duplicate_and_interleave_p is false.
22555 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
22556 (duplicate_and_interleave): New function.
22557 (vect_get_constant_vectors): Use gimple_build_vector for
22558 constant-length vectors and suitable variable-length constant
22559 vectors. Use duplicate_and_interleave for other variable-length
22560 vectors. Don't defer the update when inserting new statements.
22562 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22563 Alan Hayward <alan.hayward@arm.com>
22564 David Sherwood <david.sherwood@arm.com>
22566 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
22567 min_profitable_iters doesn't go negative.
22569 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22570 Alan Hayward <alan.hayward@arm.com>
22571 David Sherwood <david.sherwood@arm.com>
22573 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
22574 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
22575 * optabs.def (vec_mask_load_lanes_optab): New optab.
22576 (vec_mask_store_lanes_optab): Likewise.
22577 * internal-fn.def (MASK_LOAD_LANES): New internal function.
22578 (MASK_STORE_LANES): Likewise.
22579 * internal-fn.c (mask_load_lanes_direct): New macro.
22580 (mask_store_lanes_direct): Likewise.
22581 (expand_mask_load_optab_fn): Handle masked operations.
22582 (expand_mask_load_lanes_optab_fn): New macro.
22583 (expand_mask_store_optab_fn): Handle masked operations.
22584 (expand_mask_store_lanes_optab_fn): New macro.
22585 (direct_mask_load_lanes_optab_supported_p): Likewise.
22586 (direct_mask_store_lanes_optab_supported_p): Likewise.
22587 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
22589 (vect_load_lanes_supported): Likewise.
22590 * tree-vect-data-refs.c (strip_conversion): New function.
22591 (can_group_stmts_p): Likewise.
22592 (vect_analyze_data_ref_accesses): Use it instead of checking
22593 for a pair of assignments.
22594 (vect_store_lanes_supported): Take a masked_p parameter.
22595 (vect_load_lanes_supported): Likewise.
22596 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
22597 vect_store_lanes_supported and vect_load_lanes_supported.
22598 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
22599 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
22600 parameter. Don't allow gaps for masked accesses.
22601 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
22602 and vect_load_lanes_supported.
22603 (get_load_store_type): Take a masked_p parameter and update
22604 call to get_group_load_store_type.
22605 (vectorizable_store): Update call to get_load_store_type.
22606 Handle IFN_MASK_STORE_LANES.
22607 (vectorizable_load): Update call to get_load_store_type.
22608 Handle IFN_MASK_LOAD_LANES.
22610 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22611 Alan Hayward <alan.hayward@arm.com>
22612 David Sherwood <david.sherwood@arm.com>
22614 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
22616 * config/aarch64/aarch64-protos.h
22617 (aarch64_sve_struct_memory_operand_p): Declare.
22618 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
22619 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
22620 (VPRED, vpred): Handle SVE structure modes.
22621 * config/aarch64/constraints.md (Utx): New constraint.
22622 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
22623 (aarch64_sve_struct_nonimmediate_operand): New predicates.
22624 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
22625 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
22626 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
22627 structure modes. Split into pieces after RA.
22628 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
22629 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
22631 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
22632 SVE structure modes.
22633 (aarch64_classify_address): Likewise.
22634 (sizetochar): Move earlier in file.
22635 (aarch64_print_operand): Handle SVE register lists.
22636 (aarch64_array_mode): New function.
22637 (aarch64_sve_struct_memory_operand_p): Likewise.
22638 (TARGET_ARRAY_MODE): Redefine.
22640 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22641 Alan Hayward <alan.hayward@arm.com>
22642 David Sherwood <david.sherwood@arm.com>
22644 * target.def (array_mode): New target hook.
22645 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
22646 * doc/tm.texi: Regenerate.
22647 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
22648 * hooks.c (hook_optmode_mode_uhwi_none): New function.
22649 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
22650 targetm.array_mode.
22651 * stor-layout.c (mode_for_array): Likewise. Support polynomial
22654 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22655 Alan Hayward <alan.hayward@arm.com>
22656 David Sherwood <david.sherwood@arm.com>
22658 * fold-const.c (fold_binary_loc): Check the argument types
22659 rather than the result type when testing for a vector operation.
22661 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22663 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
22664 * doc/tm.texi: Regenerate.
22666 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22667 Alan Hayward <alan.hayward@arm.com>
22668 David Sherwood <david.sherwood@arm.com>
22670 * doc/invoke.texi (-msve-vector-bits=): Document new option.
22671 (sve): Document new AArch64 extension.
22672 * doc/md.texi (w): Extend the description of the AArch64
22673 constraint to include SVE vectors.
22674 (Upl, Upa): Document new AArch64 predicate constraints.
22675 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
22677 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
22678 (msve-vector-bits=): New option.
22679 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
22680 SVE when these are disabled.
22681 (sve): New extension.
22682 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
22683 modes. Adjust their number of units based on aarch64_sve_vg.
22684 (MAX_BITSIZE_MODE_ANY_MODE): Define.
22685 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
22686 aarch64_addr_query_type.
22687 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
22688 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
22689 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
22690 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
22691 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
22692 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
22693 (aarch64_simd_imm_zero_p): Delete.
22694 (aarch64_check_zero_based_sve_index_immediate): Declare.
22695 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22696 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22697 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22698 (aarch64_sve_float_mul_immediate_p): Likewise.
22699 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22700 rather than an rtx.
22701 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
22702 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
22703 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
22704 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
22705 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
22706 (aarch64_regmode_natural_size): Likewise.
22707 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
22708 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
22710 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
22711 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
22712 for VG and the SVE predicate registers.
22713 (V_ALIASES): Add a "z"-prefixed alias.
22714 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
22715 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
22716 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
22717 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
22718 (REG_CLASS_NAMES): Add entries for them.
22719 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
22720 and the predicate registers.
22721 (aarch64_sve_vg): Declare.
22722 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
22723 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
22724 (REGMODE_NATURAL_SIZE): Define.
22725 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
22727 * config/aarch64/aarch64.c: Include cfgrtl.h.
22728 (simd_immediate_info): Add a constructor for series vectors,
22729 and an associated step field.
22730 (aarch64_sve_vg): New variable.
22731 (aarch64_dbx_register_number): Handle VG and the predicate registers.
22732 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
22733 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
22734 (VEC_ANY_DATA, VEC_STRUCT): New constants.
22735 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
22736 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
22737 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
22738 (aarch64_get_mask_mode): New functions.
22739 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
22740 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
22741 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
22742 predicate modes and predicate registers. Explicitly restrict
22743 GPRs to modes of 16 bytes or smaller. Only allow FP registers
22744 to store a vector mode if it is recognized by
22745 aarch64_classify_vector_mode.
22746 (aarch64_regmode_natural_size): New function.
22747 (aarch64_hard_regno_caller_save_mode): Return the original mode
22749 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
22750 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
22751 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
22752 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
22754 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
22755 does not overlap dest if the function is frame-related. Handle
22757 (aarch64_split_add_offset): New function.
22758 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
22759 them aarch64_add_offset.
22760 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
22761 and update call to aarch64_sub_sp.
22762 (aarch64_add_cfa_expression): New function.
22763 (aarch64_expand_prologue): Pass extra temporary registers to the
22764 functions above. Handle the case in which we need to emit new
22765 DW_CFA_expressions for registers that were originally saved
22766 relative to the stack pointer, but now have to be expressed
22767 relative to the frame pointer.
22768 (aarch64_output_mi_thunk): Pass extra temporary registers to the
22770 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
22771 IP0 and IP1 values for SVE frames.
22772 (aarch64_expand_vec_series): New function.
22773 (aarch64_expand_sve_widened_duplicate): Likewise.
22774 (aarch64_expand_sve_const_vector): Likewise.
22775 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
22776 Handle SVE constants. Use emit_move_insn to move a force_const_mem
22777 into the register, rather than emitting a SET directly.
22778 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
22779 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
22780 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
22781 (offset_9bit_signed_scaled_p): New functions.
22782 (aarch64_replicate_bitmask_imm): New function.
22783 (aarch64_bitmask_imm): Use it.
22784 (aarch64_cannot_force_const_mem): Reject expressions involving
22785 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
22786 (aarch64_classify_index): Handle SVE indices, by requiring
22787 a plain register index with a scale that matches the element size.
22788 (aarch64_classify_address): Handle SVE addresses. Assert that
22789 the mode of the address is VOIDmode or an integer mode.
22790 Update call to aarch64_classify_symbol.
22791 (aarch64_classify_symbolic_expression): Update call to
22792 aarch64_classify_symbol.
22793 (aarch64_const_vec_all_in_range_p): New function.
22794 (aarch64_print_vector_float_operand): Likewise.
22795 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
22796 "vN" for FP registers with SVE modes. Handle (const ...) vectors
22797 and the FP immediates 1.0 and 0.5.
22798 (aarch64_print_address_internal): Handle SVE addresses.
22799 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
22800 (aarch64_regno_regclass): Handle predicate registers.
22801 (aarch64_secondary_reload): Handle big-endian reloads of SVE
22803 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
22804 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
22805 (aarch64_convert_sve_vector_bits): New function.
22806 (aarch64_override_options): Use it to handle -msve-vector-bits=.
22807 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22808 rather than an rtx.
22809 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
22810 Handle SVE vector and predicate modes. Accept VL-based constants
22811 that need only one temporary register, and VL offsets that require
22812 no temporary registers.
22813 (aarch64_conditional_register_usage): Mark the predicate registers
22814 as fixed if SVE isn't available.
22815 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
22816 Return true for SVE vector and predicate modes.
22817 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
22818 rather than an unsigned int. Handle SVE modes.
22819 (aarch64_preferred_simd_mode): Update call accordingly. Handle
22821 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
22823 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22824 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22825 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22826 (aarch64_sve_float_mul_immediate_p): New functions.
22827 (aarch64_sve_valid_immediate): New function.
22828 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
22829 Explicitly reject structure modes. Check for INDEX constants.
22830 Handle PTRUE and PFALSE constants.
22831 (aarch64_check_zero_based_sve_index_immediate): New function.
22832 (aarch64_simd_imm_zero_p): Delete.
22833 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
22834 vector modes. Accept constants in the range of CNT[BHWD].
22835 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
22836 ask for an Advanced SIMD mode.
22837 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
22838 (aarch64_simd_vector_alignment): Handle SVE predicates.
22839 (aarch64_vectorize_preferred_vector_alignment): New function.
22840 (aarch64_simd_vector_alignment_reachable): Use it instead of
22842 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
22843 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
22845 (MAX_VECT_LEN): Delete.
22846 (expand_vec_perm_d): Add a vec_flags field.
22847 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
22848 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
22849 (aarch64_evpc_ext): Don't apply a big-endian lane correction
22851 (aarch64_evpc_rev): Rename to...
22852 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
22853 (aarch64_evpc_rev_global): New function.
22854 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
22855 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
22857 (aarch64_evpc_sve_tbl): New function.
22858 (aarch64_expand_vec_perm_const_1): Update after rename of
22859 aarch64_evpc_rev. Handle SVE permutes too, trying
22860 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
22861 than aarch64_evpc_tbl.
22862 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
22863 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
22864 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
22865 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
22866 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
22867 (aarch64_expand_sve_vcond): New functions.
22868 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
22869 of aarch64_vector_mode_p.
22870 (aarch64_dwarf_poly_indeterminate_value): New function.
22871 (aarch64_compute_pressure_classes): Likewise.
22872 (aarch64_can_change_mode_class): Likewise.
22873 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
22874 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
22875 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
22876 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
22877 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
22878 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
22879 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
22880 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
22882 (Dn, Dl, Dr): Accept const as well as const_vector.
22883 (Dz): Likewise. Compare against CONST0_RTX.
22884 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
22885 of "vector" where appropriate.
22886 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
22887 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
22888 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
22889 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
22890 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
22891 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
22892 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
22893 (v_int_equiv): Extend to SVE modes.
22894 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
22896 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
22897 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
22898 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
22899 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
22900 (SVE_COND_FP_CMP): New int iterators.
22901 (perm_hilo): Handle the new unpack unspecs.
22902 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
22904 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
22905 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
22906 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
22907 (aarch64_equality_operator, aarch64_constant_vector_operand)
22908 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
22909 (aarch64_sve_nonimmediate_operand): Likewise.
22910 (aarch64_sve_general_operand): Likewise.
22911 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
22912 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
22913 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
22914 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
22915 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
22916 (aarch64_sve_float_arith_immediate): Likewise.
22917 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
22918 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
22919 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
22920 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
22921 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
22922 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
22923 (aarch64_sve_float_arith_operand): Likewise.
22924 (aarch64_sve_float_arith_with_sub_operand): Likewise.
22925 (aarch64_sve_float_mul_operand): Likewise.
22926 (aarch64_sve_vec_perm_operand): Likewise.
22927 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
22928 (aarch64_mov_operand): Accept const_poly_int and const_vector.
22929 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
22930 as well as const_vector.
22931 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
22932 in file. Use CONST0_RTX and CONSTM1_RTX.
22933 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
22934 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
22935 Use aarch64_simd_imm_zero.
22936 * config/aarch64/aarch64-sve.md: New file.
22937 * config/aarch64/aarch64.md: Include it.
22938 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
22939 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
22940 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
22941 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
22942 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
22943 (sve): New attribute.
22944 (enabled): Disable instructions with the sve attribute unless
22946 (movqi, movhi): Pass CONST_POLY_INT operaneds through
22947 aarch64_expand_mov_immediate.
22948 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
22949 CNT[BHSD] immediates.
22950 (movti): Split CONST_POLY_INT moves into two halves.
22951 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
22952 Split additions that need a temporary here if the destination
22953 is the stack pointer.
22954 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
22955 (*add<mode>3_poly_1): New instruction.
22956 (set_clobber_cc): New expander.
22958 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22960 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
22961 parameter and use it instead of GET_MODE_SIZE (innermode). Use
22962 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
22963 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
22964 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
22965 Change innermode from fixed_mode_size to machine_mode.
22966 (simplify_subreg): Update call accordingly. Handle a constant-sized
22967 subreg of a variable-length CONST_VECTOR.
22969 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22970 Alan Hayward <alan.hayward@arm.com>
22971 David Sherwood <david.sherwood@arm.com>
22973 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
22974 (add_offset_to_base): New function, split out from...
22975 (create_mem_ref): ...here. When handling a scale other than 1,
22976 check first whether the address is valid without the offset.
22977 Add it into the base if so, leaving the index and scale as-is.
22979 2018-01-12 Jakub Jelinek <jakub@redhat.com>
22982 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
22983 fold_for_warn before checking if arg2 is INTEGER_CST.
22985 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
22987 * config/rs6000/predicates.md (load_multiple_operation): Delete.
22988 (store_multiple_operation): Delete.
22989 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
22990 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
22991 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
22992 guarded by TARGET_STRING.
22993 (rs6000_output_load_multiple): Delete.
22994 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
22995 OPTION_MASK_STRING / TARGET_STRING handling.
22996 (print_operand) <'N', 'O'>: Add comment that these are unused now.
22997 (const rs6000_opt_masks) <"string">: Change mask to 0.
22998 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
22999 (MASK_STRING): Delete.
23000 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
23002 (load_multiple): Delete.
23009 (store_multiple): Delete.
23016 (movmemsi_8reg): Delete.
23017 (corresponding unnamed define_insn): Delete.
23018 (movmemsi_6reg): Delete.
23019 (corresponding unnamed define_insn): Delete.
23020 (movmemsi_4reg): Delete.
23021 (corresponding unnamed define_insn): Delete.
23022 (movmemsi_2reg): Delete.
23023 (corresponding unnamed define_insn): Delete.
23024 (movmemsi_1reg): Delete.
23025 (corresponding unnamed define_insn): Delete.
23026 * config/rs6000/rs6000.opt (mno-string): New.
23027 (mstring): Replace by deprecation warning stub.
23028 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
23030 2018-01-12 Jakub Jelinek <jakub@redhat.com>
23032 * regrename.c (regrename_do_replace): If replacing the same
23033 reg multiple times, try to reuse last created gen_raw_REG.
23036 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
23037 main to workaround a bug in GDB.
23039 2018-01-12 Tom de Vries <tom@codesourcery.com>
23042 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
23044 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
23046 PR rtl-optimization/80481
23047 * ira-color.c (get_cap_member): New function.
23048 (allocnos_conflict_by_live_ranges_p): Use it.
23049 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
23050 (setup_slot_coalesced_allocno_live_ranges): Ditto.
23052 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
23055 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
23056 (*saddl_se_1): Ditto.
23057 (*ssubsi_1): Ditto.
23058 (*ssubl_se_1): Ditto.
23060 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
23062 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
23063 rather than wi::to_widest for DR_INITs.
23064 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
23065 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
23066 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
23068 (vect_analyze_group_access_1): Note that here.
23070 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
23072 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
23073 polynomial type sizes.
23075 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
23077 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
23078 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
23079 (gimple_add_tmp_var): Likewise.
23081 2018-01-12 Martin Liska <mliska@suse.cz>
23083 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
23084 (gimple_alloc_sizes): Likewise.
23085 (dump_gimple_statistics): Use PRIu64 in printf format.
23086 * gimple.h: Change uint64_t to int.
23088 2018-01-12 Martin Liska <mliska@suse.cz>
23090 * tree-core.h: Use uint64_t instead of int.
23091 * tree.c (tree_node_counts): Likewise.
23092 (tree_node_sizes): Likewise.
23093 (dump_tree_statistics): Use PRIu64 in printf format.
23095 2018-01-12 Martin Liska <mliska@suse.cz>
23097 * Makefile.in: As qsort_chk is implemented in vec.c, add
23098 vec.o to linkage of gencfn-macros.
23099 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
23100 passing the info to record_node_allocation_statistics.
23101 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
23103 * ggc-common.c (struct ggc_usage): Add operator== and use
23104 it in operator< and compare function.
23105 * mem-stats.h (struct mem_usage): Likewise.
23106 * vec.c (struct vec_usage): Remove operator< and compare
23107 function. Can be simply inherited.
23109 2018-01-12 Martin Jambor <mjambor@suse.cz>
23112 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
23113 * tree-ssa-math-opts.c: Include domwalk.h.
23114 (convert_mult_to_fma_1): New function.
23115 (fma_transformation_info): New type.
23116 (fma_deferring_state): Likewise.
23117 (cancel_fma_deferring): New function.
23118 (result_of_phi): Likewise.
23119 (last_fma_candidate_feeds_initial_phi): Likewise.
23120 (convert_mult_to_fma): Added deferring logic, split actual
23121 transformation to convert_mult_to_fma_1.
23122 (math_opts_dom_walker): New type.
23123 (math_opts_dom_walker::after_dom_children): New method, body moved
23124 here from pass_optimize_widening_mul::execute, added deferring logic
23126 (pass_optimize_widening_mul::execute): Moved most of code to
23127 math_opts_dom_walker::after_dom_children.
23128 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
23129 * config/i386/i386.c (ix86_option_override_internal): Added
23130 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
23132 2018-01-12 Richard Biener <rguenther@suse.de>
23135 * dwarf2out.c (gen_variable_die): Do not reset old_die for
23136 inline instance vars.
23138 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
23141 * config/rx/rx.c (rx_is_restricted_memory_address):
23142 Handle SUBREG case.
23144 2018-01-12 Richard Biener <rguenther@suse.de>
23146 PR tree-optimization/80846
23147 * target.def (split_reduction): New target hook.
23148 * targhooks.c (default_split_reduction): New function.
23149 * targhooks.h (default_split_reduction): Declare.
23150 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
23151 target requests first reduce vectors by combining low and high
23153 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
23154 (get_vectype_for_scalar_type_and_size): Export.
23155 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
23156 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
23157 * doc/tm.texi: Regenerate.
23158 * config/i386/i386.c (ix86_split_reduction): Implement
23159 TARGET_VECTORIZE_SPLIT_REDUCTION.
23161 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
23164 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
23165 in PIC mode except for TARGET_VXWORKS_RTP.
23166 * config/sparc/sparc.c: Include cfgrtl.h.
23167 (TARGET_INIT_PIC_REG): Define.
23168 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
23169 (sparc_pic_register_p): New predicate.
23170 (sparc_legitimate_address_p): Use it.
23171 (sparc_legitimize_pic_address): Likewise.
23172 (sparc_delegitimize_address): Likewise.
23173 (sparc_mode_dependent_address_p): Likewise.
23174 (gen_load_pcrel_sym): Remove 4th parameter.
23175 (load_got_register): Adjust call to above. Remove obsolete stuff.
23176 (sparc_expand_prologue): Do not call load_got_register here.
23177 (sparc_flat_expand_prologue): Likewise.
23178 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
23179 (sparc_use_pseudo_pic_reg): New function.
23180 (sparc_init_pic_reg): Likewise.
23181 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
23182 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
23184 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
23186 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
23187 Add item for branch_cost.
23189 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
23191 PR rtl-optimization/83565
23192 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
23193 not extend the result to a larger mode for rotate operations.
23194 (num_sign_bit_copies1): Likewise.
23196 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23199 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
23201 Use values-Xc.o for -pedantic.
23202 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
23204 2018-01-12 Martin Liska <mliska@suse.cz>
23207 * ipa-devirt.c (final_warning_record::grow_type_warnings):
23209 (possible_polymorphic_call_targets): Use it.
23210 (ipa_devirt): Likewise.
23212 2018-01-12 Martin Liska <mliska@suse.cz>
23214 * profile-count.h (enum profile_quality): Use 0 as invalid
23215 enum value of profile_quality.
23217 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
23219 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
23220 -mext-string options.
23222 2018-01-12 Richard Biener <rguenther@suse.de>
23224 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
23225 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
23226 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
23228 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
23230 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
23232 * configure.ac (--with-long-double-format): Add support for the
23233 configuration option to change the default long double format on
23235 * config.gcc (powerpc*-linux*-*): Likewise.
23236 * configure: Regenerate.
23237 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
23238 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
23239 used without modification.
23241 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23243 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
23244 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
23245 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
23246 MISC_BUILTIN_SPEC_BARRIER.
23247 (rs6000_init_builtins): Likewise.
23248 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
23250 (speculation_barrier): New define_insn.
23251 * doc/extend.texi: Document __builtin_speculation_barrier.
23253 2018-01-11 Jakub Jelinek <jakub@redhat.com>
23256 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
23257 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
23258 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
23260 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
23261 integral modes instead of "ss" and "sd".
23262 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
23263 vectors with 32-bit and 64-bit elements.
23264 (vecdupssescalarmodesuffix): New mode attribute.
23265 (vec_dup<mode>): Use it.
23267 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
23270 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
23271 frame if argument is passed on stack.
23273 2018-01-11 Jakub Jelinek <jakub@redhat.com>
23276 * ree.c (combine_reaching_defs): Optimize also
23277 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
23278 reg2=any_extend(exp); reg1=reg2;, formatting fix.
23280 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
23282 PR middle-end/83189
23283 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
23285 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
23287 PR middle-end/83718
23288 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
23289 after they are computed.
23291 2018-01-11 Bin Cheng <bin.cheng@arm.com>
23293 PR tree-optimization/83695
23294 * gimple-loop-linterchange.cc
23295 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
23296 reset cached scev information after interchange.
23297 (pass_linterchange::execute): Remove call to scev_reset_htab.
23299 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23301 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
23302 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
23303 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
23304 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
23305 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
23306 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
23307 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
23308 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
23309 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
23310 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
23311 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
23312 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
23313 (V_lane_reg): Likewise.
23314 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
23316 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
23317 (vfmal_lane_low<mode>_intrinsic,
23318 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
23319 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
23320 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
23321 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
23322 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
23323 vfmsl_lane_high<mode>_intrinsic): New define_insns.
23325 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23327 * config/arm/arm-cpus.in (fp16fml): New feature.
23328 (ALL_SIMD): Add fp16fml.
23329 (armv8.2-a): Add fp16fml as an option.
23330 (armv8.3-a): Likewise.
23331 (armv8.4-a): Add fp16fml as part of fp16.
23332 * config/arm/arm.h (TARGET_FP16FML): Define.
23333 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
23335 * config/arm/arm-modes.def (V2HF): Define.
23336 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
23337 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
23338 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
23339 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
23340 vfmsl_low, vfmsl_high): New set of builtins.
23341 * config/arm/iterators.md (PLUSMINUS): New code iterator.
23342 (vfml_op): New code attribute.
23343 (VFMLHALVES): New int iterator.
23344 (VFML, VFMLSEL): New mode attributes.
23345 (V_reg): Define mapping for V2HF.
23346 (V_hi, V_lo): New mode attributes.
23347 (VF_constraint): Likewise.
23348 (vfml_half, vfml_half_selector): New int attributes.
23349 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
23351 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
23352 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
23354 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
23355 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
23356 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
23357 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
23359 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
23360 Document new effective target and option set.
23362 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23364 * config/arm/arm-cpus.in (armv8_4): New feature.
23365 (ARMv8_4a): New fgroup.
23366 (armv8.4-a): New arch.
23367 * config/arm/arm-tables.opt: Regenerate.
23368 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
23369 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
23370 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
23371 Add matching rules for -march=armv8.4-a and extensions.
23372 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
23374 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
23377 * config/rx/rx.md (BW): New mode attribute.
23378 (sync_lock_test_and_setsi): Add mode suffix to insn output.
23380 2018-01-11 Richard Biener <rguenther@suse.de>
23382 PR tree-optimization/83435
23383 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
23384 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
23385 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
23387 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23388 Alan Hayward <alan.hayward@arm.com>
23389 David Sherwood <david.sherwood@arm.com>
23391 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
23393 (aarch64_classify_address): Initialize it. Track polynomial offsets.
23394 (aarch64_print_address_internal): Use it to check for a zero offset.
23396 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23397 Alan Hayward <alan.hayward@arm.com>
23398 David Sherwood <david.sherwood@arm.com>
23400 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
23401 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
23402 Return a poly_int64 rather than a HOST_WIDE_INT.
23403 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
23404 rather than a HOST_WIDE_INT.
23405 * config/aarch64/aarch64.h (aarch64_frame): Protect with
23406 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
23407 hard_fp_offset, frame_size, initial_adjust, callee_offset and
23408 final_offset from HOST_WIDE_INT to poly_int64.
23409 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
23410 to_constant when getting the number of units in an Advanced SIMD
23412 (aarch64_builtin_vectorized_function): Check for a constant number
23414 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
23416 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
23417 attribute instead of GET_MODE_NUNITS.
23418 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
23419 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
23420 GET_MODE_SIZE for fixed-size registers.
23421 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
23422 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
23423 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
23424 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
23425 (aarch64_print_operand, aarch64_print_address_internal)
23426 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
23427 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
23428 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
23429 Handle polynomial GET_MODE_SIZE.
23430 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
23431 wider than SImode without modification.
23432 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
23433 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
23434 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
23435 passing and returning SVE modes.
23436 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
23437 rather than GEN_INT.
23438 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
23439 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
23440 (aarch64_allocate_and_probe_stack_space): Likewise.
23441 (aarch64_layout_frame): Cope with polynomial offsets.
23442 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
23443 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
23444 polynomial offsets.
23445 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
23446 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
23447 poly_int64 rather than a HOST_WIDE_INT.
23448 (aarch64_get_separate_components, aarch64_process_components)
23449 (aarch64_expand_prologue, aarch64_expand_epilogue)
23450 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
23451 (aarch64_anchor_offset): New function, split out from...
23452 (aarch64_legitimize_address): ...here.
23453 (aarch64_builtin_vectorization_cost): Handle polynomial
23454 TYPE_VECTOR_SUBPARTS.
23455 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
23457 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
23458 number of elements from the PARALLEL rather than the mode.
23459 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
23460 rather than GET_MODE_BITSIZE.
23461 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
23462 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
23463 (aarch64_expand_vec_perm_const_1): Handle polynomial
23464 d->perm.length () and d->perm elements.
23465 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
23466 Apply to_constant to d->perm elements.
23467 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
23468 polynomial CONST_VECTOR_NUNITS.
23469 (aarch64_move_pointer): Take amount as a poly_int64 rather
23471 (aarch64_progress_pointer): Avoid temporary variable.
23472 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
23473 the mode attribute instead of GET_MODE.
23475 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23476 Alan Hayward <alan.hayward@arm.com>
23477 David Sherwood <david.sherwood@arm.com>
23479 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
23480 x exists before using it.
23481 (aarch64_add_constant_internal): Rename to...
23482 (aarch64_add_offset_1): ...this. Replace regnum with separate
23483 src and dest rtxes. Handle the case in which they're different,
23484 including when the offset is zero. Replace scratchreg with an rtx.
23485 Use 2 additions if there is no spare register into which we can
23486 move a 16-bit constant.
23487 (aarch64_add_constant): Delete.
23488 (aarch64_add_offset): Replace reg with separate src and dest
23489 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
23490 Use aarch64_add_offset_1.
23491 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
23492 an rtx rather than an int. Take the delta as a poly_int64
23493 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
23494 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
23495 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
23496 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
23497 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
23498 and aarch64_add_sp.
23499 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
23500 aarch64_add_constant.
23502 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23504 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
23505 Use scalar_float_mode.
23507 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23509 * config/aarch64/aarch64-simd.md
23510 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
23511 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
23512 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
23513 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
23514 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
23515 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
23516 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
23517 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
23518 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
23519 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
23521 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23524 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
23525 targ_options->x_arm_arch_string is non NULL.
23527 2018-01-11 Tamar Christina <tamar.christina@arm.com>
23529 * config/aarch64/aarch64.h
23530 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
23532 2018-01-11 Sudakshina Das <sudi.das@arm.com>
23535 * expmed.c (emit_store_flag_force): Swap if const op0
23536 and change VOIDmode to mode of op0.
23538 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23540 PR rtl-optimization/83761
23541 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
23542 than bytes to mode_for_size.
23544 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23546 PR middle-end/83189
23547 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
23548 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
23551 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23553 PR middle-end/83575
23554 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
23555 when in layout mode.
23556 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
23557 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
23560 2018-01-10 Michael Collison <michael.collison@arm.com>
23562 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
23563 * config/aarch64/aarch64-option-extension.def: Add
23564 AARCH64_OPT_EXTENSION of 'fp16fml'.
23565 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23566 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
23567 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
23568 * config/aarch64/constraints.md (Ui7): New constraint.
23569 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
23570 (VFMLA_SEL_W): Ditto.
23573 (VFMLA16_LOW): New int iterator.
23574 (VFMLA16_HIGH): Ditto.
23575 (UNSPEC_FMLAL): New unspec.
23576 (UNSPEC_FMLSL): Ditto.
23577 (UNSPEC_FMLAL2): Ditto.
23578 (UNSPEC_FMLSL2): Ditto.
23579 (f16mac): New code attribute.
23580 * config/aarch64/aarch64-simd-builtins.def
23581 (aarch64_fmlal_lowv2sf): Ditto.
23582 (aarch64_fmlsl_lowv2sf): Ditto.
23583 (aarch64_fmlalq_lowv4sf): Ditto.
23584 (aarch64_fmlslq_lowv4sf): Ditto.
23585 (aarch64_fmlal_highv2sf): Ditto.
23586 (aarch64_fmlsl_highv2sf): Ditto.
23587 (aarch64_fmlalq_highv4sf): Ditto.
23588 (aarch64_fmlslq_highv4sf): Ditto.
23589 (aarch64_fmlal_lane_lowv2sf): Ditto.
23590 (aarch64_fmlsl_lane_lowv2sf): Ditto.
23591 (aarch64_fmlal_laneq_lowv2sf): Ditto.
23592 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
23593 (aarch64_fmlalq_lane_lowv4sf): Ditto.
23594 (aarch64_fmlsl_lane_lowv4sf): Ditto.
23595 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
23596 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
23597 (aarch64_fmlal_lane_highv2sf): Ditto.
23598 (aarch64_fmlsl_lane_highv2sf): Ditto.
23599 (aarch64_fmlal_laneq_highv2sf): Ditto.
23600 (aarch64_fmlsl_laneq_highv2sf): Ditto.
23601 (aarch64_fmlalq_lane_highv4sf): Ditto.
23602 (aarch64_fmlsl_lane_highv4sf): Ditto.
23603 (aarch64_fmlalq_laneq_highv4sf): Ditto.
23604 (aarch64_fmlsl_laneq_highv4sf): Ditto.
23605 * config/aarch64/aarch64-simd.md:
23606 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
23607 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23608 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
23609 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23610 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
23611 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
23612 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
23613 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
23614 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
23615 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
23616 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
23617 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
23618 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
23619 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
23620 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
23621 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
23622 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
23623 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
23624 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
23625 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
23626 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
23627 (vfmlsl_low_u32): Ditto.
23628 (vfmlalq_low_u32): Ditto.
23629 (vfmlslq_low_u32): Ditto.
23630 (vfmlal_high_u32): Ditto.
23631 (vfmlsl_high_u32): Ditto.
23632 (vfmlalq_high_u32): Ditto.
23633 (vfmlslq_high_u32): Ditto.
23634 (vfmlal_lane_low_u32): Ditto.
23635 (vfmlsl_lane_low_u32): Ditto.
23636 (vfmlal_laneq_low_u32): Ditto.
23637 (vfmlsl_laneq_low_u32): Ditto.
23638 (vfmlalq_lane_low_u32): Ditto.
23639 (vfmlslq_lane_low_u32): Ditto.
23640 (vfmlalq_laneq_low_u32): Ditto.
23641 (vfmlslq_laneq_low_u32): Ditto.
23642 (vfmlal_lane_high_u32): Ditto.
23643 (vfmlsl_lane_high_u32): Ditto.
23644 (vfmlal_laneq_high_u32): Ditto.
23645 (vfmlsl_laneq_high_u32): Ditto.
23646 (vfmlalq_lane_high_u32): Ditto.
23647 (vfmlslq_lane_high_u32): Ditto.
23648 (vfmlalq_laneq_high_u32): Ditto.
23649 (vfmlslq_laneq_high_u32): Ditto.
23650 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
23651 (AARCH64_FL_FOR_ARCH8_4): New.
23652 (AARCH64_ISA_F16FML): New ISA flag.
23653 (TARGET_F16FML): New feature flag for fp16fml.
23654 (doc/invoke.texi): Document new fp16fml option.
23656 2018-01-10 Michael Collison <michael.collison@arm.com>
23658 * config/aarch64/aarch64-builtins.c:
23659 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
23660 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23661 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
23662 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
23663 (AARCH64_ISA_SHA3): New ISA flag.
23664 (TARGET_SHA3): New feature flag for sha3.
23665 * config/aarch64/iterators.md (sha512_op): New int attribute.
23666 (CRYPTO_SHA512): New int iterator.
23667 (UNSPEC_SHA512H): New unspec.
23668 (UNSPEC_SHA512H2): Ditto.
23669 (UNSPEC_SHA512SU0): Ditto.
23670 (UNSPEC_SHA512SU1): Ditto.
23671 * config/aarch64/aarch64-simd-builtins.def
23672 (aarch64_crypto_sha512hqv2di): New builtin.
23673 (aarch64_crypto_sha512h2qv2di): Ditto.
23674 (aarch64_crypto_sha512su0qv2di): Ditto.
23675 (aarch64_crypto_sha512su1qv2di): Ditto.
23676 (aarch64_eor3qv8hi): Ditto.
23677 (aarch64_rax1qv2di): Ditto.
23678 (aarch64_xarqv2di): Ditto.
23679 (aarch64_bcaxqv8hi): Ditto.
23680 * config/aarch64/aarch64-simd.md:
23681 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
23682 (aarch64_crypto_sha512su0qv2di): Ditto.
23683 (aarch64_crypto_sha512su1qv2di): Ditto.
23684 (aarch64_eor3qv8hi): Ditto.
23685 (aarch64_rax1qv2di): Ditto.
23686 (aarch64_xarqv2di): Ditto.
23687 (aarch64_bcaxqv8hi): Ditto.
23688 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
23689 (vsha512h2q_u64): Ditto.
23690 (vsha512su0q_u64): Ditto.
23691 (vsha512su1q_u64): Ditto.
23692 (veor3q_u16): Ditto.
23693 (vrax1q_u64): Ditto.
23694 (vxarq_u64): Ditto.
23695 (vbcaxq_u16): Ditto.
23696 * config/arm/types.md (crypto_sha512): New type attribute.
23697 (crypto_sha3): Ditto.
23698 (doc/invoke.texi): Document new sha3 option.
23700 2018-01-10 Michael Collison <michael.collison@arm.com>
23702 * config/aarch64/aarch64-builtins.c:
23703 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
23704 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23705 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
23706 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
23707 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
23708 (AARCH64_ISA_SM4): New ISA flag.
23709 (TARGET_SM4): New feature flag for sm4.
23710 * config/aarch64/aarch64-simd-builtins.def
23711 (aarch64_sm3ss1qv4si): Ditto.
23712 (aarch64_sm3tt1aq4si): Ditto.
23713 (aarch64_sm3tt1bq4si): Ditto.
23714 (aarch64_sm3tt2aq4si): Ditto.
23715 (aarch64_sm3tt2bq4si): Ditto.
23716 (aarch64_sm3partw1qv4si): Ditto.
23717 (aarch64_sm3partw2qv4si): Ditto.
23718 (aarch64_sm4eqv4si): Ditto.
23719 (aarch64_sm4ekeyqv4si): Ditto.
23720 * config/aarch64/aarch64-simd.md:
23721 (aarch64_sm3ss1qv4si): Ditto.
23722 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
23723 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
23724 (aarch64_sm4eqv4si): Ditto.
23725 (aarch64_sm4ekeyqv4si): Ditto.
23726 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
23727 (sm3part_op): Ditto.
23728 (CRYPTO_SM3TT): Ditto.
23729 (CRYPTO_SM3PART): Ditto.
23730 (UNSPEC_SM3SS1): New unspec.
23731 (UNSPEC_SM3TT1A): Ditto.
23732 (UNSPEC_SM3TT1B): Ditto.
23733 (UNSPEC_SM3TT2A): Ditto.
23734 (UNSPEC_SM3TT2B): Ditto.
23735 (UNSPEC_SM3PARTW1): Ditto.
23736 (UNSPEC_SM3PARTW2): Ditto.
23737 (UNSPEC_SM4E): Ditto.
23738 (UNSPEC_SM4EKEY): Ditto.
23739 * config/aarch64/constraints.md (Ui2): New constraint.
23740 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
23741 * config/arm/types.md (crypto_sm3): New type attribute.
23742 (crypto_sm4): Ditto.
23743 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
23744 (vsm3tt1aq_u32): Ditto.
23745 (vsm3tt1bq_u32): Ditto.
23746 (vsm3tt2aq_u32): Ditto.
23747 (vsm3tt2bq_u32): Ditto.
23748 (vsm3partw1q_u32): Ditto.
23749 (vsm3partw2q_u32): Ditto.
23750 (vsm4eq_u32): Ditto.
23751 (vsm4ekeyq_u32): Ditto.
23752 (doc/invoke.texi): Document new sm4 option.
23754 2018-01-10 Michael Collison <michael.collison@arm.com>
23756 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
23757 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
23758 (AARCH64_FL_FOR_ARCH8_4): New.
23759 (AARCH64_FL_V8_4): New flag.
23760 (doc/invoke.texi): Document new armv8.4-a option.
23762 2018-01-10 Michael Collison <michael.collison@arm.com>
23764 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23765 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
23766 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
23767 * config/aarch64/aarch64-option-extension.def: Add
23768 AARCH64_OPT_EXTENSION of 'sha2'.
23769 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
23770 (crypto): Disable sha2 and aes if crypto disabled.
23771 (crypto): Enable aes and sha2 if enabled.
23772 (simd): Disable sha2 and aes if simd disabled.
23773 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
23775 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
23776 (TARGET_SHA2): New feature flag for sha2.
23777 (TARGET_AES): New feature flag for aes.
23778 * config/aarch64/aarch64-simd.md:
23779 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
23780 conditional on TARGET_AES.
23781 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
23782 (aarch64_crypto_sha1hsi): Make pattern conditional
23784 (aarch64_crypto_sha1hv4si): Ditto.
23785 (aarch64_be_crypto_sha1hv4si): Ditto.
23786 (aarch64_crypto_sha1su1v4si): Ditto.
23787 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
23788 (aarch64_crypto_sha1su0v4si): Ditto.
23789 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
23790 (aarch64_crypto_sha256su0v4si): Ditto.
23791 (aarch64_crypto_sha256su1v4si): Ditto.
23792 (doc/invoke.texi): Document new aes and sha2 options.
23794 2018-01-10 Martin Sebor <msebor@redhat.com>
23796 PR tree-optimization/83781
23797 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
23800 2018-01-11 Martin Sebor <msebor@gmail.com>
23801 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23803 PR tree-optimization/83501
23804 PR tree-optimization/81703
23806 * tree-ssa-strlen.c (get_string_cst): Rename...
23807 (get_string_len): ...to this. Handle global constants.
23808 (handle_char_store): Adjust.
23810 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
23811 Jim Wilson <jimw@sifive.com>
23813 * config/riscv/riscv-protos.h (riscv_output_return): New.
23814 * config/riscv/riscv.c (struct machine_function): New naked_p field.
23815 (riscv_attribute_table, riscv_output_return),
23816 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
23817 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
23818 (riscv_compute_frame_info): Only compute frame->mask if not a naked
23820 (riscv_expand_prologue): Add early return for naked function.
23821 (riscv_expand_epilogue): Likewise.
23822 (riscv_function_ok_for_sibcall): Return false for naked function.
23823 (riscv_set_current_function): New.
23824 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
23825 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
23826 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
23827 * doc/extend.texi (RISC-V Function Attributes): New.
23829 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
23831 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
23832 check for 128-bit long double before checking TCmode.
23833 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
23834 128-bit long doubles before checking TFmode or TCmode.
23835 (FLOAT128_IBM_P): Likewise.
23837 2018-01-10 Martin Sebor <msebor@redhat.com>
23839 PR tree-optimization/83671
23840 * builtins.c (c_strlen): Unconditionally return zero for the empty
23842 Use -Warray-bounds for warnings.
23843 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
23844 for non-constant array indices with COMPONENT_REF, arrays of
23845 arrays, and pointers to arrays.
23846 (gimple_fold_builtin_strlen): Determine and set length range for
23847 non-constant character arrays.
23849 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
23851 PR middle-end/81897
23852 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
23855 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
23857 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
23859 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
23862 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
23863 VECTOR_MEM_ALTIVEC_OR_VSX_P.
23864 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
23865 indexed_or_indirect_operand predicate.
23866 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
23867 (*vsx_le_perm_load_v8hi): Likewise.
23868 (*vsx_le_perm_load_v16qi): Likewise.
23869 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
23870 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
23871 (*vsx_le_perm_store_v8hi): Likewise.
23872 (*vsx_le_perm_store_v16qi): Likewise.
23873 (eight unnamed splitters): Likewise.
23875 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
23877 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
23878 * config/rs6000/emmintrin.h: Likewise.
23879 * config/rs6000/mmintrin.h: Likewise.
23880 * config/rs6000/xmmintrin.h: Likewise.
23882 2018-01-10 David Malcolm <dmalcolm@redhat.com>
23885 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
23887 * tree.c (tree_nop_conversion): Return true for location wrapper
23889 (maybe_wrap_with_location): New function.
23890 (selftest::check_strip_nops): New function.
23891 (selftest::test_location_wrappers): New function.
23892 (selftest::tree_c_tests): Call it.
23893 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
23894 (maybe_wrap_with_location): New decl.
23895 (EXPR_LOCATION_WRAPPER_P): New macro.
23896 (location_wrapper_p): New inline function.
23897 (tree_strip_any_location_wrapper): New inline function.
23899 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
23902 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
23903 stack_realign_offset for the largest alignment of stack slot
23905 (ix86_find_max_used_stack_alignment): New function.
23906 (ix86_finalize_stack_frame_flags): Use it. Set
23907 max_used_stack_alignment if we don't realign stack.
23908 * config/i386/i386.h (machine_function): Add
23909 max_used_stack_alignment.
23911 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
23913 * config/arm/arm.opt (-mbranch-cost): New option.
23914 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
23917 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
23920 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
23921 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
23923 2018-01-10 Richard Biener <rguenther@suse.de>
23926 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
23927 early out so it also covers the case where we have a non-NULL
23930 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
23932 PR tree-optimization/83753
23933 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
23934 for non-strided grouped accesses if the number of elements is 1.
23936 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23939 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
23940 * i386.h (TARGET_USE_GATHER): Define.
23941 * x86-tune.def (X86_TUNE_USE_GATHER): New.
23943 2018-01-10 Martin Liska <mliska@suse.cz>
23946 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
23947 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
23949 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
23950 CLEANUP_NO_PARTITIONING is not set.
23952 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
23954 * doc/rtl.texi: Remove documentation of (const ...) wrappers
23955 for vectors, as a partial revert of r254296.
23956 * rtl.h (const_vec_p): Delete.
23957 (const_vec_duplicate_p): Don't test for vector CONSTs.
23958 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
23959 * expmed.c (make_tree): Likewise.
23962 * common.md (E, F): Use CONSTANT_P instead of checking for
23964 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
23965 checking for CONST_VECTOR.
23967 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
23969 PR middle-end/83575
23970 * predict.c (force_edge_cold): Handle in more sane way edges
23971 with no prediction.
23973 2018-01-09 Carl Love <cel@us.ibm.com>
23975 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
23977 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
23978 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
23979 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
23980 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
23981 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
23982 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
23983 * config/rs6000/rs6000-protos.h: Add extern defition for
23984 rs6000_generate_float2_double_code.
23985 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
23987 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
23988 (float2_v2df): Add define_expand.
23990 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
23993 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
23994 op_mode in the force_to_mode call.
23996 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
23998 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
23999 instead of checking each element individually.
24000 (aarch64_evpc_uzp): Likewise.
24001 (aarch64_evpc_zip): Likewise.
24002 (aarch64_evpc_ext): Likewise.
24003 (aarch64_evpc_rev): Likewise.
24004 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
24005 instead of checking each element individually. Return true without
24007 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
24008 whether all selected elements come from the same input, instead of
24009 checking each element individually. Remove calls to gen_rtx_REG,
24010 start_sequence and end_sequence and instead assert that no rtl is
24013 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
24015 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
24016 order of HIGH and CONST checks.
24018 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
24020 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
24021 if the destination isn't an SSA_NAME.
24023 2018-01-09 Richard Biener <rguenther@suse.de>
24025 PR tree-optimization/83668
24026 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
24028 (canonicalize_loop_form): ... here, renamed from ...
24029 (canonicalize_loop_closed_ssa_form): ... this and amended to
24030 swap successor edges for loop exit blocks to make us use
24031 the RPO order we need for initial schedule generation.
24033 2018-01-09 Joseph Myers <joseph@codesourcery.com>
24035 PR tree-optimization/64811
24036 * match.pd: When optimizing comparisons with Inf, avoid
24037 introducing or losing exceptions from comparisons with NaN.
24039 2018-01-09 Martin Liska <mliska@suse.cz>
24042 * asan.c (shadow_mem_size): Add gcc_assert.
24044 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
24046 Don't save registers in main().
24049 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
24050 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
24051 * config/avr/avr.c (avr_set_current_function): Don't error if
24052 naked, OS_task or OS_main are specified at the same time.
24053 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
24055 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
24057 * common/config/avr/avr-common.c (avr_option_optimization_table):
24058 Switch on -mmain-is-OS_task for optimizing compilations.
24060 2018-01-09 Richard Biener <rguenther@suse.de>
24062 PR tree-optimization/83572
24063 * graphite.c: Include cfganal.h.
24064 (graphite_transform_loops): Connect infinite loops to exit
24065 and remove fake edges at the end.
24067 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
24069 * ipa-inline.c (edge_badness): Revert accidental checkin.
24071 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
24074 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
24075 symbols; not inline clones.
24077 2018-01-09 Jakub Jelinek <jakub@redhat.com>
24080 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
24081 hard registers. Formatting fixes.
24083 PR preprocessor/83722
24084 * gcc.c (try_generate_repro): Pass
24085 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
24086 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
24089 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
24090 Kito Cheng <kito.cheng@gmail.com>
24092 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
24093 (riscv_leaf_function_p): Delete.
24094 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
24096 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24098 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
24100 (do_ifelse): New function.
24101 (do_isel): New function.
24102 (do_sub3): New function.
24103 (do_add3): New function.
24104 (do_load_mask_compare): New function.
24105 (do_overlap_load_compare): New function.
24106 (expand_compare_loop): New function.
24107 (expand_block_compare): Call expand_compare_loop() when appropriate.
24108 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
24109 option description.
24110 (-mblock-compare-inline-loop-limit): New option.
24112 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24115 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
24116 Reverse order of second and third operands in first alternative.
24117 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
24118 of first and second elements in UNSPEC_VPERMR vector.
24119 (altivec_expand_vec_perm_le): Likewise.
24121 2018-01-08 Jeff Law <law@redhat.com>
24123 PR rtl-optimizatin/81308
24124 * tree-switch-conversion.c (cfg_altered): New file scoped static.
24125 (process_switch): If group_case_labels makes a change, then set
24127 (pass_convert_switch::execute): If a switch is converted, then
24128 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
24130 PR rtl-optimization/81308
24131 * recog.c (split_all_insns): Conditionally cleanup the CFG after
24134 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
24136 PR target/83663 - Revert r255946
24137 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
24138 generation for cases where splatting a value is not useful.
24139 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
24140 across a vec_duplicate and a paradoxical subreg forming a vector
24141 mode to a vec_concat.
24143 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24145 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
24146 -march=armv8.3-a variants.
24147 * config/arm/t-multilib: Likewise.
24148 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
24150 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24152 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
24154 (cceq_ior_compare_complement): Give it a name so I can use it, and
24155 change boolean_or_operator predicate to boolean_operator so it can
24156 be used to generate a crand.
24157 (eqne): New code iterator.
24158 (bd/bd_neg): New code_attrs.
24159 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
24160 a single define_insn.
24161 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
24162 decrement (bdnzt/bdnzf/bdzt/bdzf).
24163 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
24164 with the new names of the branch decrement patterns, and added the
24165 names of the branch decrement conditional patterns.
24167 2018-01-08 Richard Biener <rguenther@suse.de>
24169 PR tree-optimization/83563
24170 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
24173 2018-01-08 Richard Biener <rguenther@suse.de>
24175 PR middle-end/83713
24176 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
24178 2018-01-08 Richard Biener <rguenther@suse.de>
24180 PR tree-optimization/83685
24181 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
24182 references to abnormals.
24184 2018-01-08 Richard Biener <rguenther@suse.de>
24187 * dwarf2out.c (output_indirect_strings): Handle empty
24188 skeleton_debug_str_hash.
24189 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
24191 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
24193 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
24194 (emit_store_direct): Likewise.
24195 (arc_trampoline_adjust_address): Likewise.
24196 (arc_asm_trampoline_template): New function.
24197 (arc_initialize_trampoline): Use asm_trampoline_template.
24198 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
24199 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
24200 * config/arc/arc.md (flush_icache): Delete pattern.
24202 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
24204 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
24205 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
24208 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
24211 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
24212 by not USED_FOR_TARGET.
24213 (make_pass_resolve_sw_modes): Likewise.
24215 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
24217 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
24220 2018-01-08 Richard Biener <rguenther@suse.de>
24222 PR middle-end/83580
24223 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
24225 2018-01-08 Richard Biener <rguenther@suse.de>
24227 PR middle-end/83517
24228 * match.pd ((t * 2) / 2) -> t): Add missing :c.
24230 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
24232 PR middle-end/81897
24233 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
24234 basic blocks with a small number of successors.
24235 (convert_control_dep_chain_into_preds): Improve handling of
24237 (dump_predicates): Split apart into...
24238 (dump_pred_chain): ...here...
24239 (dump_pred_info): ...and here.
24240 (can_one_predicate_be_invalidated_p): Add debugging printfs.
24241 (can_chain_union_be_invalidated_p): Improve check for invalidation
24243 (uninit_uses_cannot_happen): Avoid unnecessary if
24244 convert_control_dep_chain_into_preds yielded nothing.
24246 2018-01-06 Martin Sebor <msebor@redhat.com>
24248 PR tree-optimization/83640
24249 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
24250 subtracting negative offset from size.
24251 (builtin_access::overlap): Adjust offset bounds of the access to fall
24252 within the size of the object if possible.
24254 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
24256 PR rtl-optimization/83699
24257 * expmed.c (extract_bit_field_1): Restrict the vector usage of
24258 extract_bit_field_as_subreg to cases in which the extracted
24259 value is also a vector.
24261 * lra-constraints.c (process_alt_operands): Test for the equivalence
24262 substitutions when detecting a possible reload cycle.
24264 2018-01-06 Jakub Jelinek <jakub@redhat.com>
24267 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
24268 by default if flag_selective_schedling{,2}. Formatting fixes.
24270 PR rtl-optimization/83682
24271 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
24272 if it has non-VECTOR_MODE element mode.
24273 (vec_duplicate_p): Likewise.
24275 PR middle-end/83694
24276 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
24277 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
24279 2018-01-05 Jakub Jelinek <jakub@redhat.com>
24282 * config/i386/i386-builtin.def
24283 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
24284 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
24285 Require also OPTION_MASK_ISA_AVX512F in addition to
24286 OPTION_MASK_ISA_GFNI.
24287 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
24288 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
24289 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
24290 to OPTION_MASK_ISA_GFNI.
24291 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
24292 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
24293 OPTION_MASK_ISA_AVX512BW.
24294 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
24295 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
24296 addition to OPTION_MASK_ISA_GFNI.
24297 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
24298 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
24299 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
24300 to OPTION_MASK_ISA_GFNI.
24301 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
24302 a requirement for all ISAs rather than any of them with a few
24304 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
24306 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
24307 bitmasks to be enabled with 3 exceptions, instead of requiring any
24308 enabled ISA with lots of exceptions.
24309 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
24310 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
24311 Change avx512bw in isa attribute to avx512f.
24312 * config/i386/sgxintrin.h: Add license boilerplate.
24313 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
24314 to __AVX512F__ and __AVX512VL to __AVX512VL__.
24315 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
24316 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
24318 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
24319 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
24320 temporarily sse2 rather than sse if not enabled already.
24323 * config/i386/sse.md (VI248_VLBW): Rename to ...
24324 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
24325 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
24326 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
24327 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
24328 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
24329 mode iterator instead of VI248_VLBW.
24331 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
24333 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
24334 (record_modified): Skip clobbers; add debug output.
24335 (param_change_prob): Use sreal frequencies.
24337 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
24339 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
24340 punt for user-aligned variables.
24342 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
24344 * tree-chrec.c (chrec_contains_symbols): Return true for
24347 2018-01-05 Sudakshina Das <sudi.das@arm.com>
24350 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
24351 of (x|y) == x for BICS pattern.
24353 2018-01-05 Jakub Jelinek <jakub@redhat.com>
24355 PR tree-optimization/83605
24356 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
24357 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
24360 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
24362 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
24363 * config/epiphany/rtems.h: New file.
24365 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24366 Uros Bizjak <ubizjak@gmail.com>
24369 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
24370 QIreg_operand instead of register_operand predicate.
24371 * config/i386/i386.c (ix86_rop_should_change_byte_p,
24372 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
24373 comments instead of -fmitigate[-_]rop.
24375 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24378 * cgraphunit.c (symbol_table::compile): Switch to text_section
24379 before calling assembly_start debug hook.
24380 * run-rtl-passes.c (run_rtl_passes): Likewise.
24383 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24385 * tree-vrp.c (extract_range_from_binary_expr_1): Check
24386 range_int_cst_p rather than !symbolic_range_p before calling
24387 extract_range_from_multiplicative_op_1.
24389 2018-01-04 Jeff Law <law@redhat.com>
24391 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
24392 redundant test in assertion.
24394 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24396 * doc/rtl.texi: Document machine_mode wrapper classes.
24398 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24400 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
24401 using tree_to_uhwi.
24403 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24405 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
24406 the VEC_PERM_EXPR fold to fail.
24408 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24411 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
24412 to switched_sections.
24414 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24417 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
24418 test for d.testing.
24420 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
24423 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
24424 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
24426 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24429 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
24430 is BLKmode and bitpos not zero or mode change is needed.
24432 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24435 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
24438 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
24441 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
24442 instead of MULT rtx. Update all corresponding splitters.
24443 (*saddl_se): Ditto.
24444 (*ssub<modesuffix>): Ditto.
24445 (*ssubl_se): Ditto.
24446 (*cmp_sadd_di): Update split patterns.
24447 (*cmp_sadd_si): Ditto.
24448 (*cmp_sadd_sidi): Ditto.
24449 (*cmp_ssub_di): Ditto.
24450 (*cmp_ssub_si): Ditto.
24451 (*cmp_ssub_sidi): Ditto.
24452 * config/alpha/predicates.md (const23_operand): New predicate.
24453 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
24454 Look for ASHIFT, not MULT inner operand.
24455 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
24457 2018-01-04 Martin Liska <mliska@suse.cz>
24459 PR gcov-profile/83669
24460 * gcov.c (output_intermediate_file): Add version to intermediate
24462 * doc/gcov.texi: Document new field 'version' in intermediate
24463 file format. Fix location of '-k' option of gcov command.
24465 2018-01-04 Martin Liska <mliska@suse.cz>
24468 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
24470 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24472 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
24474 2018-01-03 Martin Sebor <msebor@redhat.com>
24476 PR tree-optimization/83655
24477 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
24478 checking calls with invalid arguments.
24480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24482 * tree-vect-stmts.c (vect_get_store_rhs): New function.
24483 (vectorizable_mask_load_store): Delete.
24484 (vectorizable_call): Return false for masked loads and stores.
24485 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
24486 instead of gimple_assign_rhs1.
24487 (vectorizable_load): Handle IFN_MASK_LOAD.
24488 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
24490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24492 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
24494 (vectorizable_mask_load_store): ...here.
24495 (vectorizable_load): ...and here.
24497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24499 * tree-vect-stmts.c (vect_build_all_ones_mask)
24500 (vect_build_zero_merge_argument): New functions, split out from...
24501 (vectorizable_load): ...here.
24503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24505 * tree-vect-stmts.c (vect_check_store_rhs): New function,
24507 (vectorizable_mask_load_store): ...here.
24508 (vectorizable_store): ...and here.
24510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24512 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
24514 (vectorizable_mask_load_store): ...here.
24516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24518 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
24519 (vect_model_store_cost): Take a vec_load_store_type instead of a
24521 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
24522 (vect_model_store_cost): Take a vec_load_store_type instead of a
24524 (vectorizable_mask_load_store): Update accordingly.
24525 (vectorizable_store): Likewise.
24526 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
24528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24530 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
24531 IFN_MASK_LOAD calls here rather than...
24532 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
24534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24535 Alan Hayward <alan.hayward@arm.com>
24536 David Sherwood <david.sherwood@arm.com>
24538 * expmed.c (extract_bit_field_1): For vector extracts,
24539 fall back to extract_bit_field_as_subreg if vec_extract
24542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24543 Alan Hayward <alan.hayward@arm.com>
24544 David Sherwood <david.sherwood@arm.com>
24546 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
24547 they are variable or constant sized.
24548 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
24549 slots for constant-sized data.
24551 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24552 Alan Hayward <alan.hayward@arm.com>
24553 David Sherwood <david.sherwood@arm.com>
24555 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
24556 handling COND_EXPRs with boolean comparisons, try to find a better
24557 basis for the mask type than the boolean itself.
24559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24561 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
24562 is calculated and how it can be overridden.
24563 * genmodes.c (max_bitsize_mode_any_mode): New variable.
24564 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
24566 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
24569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24570 Alan Hayward <alan.hayward@arm.com>
24571 David Sherwood <david.sherwood@arm.com>
24573 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
24574 Remove the mode argument.
24575 (aarch64_simd_valid_immediate): Remove the mode and inverse
24577 * config/aarch64/iterators.md (bitsize): New iterator.
24578 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
24579 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
24580 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
24581 aarch64_simd_valid_immediate.
24582 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
24583 (aarch64_reg_or_bic_imm): Likewise.
24584 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
24585 with an insn_type enum and msl with a modifier_type enum.
24586 Replace element_width with a scalar_mode. Change the shift
24587 to unsigned int. Add constructors for scalar_float_mode and
24588 scalar_int_mode elements.
24589 (aarch64_vect_float_const_representable_p): Delete.
24590 (aarch64_can_const_movi_rtx_p)
24591 (aarch64_simd_scalar_immediate_valid_for_move)
24592 (aarch64_simd_make_constant): Update call to
24593 aarch64_simd_valid_immediate.
24594 (aarch64_advsimd_valid_immediate_hs): New function.
24595 (aarch64_advsimd_valid_immediate): Likewise.
24596 (aarch64_simd_valid_immediate): Remove mode and inverse
24597 arguments. Rewrite to use the above. Use const_vec_duplicate_p
24598 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
24599 and aarch64_float_const_representable_p on the result.
24600 (aarch64_output_simd_mov_immediate): Remove mode argument.
24601 Update call to aarch64_simd_valid_immediate and use of
24602 simd_immediate_info.
24603 (aarch64_output_scalar_simd_mov_immediate): Update call
24606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24607 Alan Hayward <alan.hayward@arm.com>
24608 David Sherwood <david.sherwood@arm.com>
24610 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
24611 (mode_nunits): Likewise CONST_MODE_NUNITS.
24612 * machmode.def (ADJUST_NUNITS): Document.
24613 * genmodes.c (mode_data::need_nunits_adj): New field.
24614 (blank_mode): Update accordingly.
24615 (adj_nunits): New variable.
24616 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
24618 (emit_mode_size_inline): Set need_bytesize_adj for all modes
24619 listed in adj_nunits.
24620 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
24621 listed in adj_nunits. Don't emit case statements for such modes.
24622 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
24623 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
24624 nothing if adj_nunits is nonnull.
24625 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
24626 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
24627 (emit_mode_fbit): Update use of print_maybe_const_decl.
24628 (emit_move_size): Likewise. Treat the array as non-const
24630 (emit_mode_adjustments): Handle adj_nunits.
24632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24634 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
24635 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
24636 (VECTOR_MODES): Use it.
24637 (make_vector_modes): Take the prefix as an argument.
24639 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24640 Alan Hayward <alan.hayward@arm.com>
24641 David Sherwood <david.sherwood@arm.com>
24643 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
24644 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
24645 for MODE_VECTOR_BOOL.
24646 * machmode.def (VECTOR_BOOL_MODE): Document.
24647 * genmodes.c (VECTOR_BOOL_MODE): New macro.
24648 (make_vector_bool_mode): New function.
24649 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
24651 * lto-streamer-in.c (lto_input_mode_table): Likewise.
24652 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
24654 * stor-layout.c (int_mode_for_mode): Likewise.
24655 * tree.c (build_vector_type_for_mode): Likewise.
24656 * varasm.c (output_constant_pool_2): Likewise.
24657 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
24658 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
24659 for MODE_VECTOR_BOOL.
24660 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
24661 of mode class checks.
24662 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
24663 instead of a list of mode class checks.
24664 (expand_vector_scalar_condition): Likewise.
24665 (type_for_widest_vector_mode): Handle BImode as an inner mode.
24667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24668 Alan Hayward <alan.hayward@arm.com>
24669 David Sherwood <david.sherwood@arm.com>
24671 * machmode.h (mode_size): Change from unsigned short to
24673 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
24674 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24675 or if measurement_type is not polynomial.
24676 (fixed_size_mode::includes_p): Check for constant-sized modes.
24677 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
24678 return a poly_uint16 rather than an unsigned short.
24679 (emit_mode_size): Change the type of mode_size from unsigned short
24680 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
24681 (emit_mode_adjustments): Cope with polynomial vector sizes.
24682 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24684 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24686 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
24687 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
24688 * caller-save.c (setup_save_areas): Likewise.
24689 (replace_reg_with_saved_mem): Likewise.
24690 * calls.c (emit_library_call_value_1): Likewise.
24691 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
24692 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
24693 (gen_lowpart_for_combine): Likewise.
24694 * convert.c (convert_to_integer_1): Likewise.
24695 * cse.c (equiv_constant, cse_insn): Likewise.
24696 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
24697 (cselib_subst_to_values): Likewise.
24698 * dce.c (word_dce_process_block): Likewise.
24699 * df-problems.c (df_word_lr_mark_ref): Likewise.
24700 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
24701 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
24702 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
24703 (rtl_for_decl_location): Likewise.
24704 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
24705 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
24706 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
24707 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
24708 (expand_expr_real_1): Likewise.
24709 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
24710 (pad_below): Likewise.
24711 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24712 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
24713 * ira.c (get_subreg_tracking_sizes): Likewise.
24714 * ira-build.c (ira_create_allocno_objects): Likewise.
24715 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
24716 (ira_sort_regnos_for_alter_reg): Likewise.
24717 * ira-costs.c (record_operand_costs): Likewise.
24718 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
24719 (resolve_simple_move): Likewise.
24720 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
24721 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
24722 (lra_constraints): Likewise.
24723 (CONST_POOL_OK_P): Reject variable-sized modes.
24724 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
24725 (add_pseudo_to_slot, lra_spill): Likewise.
24726 * omp-low.c (omp_clause_aligned_alignment): Likewise.
24727 * optabs-query.c (get_best_extraction_insn): Likewise.
24728 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24729 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
24730 (expand_mult_highpart, valid_multiword_target_p): Likewise.
24731 * recog.c (offsettable_address_addr_space_p): Likewise.
24732 * regcprop.c (maybe_mode_change): Likewise.
24733 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
24734 * regrename.c (build_def_use): Likewise.
24735 * regstat.c (dump_reg_info): Likewise.
24736 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
24737 (find_reloads, find_reloads_subreg_address): Likewise.
24738 * reload1.c (eliminate_regs_1): Likewise.
24739 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
24740 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
24741 (simplify_binary_operation_1, simplify_subreg): Likewise.
24742 * targhooks.c (default_function_arg_padding): Likewise.
24743 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
24744 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
24745 (verify_gimple_assign_ternary): Likewise.
24746 * tree-inline.c (estimate_move_cost): Likewise.
24747 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24748 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
24749 (get_address_cost_ainc): Likewise.
24750 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
24751 (vect_supportable_dr_alignment): Likewise.
24752 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24753 (vectorizable_reduction): Likewise.
24754 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
24755 (vectorizable_operation, vectorizable_load): Likewise.
24756 * tree.c (build_same_sized_truth_vector_type): Likewise.
24757 * valtrack.c (cleanup_auto_inc_dec): Likewise.
24758 * var-tracking.c (emit_note_insn_var_location): Likewise.
24759 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
24760 (ADDR_VEC_ALIGN): Likewise.
24762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24763 Alan Hayward <alan.hayward@arm.com>
24764 David Sherwood <david.sherwood@arm.com>
24766 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
24768 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24769 or if measurement_type is polynomial.
24770 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
24771 * combine.c (make_extraction): Likewise.
24772 * dse.c (find_shift_sequence): Likewise.
24773 * dwarf2out.c (mem_loc_descriptor): Likewise.
24774 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
24775 (extract_bit_field, extract_low_bits): Likewise.
24776 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
24777 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
24778 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
24779 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
24780 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24781 * reload.c (find_reloads): Likewise.
24782 * reload1.c (alter_reg): Likewise.
24783 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
24784 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
24785 * tree-if-conv.c (predicate_mem_writes): Likewise.
24786 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
24787 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
24788 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
24789 * valtrack.c (dead_debug_insert_temp): Likewise.
24790 * varasm.c (mergeable_constant_section): Likewise.
24791 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
24793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24794 Alan Hayward <alan.hayward@arm.com>
24795 David Sherwood <david.sherwood@arm.com>
24797 * expr.c (expand_assignment): Cope with polynomial mode sizes
24798 when assigning to a CONCAT.
24800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24801 Alan Hayward <alan.hayward@arm.com>
24802 David Sherwood <david.sherwood@arm.com>
24804 * machmode.h (mode_precision): Change from unsigned short to
24806 (mode_to_precision): Return a poly_uint16 rather than an unsigned
24808 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
24809 or if measurement_type is not polynomial.
24810 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
24811 in which the mode is already known to be a scalar_int_mode.
24812 * genmodes.c (emit_mode_precision): Change the type of mode_precision
24813 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
24815 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24816 for GET_MODE_PRECISION.
24817 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24818 for GET_MODE_PRECISION.
24819 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
24821 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
24822 (expand_field_assignment, make_extraction): Likewise.
24823 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
24824 (get_last_value): Likewise.
24825 * convert.c (convert_to_integer_1): Likewise.
24826 * cse.c (cse_insn): Likewise.
24827 * expr.c (expand_expr_real_1): Likewise.
24828 * lra-constraints.c (simplify_operand_subreg): Likewise.
24829 * optabs-query.c (can_atomic_load_p): Likewise.
24830 * optabs.c (expand_atomic_load): Likewise.
24831 (expand_atomic_store): Likewise.
24832 * ree.c (combine_reaching_defs): Likewise.
24833 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
24834 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
24835 * tree.h (type_has_mode_precision_p): Likewise.
24836 * ubsan.c (instrument_si_overflow): Likewise.
24838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24839 Alan Hayward <alan.hayward@arm.com>
24840 David Sherwood <david.sherwood@arm.com>
24842 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
24843 polynomial numbers of units.
24844 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
24845 (valid_vector_subparts_p): New function.
24846 (build_vector_type): Remove temporary shim and take the number
24847 of units as a poly_uint64 rather than an int.
24848 (build_opaque_vector_type): Take the number of units as a
24849 poly_uint64 rather than an int.
24850 * tree.c (build_vector_from_ctor): Handle polynomial
24851 TYPE_VECTOR_SUBPARTS.
24852 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
24853 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
24854 (build_vector_from_val): If the number of units is variable,
24855 use build_vec_duplicate_cst for constant operands and
24856 VEC_DUPLICATE_EXPR otherwise.
24857 (make_vector_type): Remove temporary is_constant ().
24858 (build_vector_type, build_opaque_vector_type): Take the number of
24859 units as a poly_uint64 rather than an int.
24860 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
24862 * cfgexpand.c (expand_debug_expr): Likewise.
24863 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
24864 (store_constructor, expand_expr_real_1): Likewise.
24865 (const_scalar_mask_from_tree): Likewise.
24866 * fold-const-call.c (fold_const_reduction): Likewise.
24867 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
24868 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
24869 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
24870 (fold_relational_const): Likewise.
24871 (native_interpret_vector): Likewise. Change the size from an
24872 int to an unsigned int.
24873 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
24874 TYPE_VECTOR_SUBPARTS.
24875 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
24876 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
24877 duplicating a non-constant operand into a variable-length vector.
24878 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
24879 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
24880 * ipa-icf.c (sem_variable::equals): Likewise.
24881 * match.pd: Likewise.
24882 * omp-simd-clone.c (simd_clone_subparts): Likewise.
24883 * print-tree.c (print_node): Likewise.
24884 * stor-layout.c (layout_type): Likewise.
24885 * targhooks.c (default_builtin_vectorization_cost): Likewise.
24886 * tree-cfg.c (verify_gimple_comparison): Likewise.
24887 (verify_gimple_assign_binary): Likewise.
24888 (verify_gimple_assign_ternary): Likewise.
24889 (verify_gimple_assign_single): Likewise.
24890 * tree-pretty-print.c (dump_generic_node): Likewise.
24891 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24892 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
24893 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
24894 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
24895 (vect_shift_permute_load_chain): Likewise.
24896 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
24897 (expand_vector_condition, optimize_vector_constructor): Likewise.
24898 (lower_vec_perm, get_compute_type): Likewise.
24899 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24900 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
24901 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
24902 (vect_recog_mask_conversion_pattern): Likewise.
24903 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
24904 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
24905 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24906 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
24907 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
24908 (vectorizable_shift, vectorizable_operation, vectorizable_store)
24909 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
24910 (supportable_widening_operation): Likewise.
24911 (supportable_narrowing_operation): Likewise.
24912 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
24914 * varasm.c (output_constant): Likewise.
24916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24917 Alan Hayward <alan.hayward@arm.com>
24918 David Sherwood <david.sherwood@arm.com>
24920 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
24921 so that both the length == 3 and length != 3 cases set up their
24922 own permute vectors. Add comments explaining why we know the
24923 number of elements is constant.
24924 (vect_permute_load_chain): Likewise.
24926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24927 Alan Hayward <alan.hayward@arm.com>
24928 David Sherwood <david.sherwood@arm.com>
24930 * machmode.h (mode_nunits): Change from unsigned char to
24932 (ONLY_FIXED_SIZE_MODES): New macro.
24933 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
24934 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
24935 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
24937 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
24938 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
24939 or if measurement_type is not polynomial.
24940 * genmodes.c (ZERO_COEFFS): New macro.
24941 (emit_mode_nunits_inline): Make mode_nunits_inline return a
24943 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
24944 Use ZERO_COEFFS when emitting initializers.
24945 * data-streamer.h (bp_pack_poly_value): New function.
24946 (bp_unpack_poly_value): Likewise.
24947 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24948 for GET_MODE_NUNITS.
24949 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24950 for GET_MODE_NUNITS.
24951 * tree.c (make_vector_type): Remove temporary shim and make
24952 the real function take the number of units as a poly_uint64
24953 rather than an int.
24954 (build_vector_type_for_mode): Handle polynomial nunits.
24955 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
24956 * emit-rtl.c (const_vec_series_p_1): Likewise.
24957 (gen_rtx_CONST_VECTOR): Likewise.
24958 * fold-const.c (test_vec_duplicate_folding): Likewise.
24959 * genrecog.c (validate_pattern): Likewise.
24960 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
24961 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24962 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
24963 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
24964 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
24965 * rtlanal.c (subreg_get_info): Likewise.
24966 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24967 (vect_grouped_load_supported): Likewise.
24968 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
24969 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
24970 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
24971 (simplify_const_unary_operation, simplify_binary_operation_1)
24972 (simplify_const_binary_operation, simplify_ternary_operation)
24973 (test_vector_ops_duplicate, test_vector_ops): Likewise.
24974 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
24975 instead of CONST_VECTOR_NUNITS.
24976 * varasm.c (output_constant_pool_2): Likewise.
24977 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
24978 explicit-encoded elements in the XVEC for variable-length vectors.
24980 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24982 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
24984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24985 Alan Hayward <alan.hayward@arm.com>
24986 David Sherwood <david.sherwood@arm.com>
24988 * coretypes.h (fixed_size_mode): Declare.
24989 (fixed_size_mode_pod): New typedef.
24990 * builtins.h (target_builtins::x_apply_args_mode)
24991 (target_builtins::x_apply_result_mode): Change type to
24992 fixed_size_mode_pod.
24993 * builtins.c (apply_args_size, apply_result_size, result_vector)
24994 (expand_builtin_apply_args_1, expand_builtin_apply)
24995 (expand_builtin_return): Update accordingly.
24997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24999 * cse.c (hash_rtx_cb): Hash only the encoded elements.
25000 * cselib.c (cselib_hash_rtx): Likewise.
25001 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
25002 CONST_VECTOR encoding.
25004 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25005 Jeff Law <law@redhat.com>
25008 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
25009 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
25010 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
25011 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
25014 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
25015 explicitly probe *sp in a noreturn function if there were any callee
25016 register saves or frame pointer is needed.
25018 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25021 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
25022 BLKmode for ternary, binary or unary expressions.
25025 * var-tracking.c (delete_vta_debug_insn): New inline function.
25026 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
25027 insns from get_insns () to NULL instead of each bb separately.
25028 Use delete_vta_debug_insn. No longer static.
25029 (vt_debug_insns_local, variable_tracking_main_1): Adjust
25030 delete_vta_debug_insns callers.
25031 * rtl.h (delete_vta_debug_insns): Declare.
25032 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
25033 instead of variable_tracking_main.
25035 2018-01-03 Martin Sebor <msebor@redhat.com>
25037 PR tree-optimization/83603
25038 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
25039 arguments past the endof the argument list in functions declared
25040 without a prototype.
25041 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
25042 Avoid checking when arguments are null.
25044 2018-01-03 Martin Sebor <msebor@redhat.com>
25047 * doc/extend.texi (attribute const): Fix a typo.
25048 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
25049 issuing -Wsuggest-attribute for void functions.
25051 2018-01-03 Martin Sebor <msebor@redhat.com>
25053 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
25054 offset_int::from instead of wide_int::to_shwi.
25055 (maybe_diag_overlap): Remove assertion.
25056 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
25057 * gimple-ssa-sprintf.c (format_directive): Same.
25058 (parse_directive): Same.
25059 (sprintf_dom_walker::compute_format_length): Same.
25060 (try_substitute_return_value): Same.
25062 2018-01-03 Jeff Law <law@redhat.com>
25064 PR middle-end/83654
25065 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
25066 non-constant residual for zero at runtime and avoid probing in
25067 that case. Reorganize code for trailing problem to mirror handling
25070 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
25072 PR tree-optimization/83501
25073 * tree-ssa-strlen.c (get_string_cst): New.
25074 (handle_char_store): Call get_string_cst.
25076 2018-01-03 Martin Liska <mliska@suse.cz>
25078 PR tree-optimization/83593
25079 * tree-ssa-strlen.c: Include tree-cfg.h.
25080 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
25081 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
25082 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
25084 (strlen_dom_walker::before_dom_children): Call
25085 gimple_purge_dead_eh_edges. Dump tranformation with details
25087 (strlen_dom_walker::before_dom_children): Update call by adding
25088 new argument cleanup_eh.
25089 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
25091 2018-01-03 Martin Liska <mliska@suse.cz>
25094 * cif-code.def (VARIADIC_THUNK): New enum value.
25095 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
25098 2018-01-03 Jan Beulich <jbeulich@suse.com>
25100 * sse.md (mov<mode>_internal): Tighten condition for when to use
25101 vmovdqu<ssescalarsize> for TI and OI modes.
25103 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25105 Update copyright years.
25107 2018-01-03 Martin Liska <mliska@suse.cz>
25110 * ipa-visibility.c (function_and_variable_visibility): Skip
25111 functions with noipa attribure.
25113 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25115 * gcc.c (process_command): Update copyright notice dates.
25116 * gcov-dump.c (print_version): Ditto.
25117 * gcov.c (print_version): Ditto.
25118 * gcov-tool.c (print_version): Ditto.
25119 * gengtype.c (create_file): Ditto.
25120 * doc/cpp.texi: Bump @copying's copyright year.
25121 * doc/cppinternals.texi: Ditto.
25122 * doc/gcc.texi: Ditto.
25123 * doc/gccint.texi: Ditto.
25124 * doc/gcov.texi: Ditto.
25125 * doc/install.texi: Ditto.
25126 * doc/invoke.texi: Ditto.
25128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25130 * vector-builder.h (vector_builder::m_full_nelts): Change from
25131 unsigned int to poly_uint64.
25132 (vector_builder::full_nelts): Update prototype accordingly.
25133 (vector_builder::new_vector): Likewise.
25134 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
25135 (vector_builder::operator ==): Likewise.
25136 (vector_builder::finalize): Likewise.
25137 * int-vector-builder.h (int_vector_builder::int_vector_builder):
25138 Take the number of elements as a poly_uint64 rather than an
25140 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
25141 from unsigned int to poly_uint64.
25142 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
25143 (vec_perm_indices::new_vector): Likewise.
25144 (vec_perm_indices::length): Likewise.
25145 (vec_perm_indices::nelts_per_input): Likewise.
25146 (vec_perm_indices::input_nelts): Likewise.
25147 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
25148 number of elements per input as a poly_uint64 rather than an
25149 unsigned int. Use the original encoding for variable-length
25150 vectors, rather than clamping each individual element.
25151 For the second and subsequent elements in each pattern,
25152 clamp the step and base before clamping their sum.
25153 (vec_perm_indices::series_p): Handle polynomial element counts.
25154 (vec_perm_indices::all_in_range_p): Likewise.
25155 (vec_perm_indices_to_tree): Likewise.
25156 (vec_perm_indices_to_rtx): Likewise.
25157 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
25158 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
25159 (tree_vector_builder::new_binary_operation): Handle polynomial
25160 element counts. Return false if we need to know the number
25161 of elements at compile time.
25162 * fold-const.c (fold_vec_perm): Punt if the number of elements
25163 isn't known at compile time.
25165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25167 * vec-perm-indices.h (vec_perm_builder): Change element type
25168 from HOST_WIDE_INT to poly_int64.
25169 (vec_perm_indices::element_type): Update accordingly.
25170 (vec_perm_indices::clamp): Handle polynomial element_types.
25171 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25172 (vec_perm_indices::all_in_range_p): Likewise.
25173 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
25175 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
25176 polynomial vec_perm_indices element types.
25177 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
25178 * fold-const.c (fold_vec_perm): Likewise.
25179 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
25180 * tree-vect-generic.c (lower_vec_perm): Likewise.
25181 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25182 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
25183 element type to HOST_WIDE_INT.
25185 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25186 Alan Hayward <alan.hayward@arm.com>
25187 David Sherwood <david.sherwood@arm.com>
25189 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
25190 rather than an int. Use plus_constant.
25191 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
25192 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
25194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25195 Alan Hayward <alan.hayward@arm.com>
25196 David Sherwood <david.sherwood@arm.com>
25198 * calls.c (emit_call_1, expand_call): Change struct_value_size from
25199 a HOST_WIDE_INT to a poly_int64.
25201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25202 Alan Hayward <alan.hayward@arm.com>
25203 David Sherwood <david.sherwood@arm.com>
25205 * calls.c (load_register_parameters): Cope with polynomial
25206 mode sizes. Require a constant size for BLKmode parameters
25207 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
25208 forces a parameter to be padded at the lsb end in order to
25209 fill a complete number of words, require the parameter size
25210 to be ordered wrt UNITS_PER_WORD.
25212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25213 Alan Hayward <alan.hayward@arm.com>
25214 David Sherwood <david.sherwood@arm.com>
25216 * reload1.c (spill_stack_slot_width): Change element type
25217 from unsigned int to poly_uint64_pod.
25218 (alter_reg): Treat mode sizes as polynomial.
25220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25221 Alan Hayward <alan.hayward@arm.com>
25222 David Sherwood <david.sherwood@arm.com>
25224 * reload.c (complex_word_subreg_p): New function.
25225 (reload_inner_reg_of_subreg, push_reload): Use it.
25227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25228 Alan Hayward <alan.hayward@arm.com>
25229 David Sherwood <david.sherwood@arm.com>
25231 * lra-constraints.c (process_alt_operands): Reject matched
25232 operands whose sizes aren't ordered.
25233 (match_reload): Refer to this check here.
25235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25236 Alan Hayward <alan.hayward@arm.com>
25237 David Sherwood <david.sherwood@arm.com>
25239 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
25240 that the mode size is in the set {1, 2, 4, 8, 16}.
25242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25243 Alan Hayward <alan.hayward@arm.com>
25244 David Sherwood <david.sherwood@arm.com>
25246 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
25247 Use plus_constant instead of gen_rtx_PLUS.
25249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25250 Alan Hayward <alan.hayward@arm.com>
25251 David Sherwood <david.sherwood@arm.com>
25253 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
25254 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
25255 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
25256 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
25257 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
25258 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
25259 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
25260 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
25261 * config/i386/i386.c (ix86_push_rounding): ...this new function.
25262 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
25264 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
25265 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
25266 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
25267 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
25268 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
25269 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
25270 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
25271 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
25272 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
25273 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
25275 * expr.c (emit_move_resolve_push): Treat the input and result
25276 of PUSH_ROUNDING as a poly_int64.
25277 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
25278 (emit_push_insn): Likewise.
25279 * lra-eliminations.c (mark_not_eliminable): Likewise.
25280 * recog.c (push_operand): Likewise.
25281 * reload1.c (elimination_effects): Likewise.
25282 * rtlanal.c (nonzero_bits1): Likewise.
25283 * calls.c (store_one_arg): Likewise. Require the padding to be
25284 known at compile time.
25286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25287 Alan Hayward <alan.hayward@arm.com>
25288 David Sherwood <david.sherwood@arm.com>
25290 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
25291 Use plus_constant instead of gen_rtx_PLUS.
25293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25294 Alan Hayward <alan.hayward@arm.com>
25295 David Sherwood <david.sherwood@arm.com>
25297 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
25298 rather than an int.
25300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25301 Alan Hayward <alan.hayward@arm.com>
25302 David Sherwood <david.sherwood@arm.com>
25304 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
25305 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
25306 via stack temporaries. Treat the mode size as polynomial too.
25308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25309 Alan Hayward <alan.hayward@arm.com>
25310 David Sherwood <david.sherwood@arm.com>
25312 * expr.c (expand_expr_real_2): When handling conversions involving
25313 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
25314 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
25315 as a poly_uint64 too.
25317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25318 Alan Hayward <alan.hayward@arm.com>
25319 David Sherwood <david.sherwood@arm.com>
25321 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
25323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25324 Alan Hayward <alan.hayward@arm.com>
25325 David Sherwood <david.sherwood@arm.com>
25327 * combine.c (can_change_dest_mode): Handle polynomial
25328 REGMODE_NATURAL_SIZE.
25329 * expmed.c (store_bit_field_1): Likewise.
25330 * expr.c (store_constructor): Likewise.
25331 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
25332 and polynomial REGMODE_NATURAL_SIZE.
25333 (gen_lowpart_common): Likewise.
25334 * reginfo.c (record_subregs_of_mode): Likewise.
25335 * rtlanal.c (read_modify_subreg_p): Likewise.
25337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25338 Alan Hayward <alan.hayward@arm.com>
25339 David Sherwood <david.sherwood@arm.com>
25341 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
25342 numbers of elements.
25344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25345 Alan Hayward <alan.hayward@arm.com>
25346 David Sherwood <david.sherwood@arm.com>
25348 * match.pd: Cope with polynomial numbers of vector elements.
25350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25351 Alan Hayward <alan.hayward@arm.com>
25352 David Sherwood <david.sherwood@arm.com>
25354 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
25355 in a POINTER_PLUS_EXPR.
25357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25358 Alan Hayward <alan.hayward@arm.com>
25359 David Sherwood <david.sherwood@arm.com>
25361 * omp-simd-clone.c (simd_clone_subparts): New function.
25362 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
25363 (ipa_simd_modify_function_body): Likewise.
25365 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25366 Alan Hayward <alan.hayward@arm.com>
25367 David Sherwood <david.sherwood@arm.com>
25369 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
25370 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
25371 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
25372 (expand_vector_condition, vector_element): Likewise.
25373 (subparts_gt): New function.
25374 (get_compute_type): Use subparts_gt.
25375 (count_type_subparts): Delete.
25376 (expand_vector_operations_1): Use subparts_gt instead of
25377 count_type_subparts.
25379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25380 Alan Hayward <alan.hayward@arm.com>
25381 David Sherwood <david.sherwood@arm.com>
25383 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
25384 (vect_compile_time_alias): ...this new function. Do the calculation
25385 on poly_ints rather than trees.
25386 (vect_prune_runtime_alias_test_list): Update call accordingly.
25388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25389 Alan Hayward <alan.hayward@arm.com>
25390 David Sherwood <david.sherwood@arm.com>
25392 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
25394 (vect_schedule_slp_instance): Likewise.
25396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25397 Alan Hayward <alan.hayward@arm.com>
25398 David Sherwood <david.sherwood@arm.com>
25400 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
25401 constant and extern definitions for variable-length vectors.
25402 (vect_get_constant_vectors): Note that the number of units
25403 is known to be constant.
25405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25406 Alan Hayward <alan.hayward@arm.com>
25407 David Sherwood <david.sherwood@arm.com>
25409 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
25410 of units as polynomial. Choose between WIDE and NARROW based
25413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25414 Alan Hayward <alan.hayward@arm.com>
25415 David Sherwood <david.sherwood@arm.com>
25417 * tree-vect-stmts.c (simd_clone_subparts): New function.
25418 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
25420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25421 Alan Hayward <alan.hayward@arm.com>
25422 David Sherwood <david.sherwood@arm.com>
25424 * tree-vect-stmts.c (vectorizable_call): Treat the number of
25425 vectors as polynomial. Use build_index_vector for
25426 IFN_GOMP_SIMD_LANE.
25428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25429 Alan Hayward <alan.hayward@arm.com>
25430 David Sherwood <david.sherwood@arm.com>
25432 * tree-vect-stmts.c (get_load_store_type): Treat the number of
25433 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
25434 for variable-length vectors.
25435 (vectorizable_mask_load_store): Treat the number of units as
25436 polynomial, asserting that it is constant if the condition has
25437 already been enforced.
25438 (vectorizable_store, vectorizable_load): Likewise.
25440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25441 Alan Hayward <alan.hayward@arm.com>
25442 David Sherwood <david.sherwood@arm.com>
25444 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
25445 of units as polynomial. Punt if we can't tell at compile time
25446 which vector contains the final result.
25448 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25449 Alan Hayward <alan.hayward@arm.com>
25450 David Sherwood <david.sherwood@arm.com>
25452 * tree-vect-loop.c (vectorizable_induction): Treat the number
25453 of units as polynomial. Punt on SLP inductions. Use an integer
25454 VEC_SERIES_EXPR for variable-length integer reductions. Use a
25455 cast of such a series for variable-length floating-point
25458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25459 Alan Hayward <alan.hayward@arm.com>
25460 David Sherwood <david.sherwood@arm.com>
25462 * tree.h (build_index_vector): Declare.
25463 * tree.c (build_index_vector): New function.
25464 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
25465 of units as polynomial, forcibly converting it to a constant if
25466 vectorizable_reduction has already enforced the condition.
25467 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
25468 to create a {1,2,3,...} vector.
25469 (vectorizable_reduction): Treat the number of units as polynomial.
25470 Choose vectype_in based on the largest scalar element size rather
25471 than the smallest number of units. Enforce the restrictions
25474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25475 Alan Hayward <alan.hayward@arm.com>
25476 David Sherwood <david.sherwood@arm.com>
25478 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
25479 number of units as polynomial.
25481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25482 Alan Hayward <alan.hayward@arm.com>
25483 David Sherwood <david.sherwood@arm.com>
25485 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
25486 * target.def (autovectorize_vector_sizes): Return the vector sizes
25487 by pointer, using vector_sizes rather than a bitmask.
25488 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
25489 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
25490 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
25492 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
25493 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
25494 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
25495 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
25496 * omp-general.c (omp_max_vf): Likewise.
25497 * omp-low.c (omp_clause_aligned_alignment): Likewise.
25498 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
25499 * tree-vect-loop.c (vect_analyze_loop): Likewise.
25500 * tree-vect-slp.c (vect_slp_bb): Likewise.
25501 * doc/tm.texi: Regenerate.
25502 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
25504 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
25505 the vector size as a poly_uint64 rather than an unsigned int.
25506 (current_vector_size): Change from an unsigned int to a poly_uint64.
25507 (get_vectype_for_scalar_type): Update accordingly.
25508 * tree.h (build_truth_vector_type): Take the size and number of
25509 units as a poly_uint64 rather than an unsigned int.
25510 (build_vector_type): Add a temporary overload that takes
25511 the number of units as a poly_uint64 rather than an unsigned int.
25512 * tree.c (make_vector_type): Likewise.
25513 (build_truth_vector_type): Take the number of units as a poly_uint64
25514 rather than an unsigned int.
25516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25517 Alan Hayward <alan.hayward@arm.com>
25518 David Sherwood <david.sherwood@arm.com>
25520 * target.def (get_mask_mode): Take the number of units and length
25521 as poly_uint64s rather than unsigned ints.
25522 * targhooks.h (default_get_mask_mode): Update accordingly.
25523 * targhooks.c (default_get_mask_mode): Likewise.
25524 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
25525 * doc/tm.texi: Regenerate.
25527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25528 Alan Hayward <alan.hayward@arm.com>
25529 David Sherwood <david.sherwood@arm.com>
25531 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
25532 * omp-general.c (omp_max_vf): Likewise.
25533 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
25534 (expand_omp_simd): Handle polynomial safelen.
25535 * omp-low.c (omplow_simd_context): Add a default constructor.
25536 (omplow_simd_context::max_vf): Change from int to poly_uint64.
25537 (lower_rec_simd_input_clauses): Update accordingly.
25538 (lower_rec_input_clauses): Likewise.
25540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25541 Alan Hayward <alan.hayward@arm.com>
25542 David Sherwood <david.sherwood@arm.com>
25544 * tree-vectorizer.h (vect_nunits_for_cost): New function.
25545 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
25546 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
25547 (vect_analyze_slp_cost): Likewise.
25548 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
25549 (vect_model_load_cost): Likewise.
25551 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25552 Alan Hayward <alan.hayward@arm.com>
25553 David Sherwood <david.sherwood@arm.com>
25555 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
25556 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
25557 from an unsigned int * to a poly_uint64_pod *.
25558 (calculate_unrolling_factor): New function.
25559 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
25561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25562 Alan Hayward <alan.hayward@arm.com>
25563 David Sherwood <david.sherwood@arm.com>
25565 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
25566 from an unsigned int to a poly_uint64.
25567 (_loop_vec_info::slp_unrolling_factor): Likewise.
25568 (_loop_vec_info::vectorization_factor): Change from an int
25570 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
25571 (vect_get_num_vectors): New function.
25572 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
25573 (vect_get_num_copies): Use vect_get_num_vectors.
25574 (vect_analyze_data_ref_dependences): Change max_vf from an int *
25575 to an unsigned int *.
25576 (vect_analyze_data_refs): Change min_vf from an int * to a
25578 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25579 than an unsigned HOST_WIDE_INT.
25580 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
25581 (vect_analyze_data_ref_dependence): Change max_vf from an int *
25582 to an unsigned int *.
25583 (vect_analyze_data_ref_dependences): Likewise.
25584 (vect_compute_data_ref_alignment): Handle polynomial vf.
25585 (vect_enhance_data_refs_alignment): Likewise.
25586 (vect_prune_runtime_alias_test_list): Likewise.
25587 (vect_shift_permute_load_chain): Likewise.
25588 (vect_supportable_dr_alignment): Likewise.
25589 (dependence_distance_ge_vf): Take the vectorization factor as a
25590 poly_uint64 rather than an unsigned HOST_WIDE_INT.
25591 (vect_analyze_data_refs): Change min_vf from an int * to a
25593 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
25594 vfm1 as a poly_uint64 rather than an int. Make the same change
25595 for the returned bound_scalar.
25596 (vect_gen_vector_loop_niters): Handle polynomial vf.
25597 (vect_do_peeling): Likewise. Update call to
25598 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
25599 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
25601 * tree-vect-loop.c (vect_determine_vectorization_factor)
25602 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
25603 (vect_get_known_peeling_cost): Likewise.
25604 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
25605 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
25606 (vect_transform_loop): Likewise. Use the lowest possible VF when
25607 updating the upper bounds of the loop.
25608 (vect_min_worthwhile_factor): Make static. Return an unsigned int
25609 rather than an int.
25610 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
25611 polynomial unroll factors.
25612 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
25613 (vect_make_slp_decision): Likewise.
25614 (vect_supported_load_permutation_p): Likewise, and polynomial
25616 (vect_analyze_slp_cost): Handle polynomial vf.
25617 (vect_slp_analyze_node_operations): Likewise.
25618 (vect_slp_analyze_bb_1): Likewise.
25619 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25620 than an unsigned HOST_WIDE_INT.
25621 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
25622 (vectorizable_load): Handle polynomial vf.
25623 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
25625 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
25627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25628 Alan Hayward <alan.hayward@arm.com>
25629 David Sherwood <david.sherwood@arm.com>
25631 * match.pd: Handle bit operations involving three constants
25632 and try to fold one pair.
25634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25636 * tree-vect-loop-manip.c: Include gimple-fold.h.
25637 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
25638 niters_maybe_zero parameters. Handle other cases besides a step of 1.
25639 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
25640 Add a path that uses a step of VF instead of 1, but disable it
25642 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
25643 and niters_no_overflow parameters. Update calls to
25644 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
25645 Create a new SSA name if the latter choses to use a ste other
25646 than zero, and return it via niters_vector_mult_vf_var.
25647 * tree-vect-loop.c (vect_transform_loop): Update calls to
25648 vect_do_peeling, vect_gen_vector_loop_niters and
25649 slpeel_make_loop_iterate_ntimes.
25650 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
25651 (vect_gen_vector_loop_niters): Update declarations after above changes.
25653 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
25655 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
25656 128-bit round to integer instructions.
25657 (ceil<mode>2): Likewise.
25658 (btrunc<mode>2): Likewise.
25659 (round<mode>2): Likewise.
25661 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25663 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
25664 unaligned VSX load/store on P8/P9.
25665 (expand_block_clear): Allow the use of unaligned VSX
25666 load/store on P8/P9.
25668 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25670 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
25672 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
25673 swap associated with both a load and a store.
25675 2018-01-02 Andrew Waterman <andrew@sifive.com>
25677 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
25678 * config/riscv/riscv.md (clear_cache): Use it.
25680 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
25682 * web.c: Remove out-of-date comment.
25684 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25686 * expr.c (fixup_args_size_notes): Check that any existing
25687 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
25688 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
25689 (emit_single_push_insn): ...here.
25691 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25693 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
25694 (const_vector_encoded_nelts): New function.
25695 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
25696 (const_vector_int_elt, const_vector_elt): Declare.
25697 * emit-rtl.c (const_vector_int_elt_1): New function.
25698 (const_vector_elt): Likewise.
25699 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
25700 of CONST_VECTOR_ELT.
25702 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25704 * expr.c: Include rtx-vector-builder.h.
25705 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
25706 directly on the tree encoding.
25707 (const_vector_from_tree): Likewise.
25708 * optabs.c: Include rtx-vector-builder.h.
25709 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
25710 sequence of "u" values.
25711 * vec-perm-indices.c: Include rtx-vector-builder.h.
25712 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
25713 directly on the vec_perm_indices encoding.
25715 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25717 * doc/rtl.texi (const_vector): Describe new encoding scheme.
25718 * Makefile.in (OBJS): Add rtx-vector-builder.o.
25719 * rtx-vector-builder.h: New file.
25720 * rtx-vector-builder.c: Likewise.
25721 * rtl.h (rtx_def::u2): Add a const_vector field.
25722 (CONST_VECTOR_NPATTERNS): New macro.
25723 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
25724 (CONST_VECTOR_DUPLICATE_P): Likewise.
25725 (CONST_VECTOR_STEPPED_P): Likewise.
25726 (CONST_VECTOR_ENCODED_ELT): Likewise.
25727 (const_vec_duplicate_p): Check for a duplicated vector encoding.
25728 (unwrap_const_vec_duplicate): Likewise.
25729 (const_vec_series_p): Check for a non-duplicated vector encoding.
25730 Say that the function only returns true for integer vectors.
25731 * emit-rtl.c: Include rtx-vector-builder.h.
25732 (gen_const_vec_duplicate_1): Delete.
25733 (gen_const_vector): Call gen_const_vec_duplicate instead of
25734 gen_const_vec_duplicate_1.
25735 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
25736 (gen_const_vec_duplicate): Use rtx_vector_builder.
25737 (gen_const_vec_series): Likewise.
25738 (gen_rtx_CONST_VECTOR): Likewise.
25739 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
25740 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25741 Build a new vector rather than modifying a CONST_VECTOR in-place.
25742 (handle_special_swappables): Update call accordingly.
25743 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
25744 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25745 Build a new vector rather than modifying a CONST_VECTOR in-place.
25746 (handle_special_swappables): Update call accordingly.
25748 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25750 * simplify-rtx.c (simplify_const_binary_operation): Use
25751 CONST_VECTOR_ELT instead of XVECEXP.
25753 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25755 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
25756 the selector elements to be different from the data elements
25757 if the selector is a VECTOR_CST.
25758 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
25759 ssizetype for the selector.
25761 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25763 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
25764 before testing each element individually.
25765 * tree-vect-generic.c (lower_vec_perm): Likewise.
25767 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25769 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
25770 * selftest-run-tests.c (selftest::run_tests): Call it.
25771 * vector-builder.h (vector_builder::operator ==): New function.
25772 (vector_builder::operator !=): Likewise.
25773 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
25774 (vec_perm_indices::all_from_input_p): New function.
25775 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25776 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
25777 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
25778 instead of reading the VECTOR_CST directly. Detect whether both
25779 vector inputs are the same before constructing the vec_perm_indices,
25780 and update the number of inputs argument accordingly. Use the
25781 utility functions added above. Only construct sel2 if we need to.
25783 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25785 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
25786 the broadcast of the low byte.
25787 (expand_mult_highpart): Use an explicit encoding for the permutes.
25788 * optabs-query.c (can_mult_highpart_p): Likewise.
25789 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25790 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25791 (vectorizable_bswap): Likewise.
25792 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
25793 explicit encoding for the power-of-2 permutes.
25794 (vect_permute_store_chain): Likewise.
25795 (vect_grouped_load_supported): Likewise.
25796 (vect_permute_load_chain): Likewise.
25798 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25800 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
25801 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
25802 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
25803 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25804 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
25805 (vect_gen_perm_mask_any): Likewise.
25807 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25809 * int-vector-builder.h: New file.
25810 * vec-perm-indices.h: Include int-vector-builder.h.
25811 (vec_perm_indices): Redefine as an int_vector_builder.
25812 (auto_vec_perm_indices): Delete.
25813 (vec_perm_builder): Redefine as a stand-alone class.
25814 (vec_perm_indices::vec_perm_indices): New function.
25815 (vec_perm_indices::clamp): Likewise.
25816 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
25817 (vec_perm_indices::new_vector): New function.
25818 (vec_perm_indices::new_expanded_vector): Update for new
25819 vec_perm_indices class.
25820 (vec_perm_indices::rotate_inputs): New function.
25821 (vec_perm_indices::all_in_range_p): Operate directly on the
25822 encoded form, without computing elided elements.
25823 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
25824 encoding. Update for new vec_perm_indices class.
25825 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
25826 the given vec_perm_builder.
25827 (expand_vec_perm_var): Update vec_perm_builder constructor.
25828 (expand_mult_highpart): Use vec_perm_builder instead of
25829 auto_vec_perm_indices.
25830 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
25831 vec_perm_indices instead of auto_vec_perm_indices. Use a single
25832 or double series encoding as appropriate.
25833 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
25834 vec_perm_indices instead of auto_vec_perm_indices.
25835 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25836 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25837 (vect_permute_store_chain): Likewise.
25838 (vect_grouped_load_supported): Likewise.
25839 (vect_permute_load_chain): Likewise.
25840 (vect_shift_permute_load_chain): Likewise.
25841 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
25842 (vect_transform_slp_perm_load): Likewise.
25843 (vect_schedule_slp_instance): Likewise.
25844 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25845 (vectorizable_mask_load_store): Likewise.
25846 (vectorizable_bswap): Likewise.
25847 (vectorizable_store): Likewise.
25848 (vectorizable_load): Likewise.
25849 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
25850 vec_perm_indices instead of auto_vec_perm_indices. Use
25851 tree_to_vec_perm_builder to read the vector from a tree.
25852 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
25853 vec_perm_builder instead of a vec_perm_indices.
25854 (have_whole_vector_shift): Use vec_perm_builder and
25855 vec_perm_indices instead of auto_vec_perm_indices. Leave the
25856 truncation to calc_vec_perm_mask_for_shift.
25857 (vect_create_epilog_for_reduction): Likewise.
25858 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
25859 from auto_vec_perm_indices to vec_perm_indices.
25860 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25861 instead of changing individual elements.
25862 (aarch64_vectorize_vec_perm_const): Use new_vector to install
25863 the vector in d.perm.
25864 * config/arm/arm.c (expand_vec_perm_d::perm): Change
25865 from auto_vec_perm_indices to vec_perm_indices.
25866 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25867 instead of changing individual elements.
25868 (arm_vectorize_vec_perm_const): Use new_vector to install
25869 the vector in d.perm.
25870 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
25871 Update vec_perm_builder constructor.
25872 (rs6000_expand_interleave): Likewise.
25873 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
25874 (rs6000_expand_interleave): Likewise.
25876 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25878 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
25879 to qimode could truncate the indices.
25880 * optabs.c (expand_vec_perm_var): Likewise.
25882 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25884 * Makefile.in (OBJS): Add vec-perm-indices.o.
25885 * vec-perm-indices.h: New file.
25886 * vec-perm-indices.c: Likewise.
25887 * target.h (vec_perm_indices): Replace with a forward class
25889 (auto_vec_perm_indices): Move to vec-perm-indices.h.
25890 * optabs.h: Include vec-perm-indices.h.
25891 (expand_vec_perm): Delete.
25892 (selector_fits_mode_p, expand_vec_perm_var): Declare.
25893 (expand_vec_perm_const): Declare.
25894 * target.def (vec_perm_const_ok): Replace with...
25895 (vec_perm_const): ...this new hook.
25896 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
25897 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
25898 * doc/tm.texi: Regenerate.
25899 * optabs.def (vec_perm_const): Delete.
25900 * doc/md.texi (vec_perm_const): Likewise.
25901 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
25902 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
25903 expand_vec_perm for constant permutation vectors. Assert that
25904 the mode of variable permutation vectors is the integer equivalent
25905 of the mode that is being permuted.
25906 * optabs-query.h (selector_fits_mode_p): Declare.
25907 * optabs-query.c: Include vec-perm-indices.h.
25908 (selector_fits_mode_p): New function.
25909 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
25910 is defined, instead of checking whether the vec_perm_const_optab
25911 exists. Use targetm.vectorize.vec_perm_const instead of
25912 targetm.vectorize.vec_perm_const_ok. Check whether the indices
25913 fit in the vector mode before using a variable permute.
25914 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
25915 vec_perm_indices instead of an rtx.
25916 (expand_vec_perm): Replace with...
25917 (expand_vec_perm_const): ...this new function. Take the selector
25918 as a vec_perm_indices rather than an rtx. Also take the mode of
25919 the selector. Update call to shift_amt_for_vec_perm_mask.
25920 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
25921 Use vec_perm_indices::new_expanded_vector to expand the original
25922 selector into bytes. Check whether the indices fit in the vector
25923 mode before using a variable permute.
25924 (expand_vec_perm_var): Make global.
25925 (expand_mult_highpart): Use expand_vec_perm_const.
25926 * fold-const.c: Includes vec-perm-indices.h.
25927 * tree-ssa-forwprop.c: Likewise.
25928 * tree-vect-data-refs.c: Likewise.
25929 * tree-vect-generic.c: Likewise.
25930 * tree-vect-loop.c: Likewise.
25931 * tree-vect-slp.c: Likewise.
25932 * tree-vect-stmts.c: Likewise.
25933 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
25935 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
25936 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
25937 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
25938 (aarch64_vectorize_vec_perm_const): ...this new function.
25939 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25940 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25941 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
25942 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
25943 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25944 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25945 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
25947 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
25948 check for NEON modes.
25949 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
25950 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
25951 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
25952 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
25954 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
25955 the old VEC_PERM_CONST conditions.
25956 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
25957 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
25958 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
25959 (ia64_vectorize_vec_perm_const_ok): Merge into...
25960 (ia64_vectorize_vec_perm_const): ...this new function.
25961 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
25962 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
25963 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
25964 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
25965 * config/mips/mips.c (mips_expand_vec_perm_const)
25966 (mips_vectorize_vec_perm_const_ok): Merge into...
25967 (mips_vectorize_vec_perm_const): ...this new function.
25968 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
25969 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
25970 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
25971 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
25972 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
25973 (rs6000_expand_vec_perm_const): Delete.
25974 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
25976 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25977 (altivec_expand_vec_perm_const_le): Take each operand individually.
25978 Operate on constant selectors rather than rtxes.
25979 (altivec_expand_vec_perm_const): Likewise. Update call to
25980 altivec_expand_vec_perm_const_le.
25981 (rs6000_expand_vec_perm_const): Delete.
25982 (rs6000_vectorize_vec_perm_const_ok): Delete.
25983 (rs6000_vectorize_vec_perm_const): New function.
25984 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
25985 an element count and rtx array.
25986 (rs6000_expand_extract_even): Update call accordingly.
25987 (rs6000_expand_interleave): Likewise.
25988 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
25989 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
25990 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
25991 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
25992 (rs6000_expand_vec_perm_const): Delete.
25993 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25994 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25995 (altivec_expand_vec_perm_const_le): Take each operand individually.
25996 Operate on constant selectors rather than rtxes.
25997 (altivec_expand_vec_perm_const): Likewise. Update call to
25998 altivec_expand_vec_perm_const_le.
25999 (rs6000_expand_vec_perm_const): Delete.
26000 (rs6000_vectorize_vec_perm_const_ok): Delete.
26001 (rs6000_vectorize_vec_perm_const): New function. Remove stray
26002 reference to the SPE evmerge intructions.
26003 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
26004 an element count and rtx array.
26005 (rs6000_expand_extract_even): Update call accordingly.
26006 (rs6000_expand_interleave): Likewise.
26007 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
26008 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
26010 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26012 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26014 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
26015 vector mode and that that mode matches the mode of the data
26017 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
26018 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
26019 directly using expand_vec_perm_1 when forcing selectors into
26021 (expand_vec_perm_var): New function, split out from expand_vec_perm.
26023 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26025 * optabs-query.h (can_vec_perm_p): Delete.
26026 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
26027 * optabs-query.c (can_vec_perm_p): Split into...
26028 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
26029 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
26030 particular selector is valid.
26031 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26032 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26033 (vect_grouped_load_supported): Likewise.
26034 (vect_shift_permute_load_chain): Likewise.
26035 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
26036 (vect_transform_slp_perm_load): Likewise.
26037 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26038 (vectorizable_bswap): Likewise.
26039 (vect_gen_perm_mask_checked): Likewise.
26040 * fold-const.c (fold_ternary_loc): Likewise. Don't take
26041 implementations of variable permutation vectors into account
26042 when deciding which selector to use.
26043 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
26044 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
26045 with a false third argument.
26046 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
26047 to test whether the constant selector is valid and can_vec_perm_var_p
26048 to test whether a variable selector is valid.
26050 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26052 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
26053 * optabs-query.c (can_vec_perm_p): Likewise.
26054 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
26055 instead of vec_perm_indices.
26056 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
26057 (vect_gen_perm_mask_checked): Likewise,
26058 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
26059 (vect_gen_perm_mask_checked): Likewise,
26061 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26063 * optabs-query.h (qimode_for_vec_perm): Declare.
26064 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
26065 (qimode_for_vec_perm): ...this new function.
26066 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
26068 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26070 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
26071 does not have a conditional at the top.
26073 2018-01-02 Richard Biener <rguenther@suse.de>
26075 * ipa-inline.c (big_speedup_p): Fix expression.
26077 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
26080 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
26083 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
26087 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
26088 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
26089 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
26090 cond_taken_branch_cost 3->4.
26092 2018-01-01 Jakub Jelinek <jakub@redhat.com>
26094 PR tree-optimization/83581
26095 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
26096 TODO_cleanup_cfg if any changes have been made.
26098 PR middle-end/83608
26099 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
26100 convert_modes if target mode has the right side, but different mode
26103 PR middle-end/83609
26104 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
26105 last argument when extracting from CONCAT. If either from_real or
26106 from_imag is NULL, use expansion through memory. If result is not
26107 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
26108 the parts directly to inner mode, if even that fails, use expansion
26111 PR middle-end/83623
26112 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
26113 check for bswap in mode rather than HImode and use that in expand_unop
26116 Copyright (C) 2018 Free Software Foundation, Inc.
26118 Copying and distribution of this file, with or without modification,
26119 are permitted in any medium without royalty provided the copyright
26120 notice and this notice are preserved.