1 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
2 Julian Brown <julian@codesourcery.com>
4 * builtins.c (get_builtin_sync_mem): Handle address spaces.
6 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
8 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
9 if the call takes a static chain.
11 2018-09-21 Martin Liska <mliska@suse.cz>
13 * auto-profile.c (autofdo_source_profile::read): Do not
15 (read_profile): Do not add working sets.
16 (read_autofdo_file): Remove sum_all.
17 (afdo_callsite_hot_enough_for_early_inline): Remove const
19 * coverage.c (struct counts_entry): Remove gcov_summary.
20 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
21 do not support GCOV_TAG_PROGRAM_SUMMARY.
22 (get_coverage_counts): Remove summary and expected
24 * coverage.h (get_coverage_counts): Likewise.
25 * doc/gcov-dump.texi: Remove -w option.
26 * gcov-dump.c (dump_working_sets): Remove.
27 (main): Do not support '-w' option.
28 (print_usage): Likewise.
29 (tag_summary): Likewise.
30 * gcov-io.c (gcov_write_summary): Do not dump
32 (gcov_read_summary): Likewise.
33 (gcov_histo_index): Remove.
34 (gcov_histogram_merge): Likewise.
35 (compute_working_sets): Likewise.
36 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
38 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
39 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
40 (GCOV_HISTOGRAM_SIZE): Remove.
41 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
42 (struct gcov_summary): Simplify rapidly just
43 to runs and sum_max fields.
44 (gcov_histo_index): Remove.
45 (NUM_GCOV_WORKING_SETS): Likewise.
46 (compute_working_sets): Likewise.
47 * gcov-tool.c (print_overlap_usage_message): Remove
49 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
50 (output_lines): Remove program related line.
51 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
52 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
54 (input_profile_summary): Do not read it.
55 (merge_profile_summaries): And do not merge it.
56 (input_symtab): Do not call removed function.
57 * modulo-sched.c (sms_schedule): Do not print sum_max.
58 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
59 removed when histogram method was invented.
60 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
62 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
64 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
65 and dump selected value.
66 * profile.c (add_working_set): Remove.
67 (get_working_sets): Likewise.
68 (find_working_set): Likewise.
69 (get_exec_counts): Do not work with working sets.
70 (read_profile_edge_counts): Do not inform as sum_max is removed.
71 (compute_branch_probabilities): Likewise.
72 (compute_value_histograms): Remove argument for call of
74 * profile.h: Do not make gcov_summary const.
76 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
78 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
80 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
82 PR tree-optimization/86990
83 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
84 Check that the entire merged store group is made of constants only for
87 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
89 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
90 (VTABLE_VERIFICATION_SPEC): Likewise.
91 (SANITIZER_EARLY_SPEC): Likewise.
92 (SANITIZER_SPEC): Likewise.
93 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
94 * doc/invoke.texi (Link Options): Document -r.
96 2018-09-20 Richard Biener <rguenther@suse.de>
99 * gimplify.c (gimplify_expr): Retain alignment of
100 addressable lvalue in dereference.
102 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
105 * configure.ac: Check for .loc is_stmt support.
106 * configure, config.in: Rebuilt.
107 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
110 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
112 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
113 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
116 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
118 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
120 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
123 2018-09-20 Marek Polacek <polacek@redhat.com>
125 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
127 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
129 PR tree-optimization/87288
130 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
131 into account when determining PEELING_FOR_NITERS.
133 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
135 PR tree-optimization/86877
136 * tree-vect-loop.c (vect_analyze_loop_2): Call
137 vect_verify_datarefs_alignment.
139 2018-09-19 Marek Polacek <polacek@redhat.com>
141 * doc/invoke.texi: Document -Wclass-conversion.
143 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
145 * config/pa/pa.c (pa_adjust_priority): Delete.
146 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
148 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
149 (atomic_storehi): Likewise.
150 (atomic_storesi): Likewise.
151 (atomic_loaddi): Restore compare and swap exchange loop code.
153 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
155 PR rtl-optimization/86902
156 * combine.c (try_combine): When changing the CC mode used, don't change
157 an unrelated mode in other_insn to that new CC mode.
159 2018-09-19 David Malcolm <dmalcolm@redhat.com>
161 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
162 with %T in place of calls to dump_generic_expr.
163 (prune_runtime_alias_test_list): Likewise.
164 (create_runtime_alias_checks): Likewise.
165 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
166 (vect_analyze_data_ref_dependence): Likewise.
167 (vect_slp_analyze_data_ref_dependence): Likewise.
168 (vect_record_base_alignment): Likewise. Use %G in place of call
170 (vect_compute_data_ref_alignment): Likewise.
171 (verify_data_ref_alignment): Likewise.
172 (vect_find_same_alignment_drs): Likewise.
173 (vect_analyze_group_access_1): Likewise.
174 (vect_analyze_data_ref_accesses): Likewise.
175 (dependence_distance_ge_vf): Likewise.
176 (dump_lower_bound): Likewise.
177 (vect_prune_runtime_alias_test_list): Likewise.
178 (vect_find_stmt_data_reference): Likewise.
179 (vect_analyze_data_refs): Likewise.
180 (vect_create_addr_base_for_vector_ref): Likewise.
181 (vect_create_data_ref_ptr): Likewise.
182 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
183 (vect_can_advance_ivs_p): Likewise.
184 (vect_update_ivs_after_vectorizer): Likewise.
185 (vect_gen_prolog_loop_niters): Likewise.
186 (vect_prepare_for_masked_peels): Likewise.
187 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
188 (vect_determine_vectorization_factor): Likewise.
189 (vect_is_simple_iv_evolution): Likewise.
190 (vect_analyze_scalar_cycles_1): Likewise.
191 (vect_analyze_loop_operations): Likewise.
192 (report_vect_op): Likewise.
193 (vect_is_slp_reduction): Likewise.
194 (check_reduction_path): Likewise.
195 (vect_is_simple_reduction): Likewise.
196 (vect_create_epilog_for_reduction): Likewise.
197 (vect_finalize_reduction:): Likewise.
198 (vectorizable_induction): Likewise.
199 (vect_transform_loop_stmt): Likewise.
200 (vect_transform_loop): Likewise.
201 (optimize_mask_stores): Likewise.
202 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
203 (vect_split_statement): Likewise.
204 (vect_recog_over_widening_pattern): Likewise.
205 (vect_recog_average_pattern): Likewise.
206 (vect_determine_min_output_precision_1): Likewise.
207 (vect_determine_precisions_from_range): Likewise.
208 (vect_determine_precisions_from_users): Likewise.
209 (vect_mark_pattern_stmts): Likewise.
210 (vect_pattern_recog_1): Likewise.
211 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
212 (vect_record_max_nunits): Likewise.
213 (vect_build_slp_tree_1): Likewise.
214 (vect_build_slp_tree_2): Likewise.
215 (vect_print_slp_tree): Likewise.
216 (vect_analyze_slp_instance): Likewise.
217 (vect_detect_hybrid_slp_stmts): Likewise.
218 (vect_detect_hybrid_slp_1): Likewise.
219 (vect_slp_analyze_operations): Likewise.
220 (vect_slp_analyze_bb_1): Likewise.
221 (vect_transform_slp_perm_load): Likewise.
222 (vect_schedule_slp_instance): Likewise.
223 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
224 (vect_mark_stmts_to_be_vectorized): Likewise.
225 (vect_init_vector_1): Likewise.
226 (vect_get_vec_def_for_operand): Likewise.
227 (vect_finish_stmt_generation_1): Likewise.
228 (vect_check_load_store_mask): Likewise.
229 (vectorizable_call): Likewise.
230 (vectorizable_conversion): Likewise.
231 (vectorizable_operation): Likewise.
232 (vectorizable_load): Likewise.
233 (vect_analyze_stmt): Likewise.
234 (vect_is_simple_use): Likewise.
235 (vect_get_vector_types_for_stmt): Likewise.
236 (vect_get_mask_type_for_stmt): Likewise.
237 * tree-vectorizer.c (increase_alignment): Likewise.
239 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
241 * doc/rtl.texi: Adjust vec_select description.
242 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
243 non-constant selectors.
245 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
247 * config/aarch64/aarch64-protos.h
248 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
249 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
250 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
251 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
252 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
253 (AARCH64_FL_PROFILE): Move index so flags are ordered.
254 (AARCH64_ISA_RCPC8_4): New flag.
255 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
256 to aarch64_offset_9bit_signed_unscaled_p.
257 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
259 * config/aarch64/constraints.md (Ust): New constraint.
260 * config/aarch64/predicates.md.
261 (aarch64_9bit_offset_memory_operand): New predicate.
262 (aarch64_rcpc_memory_operand): New predicate.
264 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
266 PR rtl-optimization/87361
267 * rtlanal.c (nonzero_bits1): Revert accidental change.
269 2018-09-19 Richard Biener <rguenther@suse.de>
271 PR tree-optimization/87349
272 PR tree-optimization/87342
273 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
275 2018-09-18 Marek Polacek <polacek@redhat.com>
277 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
278 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
280 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
282 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
284 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
286 PR rtl-optimization/86882
287 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
289 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
291 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
292 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
294 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
297 * doc/invoke.texi (Link Options): Fix formatting and grammar.
299 2018-09-18 Richard Biener <rguenther@suse.de>
302 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
303 (compute_samebase_partition_bases): Likewise.
304 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
305 (gimple_can_coalesce_p): Simplify.
307 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
309 Handle a library implementation of ffs calling __builtin_ffs.
310 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
311 (mmix_init_libfuncs): New function: make __builtin_ffs expand
314 2018-09-17 David Malcolm <dmalcolm@redhat.com>
316 * diagnostic-show-locus.c (class layout_range): Add field
318 (layout_range::layout_range): Add "original_idx" param and use it
319 to initialize new field.
320 (make_range): Use 0 for original_idx.
321 (layout::layout): Pass in index to calls to
322 maybe_add_location_range.
323 (layout::maybe_add_location_range): Add param "original_idx" and
324 pass it on to layout_range.
325 (layout::print_any_labels): Pass on range->m_original_idx to
327 (gcc_rich_location::add_location_if_nearby): Use 0 for
329 * gcc-rich-location.h (text_range_label::get_text): Update for new
331 (range_label_for_type_mismatch::get_text): Likewise.
333 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
335 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
337 2018-09-17 David Malcolm <dmalcolm@redhat.com>
339 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
340 format_string_diagnostic_t.
341 (fmtwarn_n): Likewise.
342 * substring-locations.c
343 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
344 (format_warning_n_va): Convert to...
345 (format_string_diagnostic_t::emit_warning_n_va): ...this.
346 (format_warning_va): Convert to...
347 (format_string_diagnostic_t::emit_warning_va): ...this.
348 (format_warning_at_substring): Convert to...
349 (format_string_diagnostic_t::emit_warning): ...this.
350 (format_warning_at_substring_n): Convert to...
351 (format_string_diagnostic_t::emit_warning_n): ...this.
352 * substring-locations.h (class format_string_diagnostic_t): New
354 (format_warning_va): Convert to
355 format_string_diagnostic_t::emit_warning_va.
356 (format_warning_n_va): Convert to
357 format_string_diagnostic_t::emit_warning_n_va.
358 (format_warning_at_substring): Convert to
359 format_string_diagnostic_t::emit_warning.
360 (format_warning_at_substring_n): Convert to
361 format_string_diagnostic_t::emit_warning_n.
363 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
364 Bernd Schmidt <bernds_cb1@t-online.de>
366 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
369 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
371 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
372 operand 0 predicate to nonimmediate operand.
373 (rint<mode>2_frndint): Remove insn pattern.
374 (rint<mode>2): Change operand 1 predicate to general_operand.
375 Extend operand 1 to XFmode and generate rintxf2 insn.
376 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
377 Do not use X87MODEF mode macro.
378 (frndintxf2_<rounding>_i387): Rename from
379 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
380 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
381 to XFmode and generate significandxf3 insn.
383 2018-09-17 Richard Biener <rguenther@suse.de>
385 PR tree-optimization/87328
386 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
387 visiting unexecutable backedges when not iterating.
388 (do_rpo_vn): Mark all edges not executable even when not
391 2018-09-17 Martin Jambor <mjambor@suse.cz>
394 * doc/invoke.texi (Warning Options): Likewise.
396 2018-09-17 Richard Biener <rguenther@suse.de>
398 PR tree-optimization/87301
399 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
400 clean EH info from leftover copy assignments.
402 2018-09-17 Martin Liska <mliska@suse.cz>
404 PR gcov-profile/85871
405 * gcov.c (output_intermediate_file): Fix out of bounds
408 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
410 * config/arc/arc.c: Object attributes for core4 not reflected
412 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
415 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
417 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
419 2018-09-17 Martin Liska <mliska@suse.cz>
421 * doc/gcov.texi: Document new option --use-hotness-colors.
422 * gcov.c (struct source_info): Declare new field.
423 (source_info::source_info): Set default for maximum_count.
424 (print_usage): Add new -q option.
425 (process_args): Process it.
426 (accumulate_line_info): Save src->maximum_count.
427 (output_line_beginning): Make color line number if
428 flag_use_hotness_colors is set.
429 (output_line_details): Pass default argument value.
430 (output_lines): Pass src->maximum_count.
432 2018-09-17 Martin Liska <mliska@suse.cz>
434 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
435 Use processor_names table.
436 * config/i386/i386.c (ix86_default_align): Use
437 processor_cost_table for alignment values.
438 (ix86_option_override_internal): Use processor_names.
439 (ix86_function_specific_print): Likewise.
440 * config/i386/i386.h (struct processor_costs):
441 Add alignment values.
442 (struct ptt): Remove and replace with const char *.
443 * config/i386/x86-tune-costs.h (struct processor_costs):
444 Declare default alignments for all costs.
446 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
448 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
449 symbolics or VR_VARYING ranges for ABS_EXPR.
450 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
451 when range will wrap.
453 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
456 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
457 before and after a JUMP_TABLE_DATA.
459 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
462 * dojump.c (do_compare_and_jump): Canonicalize function pointers
463 when one operand is a function pointer. Use POINTER_TYPE_P and
464 FUNC_OR_METHOD_TYPE_P.
465 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
466 * fold-const.c (build_range_check): Likewise.
467 * match.pd (simple_comparison): Likewise.
469 2018-09-14 David Malcolm <dmalcolm@redhat.com>
472 * spellcheck.c (get_edit_distance_cutoff): New function.
473 (selftest::test_edit_distance_unit_test_oneway): Rename to...
474 (selftest::test_get_edit_distance_one_way): ...this.
475 (selftest::test_get_edit_distance_unit): Rename to...
476 (selftest::test_get_edit_distance_both_ways): ...this.
477 (selftest::test_edit_distances): Move tests to this new function,
478 and test some more pairs of strings. Update for above renaming.
479 (selftest::get_old_cutoff): New function.
480 (selftest::test_get_edit_distance_cutoff): New function.
481 (selftest::assert_suggested_for): New function.
482 (ASSERT_SUGGESTED_FOR): New macro.
483 (selftest::assert_not_suggested_for): New function.
484 (ASSERT_NOT_SUGGESTED_FOR): New macro.
485 (selftest::test_suggestions): New function.
486 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
487 tests to selftest::test_edit_distances and call it. Add calls to
488 selftest::test_get_edit_distance_cutoff and
489 selftest::test_suggestions.
490 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
491 (best_match::consider): Replace hard-coded cutoff calculation with
493 (best_match::get_cutoff): New declaration.
494 (best_match::get_best_meaningful_candidate): Likewise.
496 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
498 * builtins.c (fold_builtin_strlen): Remove TODO comment.
500 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
503 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
505 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
506 terminated string literal.
508 2018-09-14 Martin Sebor <msebor@redhat.com>
510 * builtins.c (unterminated_array): Handle ARRAY_REF.
511 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
512 * builtins.h (unterminated_array): Declare extern.
513 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
515 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
518 2018-09-14 Martin Sebor <msebor@redhat.com>
519 Jeff Law <law@redhat.com>
521 * builtins.c (unterminated_array): New.
522 (expand_builtin_strcpy): Adjust.
523 (expand_builtin_strcpy_args): Detect unterminated arrays.
524 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
526 * gimple-fold.h (get_maxval_strlen): Add argument.
527 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
529 * gimple-fold.c (get_range_strlen): Add argument.
530 (get_maxval_strlen): Adjust.
531 * gimple-fold.h (get_range_strlen): Add argument.
533 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
535 * config/i386/movdirintrin.h: Fix copyright year.
537 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
539 * reg-stack.c: Include regs.h.
540 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
541 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
542 FIRST_STACK_REG, not DFmode.
543 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
544 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
545 (change stack): Default register mode to the reg_raw_mode of
546 FIRST_STACK_REG, not DFmode.
547 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
548 (*swapxf): Rename from swapxf.
550 2018-09-14 Carl Love <cel@us.ibm.com>
552 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
553 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
555 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
558 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
561 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
564 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
566 2018-09-14 Jason Merrill <jason@redhat.com>
568 Fix --enable-gather-detailed-mem-stats.
569 * hash-table.c (hash_table_usage): Change from variable to function.
570 * hash-table.h: Adjust.
571 * Makefile.in: Add missing dependencies on hash-table.h.
573 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
575 PR tree-optimization/87259
577 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
578 execute_cse_reciprocals_1 has tried transforming.
580 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
582 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
583 VR_VARYING for PLUS/MINUS_EXPR.
585 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
587 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
590 2018-09-14 Richard Biener <rguenther@suse.de>
593 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
594 bits for the merged partition.
596 2018-09-13 Martin Sebor <msebor@redhat.com>
597 Bernd Edlinger <bernd.edlinger@hotmail.de>
599 * builtins.h (c_srlen): Add argument.
600 * builtins.c (warn_string_no_nul): New function.
601 (c_strlen): Add argument and use it. Update recursive calls.
602 Pass DECL argument to string_constant to get info on non
603 terminated strings. Update *NONSTR as needed.
604 (fold_builtin_strlen): Add argument to calls to c_strlen.
605 Warn for unterminated arrays.
606 (warn_string_no_null): Add prototype.
607 * expr.c (string_constant): Update arguments. Update recursive
608 calls appropriately. Detect missing NUL terminator and outermost
609 declaration its missing in.
610 Improve checks for arrays with nonzero lower bound or elements
611 that are not a single byte. Simplify offset computation.
612 Simplify checks for non-NUL terminated strings.
613 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
614 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
616 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
618 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
620 * fold-const.c (c_getstr): Fix function comment. Remove unused third
621 argument. Fix range checks.
622 * fold-const.h (c_getstr): Adjust protoype.
623 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
624 string is constant but contains no NUL byte.
626 * expr.c (string_constant): Adjust function comment.
627 Remove bogus check for zero termination.
629 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
631 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
632 (get_constant_size): Don't make STRING_CSTs larger than they are.
633 (check_string_literal): New check function for STRING_CSTs.
634 (output_constant): Use it.
636 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
639 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
641 2018-09-13 Richard Biener <rguenther@suse.de>
643 PR tree-optimization/87263
644 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
645 (struct unwind_state): Add max_rpo field.
646 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
647 Compute max_rpo, the max RPO number a block can be backwards reached
648 from. Re-write non-iterating mode to a RPO ordered worklist approach,
649 separating it from the iterating mode.
651 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
653 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
654 (rfs_decision): New scheduling decision.
656 2018-09-13 Richard Biener <rguenther@suse.de>
659 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
660 (vn_nary_op_insert_pieces_predicated): Do not write useless
661 valid_dominated_by_p entry outside of the allocated storage.
663 2018-09-13 Omar Sandoval <osandov@osandov.com>
664 Tom de Vries <tdevries@suse.de>
667 * dwarf2out.c (is_c): New function.
668 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
670 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
673 * config/aarch64/aarch64.md (*aarch64_bfxil):
675 * config/aarch64/constraints.md (Ulc): Define.
676 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
678 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
681 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
683 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
684 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
685 aarch64_layout_frame call.
686 (aarch64_expand_epilogue): Likewise.
687 (aarch64_initial_elimination_offset): Likewise.
688 (aarch64_get_separate_components): Likewise.
689 (aarch64_use_return_insn_p): Likewise.
690 (aarch64_layout_frame): Remove unneeded check.
692 2018-09-13 Jakub Jelinek <jakub@redhat.com>
694 * configure.ac: Only append
695 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
696 gcc_config_arguments if it was never reconfigured or last reconfigure
697 was with different arguments.
698 * configure: Regenerated.
700 2018-09-13 Jakub Jelinek <jakub@redhat.com>
701 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
704 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
705 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
707 2018-09-13 Jakub Jelinek <jakub@redhat.com>
709 PR tree-optimization/87287
710 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
711 X % C == 0 to X % (unsigned) C == 0 optimization to ...
712 * match.pd (X % C == 0): ... here. New optimization.
714 2018-09-12 Jakub Jelinek <jakub@redhat.com>
717 * expr.h (maybe_optimize_mod_cmp): Declare.
718 * expr.c (mod_inv): New function.
719 (maybe_optimize_mod_cmp): New function.
720 (do_store_flag): Use it.
721 * cfgexpand.c (expand_gimple_cond): Likewise.
723 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
724 Julian Brown <julian@codesourcery.com>
727 * gimplify.c (gimplify_scan_omp_clauses): Set
728 target_firstprivatize_array_bases in OpenACC parallel and kernels
729 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
730 OpenACC data regions.
732 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
734 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
735 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
737 2018-09-12 Richard Biener <rguenther@suse.de>
739 PR tree-optimization/87280
740 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
741 edge but unreachable target.
742 (do_rpo_vn): For conservatively handling a PHI only mark
743 the backedge executable but not the block reachable.
745 2018-09-12 Richard Biener <rguenther@suse.de>
747 PR tree-optimization/87266
748 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
751 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
753 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
755 ("trunc<BFP:mode><DFP_ALL:mode>2")
756 ("trunc<DFP_ALL:mode><BFP:mode>2")
757 ("extend<BFP:mode><DFP_ALL:mode>2")
758 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
759 according to the target operand type.
761 2018-09-12 Jakub Jelinek <jakub@redhat.com>
762 Andreas Krebbel <krebbel@linux.ibm.com>
764 PR tree-optimization/86844
765 * gimple-ssa-store-merging.c
766 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
767 there are any overlapping stores in between them, make sure they are
768 also coalesced or we give up completely.
770 2018-09-12 Jakub Jelinek <jakub@redhat.com>
773 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
774 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
776 2018-09-12 Tom de Vries <tdevries@suse.de>
778 * common.opt (gdescribe-dies): Add option.
779 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
780 attribute for artifical and nameless decls.
781 (dwarf2out_register_external_die): Add description attribute to
782 external reference die.
783 (add_desc_attribute): New functions.
784 (gen_subprogram_die): Add description attribute to
785 DW_TAG_call_site_parameter.
786 * tree-pretty-print.c (print_generic_expr_to_str): New function.
787 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
788 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
790 (@item -gdescribe-dies): Add.
792 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
794 * tree-vrp.c (vrp_shift_undefined_p): Remove.
795 (extract_range_from_binary_expr_1: Call
796 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
797 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
800 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
802 * gimple-ssa-warn-alloca.c
803 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
804 field for ALLOCA_BOUND_*_LARGE.
806 2018-09-11 Nathan Sidwell <nathan@acm.org>
808 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
810 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
812 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
813 for clobbers. Remove obsolete comment.
815 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
817 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
818 mpxchk, mpxld and mpxst types.
819 (define_attr length_immediate): Remove all processing of mpx types.
820 (define_attr prefix_0f): Ditto.
821 (define_attr memory): Ditto.
823 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
825 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
826 (log<mode>2): Change operand 1 predicate to general_operand.
827 Extend operand 1 to XFmode and generate logxf3 insn.
828 (log10<mode>2): Change operand 1 predicate to general_operand.
829 Extend operand 1 to XFmode and generate log10xf3 insn.
830 (log2<mode>2): Change operand 1 predicate to general_operand.
831 Extend operand 1 to XFmode and generate log2xf3 insn.
832 (fyl2xp1_extend<mode>xf3_i387): Remove.
833 (log1p<mode>2): Change operand 1 predicate to general_operand.
834 Extend operand 1 to XFmode and generate log1pxf3 insn.
835 (fxtract_extend<mode>xf3_i387): Remove.
836 (logb<mode>2): Change operand 1 predicate to general_operand.
837 Extend operand 1 to XFmode and generate logbxf3 insn.
838 (ilogb<mode>2): Change operand 1 predicate to general_operand.
839 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
840 (significand<mode>2): Change operand 1 predicate to general_operand.
841 Extend operand 1 to XFmode and generate significandxf3 insn.
843 2018-09-11 Nathan Sidwell <nathan@acm.org>
845 * gcc.c (perror_with_name, pfatal_with_name): Delete.
846 (load_specs): Use fatal_error.
847 (DELETE_IF_ORDINARY, process_command): Use error.
848 (execute, run_attempt): Use fatal_error.
850 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
852 * diagnostic-core.h (sorry_at): New prototype.
853 * diagnostic.c (sorry_at): New function.
855 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
857 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
858 by zero as VR_UNDEFINED.
860 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
862 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
863 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
864 (<sincos>mode2): New expander.
865 (sincos_extend<mode>xf3_i387): Remove insn pattern.
866 (sincos -> sin, cos splitters): Remove splitter patterns.
867 (sincos<mode>3): Change operand 2 predicate to general_operand.
868 Extend operand 2 to XFmode and generate sincosxf3 insn.
869 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
870 Change operand 3 predicate to const1_operand.
871 (fptan_extend<mode>xf4_i387): Remove insn pattern.
872 (tanxf2): Update operands in the call to fptanxf4_i387.
873 (tan<mode>2): Change operand 1 predicate to general_operand.
874 Extend operand 1 to XFmode and generate tanxf3 insn.
875 (atan2xf3): Rename from *fpatanxf3_i387.
876 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
877 (atan2xf3): Remove expander.
878 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
879 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
880 (atan<mode>2): Change operand 1 predicate to general_operand.
881 Extend operand 1 to XFmode and generate atanxf3 insn.
883 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
885 * config/i386/i386.md (x87/SSE constant load splitter): Use
886 memory_operand instead of nonimmediate_operand for input operand
889 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
891 * config/i386/i386.md (float partial SSE register stall splitter): Move
892 splitter near its instruction pattern.
893 (float_extend partial SSE register stall splitter): Ditto.
894 (float_truncate partial SSE register stall splitter): Ditto.
896 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
899 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
900 to speculation_safe_value_not_needed.
903 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
904 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
905 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
906 leaf_function_p, instead use has_hard_reg_initial_val.
908 2018-09-09 Nathan Sidwell <nathan@acm.org>
910 * gcc.h (pfatal_with_name): Don't declare here.
911 * gcc.c (pfatal_with_name): Make static.
913 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
915 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
918 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
920 PR rtl-optimization/85458
921 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
922 priority hook to reduce the priority of EXPR.
924 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
926 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
927 DImode for x87 on 32bit targets. Conditionally disable x87 modes
928 with X87_ENABLE_FLOAT. Remove preparation code.
929 (*float<SWI48:mode><MODEF:mode>2): Rename from
930 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
931 math using "enabled" attribute.
932 (*floatdi<MODEF:mode>2_i387): Rename from
933 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
934 enable for 32bit targets only.
935 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
937 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
938 as operand 1 predicate. Rewrite as define_insn_and_split.
939 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
941 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
943 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
944 to fallthru to FLOAT case.
946 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
949 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
950 around folding of vec_sl to handle out of range shift values.
952 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
954 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
955 Update callers to gen_fix_trunc<mode>_i387_fisttp
956 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
957 nonimmediate_operand.
958 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
959 and corresponding splitters.
960 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
961 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
962 (fix_truncdi_i387_with_temp): Remove insn pattern
963 and corresponding splitters.
964 (fix_trunc<mode>_i387): Change operand 0 predicate to
965 nonimmediate_operand.
966 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
967 and corresponding splitters.
968 (*fistdi2_1): Remove.
970 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
971 (lrintxfdi2): Remove expander. Reimplement as define_insn.
972 (*fist<mode>2_1): Remove.
973 (fist<mode>2): Ditto.
974 (fist<mode>2_with_temp): Remove insn pattern and corresponding
976 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
977 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
978 (fistdi2_<rounding>): Change operand 0 predicate to
979 nonimmediate_operand.
980 (fistdi2_<rounding>_with_temp): Remove insn pattern
981 and corresponding splitters.
982 (fist<mode>2_<rounding>): Change operand 0 predicate to
983 nonimmediate_operand.
984 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
985 and corresponding splitters.
987 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
989 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
991 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
994 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
996 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
997 early gimple folding of vec_splat().
998 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
999 * gimple-fold.h: Add an extern define for tree_vec_extract().
1001 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
1003 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
1004 wrappers around TREE_TYPE comparisons.
1006 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
1009 * config/s390/predicates.md: Add nonsym_memory_operand.
1010 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
1011 contains a SYMBOL_REF, load it into an intermediate pseudo.
1012 (s390_emit_compare_and_swap): Legitimize operand.
1013 * config/s390/s390.md: Use the new nonsym_memory_operand
1014 with UNSPECV_CAS patterns.
1016 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
1019 * config/s390/s390-passes.def: New file.
1020 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
1022 (make_pass_s390_early_mach): Add declaration.
1023 * config/s390/s390.c (make_pass_s390_early_mach):
1024 (s390_option_override): Remove dynamic registration.
1025 * config/s390/t-s390: Add s390-passes.def.
1027 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
1029 * config/s390/s390.c (s390_decompose_constant_pool_ref):
1030 Remove UNSPEC_LTREL_BASE check.
1031 (annotate_constant_pool_refs): Likewise.
1032 (find_constant_pool_ref): Likewise.
1033 (find_ltrel_base): Removed.
1034 (replace_ltrel_base): Removed.
1035 (s390_mainpool_finish): Remove replace_ltrel_base call.
1036 (s390_chunkify_start): Remove pending LTREL_BASE logic.
1037 (s390_chunkify_finish): Remove replace_ltrel_base call.
1038 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
1040 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
1043 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1044 to speculation_safe_value_not_needed.
1046 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
1047 Bernd Schmidt <bernds_cb1@t-online.de>
1049 * config/nvptx/nvptx-opts.h: New file.
1050 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
1051 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
1053 (TARGET_SM35): New macro.
1054 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
1056 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
1058 (misa=): New option.
1059 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
1061 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
1063 * config/i386/i386.md (truncdfsf2): Remove expander.
1064 (truncdfsf2_with_temp): Ditto.
1065 (truncxf<mode>2): Ditto.
1066 (*truncdfsf_fast_mixed): Remove insn pattern.
1067 (*truncdfsf_fast_i387): Ditto.
1068 (*truncdfsf_mixed): Ditto.
1069 (*truncdfsf_i387): Ditto.
1070 (*truncdfsf2_i387_1): Ditto.
1071 (*truncxfsf2_mixed): Ditto.
1072 (*truncxfdf2_mixed): Ditto.
1073 (*truncxf<mode>2_i387_noop): Ditto. Update callers
1074 to call gen_truncxf<mode>2 instead.
1075 (*truncxf<mode>2_i387): Remove.
1076 (reg->reg splitters): Remove splitter pattern.
1077 (reg->mem splitters): Ditto.
1079 (truncdfsf2): New insn pattern.
1080 (truncxf<mode>2): Ditto.
1082 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1084 * tree-ssa-math-opts.c (is_mult_by): New function.
1085 (is_square_of): Use the above.
1086 (optimize_recip_sqrt): New function.
1087 (pass_cse_reciprocals::execute): Use the above.
1089 2018-09-05 Richard Biener <rguenther@suse.de>
1092 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
1093 to zero-init the emplaced vec.
1095 2018-09-05 Martin Liska <mliska@suse.cz>
1097 PR tree-optimization/87205
1098 * tree-switch-conversion.c (pass_lower_switch::execute):
1099 Group cases for switch statements.
1101 2018-09-05 Richard Biener <rguenther@suse.de>
1103 PR tree-optimization/87217
1104 * tree-ssa-sccvn.c (vuse_valueize): New.
1105 (vn_reference_lookup_pieces): Use it.
1106 (vn_reference_lookup): Likewise.
1108 2018-09-05 Nathan Sidwell <nathan@acm.org>
1111 * stor-layout.c (place_field): Scan forwards to check last
1112 bitfield when ms_bitfield_placement is in effect.
1114 2018-09-05 Richard Biener <rguenther@suse.de>
1117 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
1120 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
1121 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
1123 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
1124 * config.gcc (extra_objs): Build it.
1125 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
1127 * config/aarch64/aarch64-passes.def
1128 (pass_tag_collision_avoidance): New pass.
1129 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
1130 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
1131 (aarch64_classify_address): Remove static qualifier.
1132 (aarch64_address_info, aarch64_address_type): Move to...
1133 * config/aarch64/aarch64-protos.h: ... here.
1134 (make_pass_tag_collision_avoidance): New function.
1135 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
1138 2018-09-05 Martin Liska <mliska@suse.cz>
1140 * doc/gcov.texi: Update documentation of humar
1142 * gcov.c (format_count): Print one decimal place, it provides
1143 more fine number of situations like '1G' vs. '1.4G'.
1145 2018-09-05 Martin Liska <mliska@suse.cz>
1148 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
1149 * optc-gen.awk: Allow 'Var' for Deprecated options in order
1150 to generate a MASK value.
1152 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
1155 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
1156 if frame pointer isn't used.
1157 (compute_frame_pointer_to_fb_displacement): Likewise.
1159 2018-09-04 Jakub Jelinek <jakub@redhat.com>
1162 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
1163 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
1164 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
1165 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
1166 and OPTION_MASK_ISA_XSAVEC_UNSET.
1168 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
1170 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
1171 XOR operations in NAND case.
1173 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
1175 * wide-int-range.cc (wide_int_range_convert): New.
1176 * wide-int-range.h (wide_int_range_convert): New.
1177 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
1178 code into wide_int_range_convert.
1179 (extract_range_into_wide_ints): Do not munge anti range constants
1180 into the entire domain. Just return the range back.
1182 2018-09-04 Martin Liska <mliska@suse.cz>
1184 * genmatch.c (output_line_directive): Add new argument
1186 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
1188 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
1190 * doc/invoke.texi (Option Summary): Add whitespace.
1192 * doc/invoke.texi (Option Summary): Add -Waligned-new.
1194 2018-09-04 Richard Biener <rguenther@suse.de>
1196 PR tree-optimization/87211
1197 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
1198 backedge value we're supposed to treat as VARYING also number
1199 the PHI to VARYING in case it got a different value-number already.
1201 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
1203 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
1204 (extract_range_from_binary_expr_1): Do not call
1205 vrp_can_optimize_bit_op.
1206 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
1208 (wide_int_range_get_mask_and_bounds): New.
1209 (wide_int_range_optimize_bit_op): New.
1210 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
1211 (wide_int_range_bit_and): Same.
1212 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
1213 (wide_int_range_optimize_bit_op): New.
1214 (wide_int_range_get_mask_and_bounds): New.
1216 2018-09-04 Richard Biener <rguenther@suse.de>
1218 PR tree-optimization/87176
1219 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
1220 variable. When value-numbering a virtual PHI node make sure
1221 to not value-number to the backedge value.
1223 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
1225 * doc/extend.texi (Long Long, Hex Floats): Document support for
1226 long long and hex floats in more recent versions of ISO C++.
1228 2018-09-03 Richard Biener <rguenther@suse.de>
1230 PR tree-optimization/87177
1231 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
1234 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1236 * bb-reorder.c (edge_order): Convert to C-qsort-style
1237 tri-state comparator.
1238 (reorder_basic_blocks_simple): Change std::stable_sort to gcc_stablesort.
1240 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1242 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
1243 tri-state comparator.
1244 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
1246 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1248 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
1249 (mergesort): ... here as maximum count for using netsort.
1250 (gcc_qsort): Set nlim to 3 if stable sort is requested.
1251 (gcc_stablesort): New.
1252 * system.h (gcc_stablesort): Declare.
1254 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
1256 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
1257 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
1258 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
1260 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
1262 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
1263 lxsdx and stxsdx alternatives.
1264 (*mov<mode>_hardfloat64): Ditto.
1265 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
1267 2018-09-03 Richard Biener <rguenther@suse.de>
1269 PR tree-optimization/87200
1270 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
1273 2018-09-03 Martin Liska <mliska@suse.cz>
1275 PR tree-optimization/87201
1276 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
1277 Fix parenthesis in an expression.
1279 2018-09-03 Richard Biener <rguenther@suse.de>
1281 PR tree-optimization/87197
1282 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
1283 visited. CSE the VN_INFO hashtable lookup.
1285 PR tree-optimization/87169
1286 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
1287 iterating make sure there's no extra backedges from irreducible
1288 regions feeding the header. Mark the destination block
1291 2018-09-03 Martin Liska <mliska@suse.cz>
1294 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
1295 * common/common-targhooks.c (default_get_valid_option_values):
1297 * common/common-targhooks.h (default_get_valid_option_values):
1299 * common/config/i386/i386-common.c: Move processor_target_table
1301 (ix86_get_valid_option_values): New function.
1302 (TARGET_GET_VALID_OPTION_VALUES): New macro.
1303 * config/i386/i386.c (struct ptt): Move to i386-common.c.
1304 (PTA_*): Move all defined masks into i386-common.c.
1305 (ix86_function_specific_restore): Use new processor_cost_table.
1306 * config/i386/i386.h (struct ptt): Moved from i386.c.
1307 (struct pta): Likewise.
1308 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
1309 * doc/tm.texi.in: Likewise.
1310 * opt-suggestions.c (option_proposer::suggest_option):
1311 Pass prefix to build_option_suggestions.
1312 (option_proposer::get_completions): Likewise.
1313 (option_proposer::build_option_suggestions): Use the new target
1315 * opts.c (struct option_help_tuple): New struct.
1316 (print_filtered_help): Use the new target hook.
1318 2018-09-03 Martin Liska <mliska@suse.cz>
1321 * predict.c (set_even_probabilities): Add likely_edges
1322 argument and handle cases where we have precisely one
1324 (combine_predictions_for_bb): Catch also likely_edges.
1325 (tree_predict_by_opcode): Handle gswitch statements.
1326 * tree-cfg.h (find_case_label_for_value): New declaration.
1327 (find_taken_edge_switch_expr): Likewise.
1328 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
1329 Find pivot in decision tree based on probabily, not by number of
1332 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
1334 * doc/standards.texi (Standards): Update Objective-C reference.
1336 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1338 * doc/install.texi (Prerequisites): Update link for MPC.
1340 2018-09-01 Michael Matz <matz@suse.de>
1342 PR tree-optimization/87074
1343 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
1344 PHIs for outer-loop uses.
1346 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1348 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
1349 * doc/invoke.texi (C Dialect Options): Ditto.
1351 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1353 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
1355 2018-08-31 Richard Biener <rguenther@suse.de>
1357 PR tree-optimization/87168
1358 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
1359 (rpo_elim::eliminate_avail): When OP was not visited it must
1362 2018-08-31 David Malcolm <dmalcolm@redhat.com>
1364 * tree-vrp.c (copy_value_range): Convert param "from" from
1365 "value_range *" to "const value_range *".
1366 (range_is_null): Likewise for param "vr".
1367 (range_int_cst_p): Likewise.
1368 (range_int_cst_singleton_p): Likewise.
1369 (symbolic_range_p): Likewise.
1370 (value_ranges_intersect_p): Likewise for both params.
1371 (value_range_nonnegative_p): Likewise for param "vr".
1372 (value_range_constant_singleton): Likewise.
1373 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
1374 (extract_range_into_wide_ints): Likewise for param "vr".
1375 (extract_range_from_multiplicative_op): Likewise for params "vr0"
1377 (vrp_can_optimize_bit_op): Likewise.
1378 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
1380 (extract_range_from_unary_expr): Likewise.
1381 (debug_value_range): Likewise for param "vr".
1382 (value_range::dump): Add "const" qualifier.
1383 (vrp_prop::check_array_ref): Convert local "vr" from
1384 "value_range *" to "const value_range *".
1385 (vrp_prop::check_mem_ref): Likewise.
1386 (vrp_prop::visit_stmt): Likewise for local "old_vr".
1387 (vrp_intersect_ranges_1): Likewise for param "vr_1".
1388 (vrp_intersect_ranges): Likewise.
1389 (simplify_stmt_for_jump_threading): Likewise for local "vr".
1390 (vrp_prop::vrp_finalize): Likewise.
1391 * tree-vrp.h (value_range::dump): Add "const" qualifier.
1392 (vrp_intersect_ranges): Add "const" qualifier to params as above.
1393 (extract_range_from_unary_expr): Likewise.
1394 (value_range_constant_singleton): Likewise.
1395 (symbolic_range_p): Likewise.
1396 (copy_value_range): Likewise.
1397 (extract_range_from_binary_expr_1): Likewise.
1398 (range_int_cst_p): Likewise.
1399 (vrp_set_zero_nonzero_bits): Likewise.
1400 (range_int_cst_singleton_p): Likewise.
1402 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
1404 * config/aarch64/arm_neon.h (vabsd_s64): New.
1405 (vnegd_s64): Likewise.
1407 2018-08-31 Martin Jambor <mjambor@suse.cz>
1409 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
1411 2018-08-31 Martin Liska <mliska@suse.cz>
1413 * ipa-icf.c (sem_item::add_type): Use
1414 sem_item::m_type_hash_cache.
1415 * ipa-icf.h: Move the cache from sem_item_optimizer
1418 2018-08-31 Nathan Sidwell <nathan@acm.org>
1420 * doc/extend.texi (Backwards Compatibility): Remove implicit
1421 extern C leeway of () being (...).
1423 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1425 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
1427 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
1431 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
1433 2018-08-31 Jakub Jelinek <jakub@redhat.com>
1436 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
1437 gen_int_mode. Formatting fixes.
1439 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
1441 * target.def (custom_function_descriptors): Improve documentation.
1442 * doc/tm.texi.in (Trampolines): Expand discussion of function
1443 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
1444 beginning of the section.
1445 * doc/tm.texi: Regenerated.
1447 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
1449 * cfg.h (class auto_edge_flag): Spell out the template-id of the
1450 base class in the initializer list. This is a workaround for
1451 building with older GCC.
1452 (class auto_bb_flag): Likewise.
1454 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
1456 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
1457 (altivec_vcmpequ<VI_char>_p): Remove star.
1458 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
1460 (expand_strncmp_vec_sequence): New function.
1461 (emit_final_str_compare_vec): New function.
1462 (expand_strn_compare): Add support for vector strncmp.
1463 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
1464 length specification to bytes.
1465 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
1466 (vcmpnezb_p): New pattern.
1467 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
1468 for option -mstring-compare-inline-limit.
1470 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
1472 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
1473 (PTA_SKYLAKE): Add PTA_AES.
1474 (PTA_GOLDMONT): Likewise.
1476 2018-08-29 Jan Hubicka <jh@suse.cz>
1479 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
1480 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
1482 2018-08-29 Jan Hubicka <jh@suse.cz>
1484 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
1486 (hash_tree): Do not visit TYPE_STUB_DECL.
1487 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
1488 stream TYPE_STUB_DECL.
1489 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
1490 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
1491 after free_lang_data.
1492 (type_in_anonymous_namespace_p): Likewise.
1494 2018-08-29 Jan Hubicka <jh@suse.cz>
1496 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1497 comment that it has to be even number.
1498 (class sreal): Change m_sig type to int32_t.
1499 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1500 int64_t for temporary calculations.
1501 (sreal_verify_basics): Drop one bit from minimum and maximum.
1503 2018-08-30 Richard Biener <rguenther@suse.de>
1505 PR tree-optimization/87147
1506 * tree-ssa-sccvn.c (SSA_VISITED): New function.
1507 (visit_phi): When the degenerate result is from the backedge and
1508 we didn't visit its definition yet drop to VARYING.
1509 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
1511 2018-08-29 Jan Hubicka <jh@suse.cz>
1513 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
1515 (hash_tree): Likewise.
1517 2018-08-29 Jan Hubicka <jh@suse.cz>
1519 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
1520 and TYPE_NEXT_REF_TO.
1522 2018-08-29 Jan Hubicka <jh@suse.cz>
1524 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1525 comment that it has to be even number.
1526 (class sreal): Change m_sig type to int32_t.
1527 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1528 int64_t for temporary calculations.
1529 (sreal_verify_basics): Drop one bit from minimum and maximum.
1531 2018-08-30 Tamar Christina <tamar.christina@arm.com>
1533 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
1535 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
1538 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
1539 if to_add is negative.
1541 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
1544 * builtins.c (c_strlen): Improve range checks.
1546 2018-08-29 Martin Sebor <msebor@redhat.com>
1547 Jeff Law <law@redhat.com>
1549 PR tree-optimization/86714
1550 PR tree-optimization/86711
1551 * builtins.c (c_strlen): Add arguments to call to string_constant.
1552 * expr.c (string_constant): Add argument. Detect missing nul
1553 terminator and outermost declaration it's missing in.
1554 * expr.h (string_constant): Add argument.
1555 * fold-const.c (read_from_constant_string): Add arguments to call to
1557 (c_getstr): Likewise.
1558 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1560 * tree-ssa-strlen.c (get_stridx): Likewise.
1562 2018-08-29 Jan Hubicka <jh@suse.cz>
1564 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
1565 Do not stream DECL_VINDEX.
1566 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
1567 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
1568 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
1570 2018-08-29 Richard Biener <rguenther@suse.de>
1572 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
1573 virtual operands that are not default defs to honor region
1575 (rpo_vn_valueize): Remove ineffective code here.
1577 2018-08-29 Richard Biener <rguenther@suse.de>
1579 PR tree-optimization/87132
1580 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
1581 when skipping defs reachable over backedges.
1583 2018-08-29 Richard Biener <rguenther@suse.de>
1585 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
1586 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
1587 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
1588 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
1589 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
1590 (vn_reference_lookup_3): Remove use of const_parms.
1591 (free_rpo_vn): Do not free const_parms.
1592 (do_rpo_vn): Do not call init_const_parms.
1593 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
1594 SSA_NAME_POINTS_TO_READONLY_MEMORY.
1595 (call_may_clobber_ref_p_1): Likewise.
1597 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
1600 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
1602 (-ftree-scev-cprop): Document.
1604 2018-08-29 Jan Hubicka <jh@suse.cz>
1606 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
1608 (sreal constructor): Update.
1609 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
1610 sreal:operator/): Update.
1612 2018-08-29 Martin Liska <mliska@suse.cz>
1614 * tree-switch-conversion.c (switch_conversion::expand):
1615 Strenghten assumption about gswitch statements.
1617 2018-08-29 Richard Biener <rguenther@suse.de>
1619 PR tree-optimization/87117
1620 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
1621 re-value-number released SSA VDEFs.
1623 2018-08-29 Richard Biener <rguenther@suse.de>
1625 PR tree-optimization/87126
1626 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
1628 2018-08-28 Jim Wilson <jimw@sifive.com>
1630 * config/riscv/pic.md: Rewrite.
1631 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
1633 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
1634 (SOFTF, default_load, softload, softstore): New.
1636 2018-08-28 Jeff Law <law@redhat.com>
1638 * fold-const.c (fold_binary_loc): Remove recently added assert.
1640 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
1642 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
1643 to OP parmeter of generated function.
1645 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
1647 PR tree-optimization/87009
1648 * match.pd: Add boolean optimizations.
1650 2018-08-28 Martin Sebor <msebor@redhat.com>
1653 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
1654 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
1655 (pass_walloca::gate): Use it.
1656 (alloca_call_type): Same.
1657 (pass_walloca::execute): Same.
1658 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
1660 2018-08-28 David Malcolm <dmalcolm@redhat.com>
1662 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
1663 GCC_VERSION for usage of "__gcc_dump_printf__" format from
1666 2018-08-28 Richard Biener <rguenther@suse.de>
1668 PR tree-optimization/87124
1669 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
1670 constants before looking up avail.
1672 2018-08-28 Jakub Jelinek <jakub@redhat.com>
1675 * calls.c (maybe_warn_nonstring_arg): Punt early if
1676 warn_stringop_overflow is zero. Don't call get_range_strlen
1677 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
1678 Swap comparison operands to have constants on rhs. Only use
1679 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
1680 increment lenrng[0].
1682 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
1684 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
1685 use of tree_to_shwi. Remove duplicated test for the size being
1686 a whole number of bytes.
1688 2018-08-28 Richard Biener <rguenther@suse.de>
1690 PR tree-optimization/87117
1691 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
1692 Handle removed stmt without LHS (GIMPLE_NOP).
1694 2018-08-28 Richard Biener <rguenther@suse.de>
1696 PR tree-optimization/87117
1697 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
1698 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
1700 2018-08-28 Richard Biener <rguenther@suse.de>
1702 PR tree-optimization/87117
1703 * tree-ssa-pre.c (compute_avail): Do not make expressions
1704 with predicated values available.
1705 (get_expr_value_id): Assert we do not run into predicated value
1708 2018-08-28 Richard Biener <rguenther@suse.de>
1710 PR tree-optimization/87117
1711 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
1712 get virtual operands.
1713 (get_expr_operands): Handle STRING_CST like other decls.
1715 2018-08-28 Martin Liska <mliska@suse.cz>
1717 * tree.h: Update documentation of fndecl_built_in_p
1721 2018-08-27 Jeff Law <law@redhat.com>
1722 PR tree-optimization/87110
1723 * tree-ssa-dse.c (compute_trims): Handle non-constant
1726 2018-08-27 Martin Sebor <msebor@redhat.com>
1728 PR tree-optimization/86914
1729 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
1731 2018-08-27 Martin Sebor <msebor@redhat.com>
1733 PR tree-optimization/87112
1734 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
1735 the type of the bound argument.
1737 2018-08-27 Jeff Law <law@redhat.com>
1739 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
1740 type does not have a TYPE_SIZE_UNIT.
1742 2018-08-27 Steve Ellcey <sellcey@cavium.com>
1744 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
1745 with include of backend.h.
1747 2018-08-27 Richard Biener <rguenther@suse.de>
1749 PR tree-optimization/86927
1750 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
1751 use const cond reduction code.
1753 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
1755 PR tree-optimization/85758
1756 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
1758 2018-08-27 David Malcolm <dmalcolm@redhat.com>
1761 * diagnostic-show-locus.c (class layout_range): Update for
1762 conversion of show_caret_p to a tri-state.
1763 (layout_range::layout_range): Likewise.
1764 (make_range): Likewise.
1765 (layout::maybe_add_location_range): Likewise.
1766 (layout::should_print_annotation_line_p): Don't show annotation
1767 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
1768 (layout::get_state_at_point): Update for conversion of
1769 show_caret_p to a tri-state. Bail out early for
1770 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
1771 underlining or source colorization.
1772 (gcc_rich_location::add_location_if_nearby): Update for conversion
1773 of show_caret_p to a tri-state.
1774 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
1775 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
1777 (selftest::test_one_liner_labels): Likewise.
1778 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
1779 conversion of show_caret_p to a tri-state.
1780 * pretty-print.c (text_info::set_location): Likewise.
1781 * pretty-print.h (text_info::set_location): Likewise.
1782 * substring-locations.c (format_warning_n_va): Likewise.
1783 * tree-diagnostic.c (default_tree_printer): Likewise.
1784 * tree-pretty-print.c (newline_and_indent): Likewise.
1786 2018-08-27 David Malcolm <dmalcolm@redhat.com>
1789 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
1790 line above for line-insertion fix-it hints.
1791 (selftest::test_fixit_insert_containing_newline): Update the
1792 expected results, and add a test with line-numbering enabled.
1794 2018-08-27 Martin Liska <mliska@suse.cz>
1797 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
1798 params with DECL_HAS_VALUE_EXPR_P.
1800 2018-08-27 Martin Liska <mliska@suse.cz>
1802 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
1803 selected expansion strategy.
1805 2018-08-27 Martin Liska <mliska@suse.cz>
1807 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
1808 * builtins.c (is_builtin_fn): Likewise.
1809 * attribs.c (diag_attr_exclusions): Use new function
1810 fndecl_built_in_p and remove check for FUNCTION_DECL if
1812 (builtin_mathfn_code): Likewise.
1813 (fold_builtin_expect): Likewise.
1814 (fold_call_expr): Likewise.
1815 (fold_builtin_call_array): Likewise.
1816 (fold_call_stmt): Likewise.
1817 (set_builtin_user_assembler_name): Likewise.
1818 (is_simple_builtin): Likewise.
1819 * calls.c (gimple_alloca_call_p): Likewise.
1820 (maybe_warn_nonstring_arg): Likewise.
1821 * cfgexpand.c (expand_call_stmt): Likewise.
1822 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
1823 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
1824 (cgraph_node::verify_node): Likewise.
1825 * cgraphclones.c (build_function_decl_skip_args): Likewise.
1826 (cgraph_node::create_clone): Likewise.
1827 * config/arm/arm.c (arm_insert_attributes): Likewise.
1828 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
1829 * dse.c (scan_insn): Likewise.
1830 * expr.c (expand_expr_real_1): Likewise.
1831 * fold-const.c (operand_equal_p): Likewise.
1832 (fold_binary_loc): Likewise.
1833 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
1834 * gimple-low.c (lower_stmt): Likewise.
1835 * gimple-pretty-print.c (dump_gimple_call): Likewise.
1836 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Likewise.
1837 * gimple.c (gimple_build_call_from_tree): Likewise.
1838 (gimple_call_builtin_p): Likewise.
1839 (gimple_call_combined_fn): Likewise.
1840 * gimplify.c (gimplify_call_expr): Likewise.
1841 (gimple_boolify): Likewise.
1842 (gimplify_modify_expr): Likewise.
1843 (gimplify_addr_expr): Likewise.
1844 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
1845 * ipa-cp.c (determine_versionability): Likewise.
1846 * ipa-fnsummary.c (compute_fn_summary): Likewise.
1847 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
1848 * ipa-split.c (visit_bb): Likewise.
1849 (split_function): Likewise.
1850 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1851 * lto-cgraph.c (input_node): Likewise.
1852 * lto-streamer-out.c (write_symbol): Likewise.
1853 * omp-low.c (setjmp_or_longjmp_p): Likewise.
1854 (lower_omp_1): Likewise.
1855 * predict.c (strip_predict_hints): Likewise.
1856 * print-tree.c (print_node): Likewise.
1857 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
1858 * trans-mem.c (is_tm_irrevocable): Likewise.
1859 (is_tm_load): Likewise.
1860 (is_tm_simple_load): Likewise.
1861 (is_tm_store): Likewise.
1862 (is_tm_simple_store): Likewise.
1863 (is_tm_abort): Likewise.
1864 (tm_region_init_1): Likewise.
1865 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
1866 * tree-cfg.c (verify_gimple_call): Likewise.
1867 (move_stmt_r): Likewise.
1868 (stmt_can_terminate_bb_p): Likewise.
1869 * tree-eh.c (lower_eh_constructs_2): Likewise.
1870 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
1871 * tree-inline.c (remap_gimple_stmt): Likewise.
1872 (copy_bb): Likewise.
1873 (estimate_num_insns): Likewise.
1874 (fold_marked_statements): Likewise.
1875 * tree-sra.c (scan_function): Likewise.
1876 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
1877 (optimize_stack_restore): Likewise.
1878 (pass_fold_builtins::execute): Likewise.
1879 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
1880 (mark_all_reaching_defs_necessary_1): Likewise.
1881 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
1882 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1883 (pass_forwprop::execute): Likewise.
1884 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1885 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
1886 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
1887 * tree-ssa-strlen.c (get_string_length): Likewise.
1888 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
1889 (find_func_aliases_for_call): Likewise.
1890 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
1891 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
1892 * tree-tailcall.c (find_tail_calls): Likewise.
1893 * tree.c (need_assembler_name_p): Likewise.
1894 (free_lang_data_in_decl): Likewise.
1895 (get_call_combined_fn): Likewise.
1896 * ubsan.c (is_ubsan_builtin_p): Likewise.
1897 * varasm.c (incorporeal_function_p): Likewise.
1898 * tree.h (DECL_BUILT_IN): Remove and replace with
1900 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
1901 (fndecl_built_in_p): New.
1903 2018-08-27 Martin Liska <mliska@suse.cz>
1905 PR tree-optimization/86847
1906 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
1907 Dump also subtree probability.
1908 (switch_decision_tree::do_jump_if_equal): New function.
1909 (switch_decision_tree::emit_case_nodes): Handle special
1910 situations in balanced tree that can be emitted much simpler.
1911 Fix calculation of probabilities that happen in tree expansion.
1912 * tree-switch-conversion.h (struct cluster): Add
1914 (struct simple_cluster): Likewise.
1915 (struct case_tree_node): Add new function has_child.
1916 (do_jump_if_equal): New.
1918 2018-08-27 Martin Liska <mliska@suse.cz>
1920 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
1921 Add new argument to bit_test_cluster constructor.
1922 (bit_test_cluster::emit): Set bits really number of values
1924 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
1925 probability argument.
1926 * tree-switch-conversion.h (struct bit_test_cluster):
1927 Add m_handles_entire_switch.
1929 2018-08-27 Martin Liska <mliska@suse.cz>
1931 PR tree-optimization/86702
1932 * tree-switch-conversion.c (jump_table_cluster::emit):
1933 Make probabilities even for values in jump table
1934 according to number of cases handled.
1935 (switch_decision_tree::compute_cases_per_edge): Pass
1936 argument to reset_out_edges_aux function.
1937 (switch_decision_tree::analyze_switch_statement): Likewise.
1938 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
1941 2018-08-27 Martin Liska <mliska@suse.cz>
1943 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
1944 cfun argument explicitly.
1945 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
1946 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
1947 function gimple_switch_default_bb.
1948 (convert_switch_statements):
1950 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
1951 * stmt.c (label_to_block_fn): Use label_to_block and pass
1952 cfun argument explicitly and use gimple_switch_label_bb.
1953 (expand_case): Likewise.
1954 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
1955 cfun argument explicitly. Likewise.
1956 (make_edges_bb): Likewise.
1957 (make_cond_expr_edges): Likewise.
1958 (get_cases_for_edge): Likewise.
1959 (make_gimple_switch_edges): Likewise.
1960 (label_to_block_fn): Likewise.
1961 (label_to_block): Likewise.
1962 (make_goto_expr_edges): Likewise.
1963 (make_gimple_asm_edges): Likewise.
1964 (main_block_label): Likewise.
1965 (group_case_labels_stmt): Likewise.
1966 (find_taken_edge_computed_goto): Likewise.
1967 (find_taken_edge_switch_expr): Likewise.
1968 (gimple_verify_flow_info): Likewise.
1969 (gimple_redirect_edge_and_branch): Likewise.
1970 (gimple_switch_label_bb): New function.
1971 (gimple_switch_default_bb): Likewise.
1972 (gimple_switch_edge): Likewise.
1973 (gimple_switch_default_edge): Likewise.
1974 * tree-cfg.h (label_to_block_fn): Remove and replace ...
1975 (label_to_block): ... with this.
1976 (gimple_switch_label_bb): New.
1977 (gimple_switch_default_bb): Likewise.
1978 (gimple_switch_edge): Likewise.
1979 (gimple_switch_default_edge): Likewise.
1980 * tree-cfgcleanup.c (convert_single_case_switch): Use
1981 new gimple functions and pass new argument to label_to_block.
1982 (cleanup_control_flow_bb):
1983 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
1984 cfun argument explicitly.
1985 (make_eh_edges): Likewise.
1986 (redirect_eh_dispatch_edge): Likewise.
1987 (lower_resx): Likewise.
1988 (lower_eh_dispatch): Likewise.
1989 (maybe_remove_unreachable_handlers): Likewise.
1990 (unsplit_eh): Likewise.
1991 (cleanup_empty_eh): Likewise.
1992 (verify_eh_edges): Likewise.
1993 (verify_eh_dispatch_edge): Likewise.
1994 * tree-ssa-dom.c (record_edge_info): Likewise.
1995 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
1996 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
1997 (thread_through_normal_block): Likewise.
1998 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
1999 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
2000 * tree-switch-conversion.c (switch_conversion::collect): Use new
2002 (switch_conversion::check_final_bb): Likewise.
2003 (switch_conversion::gather_default_values): Pass new argument
2005 (switch_conversion::build_constructors): Likewise.
2006 (switch_decision_tree::compute_cases_per_edge): Use new
2007 gimple_switch_edge function.
2008 (switch_decision_tree::analyze_switch_statement): Pass new argument
2010 (switch_decision_tree::try_switch_expansion): Use
2011 gimple_switch_default_edge.
2012 * tree-vrp.c (find_switch_asserts): Pass new argument
2014 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
2015 (vr_values::simplify_switch_using_ranges): Likewise.
2017 2018-08-27 Richard Biener <rguenther@suse.de>
2019 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
2020 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
2022 * tree-ssa-sccvn.h (struct vn_pval): New structure.
2023 (struct vn_nary_op_s): Add unwind_to member. Add
2024 predicated_values flag and put result into a union together
2025 with a linked list of vn_pval.
2026 (struct vn_ssa_aux): Add name member to make maintaining
2027 a map of SSA name to vn_ssa_aux possible. Remove no longer
2028 needed info, dfsnum, low, visited, on_sccstack, use_processed
2029 and range_info_anti_range_p members.
2030 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
2031 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
2033 (vn_valueize): New global.
2034 (vn_context_bb): Likewise.
2035 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
2036 VN_INFO_PTR_INFO): Remove.
2037 * tree-ssa-sccvn.c: ... (rewrite)
2038 (pass_fre::execute): For -O2+ initialize loops and run
2039 RPO VN in optimistic mode (iterating). For -O1 and -Og
2040 run RPO VN in non-optimistic mode.
2041 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
2042 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
2043 * doc/invoke.texi (sccvn-max-scc-size): Remove.
2044 (rpo-vn-max-loop-depth): Document.
2045 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
2046 when valuezing the VUSE signals we walked out of the region.
2047 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
2048 (phi_translate): Set VN context block to use for availability
2050 (compute_avail): Likewise.
2051 (pre_valueize): New function.
2052 (pass_pre::execute): Adjust to the RPO VN API.
2054 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
2055 (propagate_constants_for_unrolling): Remove.
2056 (tree_unroll_loops_completely): Perform value-numbering
2057 on the unrolled bodies loop parent.
2059 2018-08-27 Richard Biener <rguenther@suse.de>
2061 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
2062 for partial antic compute.
2064 2018-08-27 Jakub Jelinek <jakub@redhat.com>
2066 PR rtl-optimization/87065
2067 * combine.c (simplify_if_then_else): Formatting fix.
2068 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
2070 (known_cond): Don't return const_true_rtx for vector modes. Use
2071 CONST0_RTX instead of const0_rtx. Formatting fixes.
2073 2018-08-27 Martin Liska <mliska@suse.cz>
2075 PR gcov-profile/87069
2076 * gcov.c (process_file): Record files already processed
2077 and warn about a file being processed multiple times.
2079 2018-08-27 Martin Liska <mliska@suse.cz>
2082 * config/aarch64/aarch64.c (aarch64_override_options_internal):
2083 Set default values for x_aarch64_*_string strings.
2084 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
2085 prefix. For -mabi do not print '=ABI' in help and use
2086 <option_value> format for -msve-vector-bits and -moverride
2089 2018-08-26 Jeff Law <law@redhat.com>
2091 * config/mips/frame-header-opt.c: Include "backend.h" rather than
2094 2018-08-26 Marek Polacek <polacek@redhat.com>
2096 PR c++/87029, Implement -Wredundant-move.
2097 * doc/invoke.texi: Document -Wredundant-move.
2099 2018-08-25 Martin Sebor <msebor@redhat.com>
2101 PR tree-optimization/87059
2102 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
2103 to the same type as the other.
2104 * fold-const.c (fold_binary_loc): Assert expectation.
2106 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
2108 * config/darwin.c (machopic_legitimize_pic_address): Clean up
2109 extraneous parentheses, dead code section and formatting.
2111 2018-08-24 David Malcolm <dmalcolm@redhat.com>
2114 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
2115 wide enough for jumps in the line-numbering to be visible.
2116 (layout::print_gap_in_line_numbering): New member function.
2117 (layout::calculate_line_spans): When using line numbering, merge
2118 line spans that are only 1 line apart.
2119 (diagnostic_show_locus): When printing line numbers, show gaps in
2120 line numbering directly, rather than printing headers.
2121 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
2122 line-numbering with multiple line spans.
2123 (selftest::test_fixit_insert_containing_newline_2): Add test of
2124 line-numbering, in which the spans are close enough to be merged.
2126 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
2128 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
2129 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
2130 * tree-vrp.c (range_is_nonnull): Remove.
2131 (range_includes_zero_p): Accept value_range instead of min/max.
2132 (extract_range_from_binary_expr_1): Do not early bail on
2134 Use range_includes_zero_p instead of range_is_nonnull.
2135 (extract_range_from_unary_expr): Use range_includes_zero_p instead
2136 of range_is_nonnull.
2137 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
2138 special case VR_ANTI_RANGE.
2139 (vrp_finalize): Same.
2140 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
2142 (range_is_nonnull): Remove.
2143 * vr-values.c (vrp_stmt_computes_nonzero): Use
2144 range_includes_zero_p instead of range_is_nonnull.
2145 (extract_range_basic): Pass value_range to range_includes_zero_p
2146 instead of range_is_nonnull.
2148 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
2150 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
2151 * emit-rtl.h (rtl_data): Remove return_bnd.
2152 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
2153 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
2154 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
2155 (POINTER_BOUNDS_MODE): Remove definition.
2156 (make_pointer_bounds_mode): Remove.
2157 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
2158 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
2159 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
2160 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
2161 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
2162 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
2163 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
2165 * config/i386/i386-modes.def (BND32, BND64): Remove.
2166 * config/i386/i386.c (dbx_register_map): Remove bound registers.
2167 (dbx64_register_map): Ditto.
2168 (svr4_dbx_register_map): Ditto.
2169 (indirect_thunk_bnd_needed): Remove.
2170 (indirect_thunks_bnd_used): Ditto.
2171 (indirect_return_bnd_needed): Ditto.
2172 (indirect_return_via_cx_bnd): Ditto.
2173 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
2174 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
2175 (output_indirect_thunk): Ditto. Remove need_prefix argument.
2176 (output_indirect_thunk_function): Remove handling of
2177 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
2178 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
2179 (ix86_save_reg): Remove handling of crtl->return_bnd.
2180 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
2181 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
2182 and UNSPEC_BNDLX_ADDR.
2183 (ix86_output_indirect_branch_via_reg): Remove handling of
2184 indirect_thunk_prefix_bnd.
2185 (ix86_output_indirect_branch_via_push): Ditto.
2186 (ix86_output_function_return): Ditto.
2187 (ix86_output_indirect_function_return): Ditto.
2188 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
2189 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
2190 (CALL_USED_REGISTERS): Ditto.
2191 (REG_ALLOC_ORDER): Update for removal of bound registers.
2192 (HI_REGISTER_NAMES): Ditto.
2193 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
2194 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
2195 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
2196 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
2197 (FIRST_PSEUDO_REG): Update.
2198 (BND): Remove mode iterator.
2199 * config/i386/predicates.md (bnd_mem_operator): Remove.
2201 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
2203 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
2206 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
2208 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
2209 the case in which the permute needs only a single element and
2210 repeats for every vector of the result. Extend that case to
2211 handle variable-length vectors.
2212 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
2214 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
2217 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
2218 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
2220 2018-08-24 Richard Biener <rguenther@suse.de>
2222 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
2223 bb_flags_allocated members.
2224 (auto_flag): New RAII class for allocating flags.
2225 (auto_edge_flag): New RAII class for allocating edge flags.
2226 (auto_bb_flag): New RAII class for allocating bb flags.
2227 * cfgloop.c (verify_loop_structure): Allocate temporary edge
2229 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
2230 in favor of temporarily allocated BB flag.
2231 * hsa-brig.c: Re-order includes.
2232 * hsa-dump.c: Likewise.
2233 * hsa-regalloc.c: Likewise.
2234 * print-rtl.c: Likewise.
2235 * profile-count.c: Likewise.
2237 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
2240 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
2243 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
2246 * wide-int-range.cc (wide_int_range_div): Do not ignore result
2247 from wide_int_range_multiplicative_op.
2249 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2251 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
2254 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
2256 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
2259 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
2261 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
2262 full GENERIC expression used for replacement.
2264 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
2266 * tree-vrp.c (abs_extent_range): Remove.
2267 (extract_range_into_wide_ints): Pass wide ints by reference.
2268 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
2269 Pass wide ints by reference in all calls to
2270 extract_range_into_wide_ints.
2271 * wide-int-range.cc (wide_int_range_div): New.
2272 * wide-int-range.h (wide_int_range_div): New.
2273 (wide_int_range_includes_zero_p): New.
2274 (wide_int_range_zero_p): New.
2276 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
2278 * config/aarch64/aarch64.md (arches): New enum.
2279 (arch): New enum attr.
2280 (arch_enabled): New attr.
2281 (enabled): Now uses arch_enabled only.
2282 (simd, sve, fp16): Removed attribute.
2283 (fp): Attr now defined in terms of 'arch'.
2284 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
2285 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
2286 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
2287 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
2288 attributes into 'arch'.
2289 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
2290 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
2291 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
2292 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
2293 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
2294 'simd' attribute into 'arch'.
2295 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
2296 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
2297 Convert use of 'fp' attribute to 'arch'.
2298 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
2299 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
2301 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
2302 (different modes) Merge 'fp' and 'simd' into 'arch'.
2303 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
2306 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
2308 PR rtl-optimization/87026
2309 * expmed.c (canonicalize_comparison): If we can no longer create
2310 pseudoregisters, don't.
2312 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
2315 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
2317 * config/arm/arm.c (speculation_barrier_libfunc): New static
2319 (arm_init_libfuncs): Initialize it.
2320 (arm_emit_speculation_barrier): New function.
2321 * config/arm/arm.md (speculation_barrier): Call
2322 arm_emit_speculation_barrier for architectures that do not have
2324 (speculation_barrier_insn): Only match on Armv7 or later.
2326 2018-08-23 Richard Biener <rguenther@suse.de>
2329 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
2332 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
2334 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
2335 of single-vector TBLs.
2336 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
2339 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
2342 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
2343 aarch64_evpc_tbl guard.
2345 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
2347 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
2350 2018-08-22 Martin Sebor <msebor@redhat.com>
2353 * tree-pretty-print.c (pretty_print_string): Add argument.
2354 (dump_generic_node): Call to pretty_print_string with string size.
2356 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
2358 PR rtl-optimization/86771
2359 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
2360 of two SETs into those two SETs, one to be placed at i2, if that SETs
2361 destination is modified between i2 and i3.
2363 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2365 PR tree-optimization/86725
2366 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
2368 (vect_analyze_scalar_cycles_1): Check it.
2370 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2372 PR tree-optimization/86725
2373 * tree-vect-loop.c (vect_is_simple_reduction): When treating
2374 an outer loop phi as a double reduction, make sure that the
2375 single user of the phi result is an inner loop phi.
2377 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2379 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
2380 grouped stores with gaps to a strided group.
2382 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2384 * tree-vect-stmts.c (get_group_load_store_type)
2385 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
2386 first statement in a group.
2388 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2390 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
2391 the sequence used in gcc/gcc.c.
2393 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2396 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
2399 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2401 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
2402 Darwin10-specific unwinder-shim.
2403 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
2404 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
2405 New to cater for Darwin10 Rosetta.
2407 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2409 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
2410 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
2412 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2417 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
2418 (dwarf2out_switch_text_section): Generate a local label for the
2419 second function sub-section and apply it as the second FDE start
2421 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
2422 second sub-section start.
2424 2018-08-22 Richard Biener <rguenther@suse.de>
2426 PR tree-optimization/86988
2427 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
2429 2018-08-22 Richard Biener <rguenther@suse.de>
2431 PR tree-optimization/86945
2432 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
2434 2018-08-22 Alexandre Oliva <oliva@adacore.com>
2436 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
2437 a comment about how uses of r2 for .sdata2 come about.
2439 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
2441 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
2443 2018-08-21 Marek Polacek <polacek@redhat.com>
2445 PR c++/86981, Implement -Wpessimizing-move.
2446 * doc/invoke.texi: Document -Wpessimizing-move.
2448 2018-08-21 Jan Hubicka <jh@suse.cz>
2450 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
2451 * tree.h (is_redundant_typedef): Remove.
2452 * dwarf2out.c (is_redundant_typedef): Turn into static function.
2454 2018-08-21 Jan Hubicka <jh@suse.cz>
2456 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
2459 2018-08-21 Tamar Christina <tamar.christina@arm.com>
2461 * expmed.c (extract_low_bits): Reject invalid subregs early.
2463 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
2466 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
2469 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2471 * config/vxworks.h: Guard vxworks_asm_out_constructor and
2472 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
2473 * config/vxworks.c: Likewise.
2475 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2477 * config/vxworks.c: Set targetm.have_ctors_dtors
2478 if HAVE_INITFINI_ARRAY_SUPPORT.
2479 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
2480 if HAVE_INITFINI_ARRAY_SUPPORT.
2482 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2484 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
2485 default search path for VxWorks < 7.
2487 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
2489 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
2490 (get_format_string): Refer to c_getstr.
2492 2018-08-21 Tom de Vries <tdevries@suse.de>
2494 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2495 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2496 (debuginfo_early_stop): Declare.
2497 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2498 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2499 (debuginfo_early_stop): New function.
2500 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
2501 and debuginfo_early_stop.
2502 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
2503 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
2504 (general_init): Call debuginfo_early_init.
2505 (finalize): Call debuginfo_fini.
2506 (do_compile): Call debuginfo_init.
2507 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
2509 (@item -fdump-debug, @item -fdump-earlydebug): Add.
2511 2018-08-21 Tom de Vries <tdevries@suse.de>
2513 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
2514 flag_dump_noaddr and flag_dump_unnumbered.
2516 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
2518 * wide-int-range.cc (wide_int_range_abs): New.
2519 (wide_int_range_order_set): Rename from wide_int_range_min_max.
2520 * wide-int-range.h (wide_int_range_abs): New.
2521 (wide_int_range_min_max): New.
2522 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
2523 case to call wide_int_range_abs.
2524 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
2525 (extract_range_from_abs_expr): Delete.
2527 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
2530 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
2531 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
2534 2018-08-20 Nathan Sidwell <nathan@acm.org>
2535 Jeff Law <law@redhat.com>
2537 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
2538 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
2540 2018-08-20 David Malcolm <dmalcolm@redhat.com>
2543 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
2544 (decl_attributes): Likewise.
2545 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
2547 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2548 * diagnostic-core.h (class auto_diagnostic_group): New class.
2549 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
2550 (diagnostic_report_diagnostic): Handle the first diagnostics within
2552 (emit_diagnostic): Add auto_diagnostic_group instance.
2554 (inform_n): Likewise.
2555 (warning): Likewise.
2556 (warning_at): Likewise.
2557 (warning_n): Likewise.
2558 (pedwarn): Likewise.
2559 (permerror): Likewise.
2561 (error_n): Likewise.
2562 (error_at): Likewise.
2564 (fatal_error): Likewise.
2565 (internal_error): Likewise.
2566 (internal_error_no_backtrace): Likewise.
2567 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
2568 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
2569 * diagnostic.h (struct diagnostic_context): Add fields
2570 "diagnostic_group_nesting_depth",
2571 "diagnostic_group_emission_count", "begin_group_cb",
2573 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
2574 Add auto_diagnostic_group instance(s).
2575 (find_explicit_erroneous_behavior): Likewise.
2576 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
2577 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
2578 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
2579 (gimplify_va_arg_expr): Likewise.
2580 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
2581 (HSA_SORRY_AT): Likewise.
2582 * ipa-devirt.c (compare_virtual_tables): Likewise.
2583 (warn_odr): Likewise.
2584 * multiple_target.c (expand_target_clones): Likewise.
2585 * opts-common.c (cmdline_handle_error): Likewise.
2586 * reginfo.c (globalize_reg): Likewise.
2587 * substring-locations.c (format_warning_n_va): Likewise.
2588 * tree-inline.c (expand_call_inline): Likewise.
2589 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
2590 * tree-ssa-loop-niter.c
2591 (do_warn_aggressive_loop_optimizations): Likewise.
2592 * tree-ssa-uninit.c (warn_uninit): Likewise.
2593 * tree.c (warn_deprecated_use): Likewise.
2595 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
2598 * config/i386/i386.md (eh_return): Always update EH return
2599 address in word_mode.
2601 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
2603 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
2604 TARGET_SPLIT_COMPLEX_ARG is defined.
2606 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
2608 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
2610 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
2613 * expr.c (expand_assignment): Assert that bitpos is positive.
2614 (store_field): Likewise
2615 (expand_expr_real_1): Make sure that bitpos is positive.
2616 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
2619 2018-08-20 Nathan Sidwell <nathan@acm.org>
2621 * Makefile.in (CPP_ID_DATA_H): Delete.
2622 (CPP_INTERNAL_H): Don't add it.
2623 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
2624 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
2626 2018-08-20 Richard Biener <rguenther@suse.de>
2628 PR tree-optimization/78655
2629 * tree-vrp.c (extract_range_from_binary_expr_1): Make
2630 pointer + offset nonnull if either operand is nonnull work.
2632 2018-08-20 Tom de Vries <tdevries@suse.de>
2634 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
2635 unless the referenced die describes the added property using
2636 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
2637 Otherwise, add a DW_AT_location to the referenced die.
2639 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
2642 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
2643 register_operand when calling ix86_set_reg_reg_cost.
2644 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
2645 Set *total to 0 for operands that satisfy x86_64_immediate_operand
2646 predicate and to 1 otherwise.
2648 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
2650 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
2651 emit a diagnostic that it is not supported and reset the option.
2652 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
2653 supported and consume the option. (ASM_FINAL_SPEC): New.
2655 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
2657 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
2660 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
2662 C-SKY port: Documentation
2664 * doc/extend.texi (C-SKY Function Attributes): New section.
2665 * doc/invoke.texi (Option Summary): Add C-SKY options.
2666 (C-SKY Options): New section.
2667 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
2669 2018-08-17 Jojo <jijie_rong@c-sky.com>
2670 Huibin Wang <huibin_wang@c-sky.com>
2671 Sandra Loosemore <sandra@codesourcery.com>
2672 Chung-Lin Tang <cltang@codesourcery.com>
2674 C-SKY port: Backend implementation
2676 * config/csky/*: New.
2677 * common/config/csky/*: New.
2679 2018-08-17 Jojo <jijie_rong@c-sky.com>
2680 Huibin Wang <huibin_wang@c-sky.com>
2681 Sandra Loosemore <sandra@codesourcery.com>
2682 Chung-Lin Tang <cltang@codesourcery.com>
2683 Andrew Jenner <andrew@codesourcery.com>
2685 C-SKY port: Configury
2687 * config.gcc (csky-*-*): New.
2688 * configure.ac: Add csky to targets for dwarf2 debug_line support.
2689 * configure: Regenerated.
2691 2018-08-17 David Malcolm <dmalcolm@redhat.com>
2693 * dump-context.h: Include "dumpfile.h".
2694 (dump_context::dump_printf_va): Convert final param from va_list
2695 to va_list *. Convert from ATTRIBUTE_PRINTF to
2696 ATTRIBUTE_GCC_DUMP_PRINTF.
2697 (dump_context::dump_printf_loc_va): Likewise.
2698 * dumpfile.c: Include "stringpool.h".
2699 (make_item_for_dump_printf_va): Delete.
2700 (make_item_for_dump_printf): Delete.
2701 (class dump_pretty_printer): New class.
2702 (dump_pretty_printer::dump_pretty_printer): New ctor.
2703 (dump_pretty_printer::emit_items): New member function.
2704 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
2706 (dump_pretty_printer::emit_item): New member function.
2707 (dump_pretty_printer::stash_item): New member function.
2708 (dump_pretty_printer::format_decoder_cb): New member function.
2709 (dump_pretty_printer::decode_format): New member function.
2710 (dump_context::dump_printf_va): Reimplement in terms of
2711 dump_pretty_printer.
2712 (dump_context::dump_printf_loc_va): Convert final param from va_list
2714 (dump_context::begin_scope): Reimplement call to
2715 make_item_for_dump_printf.
2716 (dump_printf): Update for change to dump_printf_va.
2717 (dump_printf_loc): Likewise.
2718 (selftest::test_capture_of_dump_calls): Convert "stmt" from
2719 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
2720 with %T, %E, and %G.
2721 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
2722 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
2723 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
2724 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
2725 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
2726 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
2727 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
2728 within a dump_printf_loc call to "%wu".
2729 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
2730 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
2731 missing space after "=".
2732 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
2733 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
2734 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
2735 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
2736 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
2737 duplicate "vectorized" from message.
2739 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
2741 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
2742 polyNxK_t element's TYPE_STRING_FLAG.
2744 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
2746 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
2747 (they were unnamed before). Fix comments.
2749 2018-08-17 Nathan Sidwell <nathan@acm.org>
2751 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
2753 2018-08-17 Richard Biener <rguenther@suse.de>
2755 PR tree-optimization/86841
2756 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
2758 2018-08-17 Martin Liska <mliska@suse.cz>
2760 * common.opt: Remove Warn, Init and Report for options with
2761 Ignore/Deprecated flag. Warning is done automatically for
2763 * config/i386/i386.opt: Likewise.
2764 * config/ia64/ia64.opt: Likewise.
2765 * config/rs6000/rs6000.opt: Likewise.
2766 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
2767 Remove usage of flag_check_pointer_bounds.
2768 * lto-wrapper.c (merge_and_complain): Do not handle
2769 OPT_fcheck_pointer_bounds.
2770 (append_compiler_options): Likewise.
2771 * opt-functions.awk: Do not handle Deprecated.
2772 * optc-gen.awk: Check that Var, Report and Init are not
2773 used for an option with Ignore/Deprecated flag.
2774 * opts-common.c (decode_cmdline_option): Do not report
2776 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
2778 * opts.h (struct cl_option): Remove cl_deprecated flag.
2779 (CL_ERR_DEPRECATED): Remove error enum value.
2781 2018-08-17 Richard Biener <rguenther@suse.de>
2784 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
2785 across a va-arg-pack using call adjust its return value accordingly.
2787 2018-08-16 Martin Sebor <msebor@redhat.com>
2789 PR tree-optimization/86853
2790 * gimple-ssa-sprintf.c (struct format_result): Rename member.
2791 (struct fmtresult): Add member and initialize it in ctors.
2792 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
2793 (format_string): Handle %S the same as %ls. Set MAYFAIL.
2794 (format_directive): Set POSUNDER4K when MAYFAIL is set.
2795 (parse_directive): Handle %C same as %c.
2796 (sprintf_dom_walker::compute_format_length): Adjust.
2797 (is_call_safe): Adjust.
2799 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
2801 * builtins.c (c_strlen): Add new parameter eltsize. Use it
2802 for determining how to count the elements.
2803 * builtins.h (c_strlen): Adjust prototype.
2804 * expr.c (string_constant): Add new parameter mem_size.
2805 Set *mem_size appropriately.
2806 * expr.h (string_constant): Adjust protoype.
2807 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
2808 * gimple-fold.h (get_range_strlen): Adjust prototype.
2809 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
2810 (format_string): Call get_string_length with eltsize.
2812 2018-08-16 David Malcolm <dmalcolm@redhat.com>
2814 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
2815 to emit the span, rather than setting it as the prefix.
2817 2018-08-16 David Malcolm <dmalcolm@redhat.com>
2819 * diagnostic-show-locus.c (layout::start_annotation_line): Add
2820 "margin_char" parameter, defaulting to space. Use it in place
2821 of pp_space for the initial part of the margin.
2822 (layout::print_leading_fixits): Use '+' when filling the margin
2823 of line-insertion fix-it hints.
2825 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2827 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
2830 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2832 * config/rs6000/altivec.md: Don't set length attribute to the default
2834 * config/rs6000/darwin.md: Ditto.
2835 * config/rs6000/dfp.md: Ditto.
2836 * config/rs6000/htm.md: Ditto.
2837 * config/rs6000/rs6000.md: Ditto.
2838 * config/rs6000/sync.md: Ditto.
2839 * config/rs6000/vsx.md: Ditto.
2841 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2843 * config/rs6000/altivec.md: Don't set length attribute to the default
2844 value, for branch instructions.
2845 * config/rs6000/darwin.md: Ditto.
2846 * config/rs6000/rs6000.md: Ditto.
2848 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2850 * config/rs6000/rs6000.md (length): Always define as const_int 4.
2851 (unnamed conditional branch define_insn): Set length to 4 or 8
2852 depending on offset.
2853 (<bd>_<mode>): Similar, for alternative 0.
2854 (<bd>tf_<mode>): Ditto.
2856 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2858 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
2860 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
2862 * doc/rtl.texi: Replace old RTX class names with new names.
2865 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
2867 * expmed.h (canonicalize_comparison): New declaration.
2868 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
2869 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
2870 * optabs.c (prepare_cmp_insn): Likewise.
2871 * rtl.h (unsigned_condition_p): New function which checks if a
2872 comparison operator is unsigned.
2874 2018-08-16 Nathan Sidwell <nathan@acm.org>
2876 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
2877 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
2879 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2882 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
2883 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
2884 (mov<mov>): ..this and enable unconditionally.
2886 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2888 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
2890 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
2892 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
2893 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
2894 "Common" with "Target".
2896 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
2898 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
2899 * doc/invoke.texi (mmitigate-rop): Remove.
2900 * config/i386/i386.c: Do not include "regrename.h".
2901 (ix86_rop_should_change_byte_p, reg_encoded_number)
2902 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
2904 (ix86_reorg): Remove call to ix86_mitigate_rop.
2905 * config/i386/i386.md (attr "modrm_class"): Remove.
2906 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
2907 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
2908 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
2910 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
2912 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
2913 allow folding of mergeh() and mergel() for the float and double types.
2914 (fold_mergehl_helper): Rework to handle building a permute tree
2917 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
2919 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
2922 2018-08-15 David Malcolm <dmalcolm@redhat.com>
2924 * common.opt (fdiagnostics-show-labels): New option.
2925 * diagnostic-show-locus.c (class layout_range): Add field
2927 (class layout): Add field "m_show_labels_p".
2928 (layout_range::layout_range): Add param "label" and use it to
2930 (make_range): Pass in NULL for new "label" param of layout_range's
2932 (layout::layout): Initialize m_show_labels_p.
2933 (layout::maybe_add_location_range): Pass in loc_range->m_label
2934 when constructing layout_range instances.
2935 (struct line_label): New struct.
2936 (layout::print_any_labels): New member function.
2937 (layout::print_line): Call it if label-printing is enabled.
2938 (selftest::test_one_liner_labels): New test.
2939 (selftest::test_diagnostic_show_locus_one_liner): Call it.
2940 * diagnostic.c (diagnostic_initialize): Initialize
2941 context->show_labels_p.
2942 * diagnostic.h (struct diagnostic_context): Add field
2944 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
2945 -fno-diagnostics-show-labels.
2946 * dwarf2out.c (gen_producer_string): Add
2947 OPT_fdiagnostics_show_labels to the ignored options.
2948 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
2950 (gcc_rich_location::maybe_add_expr): Likewise.
2951 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
2952 label" param, defaulting to NULL.
2953 (gcc_rich_location::add_expr): Add "label" param.
2954 (gcc_rich_location::maybe_add_expr): Likewise.
2955 (class text_range_label): New class.
2956 (class range_label_for_type_mismatch): New class.
2957 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
2958 of format_warning_va.
2959 (fmtwarn_n): Likewise for new params of format_warning_n_va.
2960 * lto-wrapper.c (merge_and_complain): Add
2961 OPT_fdiagnostics_show_labels to the "pick one setting" options.
2962 (append_compiler_options): Likewise to the dropped options.
2963 (append_diag_options): Likewise to the passed-on options.
2964 * opts.c (common_handle_option): Handle the new option.
2965 * selftest-diagnostic.c
2966 (test_diagnostic_context::test_diagnostic_context): Enable
2968 * substring-locations.c: Include "gcc-rich-location.h".
2969 (format_warning_n_va): Add "fmt_label" and "param_label" params
2970 and use them as appropriate.
2971 (format_warning_va): Add "fmt_label" and "param_label" params,
2972 passing them on to format_warning_n_va.
2973 (format_warning_at_substring): Likewise.
2974 (format_warning_at_substring_n): Likewise.
2975 * substring-locations.h (format_warning_va): Add "fmt_label" and
2976 "param_label" params.
2977 (format_warning_n_va): Likewise.
2978 (format_warning_at_substring): Likewise.
2979 (format_warning_at_substring_n): Likewise.
2980 * toplev.c (general_init): Initialize global_dc->show_labels_p.
2982 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
2985 * builtins.c (expand_builtin_memcmp): Do not expand the call
2986 when overflow is detected.
2988 2018-08-15 Martin Sebor <msebor@redhat.com>
2990 PR tree-optimization/71625
2991 * config/aarch64/aarch64-builtins.c
2992 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
2994 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
2996 * config/s390/s390.c (s390_reorg): Remove loop.
2998 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
3001 (darwin_function_switched_text_sections): Delete.
3002 * gcc/config/darwin.h
3003 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
3005 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
3008 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
3009 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
3010 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
3012 2018-08-15 Martin Liska <mliska@suse.cz>
3014 PR tree-optimization/86925
3015 * predict.c (expr_expected_value_1): When taking
3016 later predictor, assign also probability.
3017 Use fold_build2_initializer_loc in order to fold
3018 the expression in -frounding-math.
3020 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
3022 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
3024 (expand_vec_perm_1): Try the new method.
3026 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
3029 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
3030 Check whether lra_live_max_point is 0 before dividing.
3032 2018-08-14 Martin Sebor <msebor@redhat.com>
3034 PR tree-optimization/86650
3035 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
3036 (vrp_prop::check_mem_ref): Same.
3038 2018-08-13 Liu Hao <lh_mouse@126.com>
3040 * pretty-print.c (eat_esc_sequence): Swap the foreground and
3041 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
3042 and clear it thereafter, as it only works for DBCS.
3044 2018-08-13 Liu Hao <lh_mouse@126.com>
3046 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
3047 handle returned by _get_osf_handle().
3049 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3051 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
3052 for folding vec_perm.
3054 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3056 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
3057 Add support for gimple-folding of vec_pack() and vec_unpack()
3060 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3062 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
3063 vec_xst variants to the list.
3064 (rs6000_gimple_fold_builtin): Add support for folding unaligned
3065 vector loads and stores.
3067 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
3069 * config.gcc (rs6000-ibm-aix4.x): Delete.
3070 (rs6000-ibm-aix5.1): Delete.
3071 (rs6000-ibm-aix5.2): Delete.
3072 (rs6000-ibm-aix5.3): Delete.
3073 * config/rs6000/aix43.h: Delete.
3074 * config/rs6000/aix51.h: Delete.
3075 * config/rs6000/aix52.h: Delete.
3076 * config/rs6000/t-aix43: Delete.
3078 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
3080 * config/s390/s390.c (s390_decompose_constant_pool_ref):
3082 (s390_decompose_address): Factor out constant pool ref
3085 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3087 * config/nds32/nds32-predicates.c
3088 (nds32_can_use_bclr_p): Change return type as bool.
3089 (nds32_can_use_bset_p): Ditto.
3090 (nds32_can_use_btgl_p): Ditto.
3091 (nds32_can_use_bitci_p): Ditto.
3092 * config/nds32/nds32-protos.h
3093 (nds32_can_use_bclr_p): Change declaration.
3094 (nds32_can_use_bset_p): Ditto.
3095 (nds32_can_use_btgl_p): Ditto.
3096 (nds32_can_use_bitci_p): Ditto.
3098 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3100 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
3101 Support -msched-prolog-epilog option.
3102 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
3104 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3106 * common/config/nds32/nds32-common.c
3107 (nds32_option_optimization_table): Enalbe -malways-align.
3109 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3111 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
3113 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
3114 OPT_misr_secure_ case.
3115 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
3116 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
3117 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
3119 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
3120 (nds32_isr_info): New field security_level.
3121 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
3122 * config/nds32/nds32.md (return_internal): Consider critical attribute.
3123 * config/nds32/nds32.opt (misr-secure): New option.
3124 * config/nds32/nds32_init.inc: New file.
3125 * config/nds32/nds32_isr.h: New file.
3127 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
3129 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
3130 Update comment for atomic instructions.
3131 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
3132 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
3134 (atomic_loaddi): Revise fence expansion to only emit fence prior to
3135 load for __ATOMIC_SEQ_CST model.
3136 (atomic_loaddi_1): Remove float register target.
3137 (atomic_storedi): Handle CONST_INT values.
3138 (atomic_storedi_1): Remove float register source. Add special case
3140 (memory_barrier): New expander and insn.
3142 2018-08-11 Jakub Jelinek <jakub@redhat.com>
3144 PR tree-optimization/86835
3145 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
3146 new_stmt after def_gsi, make sure to insert new_square_stmt after
3147 that stmt, not 2 stmts before it.
3149 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
3152 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
3155 2018-08-10 Martin Liska <mliska@suse.cz>
3158 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
3160 * builtins.c (expand_builtin_expect_with_probability):
3162 (expand_builtin_expect_with_probability): New function.
3163 (build_builtin_expect_predicate): Add new argumnet probability
3164 for BUILT_IN_EXPECT_WITH_PROBABILITY.
3165 (fold_builtin_expect):
3168 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
3169 * builtins.h (fold_builtin_expect): Set new argument.
3170 * doc/extend.texi: Document __builtin_expect_with_probability.
3171 * doc/invoke.texi: Likewise.
3172 * gimple-fold.c (gimple_fold_call): Pass new argument.
3173 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
3174 also BUILT_IN_EXPECT_WITH_PROBABILITY.
3175 * predict.c (get_predictor_value): New function.
3176 (expr_expected_value): Add new argument probability. Assume
3177 that predictor and probability are always non-null.
3178 (expr_expected_value_1): Likewise. For __builtin_expect and
3179 __builtin_expect_with_probability set probability. Handle
3180 combination in binary expressions.
3181 (tree_predict_by_opcode): Simplify code by simply calling
3182 get_predictor_value.
3183 (pass_strip_predict_hints::execute): Add handling of
3184 BUILT_IN_EXPECT_WITH_PROBABILITY.
3185 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
3187 * tree.h (DECL_BUILT_IN_P): New function.
3189 2018-08-10 Martin Liska <mliska@suse.cz>
3191 PR tree-optimization/85799
3192 * passes.def: Add argument for pass_strip_predict_hints.
3193 * predict.c (class pass_strip_predict_hints): Add new argument
3195 (strip_predictor_early): New function.
3196 (pass_strip_predict_hints::execute): Call the function to
3198 (strip_predict_hints): New function.
3199 * predict.def: Fix comment.
3201 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
3203 * Makefile.in: Clarify which tm.texi to copy over to assert the
3204 right to grant a GFDL license for all.
3206 2018-08-09 Jeff Law <law@redhat.com>
3208 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
3211 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
3213 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
3216 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
3218 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
3219 reductions for variable-length vectors.
3221 2018-08-09 David Malcolm <dmalcolm@redhat.com>
3224 * common.opt (fdiagnostics-show-line-numbers): New option.
3225 * diagnostic-show-locus.c (class layout): Add fields
3226 "m_show_line_numbers_p" and "m_linenum_width";
3227 (num_digits): New function.
3228 (test_num_digits): New function.
3229 (layout::layout): Initialize new fields. Update m_x_offset
3230 logic to handle any left margin.
3231 (layout::print_source_line): Print line number when requested.
3232 (layout::start_annotation_line): New member function.
3233 (layout::print_annotation_line): Call it.
3234 (layout::print_leading_fixits): Likewise.
3235 (layout::print_trailing_fixits): Likewise. Update calls to
3236 move_to_column for new parameter.
3237 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
3238 it to potentially call start_annotation_line.
3239 (layout::show_ruler): Call start_annotation_line.
3240 (selftest::test_line_numbers_multiline_range): New selftest.
3241 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
3242 and selftest::test_line_numbers_multiline_range.
3243 * diagnostic.c (diagnostic_initialize): Initialize
3244 show_line_numbers_p.
3245 * diagnostic.h (struct diagnostic_context): Add field
3246 "show_line_numbers_p".
3247 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
3248 -fno-diagnostics-show-line-numbers.
3249 * dwarf2out.c (gen_producer_string): Add
3250 OPT_fdiagnostics_show_line_numbers to the ignored options.
3251 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
3252 one setting" options.
3253 (append_compiler_options): Likewise to the dropped options.
3254 (append_diag_options): Likewise to the passed-on options.
3255 * opts.c (common_handle_option): Handle the new option.
3256 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
3258 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
3260 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
3261 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
3262 third argument of type "const signed char" to descriptions of
3263 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
3264 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
3265 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
3266 __builtin_bcdsub_ov functions.
3268 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
3270 PR tree-optimization/86858
3271 * tree-vect-loop.c (vect_is_simple_reduction): Restore
3272 flow_bb_inside_loop_p calls.
3274 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
3276 PR tree-optimization/86871
3277 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
3278 instead of gimple_assign_lhs.
3280 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
3283 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
3284 register constraint to operand 0.
3285 (add<mode>3_carryinC): Likewise.
3286 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
3288 2018-08-09 Martin Liska <mliska@suse.cz>
3291 * common.opt: Remove extra line.
3293 2018-08-09 Martin Liska <mliska@suse.cz>
3295 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
3296 at the end of a line, make first letter capital and end up
3297 a sentence with a dot.
3298 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
3299 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
3300 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
3301 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
3302 (PARAM_MAX_ISL_OPERATIONS): Likewise.
3303 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
3304 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
3305 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
3306 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
3307 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
3308 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
3309 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
3310 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
3311 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
3312 (PARAM_TREE_REASSOC_WIDTH): Likewise.
3313 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
3314 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
3315 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
3317 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
3320 * config/s390/s390.c (s390_option_override_internal): Reduce the
3321 stack-clash-protection-probe-interval param if it would be too big
3324 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
3327 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
3328 * config/m68k/m68k.c (handle_move_double): Don't call
3329 m68k_final_prescan_insn.
3330 (m68k_adjust_decorated_operand): Renamed from
3331 m68k_final_prescan_insn, remove first and third operand and
3333 (print_operand): Call it.
3334 (print_operand_address): Call it.
3336 2018-08-08 Nathan Sidwell <nathan@acm.org>
3338 * diagnostic.c (diagnostic_report_current_module): Use
3339 linemap_included_from & linemap_included_from_linemap.
3341 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
3343 * config/aarch64/aarch64-cores.def: Add phecda core.
3344 * config/aarch64/aarch64-tune.md: Regenerate.
3345 * doc/invoke.texi: Add phecda core.
3347 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
3350 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
3352 * config/s390/s390.md ("movti"): Add more alternatives for
3353 constant to GPR copies.
3355 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
3357 * config/s390/s390.c: Fix whitespace damage throughout the file.
3358 * config/s390/s390.h: Likewise.
3359 * config/s390/tpf.h: Likewise.
3361 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
3363 * config/s390/s390.c (s390_loadrelative_operand_p):
3364 Remove TARGET_CPU_ZARCH usages.
3365 (s390_rtx_costs): Likewise.
3366 (s390_legitimate_constant_p): Likewise.
3367 (s390_cannot_force_const_mem): Likewise.
3368 (legitimate_reload_constant_p): Likewise.
3369 (s390_preferred_reload_class): Likewise.
3370 (legitimize_pic_address): Likewise.
3371 (legitimize_tls_address): Likewise.
3372 (s390_split_branches): Removed.
3373 (s390_add_execute): Removed.
3374 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
3375 (s390_mainpool_start): Likewise.
3376 (s390_mainpool_finish): Likewise.
3377 (s390_mainpool_cancel): Removed.
3378 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
3379 (s390_chunkify_cancel): Likewise.
3380 (s390_return_addr_rtx): Likewise.
3381 (s390_register_info): Remove split_branches_pending_p uages.
3382 (s390_optimize_register_info): Likewise.
3383 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
3384 split_branches_pending_p usages.
3385 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
3386 (s390_load_got): Likewise.
3387 (s390_expand_split_stack_prologue): Likewise.
3388 (output_asm_nops): Likewise.
3389 (s390_function_profiler): Likewise.
3390 (s390_emit_call): Likewise.
3391 (s390_conditional_register_usage): Likewise.
3392 (s390_optimize_prologue): Likewise.
3393 (s390_reorg): Remove TARGET_CPU_ZARCH and
3394 split_branches_pending_p usages.
3395 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
3397 (s390_output_indirect_thunk_function): Likewise.
3398 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
3399 (TARGET_CPU_ZARCH_P): Removed.
3400 (struct machine_function): Remove split_branches_pending_p.
3401 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
3403 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
3405 * common/config/s390/s390-common.c (processor_flags_table):
3407 * config.gcc: Remove with_arch/with_tune support.
3408 * config/s390/2064.md: Remove cpu attribute comparisons.
3409 * config/s390/driver-native.c (s390_host_detect_local_cpu):
3411 * config/s390/linux.h (ASM_SPEC):
3412 Remove -march support.
3413 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
3414 Use a table to get an arch level.
3415 * config/s390/s390-opts.h (enum processor_type):
3417 * config/s390/s390.c
3418 (processor_table): Remove entries, add arch_level values.
3419 (s390_issue_rate): Remove cases.
3420 (s390_option_override): Adjust
3421 s390_option_override_internal() call.
3422 (s390_option_override_internal): Remove deprecation warning.
3423 (s390_valid_target_attribute_tree): Adjust
3424 s390_option_override_internal() call.
3425 * config/s390/s390.h (struct s390_processor):
3426 Share with s390-c.c, add arch_level field.
3427 * config/s390/s390.md:
3428 Remove occurrences in cpu attribute.
3429 * config/s390/s390.opt: Remove -march/-mtune support.
3430 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
3431 * doc/invoke.texi: Remove deprecation warning.
3433 2018-08-08 Luis Machado <luis.machado@linaro.org>
3435 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
3437 (qdf24xx_tunings): Set vector cost structure to
3438 qdf24xx_vector_cost.
3440 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
3441 <register_sextend>: Set to 3.
3443 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
3446 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
3447 * config/aarch64/aarch64-simd.md
3448 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
3449 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
3450 (aarch64_frecpx<mode>): ...this new pattern.
3451 * config/aarch64/aarch64-simd-builtins.def: Remove comment
3452 about aarch64_frecp<FRECP:frecp_suffix><mode>.
3454 2018-08-07 Martin Liska <mliska@suse.cz>
3457 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
3458 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
3459 * predict.def (PRED_MALLOC_NONNULL): New predictor.
3460 * doc/extend.texi: Document that malloc attribute adds
3463 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
3466 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3467 Define to speculation_safe_value_not_needed.
3469 2018-08-06 Jeff Law <law@redhat.com>
3471 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
3472 the vr_values instance to cprop_into_stmt.
3473 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
3474 (cprop_operand): Also query EVRP to determine if OP is a constant.
3476 2018-08-06 Nathan Sidwell <nathan@acm.org>
3478 * diagnostic.c (diagnostic_report_current_module): Reroll
3479 included-at loop. Translate text.
3481 2018-08-06 David Malcolm <dmalcolm@redhat.com>
3483 * function-tests.c (selftest::test_expansion_to_rtl): Call
3484 free_after_compilation.
3486 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3488 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
3490 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
3492 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
3493 loops with memory block operations from getting unrolled.
3495 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
3498 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3499 Define to speculation_safe_value_not_needed.
3501 2018-08-06 Jeff Law <law@redhat.com>
3503 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
3506 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3509 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
3510 with all enabled __intN types.
3512 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
3514 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3516 * alias.c (record_set): Check for clobber high.
3517 * cfgexpand.c (expand_gimple_stmt): Likewise.
3518 * combine-stack-adj.c (single_set_for_csa): Likewise.
3519 * combine.c (find_single_use_1): Likewise.
3520 (set_nonzero_bits_and_sign_copies): Likewise.
3521 (get_combine_src_dest): Likewise.
3522 (is_parallel_of_n_reg_sets): Likewise.
3523 (try_combine): Likewise.
3524 (record_dead_and_set_regs_1): Likewise.
3525 (reg_dead_at_p_1): Likewise.
3526 (reg_dead_at_p): Likewise.
3527 * dce.c (deletable_insn_p): Likewise.
3528 (mark_nonreg_stores_1): Likewise.
3529 (mark_nonreg_stores_2): Likewise.
3530 * df-scan.c (df_find_hard_reg_defs): Likewise.
3531 (df_uses_record): Likewise.
3532 (df_get_call_refs): Likewise.
3533 * dwarf2out.c (mem_loc_descriptor): Likewise.
3534 * haifa-sched.c (haifa_classify_rtx): Likewise.
3535 * ira-build.c (create_insn_allocnos): Likewise.
3536 * ira-costs.c (scan_one_insn): Likewise.
3537 * ira.c (equiv_init_movable_p): Likewise.
3538 (rtx_moveable_p): Likewise.
3539 (interesting_dest_for_shprep): Likewise.
3540 * jump.c (mark_jump_label_1): Likewise.
3541 * postreload-gcse.c (record_opr_changes): Likewise.
3542 * postreload.c (reload_cse_simplify): Likewise.
3543 (struct reg_use): Add source expr.
3544 (reload_combine): Check for clobber high.
3545 (reload_combine_note_use): Likewise.
3546 (reload_cse_move2add): Likewise.
3547 (move2add_note_store): Likewise.
3548 * print-rtl.c (print_pattern): Likewise.
3549 * recog.c (decode_asm_operands): Likewise.
3550 (store_data_bypass_p): Likewise.
3551 (if_test_bypass_p): Likewise.
3552 * regcprop.c (kill_clobbered_value): Likewise.
3553 (kill_set_value): Likewise.
3554 * reginfo.c (reg_scan_mark_refs): Likewise.
3555 * reload1.c (maybe_fix_stack_asms): Likewise.
3556 (eliminate_regs_1): Likewise.
3557 (elimination_effects): Likewise.
3558 (mark_not_eliminable): Likewise.
3559 (scan_paradoxical_subregs): Likewise.
3560 (forget_old_reloads_1): Likewise.
3561 * reorg.c (find_end_label): Likewise.
3562 (try_merge_delay_insns): Likewise.
3563 (redundant_insn): Likewise.
3564 (own_thread_p): Likewise.
3565 (fill_simple_delay_slots): Likewise.
3566 (fill_slots_from_thread): Likewise.
3567 (dbr_schedule): Likewise.
3568 * resource.c (update_live_status): Likewise.
3569 (mark_referenced_resources): Likewise.
3570 (mark_set_resources): Likewise.
3571 * rtl.c (copy_rtx): Likewise.
3572 * rtlanal.c (reg_referenced_p): Likewise.
3573 (single_set_2): Likewise.
3574 (noop_move_p): Likewise.
3575 (note_stores): Likewise.
3576 * sched-deps.c (sched_analyze_reg): Likewise.
3577 (sched_analyze_insn): Likewise.
3579 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3581 * cse.c (invalidate_reg): New function extracted from...
3582 (invalidate): ...here.
3583 (canonicalize_insn): Check for clobber high.
3584 (invalidate_from_clobbers): invalidate clobber highs.
3585 (invalidate_from_sets_and_clobbers): Likewise.
3586 (count_reg_usage): Check for clobber high.
3587 (insn_live_p): Likewise.
3588 * cselib.c (cselib_expand_value_rtx_1):Likewise.
3589 (cselib_invalidate_regno): Check for clobber in setter.
3590 (cselib_invalidate_rtx): Pass through setter.
3591 (cselib_invalidate_rtx_note_stores):
3592 (cselib_process_insn): Check for clobber high.
3593 * cselib.h (cselib_invalidate_rtx): Add operand.
3595 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3597 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
3598 (mark_not_eliminable): Likewise.
3599 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
3600 * lra-lives.c (process_bb_lives): Check for clobber high.
3601 * lra.c (new_insn_reg): Remember clobber highs.
3602 (collect_non_operand_hard_regs): Check for clobber high.
3603 (lra_set_insn_recog_data): Likewise.
3604 (add_regs_to_insn_regno_info): Likewise.
3605 (lra_update_insn_regno_info): Likewise.
3607 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3609 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
3610 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
3612 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3614 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
3615 (copy_insn_1): Likewise.
3616 (gen_hard_reg_clobber_high): New gen function.
3617 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
3618 * genemit.c (gen_exp): Likewise.
3619 (gen_emit_seq): Pass through info.
3620 (gen_insn): Check for CLOBBER_HIGH.
3621 (gen_expand): Pass through info.
3622 (gen_split): Likewise.
3623 (output_add_clobbers): Likewise.
3624 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
3625 (remove_clobbers): Likewise.
3626 * rtl.h (gen_hard_reg_clobber_high): New declaration.
3628 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3630 * doc/rtl.texi (clobber_high): Add.
3631 (parallel): Add in clobber high
3632 * rtl.c (rtl_check_failed_code3): Add function.
3633 * rtl.def (CLOBBER_HIGH): Add expression.
3634 * rtl.h (RTL_CHECKC3): Add macro.
3635 (rtl_check_failed_code3): Add declaration.
3636 (XC3EXP): Add macro.
3638 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
3641 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
3642 cfun->machine->max_used_stack_alignment if needed.
3644 2018-08-04 Martin Sebor <msebor@redhat.com>
3646 PR tree-optimization/86571
3647 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
3650 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
3653 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3656 2018-08-03 Jeff Law <law@redhat.com>
3659 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3660 Define to speculation_safe_value_not_needed.
3662 2018-08-03 David Malcolm <dmalcolm@redhat.com>
3664 * doc/gcov.texi (-x): Remove duplicate "to".
3665 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
3666 (-Wif-not-aligned): Remove duplicate "is".
3667 (-flto): Remove duplicate "the".
3668 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
3669 duplicate "v5.00.b".
3670 (MSP430 Options): Remove duplicate "and" from the description
3671 of "-mgprel-sec=regexp".
3672 (x86 Options): Remove duplicate copies of "vmldLog102" and
3673 vmlsLog104 from description of "-mveclibabi=type".
3675 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
3677 * internal-fn.h (first_commutative_argument): Declare.
3678 * internal-fn.c (first_commutative_argument): New function.
3679 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
3680 restrictions for pattern statements. Use first_commutative_argument
3681 to look for commutative operands in calls to internal functions.
3683 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
3685 * Makefile.in (wide-int-range.o): New.
3686 * tree-vrp.c: Move all the wide_int_* functions to...
3687 * wide-int-range.cc: ...here.
3688 * tree-vrp.h: Move all the wide_int_* prototypes to...
3689 * wide-int-range.h: ...here.
3691 2018-08-03 Tom de Vries <tdevries@suse.de>
3693 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
3695 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
3696 * except.c (output_function_exception_table): Do early exit if
3697 targetm_common.except_unwind_info (&global_options) == UI_NONE.
3699 2018-08-03 Martin Liska <mliska@suse.cz>
3701 * predict.c (dump_prediction): Change to 2 digits
3704 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
3706 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
3708 (falkor_am_1_gtov_gtov): ... a new insn reservation.
3710 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
3712 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
3713 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
3714 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
3715 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
3716 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
3718 2018-08-02 David Malcolm <dmalcolm@redhat.com>
3720 * diagnostic-show-locus.c (diagnostic_show_locus): Use
3721 pp_take_prefix when saving the existing prefix.
3722 * diagnostic.c (diagnostic_append_note): Likewise.
3723 * langhooks.c (lhd_print_error_function): Likewise.
3724 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
3725 param's type. Free the existing prefix.
3726 (pp_take_prefix): New function.
3727 (pretty_printer::pretty_printer): Drop the prefix parameter.
3728 Rename the length parameter to match the comment.
3729 (pretty_printer::~pretty_printer): Free the prefix.
3730 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
3732 (struct pretty_printer): Drop the "const" from "prefix" field's
3733 type and clarify memory management.
3734 (pp_set_prefix): Drop the "const" from the 2nd param.
3735 (pp_take_prefix): New decl.
3737 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
3739 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
3740 for word_mode_ok here instead of passing as argument.
3741 (expand_block_compare): Change select_block_compare_mode() call.
3742 (expand_strncmp_gpr_sequence): New function.
3743 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
3745 2018-08-02 Jeff Law <law@redhat.com>
3748 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3749 Define to speculation_safe_value_not_needed.
3752 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3753 Define to speculation_safe_value_not_needed.
3755 2018-08-02 Tom de Vries <tdevries@suse.de>
3758 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
3759 function. Return UI_TARGET unconditionally.
3760 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
3761 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
3763 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3765 * genemit.c (print_overload_test): Fix typo.
3767 2018-08-02 Richard Biener <rguenther@suse.de>
3769 PR tree-optimization/86816
3770 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
3771 which checks for value availability before querying it.
3772 (gvn_uses_equal): Use it.
3773 (same_succ_hash): Likewise.
3774 (gimple_equal_p): Likewise.
3776 2018-08-02 Nick Clifton <nickc@redhat.com>
3779 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3780 Define to speculation_safe_value_not_needed.
3783 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3784 Define to speculation_safe_value_not_needed.
3787 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3788 Define to speculation_safe_value_not_needed.
3791 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3792 Define to speculation_safe_value_not_needed.
3795 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3796 Define to speculation_safe_value_not_needed.
3799 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3800 Define to speculation_safe_value_not_needed.
3803 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3804 Define to speculation_safe_value_not_needed.
3807 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3808 Define to speculation_safe_value_not_needed.
3811 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
3812 speculation_safe_value_not_needed.
3815 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
3816 to speculation_safe_value_not_needed.
3818 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3820 * doc/md.texi: Expand the documentation of instruction names
3821 to mention port-local uses. Document '@' in pattern names.
3822 * read-md.h (overloaded_instance, overloaded_name): New structs.
3824 (md_reader::handle_overloaded_name): New member function.
3825 (md_reader::get_overloads): Likewise.
3826 (md_reader::m_first_overload): New member variable.
3827 (md_reader::m_next_overload_ptr): Likewise.
3828 (md_reader::m_overloads_htab): Likewise.
3829 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
3830 m_next_overload_ptr and m_overloads_htab.
3831 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
3832 (get_mode_token, get_code_token, get_int_token): New functions.
3833 (map_attr_string): Add an optional argument that passes back
3834 the associated iterator.
3835 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
3836 (md_reader::handle_overloaded_name, add_overload_instance): New
3838 (apply_iterators): Handle '@' names. Report an error if '@'
3839 is used without iterators.
3840 (initialize_iterators): Initialize the new iterator_group fields.
3841 * genopinit.c (handle_overloaded_code_for)
3842 (handle_overloaded_gen): New functions.
3843 (main): Use them to print declarations of maybe_code_for_* and
3844 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
3845 * genemit.c (print_overload_arguments, print_overload_test)
3846 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
3847 (main): Use it to print definitions of maybe_code_for_* and
3848 maybe_gen_* functions.
3849 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
3850 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
3851 instead of explicit mode checks.
3852 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
3853 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
3854 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
3855 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
3856 (aarch64_expand_compare_and_swap): Likewise
3857 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
3858 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
3859 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
3860 (aarch64_constant_pool_reload_icode): Delete.
3861 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
3862 instead of aarch64_constant_pool_reload_icode. Use
3863 code_for_aarch64_reload_mov instead of explicit mode checks.
3864 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
3865 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
3866 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
3867 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
3868 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
3869 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
3870 (aarch64_atomic_load_op_code): Delete.
3871 (aarch64_emit_atomic_load_op): Likewise.
3872 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
3873 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
3874 instead of aarch64_emit_atomic_load_op.
3875 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
3876 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
3877 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
3878 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
3879 character before the pattern name.
3880 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
3881 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
3882 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
3883 (aarch64_frecps<mode>): Likewise.
3884 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
3885 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
3886 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
3887 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
3888 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
3890 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3892 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
3893 Allow HFmode constants if TARGET_FP_F16INST.
3895 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
3898 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
3899 No longer check last store for clobber of address register.
3901 2018-08-02 Martin Liska <mliska@suse.cz>
3903 PR gcov-profile/86817
3904 * gcov.c (process_all_functions): New function.
3906 (process_file): Move functions processing to
3907 process_all_functions.
3909 2018-08-02 David Malcolm <dmalcolm@redhat.com>
3911 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
3912 "const" to the "gimple *" and "rtx_insn *" parameters.
3913 * dumpfile.h (dump_user_location_t::dump_user_location_t):
3915 (dump_location_t::dump_location_t): Likewise.
3917 2018-08-01 Martin Sebor <msebor@redhat.com>
3919 PR tree-optimization/86650
3920 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
3921 rather than a "gcall *". Directly pass the data of interest
3922 to percent_K_format, rather than building a temporary CALL_EXPR
3924 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
3925 (gimple_fold_builtin_strncat): Adjust.
3926 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
3927 gcall* argument with gimple*.
3928 * gimple-ssa-warn-restrict.c (check_call): Same.
3929 (wrestrict_dom_walker::before_dom_children): Same.
3930 (builtin_access::builtin_access): Same.
3931 (check_bounds_or_overlap): Same
3932 (maybe_diag_overlap): Same.
3933 (maybe_diag_offset_bounds): Same.
3934 * tree-diagnostic.c (default_tree_printer): Move usage of
3935 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
3937 * tree-pretty-print.c (percent_K_format): Add argument.
3938 * tree-pretty-print.h: Add argument.
3939 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
3940 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
3941 (maybe_diag_stxncpy_trunc): Same.
3942 (handle_builtin_stxncpy): Same.
3943 (handle_builtin_strcat): Same.
3945 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3947 * match.pd: Optimise pointer range checks.
3949 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3951 PR tree-optimization/86758
3952 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
3953 to remove pattern statements.
3955 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3957 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
3958 result of dfs_enumerate_from when constructing stmt_vec_infos,
3959 instead of additionally calling get_loop_body.
3961 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3963 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
3965 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
3966 When creating an iv, assert that the step is not known to be zero.
3967 (vect_setup_realignment): Update call accordingly.
3968 * tree-vect-stmts.c (vectorizable_store): Likewise.
3969 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
3971 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3973 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
3974 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
3975 (vectorizable_reduction): Likewise.
3976 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3977 (vect_detect_hybrid_slp_stmts): Likewise.
3978 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
3980 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
3982 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
3983 (wide_int_set_zero_nonzero_bits): ...this.
3984 (zero_nonzero_bits_from_vr): Rename to...
3985 (vrp_set_zero_nonzero_bits): ...this.
3986 (extract_range_from_multiplicative_op_1): Abstract wide int
3988 (wide_int_range_multiplicative_op): ...here.
3989 (extract_range_from_binary_expr_1): Extract wide int binary
3990 operations into their own functions.
3991 (wide_int_range_lshift): New.
3992 (wide_int_range_can_optimize_bit_op): New.
3993 (wide_int_range_shift_undefined_p): New.
3994 (wide_int_range_bit_xor): New.
3995 (wide_int_range_bit_ior): New.
3996 (wide_int_range_bit_and): New.
3997 (wide_int_range_trunc_mod): New.
3998 (extract_range_into_wide_ints): New.
3999 (vrp_shift_undefined_p): New.
4000 (extract_range_from_multiplicative_op): New.
4001 (vrp_can_optimize_bit_op): New.
4002 * tree-vrp.h (value_range::dump): New.
4003 (wide_int_range_multiplicative_op): New.
4004 (wide_int_range_lshift):New.
4005 (wide_int_range_shift_undefined_p): New.
4006 (wide_int_range_bit_xor): New.
4007 (wide_int_range_bit_ior): New.
4008 (wide_int_range_bit_and): New.
4009 (wide_int_range_trunc_mod): New.
4010 (zero_nonzero_bits_from_bounds): Rename to...
4011 (wide_int_set_zero_nonzero_bits): ...this.
4012 (zero_nonzero_bits_from_vr): Rename to...
4013 (vrp_set_zero_nonzero_bits): ...this.
4014 (range_easy_mask_min_max): Rename to...
4015 (wide_int_range_can_optimize_bit_op): this.
4016 * vr-values.c (simplify_bit_ops_using_ranges): Rename
4017 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
4019 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4021 * tree-vectorizer.h (vect_orig_stmt): New function.
4022 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
4023 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
4024 (vect_create_epilog_for_reduction): Likewise.
4025 (vectorizable_live_operation): Likewise.
4026 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4027 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
4028 * tree-vect-stmts.c (vectorizable_call): Likewise.
4029 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
4031 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4033 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
4035 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
4036 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
4037 (vect_transform_loop): Likewise.
4038 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
4040 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4042 * tree-vectorizer.h (vect_schedule_slp): Return void.
4043 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
4044 (vect_schedule_slp): Likewise.
4046 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4048 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
4050 (vect_transform_loop): Update calls accordingly. Schedule SLP
4051 instances before the main loop, if any exist.
4053 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4055 PR tree-optimization/86749
4056 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
4057 If the lhs is used in a COND_EXPR, check that it is being used
4058 as the "then" or "else" value.
4060 2018-08-01 Tom de Vries <tdevries@suse.de>
4063 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
4064 speculation_safe_value_not_needed.
4066 2018-08-01 Richard Biener <rguenther@suse.de>
4068 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
4071 2018-08-01 Martin Liska <mliska@suse.cz>
4073 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
4074 format how successful transformation is dumped.
4075 (gimple_mod_pow2_value_transform): Likewise.
4076 (gimple_mod_subtract_transform): Likewise.
4077 (gimple_stringops_transform): Likewise.
4079 2018-08-01 Martin Liska <mliska@suse.cz>
4082 * value-prof.c (interesting_stringop_to_profile_p):
4083 Simplify the code and add BUILT_IN_MEMMOVE.
4084 (gimple_stringops_transform): Likewise.
4086 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
4088 * config/aarch64/aarch64-simd.md
4089 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
4090 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
4091 use GPI iterator instead of SI mode.
4093 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
4095 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
4096 rs6000_speculation_barrier.
4097 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
4098 new barrier pattern name.
4100 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
4102 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
4103 (speculation_barrier): New insn.
4105 2018-08-01 Richard Biener <rguenther@suse.de>
4108 * graphite.h: Include isl/id.h and isl/space.h to allow build
4111 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
4114 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4115 mode for COFF targets.
4116 * defaults.h (TARGET_COFF): Define.
4117 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
4118 TARGET_COFF): Define.
4119 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
4120 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
4121 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
4123 2018-07-31 Alexandre Oliva <oliva@adacore.com>
4125 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
4126 * gimple-streamer-out.c (output_bb): Save it.
4127 * lto-streamer-in.c (input_struct_function_base): Restore
4128 instance discriminator if available. Create map on demand.
4129 * lto-streamer-out.c (output_struct_function_base): Save it if
4131 * final.c (decl_to_instance_map): Document LTO strategy.
4133 2018-07-31 Alexandre Oliva <oliva@adacore.com>
4134 Olivier Hainque <hainque@adacore.com>
4136 * debug.h (decl_to_instance_map_t): New type.
4137 (decl_to_instance_map): Declare.
4138 (maybe_create_decl_to_instance_map): New inline function.
4139 * final.c (bb_discriminator, last_bb_discriminator): New statics,
4140 to track basic block discriminators.
4141 (final_start_function_1): Initialize them.
4142 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
4144 (decl_to_instance_map): New variable.
4145 (map_decl_to_instance, maybe_set_discriminator): New functions.
4146 (notice_source_line): Set discriminator.
4148 2018-07-31 Ian Lance Taylor <iant@golang.org>
4150 * targhooks.c (default_have_speculation_safe_value): Add
4153 2018-07-31 David Malcolm <dmalcolm@redhat.com>
4155 * dump-context.h: Include "pretty-print.h".
4156 (dump_context::refresh_dumps_are_enabled): New decl.
4157 (dump_context::emit_item): New decl.
4158 (class dump_context): Add fields "m_test_pp" and
4160 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
4161 (temp_dump_context::get_dumped_text): New decl.
4162 (class temp_dump_context): Add field "m_pp".
4163 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
4164 (dump_context::refresh_dumps_are_enabled): ...and add a test for
4166 (set_dump_file): Update for above change.
4167 (set_alt_dump_file): Likewise.
4168 (dump_loc): New overload, taking a pretty_printer *.
4169 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
4170 to any test pretty-printer.
4171 (make_item_for_dump_gimple_stmt): New function, adapted from
4172 optinfo::add_gimple_stmt.
4173 (dump_context::dump_gimple_stmt): Call it, and use the result,
4174 eliminating the direct usage of dump_file and alt_dump_file in
4175 favor of indirectly using them via emit_item.
4176 (make_item_for_dump_gimple_expr): New function, adapted from
4177 optinfo::add_gimple_expr.
4178 (dump_context::dump_gimple_expr): Call it, and use the result,
4179 eliminating the direct usage of dump_file and alt_dump_file in
4180 favor of indirectly using them via emit_item.
4181 (make_item_for_dump_generic_expr): New function, adapted from
4183 (dump_context::dump_generic_expr): Call it, and use the result,
4184 eliminating the direct usage of dump_file and alt_dump_file in
4185 favor of indirectly using them via emit_item.
4186 (make_item_for_dump_printf_va): New function, adapted from
4187 optinfo::add_printf_va.
4188 (make_item_for_dump_printf): New function.
4189 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
4190 and use the result, eliminating the direct usage of dump_file and
4191 alt_dump_file in favor of indirectly using them via emit_item.
4192 (make_item_for_dump_dec): New function.
4193 (dump_context::dump_dec): Call it, and use the result,
4194 eliminating the direct usage of dump_file and alt_dump_file in
4195 favor of indirectly using them via emit_item.
4196 (make_item_for_dump_symtab_node): New function, adapted from
4197 optinfo::add_symtab_node.
4198 (dump_context::dump_symtab_node): Call it, and use the result,
4199 eliminating the direct usage of dump_file and alt_dump_file in
4200 favor of indirectly using them via emit_item.
4201 (dump_context::begin_scope): Reimplement, avoiding direct usage
4202 of dump_file and alt_dump_file in favor of indirectly using them
4204 (dump_context::emit_item): New member function.
4205 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
4206 Set up test pretty-printer on the underlying context. Call
4207 refresh_dumps_are_enabled.
4208 (temp_dump_context::~temp_dump_context): Call
4209 refresh_dumps_are_enabled.
4210 (temp_dump_context::get_dumped_text): New member function.
4211 (selftest::verify_dumped_text): New function.
4212 (ASSERT_DUMPED_TEXT_EQ): New macro.
4213 (selftest::test_capture_of_dump_calls): Run all tests twice, with
4214 and then without optinfo enabled. Add uses of
4215 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
4216 * dumpfile.h: Update comment for the dump_* API.
4217 * optinfo-emit-json.cc
4218 (selftest::test_building_json_from_dump_calls): Update for new
4219 param for temp_dump_context ctor.
4220 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
4221 and "m_owned" field.
4222 (optinfo_item::~optinfo_item): Likewise.
4223 (optinfo::add_item): New member function.
4224 (optinfo::emit): Update comment.
4225 (optinfo::add_string): Delete.
4226 (optinfo::add_printf): Delete.
4227 (optinfo::add_printf_va): Delete.
4228 (optinfo::add_gimple_stmt): Delete.
4229 (optinfo::add_gimple_expr): Delete.
4230 (optinfo::add_tree): Delete.
4231 (optinfo::add_symtab_node): Delete.
4232 (optinfo::add_dec): Delete.
4233 * optinfo.h (class dump_context): New forward decl.
4234 (optinfo::add_item): New decl.
4235 (optinfo::add_string): Delete.
4236 (optinfo::add_printf): Delete.
4237 (optinfo::add_printf_va): Delete.
4238 (optinfo::add_gimple_stmt): Delete.
4239 (optinfo::add_gimple_expr): Delete.
4240 (optinfo::add_tree): Delete.
4241 (optinfo::add_symtab_node): Delete.
4242 (optinfo::add_dec): Delete.
4243 (optinfo::add_poly_int): Delete.
4244 (optinfo_item::optinfo_item): Remove "owned" param.
4245 (class optinfo_item): Remove field "m_owned".
4247 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4250 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
4251 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
4252 requested variable alignment.
4253 (expand_one_ssa_partition): Likewise.
4254 (expand_one_var): Likewise.
4256 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4258 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
4259 to speculation_safe_value_not_needed.
4261 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4263 * targhooks.h (speculation_safe_value_not_needed): New prototype.
4264 * targhooks.c (speculation_safe_value_not_needed): New function.
4265 * target.def (have_speculation_safe_value): Update documentation.
4266 * doc/tm.texi: Regenerated.
4268 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4270 * config/aarch64/iterators.md (ALLI_TI): New iterator.
4271 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
4273 (despeculate_copy<ALLI:mode>_insn): New insn.
4274 (despeculate_copyti_insn): New insn.
4275 (despeculate_simple<ALLI:mode>): New insn
4276 (despeculate_simpleti): New insn.
4277 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
4279 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
4280 aarch64_speculation_safe_value.
4281 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
4283 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4285 * config/aarch64/aarch64-speculation.cc: New file.
4286 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
4287 before pass_reorder_blocks.
4288 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
4290 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
4291 X14 and X15 when tracking speculation.
4292 * config/aarch64/aarch64.md (register name constants): Add
4293 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
4294 (unspec): Add UNSPEC_SPECULATION_TRACKER.
4295 (speculation_barrier): New insn attribute.
4296 (cmp<mode>): Allow SP in comparisons.
4297 (speculation_tracker): New insn.
4298 (speculation_barrier): Add speculation_barrier attribute.
4299 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
4300 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
4301 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
4303 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4305 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
4306 aarch64_track_speculation is true.
4307 (tb<optab><mode>1): Likewise.
4308 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
4309 generate CB[N]Z when tracking speculation.
4310 (aarch64_split_compare_and_swap): Likewise.
4311 (aarch64_split_atomic_op): Likewise.
4313 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4315 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
4317 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4319 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
4320 (speculation_barrier): New insn.
4322 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4324 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
4325 * config/arm/arm.md (speculation_barrier): New expand.
4326 (speculation_barrier_insn): New pattern.
4328 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
4330 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
4331 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
4332 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
4333 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
4335 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
4336 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
4337 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
4338 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
4339 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
4340 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
4341 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
4342 * builtins.c (expand_speculation_safe_value): New function.
4343 (expand_builtin): Call it.
4344 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
4345 * doc/extend.texi: Document __builtin_speculation_safe_value.
4346 * doc/md.texi: Document "speculation_barrier" pattern.
4347 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
4348 TARGET_HAVE_SPECULATION_SAFE_VALUE.
4349 * doc/tm.texi: Regenerated.
4350 * target.def (have_speculation_safe_value, speculation_safe_value): New
4352 * targhooks.c (default_have_speculation_safe_value): New function.
4353 (default_speculation_safe_value): New function.
4354 * targhooks.h (default_have_speculation_safe_value): Add prototype.
4355 (default_speculation_safe_value): Add prototype.
4357 2018-07-31 David Malcolm <dmalcolm@redhat.com>
4359 * dump-context.h (dump_context::dump_loc): New decl.
4360 * dumpfile.c (dump_context::dump_loc): New member function.
4361 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
4362 and dump_gimple_stmt.
4363 (dump_context::dump_gimple_expr_loc): Likewise, using
4365 (dump_context::dump_generic_expr_loc): Likewise, using
4367 (dump_context::dump_printf_loc_va): Likewise, using
4369 (dump_context::begin_scope): Explicitly using the global function
4370 "dump_loc", rather than the member function.
4372 2018-07-31 Martin Sebor <msebor@redhat.com>
4374 PR tree-optimization/86741
4375 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
4377 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
4379 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
4380 depend on whether prefetch instructions will be emitted or not.
4381 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
4382 will be emitted or not.
4383 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
4384 (TARGET_SETMEM_PFD): New macros.
4386 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4388 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
4389 (NULL_STMT_VEC_INFO): Delete.
4390 (stmt_vec_info::operator*): Likewise.
4391 (stmt_vec_info::operator gimple *): Likewise.
4392 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
4393 of NULL_STMT_VEC_INFO.
4394 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4395 (vect_reassociating_reduction_p): Likewise.
4396 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
4397 (vectorizable_store): Likewise.
4398 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
4399 (vec_info::free_stmt_vec_infos): Likewise.
4401 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4403 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
4404 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
4406 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4408 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
4409 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4410 (vec_info::free_stmt_vec_info): New private member functions.
4411 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
4412 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
4413 * tree-parloops.c (gather_scalar_reductions): Remove calls to
4414 set_stmt_vec_info_vec and free_stmt_vec_infos.
4415 * tree-vect-loop.c (_loop_vec_info): Remove call to
4416 set_stmt_vec_info_vec.
4417 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
4418 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
4419 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
4420 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4421 (vec_info::free_stmt_vec_info): ...these new functions. Remove
4422 assignments in {vec_info::,}new_stmt_vec_info that are redundant
4423 with the clearing in the xcalloc.
4424 (stmt_vec_info_vec): Delete.
4425 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
4426 (vectorize_loops): Likewise.
4427 (vec_info::~vec_info): Remove argument from call to
4428 free_stmt_vec_infos.
4429 (vec_info::add_stmt): Remove vinfo argument from call to
4432 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4434 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
4435 rather than a gimple stmt.
4436 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
4437 information for pattern statements when passed the original
4438 statement; instead wait to be passed the pattern statement itself.
4439 Don't call set_vinfo_for_stmt here.
4440 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
4441 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
4442 stmt_vec_infos here.
4443 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
4444 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
4445 stmt_vec_infos entry.
4447 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4449 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
4450 * tree-vectorizer.c (vec_info::replace_stmt): New function.
4451 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
4452 * tree-vect-stmts.c (vectorizable_call): Likewise.
4453 (vectorizable_simd_clone_call): Likewise.
4455 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4457 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
4458 * tree-vectorizer.c (vec_info::remove_stmt): New function.
4459 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
4460 * tree-vect-loop.c (vect_transform_loop): Likewise.
4461 * tree-vect-slp.c (vect_schedule_slp): Likewise.
4462 * tree-vect-stmts.c (vect_remove_stores): Likewise.
4464 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4466 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
4467 (vect_dr_stmt): Delete.
4468 * tree-vectorizer.c (vec_info::lookup_dr): New function.
4469 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
4471 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
4472 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
4473 (vect_verify_datarefs_alignment, vect_peeling_supportable)
4474 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4475 (vect_analyze_data_refs): Likewise.
4476 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
4478 (vect_find_same_alignment_drs): Likewise.
4479 (vect_slp_analyze_node_dependences): Update calls accordingly.
4480 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
4481 instead of DR_VECT_AUX.
4482 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
4483 of a vector data references. Use vec_info::lookup_dr instead of
4485 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
4486 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
4487 instead of DR_VECT_AUX.
4489 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4491 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
4493 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4495 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4496 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4497 (vect_gen_prolog_loop_niters): Likewise.
4499 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4501 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
4502 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
4503 (vect_known_alignment_in_bytes, vect_dr_behavior)
4504 (vect_get_scalar_dr_size): Take references as dr_vec_infos
4505 instead of data_references. Update calls to other routines for
4506 which the same change has been made.
4507 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
4508 dr_vec_infos instead of stmt_vec_infos.
4509 (vect_analyze_data_ref_dependence): Update call accordingly.
4510 (vect_slp_analyze_data_ref_dependence)
4511 (vect_record_base_alignments): Use DR_VECT_AUX.
4512 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
4513 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
4514 (vector_alignment_reachable_p, vect_get_data_access_cost)
4515 (vect_peeling_supportable, vect_analyze_group_access_1)
4516 (vect_analyze_group_access, vect_analyze_data_ref_access)
4517 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
4518 (vect_compile_time_alias, vect_small_gap_p)
4519 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
4520 (vect_supportable_dr_alignment): Take references as dr_vec_infos
4521 instead of data_references. Update calls to other routines for
4522 which the same change has been made.
4523 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
4524 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4525 (vect_slp_analyze_and_verify_node_alignment)
4526 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4527 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4528 (vect_setup_realignment): Use dr_vec_infos. Update calls after
4530 (_vect_peel_info::dr): Replace with...
4531 (_vect_peel_info::dr_info): ...this new field.
4532 (vect_peeling_hash_get_most_frequent)
4533 (vect_peeling_hash_choose_best_peeling): Update accordingly.
4534 (vect_peeling_hash_get_lowest_cost):
4535 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
4536 routines for which the same change has been made.
4537 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
4539 * tree-vect-loop-manip.c (get_misalign_in_elems)
4540 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
4542 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4543 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
4544 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
4545 (get_group_load_store_type, get_negative_load_store_type)
4546 (vect_get_data_ptr_increment, vectorizable_store)
4547 (vectorizable_load): Likewise.
4548 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
4549 Update calls to other routines for which the same change has been made.
4551 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4553 * tree-vectorizer.h (vec_info::move_dr): New member function.
4554 (dataref_aux): Rename to...
4555 (dr_vec_info): ...this and add "dr" and "stmt" fields.
4556 (_stmt_vec_info::dr_aux): Update accordingly.
4557 (_stmt_vec_info::data_ref_info): Delete.
4558 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
4559 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
4560 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
4561 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
4563 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
4564 (STMT_VINFO_DR_INFO): New macro.
4565 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
4566 (set_dr_misalignment): Update after rename of dataref_aux.
4567 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
4568 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
4569 initialization of STMT_VINFO_DATA_REF.
4570 * tree-vectorizer.c (vec_info::move_dr): New function.
4571 * tree-vect-patterns.c (vect_recog_bool_pattern)
4572 (vect_recog_mask_conversion_pattern)
4573 (vect_recog_gather_scatter_pattern): Use it.
4574 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
4575 the "dr" and "stmt" fields of dr_vec_info instead of
4576 STMT_VINFO_DATA_REF.
4578 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4580 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
4581 (is_pattern_stmt_p): Use it.
4582 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
4583 on pattern statements.
4585 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4587 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
4588 original stmt as a stmt_vec_info rather than a gimple stmt.
4589 (vect_pattern_recog_1): Take the statement directly as a
4590 stmt_vec_info, rather than via a gimple_stmt_iterator.
4591 Update call to vect_mark_pattern_stmts.
4592 (vect_pattern_recog): Update calls accordingly.
4594 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4596 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
4597 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
4598 a vect_def_type for the first argument.
4599 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
4600 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
4601 operand if it isn't defined by a vectorized statement.
4602 (vect_build_gather_load_calls): Remove the mask_dt argument and
4603 update calls to vect_get_vec_def_for_stmt_copy.
4604 (vectorizable_bswap): Likewise the dt argument.
4605 (vectorizable_call): Update calls to vectorizable_bswap and
4606 vect_get_vec_def_for_stmt_copy.
4607 (vectorizable_simd_clone_call, vectorizable_assignment)
4608 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
4609 (vectorizable_comparison): Update calls to
4610 vect_get_vec_def_for_stmt_copy.
4611 (vectorizable_store): Likewise. Remove now-unnecessary calls to
4613 (vect_get_loop_based_defs): Remove dt argument and update call
4614 to vect_get_vec_def_for_stmt_copy.
4615 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
4616 and vect_get_vec_def_for_stmt_copy.
4617 (vectorizable_load): Update calls to vect_build_gather_load_calls
4618 and vect_get_vec_def_for_stmt_copy.
4619 * tree-vect-loop.c (vect_create_epilog_for_reduction)
4620 (vectorizable_reduction, vectorizable_live_operation): Update calls
4621 to vect_get_vec_def_for_stmt_copy.
4623 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4625 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
4626 and gimple stmt arguments with a stmt_vec_info.
4627 (vect_record_base_alignments): Update calls accordingly.
4628 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
4629 and gimple stmt arguments with a stmt_vec_info.
4630 (vect_build_slp_tree_1): Remove vinfo argument and update call
4631 to vect_record_max_nunits.
4632 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
4633 and vect_record_max_nunits.
4635 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4637 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
4638 file and take a stmt_vec_info instead of a gimple stmt.
4639 (supportable_widening_operation, vect_finish_replace_stmt)
4640 (vect_finish_stmt_generation, vect_get_store_rhs)
4641 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
4642 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
4643 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
4644 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
4645 (vect_create_data_ref_ptr, bump_vector_ptr)
4646 (vect_permute_store_chain, vect_setup_realignment)
4647 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
4648 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
4649 (vectorizable_reduction, vectorizable_induction)
4650 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
4651 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
4652 than gimple stmts as arguments.
4653 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
4654 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
4655 (can_group_stmts_p, vect_check_gather_scatter)
4656 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4657 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
4658 (vect_permute_load_chain, vect_shift_permute_load_chain)
4659 (vect_transform_grouped_load)
4660 (vect_record_grouped_load_vectors): Likewise.
4661 * tree-vect-loop.c (vect_fixup_reduc_chain)
4662 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
4663 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
4664 (vectorizable_reduction, vectorizable_induction)
4665 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
4666 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
4667 (vect_get_load_store_mask): Likewise.
4668 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
4669 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
4670 * tree-vect-stmts.c (vect_mark_relevant)
4671 (is_simple_and_all_uses_invariant)
4672 (exist_non_indexing_operands_for_use_p, process_use)
4673 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
4674 (vect_get_vec_def_for_operand, vect_get_vec_defs)
4675 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
4676 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
4677 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
4678 (get_negative_load_store_type, get_load_store_type)
4679 (vect_check_load_store_mask, vect_check_store_rhs)
4680 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
4681 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4682 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4683 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4684 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
4685 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
4686 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
4687 (supportable_widening_operation): Likewise.
4689 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4691 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
4692 a stmt_vec_info instead of a gcall.
4693 (vect_check_gather_scatter): Update call accordingly.
4694 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
4696 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
4697 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
4698 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
4699 instead of a gimple stmt.
4700 (vect_transform_loop): Update calls accordingly.
4701 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
4702 stmt_vec_infos instead of gimple stmts.
4703 (vect_analyze_slp_instance): Update use accordingly.
4704 * tree-vect-stmts.c (read_vector_array, write_vector_array)
4705 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
4706 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
4707 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
4708 (vect_gen_widened_results_half, vect_get_loop_based_defs)
4709 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
4710 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
4711 down to subroutines.
4713 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4715 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
4716 of the worklist from a vector of gimple stmts to a vector of
4718 * tree-vect-stmts.c (vect_mark_relevant, process_use)
4719 (vect_mark_stmts_to_be_vectorized): Likewise
4721 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4723 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
4724 statement before passing it to vect_analyze_stmt.
4725 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
4726 the chain of phi vector definitions. Track the exit phi via its
4728 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
4729 STMT_VINFO_REDUC_DEF.
4730 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
4731 stmt_vec_infos to handle the statement chains.
4732 (vect_get_slp_defs): Record the first statement in the node
4733 using a stmt_vec_info.
4734 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
4735 statements here and pass their stmt_vec_info down to subroutines.
4736 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
4737 down to vect_finish_stmt_generation.
4738 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
4739 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
4740 stmt_vec_infos to subroutines.
4741 (vect_remove_stores): Use stmt_vec_infos to handle the statement
4744 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4746 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
4747 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
4748 (vect_permute_store_chain, vect_setup_realignment)
4749 (vect_permute_load_chain, vect_shift_permute_load_chain)
4750 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
4751 stmts internally, and when passing values to other vectorizer routines.
4752 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4753 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
4754 (vect_analyze_loop_operations, get_initial_def_for_reduction)
4755 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
4756 (vectorizable_reduction, vectorizable_induction)
4757 (vectorizable_live_operation, vect_transform_loop_stmt)
4758 (vect_transform_loop): Likewise.
4759 * tree-vect-patterns.c (vect_reassociating_reduction_p)
4760 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
4761 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
4762 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4763 (vect_slp_analyze_node_operations_1): Likewise.
4764 * tree-vect-stmts.c (vect_mark_relevant, process_use)
4765 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
4766 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
4767 (vect_finish_stmt_generation_1, get_group_load_store_type)
4768 (get_load_store_type, vect_build_gather_load_calls)
4769 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4770 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4771 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4772 (vectorizable_store, vectorizable_load, vectorizable_condition)
4773 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
4774 (supportable_widening_operation): Likewise.
4775 (vect_get_vector_types_for_stmt): Likewise.
4776 * tree-vectorizer.h (vect_dr_behavior): Likewise.
4778 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4780 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4781 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
4782 (vect_permute_store_chain, vect_permute_load_chain)
4783 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
4784 repeated stmt_vec_info lookups.
4785 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4786 (vect_update_ivs_after_vectorizer): Likewise.
4787 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
4788 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4789 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
4790 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4791 (vect_bb_slp_scalar_cost): Likewise.
4792 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
4794 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4796 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
4797 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
4798 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
4799 to get gassigns and gcalls, rather than operating on generc gimple
4801 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
4802 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
4803 (vectorizable_load, vect_analyze_stmt): Likewise.
4804 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
4806 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4808 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
4809 return stmt_vec_infos rather than gimple stmts. Do not accept
4811 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
4813 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4814 Update use of get_later_stmt.
4815 (vect_get_constant_vectors): Update call accordingly.
4816 (vect_schedule_slp_instance): Likewise
4817 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
4818 (vect_slp_analyze_instance_dependence): Likewise.
4819 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
4821 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4823 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
4824 (stmt_info_for_cost::stmt_info): ...this new field.
4825 (add_stmt_costs): Update accordingly.
4826 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
4827 (vect_get_known_peeling_cost): Likewise.
4828 (vect_estimate_min_profitable_iters): Likewise.
4829 * tree-vect-stmts.c (record_stmt_cost): Likewise.
4831 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4833 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
4834 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4835 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4837 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
4839 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4841 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
4842 a gimple stmt to a stmt_vec_info.
4843 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
4845 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4847 * tree-vectorizer.h (vec_info::grouped_stores): Change from
4848 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4849 (_loop_vec_info::reduction_chains): Likewise.
4850 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
4852 * tree-vect-slp.c (vect_analyze_slp): Likewise.
4854 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4856 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
4857 a gimple stmt to a stmt_vec_info.
4858 (_stmt_vec_info::next_element): Likewise.
4859 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
4860 (vect_slp_analyze_and_verify_node_alignment)
4861 (vect_analyze_group_access_1, vect_analyze_group_access)
4862 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
4863 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
4864 (vect_supportable_dr_alignment): Update accordingly.
4865 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
4866 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
4867 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
4868 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4869 * tree-vect-slp.c (vect_build_slp_tree_1)
4870 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
4871 (vect_split_slp_store_group, vect_analyze_slp_instance)
4872 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
4873 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
4874 (get_group_load_store_type, get_load_store_type)
4875 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
4876 (vect_transform_stmt, vect_remove_stores): Likewise.
4878 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4880 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
4882 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4883 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
4884 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
4885 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
4886 (vector_alignment_reachable_p, vect_get_data_access_cost)
4887 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
4888 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
4889 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4890 (vect_analyze_group_access_1, vect_analyze_group_access)
4891 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
4892 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
4893 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
4894 result of vect_dr_stmt and use the stmt_vec_info instead of
4895 the associated gimple stmt.
4896 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4897 (vect_gen_prolog_loop_niters): Likewise.
4898 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4900 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4902 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
4903 to a vec<stmt_vec_info>.
4904 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
4905 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
4907 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
4908 to a vec<stmt_vec_info>.
4909 (bst_traits::value_type, bst_traits::value_type): Likewise.
4910 (bst_traits::hash): Update accordingly.
4911 (vect_get_and_check_slp_defs): Change the stmts parameter from
4912 a vec<gimple *> to a vec<stmt_vec_info>.
4913 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
4914 (vect_build_slp_tree): Likewise.
4915 (vect_build_slp_tree_2): Likewise. Update uses of
4916 SLP_TREE_SCALAR_STMTS.
4917 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
4918 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
4919 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
4920 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
4921 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
4922 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
4923 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
4924 (vect_get_constant_vectors, vect_get_slp_defs)
4925 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
4926 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
4927 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
4928 instead of gimple stmts.
4929 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
4930 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
4931 (vect_slp_analyze_instance_dependence): Update uses of
4932 SLP_TREE_SCALAR_STMTS.
4933 (vect_slp_analyze_and_verify_node_alignment): Likewise.
4934 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
4935 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
4936 (get_initial_defs_for_reduction): Likewise.
4937 (vect_create_epilog_for_reduction): Likewise.
4938 (vectorize_fold_left_reduction): Likewise.
4939 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
4940 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
4941 (can_vectorize_live_stmts): Likewise.
4943 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4945 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
4946 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4947 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
4949 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
4951 (gather_scalar_reductions): Update after above interface changes.
4952 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
4953 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
4955 (vect_force_simple_reduction): Likewise.
4956 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
4957 LOOP_VINFO_REDUCTIONS.
4958 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4960 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4962 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
4963 a gimple stmt to a stmt_vec_info.
4964 * tree-vect-loop.c (vect_active_double_reduction_p)
4965 (vect_force_simple_reduction, vectorizable_reduction): Update
4968 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4970 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
4971 vec<gimple *> to a vec<stmt_vec_info>.
4972 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
4973 the reduction_phis argument from a vec<gimple *> to a
4975 (vectorizable_reduction): Likewise the phis local variable that
4976 is passed to vect_create_epilog_for_reduction. Update for new type
4977 of SLP_TREE_VEC_STMTS.
4978 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
4979 (vectorizable_live_operation): Likewise.
4980 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
4981 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
4983 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4985 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
4986 a gimple stmt to a stmt_vec_info.
4987 (vectorizable_condition, vectorizable_live_operation)
4988 (vectorizable_reduction, vectorizable_induction): Pass back the
4989 vectorized statement as a stmt_vec_info.
4990 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
4991 use of STMT_VINFO_VEC_STMT.
4992 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
4993 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
4994 as stmt_vec_infos rather than gimple stmts.
4995 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
4997 (vectorizable_live_operation): Likewise.
4998 (vectorizable_reduction, vectorizable_induction): Likewise,
4999 updating use of STMT_VINFO_VEC_STMT.
5000 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
5001 of STMT_VINFO_VEC_STMT.
5002 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
5003 (vectorizable_simd_clone_call, vectorizable_conversion)
5004 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5005 (vectorizable_store, vectorizable_load, vectorizable_condition)
5006 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
5007 from a gimple stmt to a stmt_vec_info.
5008 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
5009 pointer to a stmt_vec_info to the vectorizable_* routines.
5011 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5013 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
5014 a gimple stmt to a stmt_vec_info.
5015 (is_pattern_stmt_p): Update accordingly.
5016 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
5017 (vect_record_grouped_load_vectors): Likewise.
5018 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
5019 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
5020 (vect_model_reduction_cost): Likewise.
5021 (vect_create_epilog_for_reduction): Likewise.
5022 (vectorizable_reduction, vectorizable_induction): Likewise.
5023 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5024 Return the stmt_vec_info for the pattern statement.
5025 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
5026 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
5027 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
5028 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
5029 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
5030 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
5031 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
5032 (free_stmt_vec_info, vect_is_simple_use): Likewise.
5034 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5036 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
5037 (vect_finish_stmt_generation): Likewise.
5038 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
5039 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
5040 (vect_build_gather_load_calls): Use the return value of the above
5041 functions instead of a separate call to vinfo_for_stmt. Use narrow
5042 scopes for the input gimple stmt and wider scopes for the associated
5043 stmt_vec_info. Use vec_info::lookup_def when setting these
5044 stmt_vec_infos from an SSA_NAME definition.
5045 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
5046 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
5047 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5048 (vectorizable_store, vectorizable_load, vectorizable_condition)
5049 (vectorizable_comparison): Likewise.
5050 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
5051 (vectorizable_reduction): Likewise.
5053 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5055 * tree-vectorizer.h (vect_is_simple_use): Add an optional
5056 stmt_vec_info * parameter before the optional gimple **.
5057 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5058 (process_use, vect_get_vec_def_for_operand_1): Update callers.
5059 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
5060 * tree-vect-loop.c (vectorizable_reduction): Likewise.
5061 (vectorizable_live_operation): Likewise.
5062 * tree-vect-patterns.c (type_conversion_p): Likewise.
5063 (vect_look_through_possible_promotion): Likewise.
5064 (vect_recog_rotate_pattern): Likewise.
5065 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
5067 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5069 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
5070 a typedef to a wrapper class.
5071 (NULL_STMT_VEC_INFO): New macro.
5072 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
5073 (stmt_vec_info::operator*): New function.
5074 (stmt_vec_info::operator gimple *): Likewise.
5075 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
5076 (add_stmt_costs): Likewise.
5077 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
5078 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
5079 (vect_get_known_peeling_cost): Likewise.
5080 (vect_estimate_min_profitable_iters): Likewise.
5081 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5082 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
5083 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
5084 (vectorizable_store, free_stmt_vec_infos): Likewise.
5085 (new_stmt_vec_info): Change return type of xcalloc to
5088 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5090 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
5091 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
5092 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
5093 a single_imm_use-based sequence.
5094 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
5096 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5098 * tree-vectorizer.h (vec_info::lookup_def): Declare.
5099 * tree-vectorizer.c (vec_info::lookup_def): New function.
5100 * tree-vect-patterns.c (vect_get_internal_def): Use it.
5101 (vect_widened_op_tree): Likewise.
5102 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5103 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
5104 (vectorizable_reduction): Likewise.
5105 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
5107 (vect_is_slp_reduction): Update calls accordingly. Use
5108 vec_info::lookup_def.
5109 (vect_is_simple_reduction): Likewise
5110 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
5112 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5114 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
5115 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
5116 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
5118 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
5119 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
5120 (vect_update_vf_for_slp, vect_analyze_loop_operations)
5121 (vect_is_slp_reduction, vectorizable_induction)
5122 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
5123 * tree-vect-patterns.c (vect_init_pattern_stmt):
5124 (vect_determine_min_output_precision_1, vect_determine_precisions)
5125 (vect_pattern_recog): Likewise.
5126 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
5127 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
5128 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
5129 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
5130 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
5131 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
5132 info field from a loop to a loop_vec_info.
5134 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5136 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
5137 (vec_info::add_stmt): Declare.
5138 * tree-vectorizer.c (vec_info::add_stmt): New function.
5139 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
5140 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
5141 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
5142 (vectorizable_induction): Likewise.
5143 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
5144 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
5145 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
5146 (vectorizable_load): Likewise.
5147 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5148 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
5149 (vect_recog_gather_scatter_pattern): Likewise.
5150 (append_pattern_def_seq): Likewise. Remove a check that is
5151 performed by add_stmt itself.
5153 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5155 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
5156 which make_ssa_name was called with new_stmt before new_stmt
5159 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5161 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
5163 (vect_is_slp_reduction): ...here...
5164 (vect_is_simple_reduction): ...and here. Remove repetition of tests
5165 that are already known to be false.
5167 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5169 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
5170 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
5171 STMT_VINFO_NUM_SLP_USES when it's true.
5172 (vect_free_slp_instance): Add a final_p parameter and pass it to
5174 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
5175 (vect_analyze_slp_instance): Likewise.
5176 (vect_slp_analyze_operations): Likewise.
5177 (vect_slp_analyze_bb_1): Likewise.
5178 * tree-vectorizer.c (vec_info): Likewise.
5179 * tree-vect-loop.c (vect_transform_loop): Likewise.
5181 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5183 * tree-vect-loop.c (vectorizable_reduction): Assert that the
5184 function is not called for second and subsequent members of
5187 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5189 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
5190 cases for nested loops from here to ...
5191 (vect_create_epilog_for_reduction): ...here. Only call
5192 vect_is_simple_use for inner-loop reductions.
5194 2018-07-31 Martin Liska <mliska@suse.cz>
5196 PR gcov-profile/85338
5197 PR gcov-profile/85350
5198 PR gcov-profile/85372
5199 * profile.c (struct location_triplet): New.
5200 (struct location_triplet_hash): Likewise.
5201 (output_location): Do not output a BB that
5202 is already recorded for a line.
5203 (branch_prob): Use streamed_locations.
5205 2018-07-31 Martin Liska <mliska@suse.cz>
5207 PR gcov-profile/85370
5208 * coverage.c (coverage_begin_function): Do not mark target
5209 clones as artificial functions.
5211 2018-07-31 Martin Liska <mliska@suse.cz>
5213 PR gcov-profile/83813
5214 PR gcov-profile/84758
5215 PR gcov-profile/85217
5216 PR gcov-profile/85332
5217 * profile.c (branch_prob): Do not record GOTO expressions
5218 for GIMPLE statements which locations are already streamed.
5220 2018-07-31 Olivier Hainque <hainque@adacore.com>
5222 * gcc.c (handle_spec_function): Accept a soft_matched_part
5223 argument, as do_spec_1. Pass it down to ...
5224 (eval_spec_function): Accept a soft_matched_part argument,
5225 and pass it down to ...
5226 (do_spec_2): Accept a soft_matched_part argument, and pass
5227 it down to do_spec_1.
5228 (do_spec_1): Pass soft_matched_part to handle_spec_function.
5229 (handle_braces): Update call to handle_spec_function.
5230 (driver::set_up_specs): Update calls to do_spec_2.
5231 (compare_debug_dump_opt_spec_function): Likewise.
5232 (compare_debug_self_opt_spec_function): Likewise.
5234 2018-07-31 Olivier Hainque <hainque@adacore.com>
5236 * common.opt (nolibc): New option.
5237 * doc/invoke.texi (Link Options): Document it.
5238 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
5239 * config/alpha/linux.h: Likewise.
5240 * config/arc/elf.h: Likewise.
5241 * config/arm/uclinux-elf.h: Likewise.
5242 * config/arm/unknown-elf.h: Likewise.
5243 * config/avr/avrlibc.h: Likewise.
5244 * config/bfin/bfin.h: Likewise.
5245 * config/bfin/linux.h: Likewise.
5246 * config/bfin/uclinux.h: Likewise.
5247 * config/darwin.h: Likewise.
5248 * config/darwin10.h: Likewise.
5249 * config/darwin12.h: Likewise.
5250 * config/gnu-user.h: Likewise.
5251 * config/lm32/uclinux-elf.h: Likewise.
5252 * config/pa/pa-hpux11.h: Likewise.
5253 * config/pa/pa64-hpux.h: Likewise.
5254 * config/sparc/sparc.h: Likewise.
5256 2018-07-31 Olivier Hainque <hainque@adacore.com>
5258 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
5259 undefined variables.
5261 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
5264 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
5267 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
5269 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
5270 terminated string literal.
5272 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
5274 PR rtl-optimization/85160
5275 * combine.c (is_just_move): New function.
5276 (try_combine): Allow combining two instructions into two if neither of
5277 the original instructions was a move.
5279 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
5282 * doc/extend.texi (Global Register Variables): Discourage use of type
5284 (Local Register Variables): Likewise.
5286 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
5288 PR tree-optimization/86506
5289 * hwint.h (ceil_log2): Resync with hwint.c implementation.
5291 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
5294 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
5295 hard_regno, make sure no insn between `from` and `to` clobbers it.
5297 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
5298 Tom de Vries <tdevries@suse.de>
5300 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
5301 (PTX_DEFAULT_RUNTIME_DIM): ... this.
5302 (nvptx_goacc_validate_dims): Set default worker and gang dims to
5303 PTX_DEFAULT_RUNTIME_DIM.
5304 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
5306 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
5308 * config/pa/pa.c (pa_output_addr_vec): Align address table.
5309 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
5310 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
5312 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
5314 * config/rs6000/constraints.md (wG constraint): Delete, no longer
5316 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
5317 predicate to reflect toc fusion has been deleted.
5318 (toc_fusion_mem_raw): Delete, no longer used.
5319 (toc_fusion_mem_wrapped): Likewise.
5320 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
5322 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
5323 Delete, no longer used.
5324 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
5325 meant to be used for toc fusion.
5326 (rs6000_debug_print_mode): Delete toc fusion debugging.
5327 (rs6000_debug_reg_global): Likewise.
5328 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
5329 fusion and secondary reload support that were never used.
5330 (rs6000_option_override_internal): Delete TOC fusion, that was only
5331 partially defined, and it did not work unless you also used the
5333 (rs6000_legitimate_address_p): Delete TOC fusion support.
5334 (rs6000_opt_masks): Likewise.
5335 (fusion_wrap_memory_address): Delete function, no longer used.
5336 (fusion_split_address); Delete TOC fusion support.
5337 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
5338 longer used with toc fusion being deleted.
5339 (TARGET_TOC_FUSION_FP): Likewise.
5340 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
5342 (toc fusion spliter): Delete TOC fusion support.
5343 (toc_fusionload_<mode>): Likewise.
5344 (toc_fusionload_di): Likewise.
5345 (fusion_gpr_load_<mode>): Delete generator function, this insn no
5346 longer needs to be named. Rename predicate to delete TOC fusion.
5347 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5348 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5349 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5350 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5351 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
5353 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
5355 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
5356 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
5357 __int128 in built-in function prototypes.
5358 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
5359 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
5361 2018-07-27 Martin Sebor <msebor@redhat.com>
5363 PR tree-optimization/86696
5364 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
5365 types, including enums.
5366 (handle_char_store): Be prepared for the above function to fail.
5368 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
5370 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
5371 when optimization level is lower than 2 or optimize for size.
5373 2018-07-26 Martin Sebor <msebor@redhat.com>
5375 PR tree-optimization/86043
5376 PR tree-optimization/86042
5377 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
5378 (get_string_cst_length): Rename...
5379 (get_min_string_length): ...to this. Add argument.
5380 (handle_char_store): Extend to handle multi-character stores by
5382 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
5383 * tree.h (initializer_zerop): Add argument.
5385 2018-07-26 Jakub Jelinek <jakub@redhat.com>
5388 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
5389 declare target to variables if they have always,{to,from,tofrom} map
5392 2018-07-26 Martin Liska <mliska@suse.cz>
5395 * lto-wrapper.c: Add linker_output as prefix
5396 for ltrans_output_file.
5398 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
5400 PR rtl-optimization/85805
5401 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
5402 value for hard registers if that was written in the same mode.
5404 2018-07-26 Martin Liska <mliska@suse.cz>
5406 PR gcov-profile/86536
5407 * gcov.c (format_gcov): Use printf format %.*f directly
5408 and do not handle special values.
5410 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5412 * common/config/arc/arc-common.c (arc_option_optimization_table):
5413 Update default optimizations for size.
5415 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5417 * config/arc/arc.md (movsf_insn): Add short instruction selection.
5418 * config/arc/constraints.md (CfZ): New constraint.
5419 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
5420 (subsf3_fpu): Likewise.
5421 (cmpsf_fpu): Likewise.
5422 (cmpsf_fpu_uneq): Likewise.
5424 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5426 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
5428 (arc_is_uncached_mem_p): uncached applies to both the variable and
5431 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5433 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
5436 2018-07-25 David Malcolm <dmalcolm@redhat.com>
5438 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
5439 field "m_scopes" from vec to auto_vec.
5441 2018-07-25 Martin Liska <mliska@suse.cz>
5443 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
5446 2018-07-25 Richard Biener <rguenther@suse.de>
5449 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
5450 special wrt context_die late.
5451 (gen_subprogram_die): Re-use DIEs in local scope.
5453 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
5455 PR tree-optimization/86644
5456 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
5458 2018-07-25 Martin Liska <mliska@suse.cz>
5461 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
5462 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
5464 2018-07-25 Martin Liska <mliska@suse.cz>
5467 * params.def: Explain ASan abbreviation and provide
5468 a documentation link.
5470 2018-07-24 Martin Sebor <msebor@redhat.com>
5472 PR tree-optimization/86622
5473 PR tree-optimization/86532
5474 * builtins.h (string_length): Declare.
5475 * builtins.c (c_strlen): Correct handling of non-constant offsets.
5476 (check_access): Be prepared for non-constant length ranges.
5477 (string_length): Make extern.
5478 * expr.c (string_constant): Only handle the minor non-constant
5479 array index. Use string_constant to compute the length of
5480 a generic string constant.
5482 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
5484 PR tree-optimization/86618
5485 * tree-vect-stmts.c (vectorizable_call): Don't take the address
5486 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
5488 2018-07-24 David Malcolm <dmalcolm@redhat.com>
5490 PR tree-optimization/86636
5491 * json.cc (json::object::set): Fix comment. Add assertions.
5492 (json::array::append): Move here from json.h. Add comment and an
5494 (json::string::string): Likewise.
5495 * json.h (json::array::append): Move to json.cc.
5496 (json::string::string): Likewise.
5497 * optinfo-emit-json.cc
5498 (optrecord_json_writer::impl_location_to_json): Assert that we
5499 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
5500 wrapper around it. Expand the location once, rather than three
5502 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
5503 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
5505 (optrecord_json_writer::optinfo_to_json): Likewise, in four
5506 places. Fix some overlong lines.
5508 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
5510 * config/aarch64/aarch64-simd.md
5511 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
5512 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
5513 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
5514 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
5515 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
5516 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
5517 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
5518 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
5519 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
5521 2018-07-24 Jakub Jelinek <jakub@redhat.com>
5524 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
5525 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
5526 and abs_d == d, do the power of two handling if profitable.
5528 2018-07-24 Richard Biener <rguenther@suse.de>
5530 * match.pd: Add BIT_FIELD_REF canonicalizations.
5532 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5535 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
5537 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5539 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
5540 terminated STRING_CST object.
5542 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5544 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
5546 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
5548 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
5549 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
5550 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
5551 the elements into a register.
5552 (rs6000_split_v4si_init_di_reg): Delete.
5553 (rs6000_split_v4si_init): Delete.
5554 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
5555 (vsx_init_v4si): Rewrite as a define_expand.
5557 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
5559 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
5560 zero_extend argument from memory): New.
5562 2018-07-22 Martin Sebor <msebor@redhat.com>
5565 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
5566 diagnosing calls with unknown arguments unless -Walloca-larger-than
5567 is restricted to less than PTRDIFF_MAX bytes.
5569 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
5571 * doc/gcov.texi (Invoking Gcov): Editorial changes.
5573 2018-07-20 David Malcolm <dmalcolm@redhat.com>
5575 * pretty-print.c (text_info::set_location): Remove redundant
5576 "line_table" parameter from call to rich_location::set_range.
5578 2018-07-20 Martin Sebor <msebor@redhat.com>
5581 * builtins.c (expand_builtin_alloca): Adjust.
5582 * calls.c (alloc_max_size): Simplify.
5583 * cgraphunit.c (cgraph_node::expand): Adjust.
5584 * common.opt (larger_than_size, warn_frame_larger_than): Remove
5586 (frame_larger_than_size): Same.
5587 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
5588 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
5590 * doc/invoke.texi (GCC Command Options): Document option arguments.
5591 Explain byte-size arguments and suffixes.
5592 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
5593 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
5594 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
5595 * doc/options.texi (UInteger): Expand.
5596 (Host_Wide_Int, ByteSize): Document new properties.
5597 * final.c (final_start_function_1): Include sizes in an error message.
5598 * function.c (frame_offset_overflow): Same.
5599 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
5600 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
5601 Diagnose unbounded alloca calls only for limits of less than
5603 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
5604 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
5606 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
5607 only for limits of less than PTRDIFF_MAX.
5608 * langhooks-def.h (lhd_handle_option): Change function argument
5610 * langhooks.c (lhd_handle_option): Same.
5611 * langhooks.h (handle_option): Same.
5612 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
5614 (var_type, var_type_struct): Same.
5615 (var_set): Handle ByteSize flag.
5616 * optc-gen.awk: Add comments to output to ease debugging. Make
5617 use of HOST_WIDE_INT where appropriate.
5618 * opts-gen-save.awk: Use %lx to format unsigned long.
5619 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
5620 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
5621 arguments. Parse bytes-size suffixes.
5622 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
5623 (enum_value_to_arg): Same.
5624 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
5625 (handle_option): Adjust.
5626 (generate_option): Change function argument to HOST_WIDE_INT.
5627 (cmdline_handle_error): Adjust.
5628 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
5629 (set_option): Change function argument to HOST_WIDE_INT.
5630 (option_enabled): Handle cl_host_wide_int.
5631 (get_option_state): Handle CLVC_SIZE.
5632 (control_warning_option): Same.
5633 * opts.c (common_handle_option): Change function argument to
5634 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
5635 OPT_Wvla_larger_than_.
5636 * opts.h (enum cl_var_type): Add an enumerator.
5637 * stor-layout.c (layout_decl): Print a more meaningful warning.
5638 * toplev.c (output_stack_usage): Adjust.
5640 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
5642 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
5643 call to inline_expand_builtin_string_cmp.
5644 (expand_builtin_strcmp): Likewise.
5645 (expand_builtin_strncmp): Likewise.
5646 (inline_string_cmp): Delete the last parameter, change char_type_node
5647 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
5649 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
5650 the inlining expansion on target where the type of the call has same or
5651 narrower precision than unsigned char.
5653 2018-07-20 David Malcolm <dmalcolm@redhat.com>
5655 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
5656 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
5657 * common.opt (fsave-optimization-record): New option.
5658 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
5659 * doc/invoke.texi (-fsave-optimization-record): New option.
5660 * dumpfile.c: Include "optinfo-emit-json.h".
5661 (struct kv_pair): Move to coretypes.h.
5662 (optgroup_options): Make non-static.
5663 (dump_context::end_scope): Call
5664 optimization_records_maybe_pop_dump_scope.
5665 * dumpfile.h (optgroup_options): New decl.
5666 * json.cc: New file.
5668 * optinfo-emit-json.cc: New file.
5669 * optinfo-emit-json.h: New file.
5670 * optinfo.cc: Include "optinfo-emit-json.h".
5671 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
5672 (optinfo_enabled_p): Check optimization_records_enabled_p.
5673 (optinfo_wants_inlining_info_p): Likewise.
5674 * optinfo.h: Update comment.
5675 * profile-count.c (profile_quality_as_string): New function.
5676 * profile-count.h (profile_quality_as_string): New decl.
5677 (profile_count::quality): New accessor.
5678 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
5679 and optinfo_emit_json_cc_tests.
5680 * selftest.h (selftest::json_cc_tests): New decl.
5681 (selftest::optinfo_emit_json_cc_tests): New decl.
5682 * toplev.c: Include "optinfo-emit-json.h".
5683 (compile_file): Call optimization_records_finish.
5684 (do_compile): Call optimization_records_start.
5685 * tree-ssa-live.c: Include optinfo.h.
5686 (remove_unused_scope_block_p): Retain inlining information if
5687 optinfo_wants_inlining_info_p returns true.
5689 2018-07-20 Richard Biener <rguenther@suse.de>
5692 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
5693 to cover -flto-partition=none.
5695 2018-07-20 Martin Liska <mliska@suse.cz>
5697 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
5698 (get_decl_source_range): Remove unused function.
5700 2018-07-20 Richard Biener <rguenther@suse.de>
5702 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
5703 (struct vn_phi_s): Likewise.
5704 (struct vn_reference_s): Likewise.
5705 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
5706 for searching the slot of an entry known to be in the hash itself.
5707 (vn_phi_hasher::equal): Likewise.
5708 (vn_reference_hasher::equal): Likewise.
5709 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
5711 (optimistic_info, current_info): Remove, keeping only valid_info.
5712 (vn_reference_lookup_1): Remove fallback lookup.
5713 (vn_reference_lookup_2): Likewise.
5714 (vn_nary_op_lookup_1): Likewise.
5715 (vn_phi_lookup): Likewise.
5716 (vn_nary_build_or_lookup_1): Make sure to not chain the built
5718 (vn_reference_insert): Adjust, chain the inserted hash element
5719 at last_inserted_ref.
5720 (vn_reference_insert_pieces): Likewise.
5721 (visit_reference_op_call): Likewise.
5722 (vn_nary_op_insert_into): Chain the inserted hash element at
5724 (vn_nary_op_insert_pieces): Adjust.
5725 (vn_nary_op_insert): Likewise.
5726 (vn_nary_op_insert_stmt): Likewise.
5727 (vn_phi_insert): Adjust, chain the inserted hash element at
5729 (process_scc): Remove clearing and copying the optimistic
5730 table. Instead remove elements inserted during an optimistic
5731 iteration from the single table we maintain.
5732 (init_scc_vn): Adjust.
5733 (free_scc_vn): Likewise.
5734 (sccvn_dom_walker::record_cond): Likewise.
5735 (sccvn_dom_walker::after_dom_children): Likewise.
5737 2018-07-19 Martin Sebor <msebor@redhat.com>
5739 PR tree-optimization/84047
5740 PR tree-optimization/83776
5741 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
5742 (check_array_bounds): Call it.
5744 2018-07-19 Martin Sebor <msebor@redhat.com>
5746 * align.h (align_flags): Use member initialization.
5748 2018-07-19 David Malcolm <dmalcolm@redhat.com>
5750 * Makefile.in (OBJS): Add optinfo.o.
5751 * coretypes.h (class symtab_node): New forward decl.
5752 (struct cgraph_node): New forward decl.
5753 (class varpool_node): New forward decl.
5754 * dump-context.h: New file.
5755 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
5757 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
5758 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
5759 (set_alt_dump_file): Likewise.
5760 (dump_context::~dump_context): New dtor.
5761 (dump_gimple_stmt): Move implementation to...
5762 (dump_context::dump_gimple_stmt): ...this new member function.
5763 Add the stmt to any pending optinfo, creating one if need be.
5764 (dump_gimple_stmt_loc): Move implementation to...
5765 (dump_context::dump_gimple_stmt_loc): ...this new member function.
5766 Start a new optinfo and add the stmt to it.
5767 (dump_gimple_expr): Move implementation to...
5768 (dump_context::dump_gimple_expr): ...this new member function.
5769 Add the stmt to any pending optinfo, creating one if need be.
5770 (dump_gimple_expr_loc): Move implementation to...
5771 (dump_context::dump_gimple_expr_loc): ...this new member function.
5772 Start a new optinfo and add the stmt to it.
5773 (dump_generic_expr): Move implementation to...
5774 (dump_context::dump_generic_expr): ...this new member function.
5775 Add the tree to any pending optinfo, creating one if need be.
5776 (dump_generic_expr_loc): Move implementation to...
5777 (dump_context::dump_generic_expr_loc): ...this new member
5778 function. Add the tree to any pending optinfo, creating one if
5780 (dump_printf): Move implementation to...
5781 (dump_context::dump_printf_va): ...this new member function. Add
5782 the text to any pending optinfo, creating one if need be.
5783 (dump_printf_loc): Move implementation to...
5784 (dump_context::dump_printf_loc_va): ...this new member function.
5785 Start a new optinfo and add the stmt to it.
5786 (dump_dec): Move implementation to...
5787 (dump_context::dump_dec): ...this new member function. Add the
5788 value to any pending optinfo, creating one if need be.
5789 (dump_context::dump_symtab_node): New member function.
5790 (dump_context::get_scope_depth): New member function.
5791 (dump_context::begin_scope): New member function.
5792 (dump_context::end_scope): New member function.
5793 (dump_context::ensure_pending_optinfo): New member function.
5794 (dump_context::begin_next_optinfo): New member function.
5795 (dump_context::end_any_optinfo): New member function.
5796 (dump_context::s_current): New global.
5797 (dump_context::s_default): New global.
5798 (dump_scope_depth): Delete global.
5799 (dumpfile_ensure_any_optinfo_are_flushed): New function.
5800 (dump_symtab_node): New function.
5801 (get_dump_scope_depth): Reimplement in terms of dump_context.
5802 (dump_begin_scope): Likewise.
5803 (dump_end_scope): Likewise.
5804 (selftest::temp_dump_context::temp_dump_context): New ctor.
5805 (selftest::temp_dump_context::~temp_dump_context): New dtor.
5806 (selftest::verify_item): New function.
5807 (ASSERT_IS_TEXT): New macro.
5808 (ASSERT_IS_TREE): New macro.
5809 (ASSERT_IS_GIMPLE): New macro.
5810 (selftest::test_capture_of_dump_calls): New test.
5811 (selftest::dumpfile_c_tests): Call it.
5812 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
5813 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
5814 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
5815 descriptive comment.
5816 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
5817 (dump_node, dump_bb): Move these unrelated decls.
5818 (class dump_manager): Add leading comment.
5819 * optinfo.cc: New file.
5820 * optinfo.h: New file.
5822 2018-07-19 Michael Collison <michael.collison@arm.com>
5823 Richard Henderson <rth@redhat.com>
5825 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
5826 (subti): Handle op1 zero.
5827 (subvti4, usub4ti4): New.
5828 (*sub<GPI>3_compare1_imm): New.
5829 (sub<GPI>3_carryinCV): New.
5830 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
5831 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
5833 2018-07-19 Michael Collison <michael.collison@arm.com>
5834 Richard Henderson <rth@redhat.com>
5836 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
5837 (addti3): Create simpler code if low part is already known to be 0.
5838 (addvti4, uaddvti4): New.
5839 (*add<GPI>3_compareC_cconly_imm): New.
5840 (*add<GPI>3_compareC_cconly): New.
5841 (*add<GPI>3_compareC_imm): New.
5842 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
5843 handle constants within this pattern..
5844 (*add<GPI>3_compareV_cconly_imm): New.
5845 (*add<GPI>3_compareV_cconly): New.
5846 (*add<GPI>3_compareV_imm): New.
5847 (add<GPI>3_compareV): New.
5848 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
5849 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
5850 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
5851 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
5852 with 'comparison' operator.
5853 (*add<GPI>3_compareV_cconly_imm): Ditto.
5854 (*add<GPI>3_compareV_cconly): Ditto.
5855 (*add<GPI>3_compareV_imm): Ditto.
5856 (add<GPI>3_compareV): Ditto.
5857 (add<mode>3_carryinC): Ditto.
5858 (*add<mode>3_carryinC_zero): Ditto.
5859 (*add<mode>3_carryinC): Ditto.
5860 (add<mode>3_carryinV): Ditto.
5861 (*add<mode>3_carryinV_zero): Ditto.
5862 (*add<mode>3_carryinV): Ditto.
5864 2018-07-19 Michael Collison <michael.collison@arm.com>
5865 Richard Henderson <rth@redhat.com>
5867 * config/aarch64/aarch64-modes.def (CC_V): New.
5868 * config/aarch64/aarch64-protos.h
5869 (aarch64_addti_scratch_regs): Declare
5870 (aarch64_subvti_scratch_regs): Declare.
5871 (aarch64_expand_subvti): Declare.
5872 (aarch64_gen_unlikely_cbranch): Declare
5873 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
5874 for signed overflow using CC_Vmode.
5875 (aarch64_get_condition_code_1): Handle CC_Vmode.
5876 (aarch64_gen_unlikely_cbranch): New function.
5877 (aarch64_addti_scratch_regs): New function.
5878 (aarch64_subvti_scratch_regs): New function.
5879 (aarch64_expand_subvti): New function.
5881 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5883 * config/aarch64/aarch64-option-extensions.def: New entry for profile
5885 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
5886 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
5889 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5892 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5893 address check not strict.
5895 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5897 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
5899 (store_pair_lanes<mode>): Likewise.
5900 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
5901 enum value 'ADDR_QUERY_LDP_STP_N'.
5902 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
5903 (aarch64_print_address_internal): Add declaration.
5904 (aarch64_print_ldpstp_address): Remove.
5905 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
5906 (aarch64_print_operand): Change printing of 'y'.
5907 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
5908 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
5909 'true' rather than '1'.
5910 * config/aarch64/constraints.md (Uml): Likewise.
5911 (Uml): Rename to Umn.
5914 2018-07-19 Richard Biener <rguenther@suse.de>
5916 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
5918 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
5919 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
5920 (vn_reference_hasher): Likewise.
5921 (struct vn_tables_s): Remove obstack and alloc-pool members.
5922 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
5923 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
5924 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
5925 (vn_reference_insert_pieces): Likewise.
5926 (alloc_vn_nary_op_noinit): Adjust.
5927 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
5928 (vn_phi_eq): Adjust.
5929 (shared_lookup_phiargs): Remove.
5930 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
5931 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
5932 (visit_reference_op_call): Likewise.
5933 (copy_nary, copy_phi, copy_reference): Remove.
5934 (process_scc): Rewind the obstack when iterating. Do not
5935 copy the elements to valid_info but just move them from one
5936 hashtable to the other.
5937 (allocate_vn_table): Adjust.
5938 (free_vn_table): Likewise.
5939 (init_scc_vn): Likewise.
5940 (free_scc_vn): Likewise.
5942 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
5945 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
5946 indirect_return as function type attribute.
5947 (ix86_attribute_table): Change indirect_return to function
5949 * doc/extend.texi: Update indirect_return attribute.
5951 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
5953 * wide-int.h (widest2_int): New.
5954 * gimple-fold.c (arith_overflowed_p): Use it.
5955 * tree.h (widest2_int_cst): New.
5956 * tree-vrp.c (wide_int_binop_overflow): Rename from
5957 vrp_int_const_binop.
5958 Rewrite to work on trees.
5959 (extract_range_from_multiplicative_op_1): Abstract code to...
5960 (wide_int_range_min_max): ...here.
5961 (wide_int_range_cross_product): ...and here.
5962 (extract_range_from_binary_expr_1): Abstract overflow code to...
5963 (wide_int_range_mult_wrapping): ...here.
5964 * tree-vrp.h (wide_int_range_cross_product): New.
5965 (wide_int_range_mult_wrapping): New.
5967 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
5968 Julia Koval <julia.koval@intel.com>
5970 * config/i386/x86-tune-costs.h (skylake_memcpy,
5971 skylake_memset): Replace rep_prefix with unrolling for size 512.
5973 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
5976 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
5977 comparision with EQ_EXPR in last stmt.
5979 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
5981 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
5982 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
5983 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
5984 previously known as "PowerPC AltiVec Built-in Functions". Move
5985 some material to new subsubsections "PowerPC AltiVec Built-in
5986 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
5988 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
5989 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
5990 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
5991 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
5993 2018-07-18 Richard Biener <rguenther@suse.de>
5995 PR tree-optimization/86557
5996 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
5999 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
6001 * config/s390/s390.c (s390_function_profiler): Generate CFI.
6003 2018-07-17 Jeff Law <law@redhat.com>
6005 * config/arm/arm.c (get_label_padding): Update for recent
6006 changes to label_to_alignment.
6008 PR tree-optimization/86010
6009 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
6011 * config/mips/mips.c (vr4130_align_insns): Update for recent
6012 changes to label_to_alignment.
6014 * config/frv/frv.c (frv_label_align): Update for recent changes
6015 to label_to_alignment.
6017 * config/nios2/nios2.c (nios2_label_align): Update for recent
6018 changes which dropped ALIGN_LABELS_LOG.
6020 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
6022 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
6023 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
6025 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
6027 * config/arc/arc.c (arc_label_align): Use align_labels instead of
6028 deprecated align_labels_log.
6030 2018-07-17 Richard Biener <rguenther@suse.de>
6033 * dwarf2out.c (init_sections_and_labels): Always generate
6034 a debug_line_str_section for early LTO debug.
6035 (dwarf2out_finish): Reset debug_line_str_hash output early.
6036 Bump counter for extra dwarf5 .debug_loc labels to not conflict
6038 (dwarf2out_early_finish): Output debug_line_str.
6040 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
6042 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
6043 index register on z196 or later.
6045 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
6047 * config/s390/s390.c (s390_default_align): Set default function
6049 (s390_override_options_after_change): Call s390_default align.
6050 (s390_option_override_internal): Call s390_default_align.
6051 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
6053 2018-07-17 Jakub Jelinek <jakub@redhat.com>
6056 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
6057 to _looptemp_ clauses, other than the first two.
6059 2018-07-17 Martin Liska <mliska@suse.cz>
6061 * opts.c: Do not enable OPT_falign_* for -Os.
6063 2018-07-17 Martin Liska <mliska@suse.cz>
6065 * align.h (MAX_CODE_ALIGN): New.
6066 (MAX_CODE_ALIGN_VALUE): New.
6067 * common/config/i386/i386-common.c (ix86_handle_option):
6068 (MAX_CODE_ALIGN): Moved to align.h.
6069 * final.c (MAX_CODE_ALIGN): Likewise.
6070 * opts.c (parse_and_check_align_values):
6071 (MAX_CODE_ALIGN): Likewise.
6072 (MAX_CODE_ALIGN_VALUE): Likewise.
6074 2018-07-17 Martin Liska <mliska@suse.cz>
6076 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
6077 in order to fulfil coding style.
6078 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
6079 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6080 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6081 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
6082 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
6083 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
6084 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
6085 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
6086 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6088 2018-07-17 Martin Liska <mliska@suse.cz>
6090 * align.h: New file.
6091 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
6093 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
6094 align_flags of label_to_alignment.
6095 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
6097 * config/m68k/m68k.c: Do not use removed align_labels_value and
6099 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
6100 (LOOP_ALIGN): Likewise.
6101 (LABEL_ALIGN): Likewise.
6102 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
6103 Remove not used macro.
6104 (rs6000_loop_align): Change return type to align_flags.
6105 (rs6000_loop_align_max_skip): Remove.
6106 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
6107 Change return type to align_flags.
6108 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
6109 Remove not used macro.
6110 (rs6000_loop_align): Change return type to align_flags.
6111 (rs6000_loop_align_max_skip): Remove.
6112 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
6113 * config/rx/rx-protos.h (rx_align_for_label): Make it
6115 * config/rx/rx.c (rx_align_for_label): Change return type
6117 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
6119 into align_flags class.
6120 (LABEL_ALIGN): Likewise.
6121 (LOOP_ALIGN): Likewise.
6122 * config/s390/s390.c (s390_label_align): Use align_flags
6124 (s390_asm_output_function_label): Likewise.
6125 * config/sh/sh.c (sh_override_options_after_change):
6126 Use align_flags class directly without macros.
6127 (find_barrier): Likewise.
6128 (barrier_align): Likewise.
6129 (sh_loop_align): Likewise.
6130 * config/spu/spu.c (spu_option_override):
6131 Use align_flags_tuple::get_value instead of removed macros.
6132 (spu_sched_init): Likewise.
6133 * config/spu/spu.h (GTY): Likewise.
6134 * config/visium/visium.c (visium_option_override):
6135 Set "8" as default secondary alignment.
6136 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
6137 in order to guarantee secondary alignment of 8.
6138 * coretypes.h: Include align.h header file.
6139 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
6140 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
6141 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
6142 * doc/tm.texi.in: Likewise.
6143 * final.c (struct label_alignment): Remove not used structure.
6144 (LABEL_ALIGN): Change type to align_flags.
6145 (LOOP_ALIGN): Likewise.
6146 (JUMP_ALIGN): Likewise.
6147 (default_loop_align_max_skip): Remove.
6148 (default_label_align_max_skip): Likewise.
6149 (default_jump_align_max_skip): Likewise.
6150 (default_label_align_after_barrier_max_skip):
6151 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
6152 (LABEL_TO_MAX_SKIP): Remove.
6153 (label_to_alignment): Return align_flags type instead of integer.
6154 (label_to_max_skip): Remove.
6155 (align_fuzz): Use align_flags type.
6156 (compute_alignments): Use align_flags type and use align_flags::max
6157 to combine multiple alignments.
6158 (grow_label_align): Grow vec instead of C array.
6159 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
6160 (shorten_branches): Use align_flags type and use align_flags::max
6161 to combine multiple alignments.
6162 (final_scan_insn_1): Remove usage of secondary alignment that comes
6163 from label alignment, but instead use proper secondary alignment
6164 which is computed in grow_label_align.
6165 * flags.h (struct align_flags_tuple): Move to align.h.
6166 (struct align_flags): Likewise.
6167 (state_align_loops): Rename to align_loops.
6168 (state_align_jumps): Rename to align_jumps.
6169 (state_align_labels): Rename to align_labels.
6170 (state_align_functions): Rename to align_functions.
6171 (align_loops_log): Remove.
6172 (align_jumps_log): Remove.
6173 (align_labels_log): Remove.
6174 (align_functions_log): Remove.
6175 (align_loops_max_skip): Remove.
6176 (align_jumps_max_skip): Remove.
6177 (align_labels_max_skip): Remove.
6178 (align_functions_max_skip): Remove.
6179 (align_loops_value): Remove.
6180 (align_jumps_value): Remove.
6181 (align_labels_value): Remove.
6182 (align_functions_value): Remove.
6183 * output.h (label_to_alignment): Change return type to align_flags.
6184 (label_to_max_skip): Remove.
6185 * target.def: Remove loop_align_max_skip, label_align_max_skip,
6186 jump_align_max_skip macros.
6187 * targhooks.h (default_loop_align_max_skip): Remove.
6188 (default_label_align_max_skip): Likewise.
6189 (default_jump_align_max_skip): Likewise.
6190 (default_label_align_after_barrier_max_skip): Remove.
6191 * toplev.c (read_log_maxskip): Use ::normalize function.
6192 (parse_N_M): Remove not used argument and also call ::normalize.
6193 (parse_alignment_opts): Do not pass unused arguments.
6194 * varasm.c (assemble_start_function): Use directly align_functions
6195 instead of removed macros.
6196 * system.h: Do not poison removed macros.
6198 2018-07-17 Jakub Jelinek <jakub@redhat.com>
6201 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
6202 and cond temporaries don't have reference type if iterator has
6203 pointer type. For init use &for_pre_body instead of pre_p if
6204 for_pre_body is non-empty.
6206 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
6208 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
6209 double-double modes to SFmode directly directly.
6210 (trunc<mode>sf2_fprs): Delete.
6212 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
6214 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
6215 for conversions between IFmode and the decimal floating point modes.
6216 (init_float128_ieee): Use the correct names for conversions between
6217 KFmode and the decimal floating point modes.
6219 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
6221 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
6222 for the conversions between TDmode and IFmode.
6223 (init_float128_ieee): Use more correct names for the conversions
6224 between TDmode and KFmode.
6226 2018-07-16 Jakub Jelinek <jakub@redhat.com>
6228 PR tree-optimization/86526
6229 * builtins.c (expand_builtin_memcmp): Formatting fixes.
6230 (inline_expand_builtin_string_cmp): Likewise.
6231 (inline_string_cmp): Likewise. Use c_readstr instead of
6232 builtin_memcpy_read_str. Add unit_mode temporary.
6234 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
6237 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
6238 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
6240 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6242 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
6243 Alphabetize prototypes of built-in functions, separating out
6244 built-in functions that are listed in this section but should be
6245 described elsewhere.
6247 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
6250 * expmed.c (emit_store_flag): Do not emit setcc followed by a
6251 conditional move when trapping comparison was split to a
6252 non-trapping one (and vice versa).
6254 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
6256 * config/s390/s390.c (s390_function_profiler): Generate nops
6257 instead of profiler call sequences.
6258 * config/s390/s390.opt: Add the new option.
6260 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
6262 * config/s390/s390.c (s390_function_profiler): Generate
6263 __mcount_loc section.
6264 * config/s390/s390.opt: Add the new option.
6266 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
6268 * common.opt: Add the new warning.
6269 * config/s390/s390.c (s390_function_profiler): Emit "brasl
6270 %r0,__fentry__" when -mfentry is specified.
6271 (s390_option_override_internal): Disallow -mfentry for 31-bit
6273 * config/s390/s390.opt: Add the new option.
6275 2018-07-16 Richard Biener <rguenther@suse.de>
6278 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
6279 for function-local FUNCTION_DECL and RESULT_DECL immediately.
6281 2018-07-16 Martin Liska <mliska@suse.cz>
6284 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
6287 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
6289 * config/arc/arcHS.md: Update ARCHS scheduling rules.
6291 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
6293 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
6295 * config/arc/arc-cpus.def (hs4x): New cpu.
6297 * config/arc/arc-tables.opt: Regenerate.
6298 * config/arc/arc.c (arc_sched_issue_rate): New function.
6299 (TARGET_SCHED_ISSUE_RATE): Define.
6300 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
6301 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
6303 (attr tune): Add ARCHS4x tune values.
6304 (attr tune_dspmpy): Define.
6305 (*tst): Correct instruction type.
6306 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
6307 * config/arc/arcHS4x.md: New file.
6308 * config/arc/fpu.md: Update instruction type attributes.
6309 * config/arc/t-multilib: Regenerate.
6311 2018-07-16 Tom de Vries <tdevries@suse.de>
6314 * var-tracking.c (vt_initialize): Fix pre_dec handling.
6316 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6318 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
6321 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
6323 PR tree-optimization/86514
6324 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
6325 conversion to a boolean type from a type with greater precision.
6327 2018-07-16 Tom de Vries <tdevries@suse.de>
6329 * var-tracking.c (vt_initialize): Print adjusted insn slim if
6330 dump_flags request TDF_SLIM.
6332 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
6334 * fold-const.c (int_const_binop_1): Abstract...
6335 (wide_int_binop): ...wide int code here.
6336 (poly_int_binop): ...poly int code here.
6337 Abstract the rest of int_const_binop_1 into int_const_binop.
6338 * fold-const.h (wide_int_binop): New.
6339 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
6340 Remove useless PLUS/MINUS_EXPR case.
6341 (zero_nonzero_bits_from_vr): Move wide int code...
6342 (zero_nonzero_bits_from_bounds): ...here.
6343 (extract_range_from_binary_expr_1): Move mask optimization code...
6344 (range_easy_mask_min_max): ...here.
6345 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
6346 (range_easy_mask_min_max): New.
6348 2018-07-15 Jeff Law <law@redhat.com>
6351 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
6354 2018-07-14 Jim Wilson <jimw@sifive.com>
6356 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
6358 2018-07-14 Paul Koning <ni1d@arrl.net>
6360 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
6362 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
6364 * lto-streamer-out.c (copy_function_or_variable): Dump info about
6367 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
6368 Steve Munroe <munroesj52@gmail.com>
6370 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
6371 (_mm_andnot_si128): Likewise.
6372 (_mm_or_si128): Likewise.
6373 (_mm_xor_si128): Likewise.
6375 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
6378 * builtins.c (expand_builtin_memcmp): Inline the calls first
6379 when result_eq is false.
6380 (expand_builtin_strcmp): Inline the calls first.
6381 (expand_builtin_strncmp): Likewise.
6382 (inline_string_cmp): New routine. Expand a string compare
6383 call by using a sequence of char comparison.
6384 (inline_expand_builtin_string_cmp): New routine. Inline expansion
6385 a call to str(n)cmp/memcmp.
6386 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
6388 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
6390 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
6392 * config/arm/driver-arm.c: Include arm-native.h.
6393 (host_detect_local_cpu): Use auto-generated data tables.
6394 (vendors, arm_cpu_table): Delete. Move part information to ...
6395 * config/arm/arm-cpus.in: ... here.
6396 * config/arm/parsecpu.awk (gen_native): New function.
6397 (vendor, part): New CPU fields.
6398 (END): Add support for building the native CPU detection tables.
6399 * config/arm/t-arm (arm-native.h): Add build rule.
6400 (driver-arm.o): Add dependency on arm-native.h.
6402 2018-07-13 Richard Biener <rguenther@suse.de>
6405 * match.pd (addr1 - addr2): Allow either of the operand to
6408 2018-07-13 Tom de Vries <tdevries@suse.de>
6410 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
6413 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
6415 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
6416 arrays instead of numbered variables.
6418 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
6420 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
6421 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
6423 2018-07-13 Richard Biener <rguenther@suse.de>
6426 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
6427 instead of get_context_die.
6429 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
6430 Richard Biener <rguenther@suse.de>
6433 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
6434 that the loop latch destination where phi is defined.
6436 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
6438 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
6439 (riscv_expand_epilogue): Add assertion to check interrupt mode.
6440 (riscv_set_current_function): Extract getting interrupt type to new
6442 (riscv_get_interrupt_type): New function.
6443 (riscv_merge_decl_attributes): New function, checking interrupt type is
6445 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
6447 2018-07-12 Paul Koning <ni1d@arrl.net>
6449 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
6452 2018-07-12 Paul Koning <ni1d@arrl.net>
6454 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
6455 zero reference, add doloop_end instead.
6456 * doc/md.texi (decrement_and_branch_until_zero): Remove.
6457 (Looping patterns): Remove decrement_and_branch_until_zero. Add
6458 detail for doloop_end.
6460 2018-07-12 Martin Sebor <msebor@redhat.com>
6463 * attribs.c (decl_attributes): Reject conflicting attributes before
6464 calling attribute handlers.
6466 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
6468 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
6470 (gcc::dump_manager::get_dump_file_name): likewise.
6471 (dump_begin): Likewise.
6472 * dumpfile.h (dump_begin): Update prototype.
6473 (gcc::dump_manager::get_dump_file_name,
6474 gcc::dump_manager::get_dump_file_name): Update prototype.
6476 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6478 * internal-fn.h (vectorizable_internal_fn_p): New function.
6479 * tree-vect-slp.c (compatible_calls_p): Likewise.
6480 (vect_build_slp_tree_1): Remove nops argument. Handle calls
6481 to internal functions.
6482 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
6484 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6486 * fold-const.h (inverse_conditions_p): Declare.
6487 * fold-const.c (inverse_conditions_p): New function.
6488 * match.pd: Use inverse_conditions_p. Add folds of view_converts
6489 that test the inverse condition of a conditional internal function.
6490 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
6491 * internal-fn.c (internal_fn_mask_index): Handle conditional
6493 (vectorized_internal_fn_supported_p): New function.
6494 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
6495 (any_pred_load_store): Replace with...
6496 (need_to_predicate): ...this new variable.
6497 (redundant_ssa_names): New variable.
6498 (ifcvt_can_use_mask_load_store): Move initial checks to...
6499 (ifcvt_can_predicate): ...this new function. Handle tree codes
6500 for which a conditional internal function exists.
6501 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
6502 instead of ifcvt_can_use_mask_load_store. Update after variable
6504 (predicate_load_or_store): New function, split out from
6505 predicate_mem_writes.
6506 (check_redundant_cond_expr): New function.
6507 (value_available_p): Likewise.
6508 (predicate_rhs_code): Likewise.
6509 (predicate_mem_writes): Rename to...
6510 (predicate_statements): ...this. Use predicate_load_or_store
6511 and predicate_rhs_code.
6512 (combine_blocks, tree_if_conversion): Update after above name changes.
6513 (ifcvt_local_dce): Handle redundant_ssa_names.
6514 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
6515 general conditional functions.
6516 * tree-vect-stmts.c (vectorizable_call): Likewise.
6518 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6519 Alan Hayward <alan.hayward@arm.com>
6520 David Sherwood <david.sherwood@arm.com>
6522 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
6523 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
6524 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
6525 plus and minus and convert them into IFN_COND_FMA-based sequences.
6526 (convert_mult_to_fma): Handle conditional plus and minus.
6528 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6530 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
6531 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
6532 (cond_fnms_optab): New optabs.
6533 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
6535 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
6536 * internal-fn.h (get_conditional_internal_fn): Declare.
6537 (get_unconditional_internal_fn): Likewise.
6538 * internal-fn.c (cond_ternary_direct): New macro.
6539 (expand_cond_ternary_optab_fn): Likewise.
6540 (direct_cond_ternary_optab_supported_p): Likewise.
6541 (FOR_EACH_COND_FN_PAIR): Likewise.
6542 (get_conditional_internal_fn): New function.
6543 (get_unconditional_internal_fn): Likewise.
6544 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
6545 (gimple_match_op::gimple_match_op): Add a new overload for 5
6547 (gimple_match_op::set_op): Likewise.
6548 (gimple_resimplify5): Declare.
6549 * genmatch.c (decision_tree::gen): Generate simplifications for
6551 * gimple-match-head.c (gimple_simplify): Define an overload for
6552 5 operands. Handle calls with 5 arguments in the top-level overload.
6553 (convert_conditional_op): Handle conversions from unconditional
6554 internal functions to conditional ones.
6555 (gimple_resimplify5): New function.
6556 (build_call_internal): Pass a fifth operand.
6557 (maybe_push_res_to_seq): Likewise.
6558 (try_conditional_simplification): Try converting conditional
6559 internal functions to unconditional internal functions.
6560 Handle 3-operand unconditional forms.
6561 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
6562 Define ternary equivalents of the current rules for binary conditional
6564 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
6566 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
6567 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
6568 (optab): Handle them.
6569 (SVE_COND_FP_TERNARY): New int iterator.
6570 (sve_fmla_op, sve_fmad_op): New int attributes.
6571 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
6572 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
6573 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
6575 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6577 * target.def (preferred_else_value): New target hook.
6578 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
6579 * doc/tm.texi: Regenerate.
6580 * targhooks.h (default_preferred_else_value): Declare.
6581 * targhooks.c (default_preferred_else_value): New function.
6582 * internal-fn.h (conditional_internal_fn_code): Declare.
6583 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
6584 (get_conditional_internal_fn): Use it.
6585 (conditional_internal_fn_code): New function.
6586 * gimple-match.h (gimple_match_cond): New struct.
6587 (gimple_match_op): Add a cond member function.
6588 (gimple_match_op::gimple_match_op): Update all forms to take a
6590 * genmatch.c (expr::gen_transform): Use the same condition as res_op
6591 for the suboperation, but don't specify a particular else_value.
6592 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
6593 (visit_nary_op, visit_reference_op_load): Pass
6594 gimple_match_cond::UNCOND to the gimple_match_op constructor.
6595 * gimple-match-head.c: Include tree-eh.h
6596 (convert_conditional_op): New function.
6597 (maybe_resimplify_conditional_op): Likewise.
6598 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
6599 (gimple_resimplify2): Likewise.
6600 (gimple_resimplify3): Likewise.
6601 (gimple_resimplify4): Likewise.
6602 (maybe_push_res_to_seq): Return null for conditional operations.
6603 (try_conditional_simplification): New function.
6604 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
6606 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
6608 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
6610 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
6612 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
6614 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
6616 (hash_tree): Do not hash DECL_FCONTEXT
6617 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
6618 Do not stream DECL_FCONTEXT.
6619 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
6620 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
6622 2018-07-12 Richard Biener <rguenther@suse.de>
6625 * dwarf2out.c (gen_block_die): Only output blocks when they have
6626 at least one !DECL_IGNORED_P variable.
6628 2018-07-12 Richard Biener <rguenther@suse.de>
6631 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
6632 Remove -mieee-fp handling.
6634 2018-07-12 Richard Biener <rguenther@suse.de>
6636 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
6637 left-over from last patch.
6639 2018-07-12 Jakub Jelinek <jakub@redhat.com>
6641 PR tree-optimization/86492
6642 * gimple-ssa-store-merging.c
6643 (imm_store_chain_info::coalesce_immediate_stores): Call
6644 check_no_overlap even for the merge_overlapping case. Formatting fix.
6646 2018-07-12 Richard Biener <rguenther@suse.de>
6649 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
6650 move possibly trapping operations into the conditional.
6652 2018-07-12 Richard Biener <rguenther@suse.de>
6654 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
6655 (vn_lookup_simplify_result): Remove recursion limit applied
6657 (vn_nary_build_or_lookup_1): Adjust.
6658 (try_to_simplify): Likewise.
6659 * gimple-match-head.c (gimple_resimplify1): Instead apply one
6661 (gimple_resimplify2): Likewise.
6662 (gimple_resimplify3): Likewise.
6663 (gimple_resimplify4): Likewise.
6665 2018-07-11 Jakub Jelinek <jakub@redhat.com>
6667 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
6668 Use __mmask64 type instead of __mmask8 for __M argument.
6669 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
6670 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
6672 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
6673 __mmask16 for __M argument.
6674 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
6675 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
6676 to __mmask16 instead of __mmask8.
6677 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
6678 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
6679 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
6680 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
6681 instead of __mmask16 for __U argument.
6682 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
6683 __mmask16 instead of __mmask8 for __U argument.
6684 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
6686 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
6688 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
6690 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
6692 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
6694 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
6696 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
6698 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
6700 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
6702 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
6704 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
6705 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
6706 return type as well as __M argument type and all casts from __mmask8
6708 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
6709 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
6710 return type as well as __M argument type and all casts from __mmask8
6712 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
6713 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
6714 return type as well as __M argument type and all casts from __mmask8
6716 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
6717 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
6718 return type as well as __M argument type and all casts from __mmask8
6720 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
6721 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
6724 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
6726 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
6727 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
6730 2018-07-11 Paul Koning <ni1d@arrl.net>
6732 * doc/md.texi (define_subst): Document how multiple occurrences of
6733 the same argument in the replacement pattern are handled.
6735 2018-07-11 Paul Koning <ni1d@arrl.net>
6737 * doc/extend.texi (Common Variable Attributes): Move "mode" into
6739 (Common Type Attributes): Add "mode" attribute.
6741 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
6743 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
6744 stream DECL_ORIGINAL_TYPE.
6745 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
6746 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
6747 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
6748 Do not walk original type.
6749 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
6751 (write_ts_decl_non_common_tree_pointers): Do not stream
6753 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
6754 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
6756 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
6758 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
6759 thread twice from the same starting edge.
6761 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
6763 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
6764 * gimple.c (gimple_call_nonnull_result_p): ...here...
6765 (gimple_call_nonnull_arg): ...and here.
6766 * gimple.h (gimple_call_nonnull_result_p): New.
6767 (gimple_call_nonnull_arg): New.
6769 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
6771 * config/arm/arm-cpus.in: Move information from fpu field of each
6772 cpu definition to the isa field.
6773 * config/arm/parsecpu.awk (fpu): Delete match rule.
6774 (gen_comm_data): Don't add bits from the CPU's FPU entry.
6776 2018-07-11 Richard Biener <rguenther@suse.de>
6779 * dwarf2out.c (init_sections_and_labels): Use
6780 output_asm_line_debug_info consistently.
6781 (dwarf2out_early_finish): Likewise.
6782 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
6785 2018-07-11 Richard Biener <rguenther@suse.de>
6787 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
6788 Rework father_bb setting in a way to avoid propagating constants
6789 multiple times on a loop body.
6791 2018-07-10 Mark Wielaard <mark@klomp.org>
6794 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
6796 2018-07-10 Richard Biener <rguenther@suse.de>
6798 * hash-map.h (hash_map::iterator::operator*): Return
6799 references to key and value.
6801 2018-07-10 Jakub Jelinek <jakub@redhat.com>
6804 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
6805 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
6806 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
6807 (gimplify_omp_for): For composite loops, move outer
6808 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
6809 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
6810 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
6811 TREE_LIST for both the original class iterator and the "last" helper
6812 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
6813 loop, remember has_decl_expr from outer composite loops for the
6814 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
6816 2018-07-09 Martin Sebor <msebor@redhat.com>
6820 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
6821 accessing implicitly initialized array elements.
6822 * expr.c (string_constant): Handle string initializers of
6823 character arrays within aggregates.
6824 * gimple-fold.c (fold_array_ctor_reference): Add argument.
6825 Store element offset. As a special case, handle zero size.
6826 (fold_nonarray_ctor_reference): Same.
6827 (fold_ctor_reference): Add argument. Store subobject offset.
6828 * gimple-fold.h (fold_ctor_reference): Add argument.
6830 2018-07-09 Paul Koning <ni1d@arrl.net>
6832 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
6833 (pdp11_insn_cost): New function.
6834 (pdp11_md_asm_adjust): New function.
6835 (TARGET_INVALID_WITHIN_DOLOOP): Define.
6836 (pdp11_rtx_costs): Update to match machine better.
6837 (output_addr_const_pdp11): Correct format mismatch warnings.
6838 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
6839 * config/pdp11/pdp11.md: General change to add base_cost and/or
6840 length attributes for use by new pdp11_insn_cost function.
6841 (MIN_BRANCH): Correct definition.
6843 (doloop_end): Use standard pattern name for looping pattern.
6844 (doloop_end_nocc): New.
6845 (movsf): Add another constraint alternative.
6846 (zero_extendqihi2): Add constraint alternatives for not in place
6848 (zero_extendhisi2): Remove.
6849 (shift patterns): Add CC handling variants.
6853 (define_peephole2): New peephole to recognize mov that sets CC for
6856 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6858 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
6859 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
6862 2018-07-09 Jakub Jelinek <jakub@redhat.com>
6865 * real.c (real_nextafter): Return true if result is denormal.
6867 2018-07-09 Martin Liska <mliska@suse.cz>
6869 * common.opt: Add back wrongly removed attribute.
6871 2018-07-09 Richard Biener <rguenther@suse.de>
6874 * dwarf2out.c (gen_block_die): For an early generated DIE
6875 always output high/low PC attributes.
6877 2018-07-09 Tom de Vries <tdevries@suse.de>
6879 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
6880 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
6881 onto VAR_DECL with abstract origin.
6883 2018-07-07 Jim Wilson <jimw@sifive.com>
6885 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
6887 2018-07-07 Tom de Vries <tdevries@suse.de>
6889 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
6891 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
6893 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
6895 (combine_bound): Use wide-int overflow calculation instead of
6897 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
6899 * fold-const.c (int_const_binop_2): Same.
6900 (extract_muldiv_1): Same.
6901 (fold_div_compare): Same.
6902 (fold_abs_const): Same.
6904 * poly-int.h (add): Same.
6908 * predict.c (predict_iv_comparison): Same.
6909 * profile-count.c (slow_safe_scale_64bit): Same.
6910 * simplify-rtx.c (simplify_const_binary_operation): Same.
6911 * tree-chrec.c (tree_fold_binomial): Same.
6912 * tree-data-ref.c (split_constant_offset_1): Same.
6913 * tree-if-conv.c (idx_within_array_bound): Same.
6914 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
6915 * tree-ssa-phiopt.c (minmax_replacement): Same.
6916 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
6917 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
6918 * vr-values.c (vr_values::adjust_range_with_scev): Same.
6919 * wide-int.cc (wi::add_large): Same.
6920 (wi::mul_internal): Same.
6921 (wi::sub_large): Same.
6922 (wi::divmod_internal): Same.
6923 * wide-int.h: Change overflow type to overflow_type for neg, add,
6924 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
6925 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
6926 mul_internal, divmod_internal.
6927 (overflow_type): New enum.
6928 (accumulate_overflow): New.
6930 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
6932 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
6933 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
6935 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
6937 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
6938 argument is checked for zero before entering loop, avoid checking again.
6940 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
6942 * gimplify.h (generic_expr_could_trap_p): Set as global function.
6943 * gimplify.h (generic_expr_could_trap_p): Likwise.
6944 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
6946 2018-07-06 Jakub Jelinek <jakub@redhat.com>
6948 PR tree-optimization/86401
6949 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
6950 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
6951 (fold_bit_and_mask): ... here. New helper function for match.pd.
6952 * fold-const.h (fold_bit_and_mask): Declare.
6953 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
6955 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
6958 * target.def (translate_mode_attribute): New hook.
6959 * targhooks.h (default_translate_mode_attribute): Declare.
6960 * targhooks.c (default_translate_mode_attribute): New function.
6961 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
6962 * doc/tm.texi: Regenerate.
6963 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
6964 (rs6000_translate_mode_attribute): New function.
6966 2018-07-06 Paul Koning <ni1d@arrl.net>
6968 * doc/md.texi (define_split): Document DONE and FAIL.
6969 (define_peephole2): Ditto.
6971 2018-07-05 Jeff Law <law@redhat.com>
6973 PR tree-optimization/86010
6974 * tree-ssa-dse.c (compute_trims): More aggressively trim at
6975 both the head and tail of mem* and str* calls.
6977 2018-07-05 Jim Wilson <jimw@sifive.com>
6979 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
6981 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
6983 * config/aarch64/aarch64-simd.md: correct flags text for
6984 MIN_EXPR replacement.
6986 2018-07-05 James Clarke <jrtc27@jrtc27.com>
6988 * configure: Regenerated.
6990 2018-07-05 Carl Love <cel@us.ibm.com>
6992 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
6993 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
6994 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
6995 VSX_BUILTIN_DOUBLEL_V4SF.
6997 2018-07-05 Martin Sebor <msebor@redhat.com>
7000 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
7001 than its domain to compute its the upper bound of a char array.
7003 2018-07-05 Nathan Sidwell <nathan@acm.org>
7005 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
7006 * doc/cpp.texi: Update comment.
7007 * doc/tm.texi: Rebuilt.
7008 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
7009 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
7010 * doc/extend.texi (Backwards Compatibility): Clarify it is system
7011 headers affected by extern "C".
7012 * system.h: Poison NO_IMPLICIT_EXTERN_C.
7013 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
7014 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
7015 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
7016 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
7017 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
7018 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
7019 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
7020 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
7021 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
7022 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
7023 config/sparc/sp64-elf.h, config/spu/spu.h,
7024 config/stormy16/stormy16.h, config/v850/v850.h,
7025 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
7026 define NO_IMPLICIT_EXTERN_C.
7027 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
7029 2018-07-05 Tamar Christina <tamar.christina@arm.com>
7032 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7033 instead of GET_MODE_SIZE when comparing Units.
7035 2018-07-05 Tamar Christina <tamar.christina@arm.com>
7038 * rtlanal.c (set_noop_p): Constrain on mode change,
7039 include hard-reg-set.h
7041 2018-07-05 Tamar Christina <tamar.christina@arm.com>
7043 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
7045 2018-07-05 Jakub Jelinek <jakub@redhat.com>
7048 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
7051 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
7053 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
7055 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
7058 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
7060 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
7062 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
7065 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
7067 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
7069 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
7070 neg_*_op* variables.
7072 2018-07-04 Martin Liska <mliska@suse.cz>
7074 * tree-switch-conversion.c: Define
7075 max_ratio_for_speed and max_ratio_for_size constants.
7077 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
7078 Martin Liska <mliska@suse.cz>
7083 * common.opt: Rename align options with 'str_' prefix.
7084 * common/config/i386/i386-common.c (set_malign_value): New
7086 (ix86_handle_option): Use it to set -falign-* options/
7087 * config/aarch64/aarch64-protos.h (struct tune_params): Change
7088 type from int to string.
7089 * config/aarch64/aarch64.c: Update default values from int
7091 * config/alpha/alpha.c (alpha_override_options_after_change):
7093 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
7094 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7095 max skip conditionally.
7096 * config/i386/freebsd.h (SUBALIGN_LOG): New.
7097 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7098 max skip conditionally.
7099 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7100 max skip conditionally.
7101 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
7102 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7103 max skip conditionally.
7104 * config/i386/i386.c (struct ptt): Change type from int to
7106 (ix86_default_align): Set default values.
7107 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
7108 max skip conditionally.
7109 * config/i386/iamcu.h (SUBALIGN_LOG): New.
7110 (ASM_OUTPUT_MAX_SKIP_ALIGN):
7111 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
7112 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7113 max skip conditionally.
7114 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
7115 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
7116 * config/i386/x86-64.h (SUBALIGN_LOG): New.
7117 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7118 max skip conditionally.
7119 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7120 * config/ia64/ia64.c (ia64_option_override): Set default values
7121 for alignment options.
7122 * config/m68k/m68k.c: Handle new str_align_* options.
7123 * config/mips/mips.c (mips_set_compression_mode): Change
7125 (mips_option_override): Set default values for options.
7126 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
7128 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7130 * config/rx/rx.c (rx_option_override): Likewise.
7131 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
7132 (LABEL_ALIGN): Use align_labels_log.
7133 (LOOP_ALIGN): Use align_loops_align.
7134 * config/s390/s390.c (s390_asm_output_function_label): Use new
7136 * config/sh/sh.c (sh_override_options_after_change):
7137 Change type of constants.
7138 * config/spu/spu.c (spu_sched_init): Likewise.
7139 * config/sparc/sparc.c (sparc_option_override): Set default
7141 * config/visium/visium.c (visium_option_override): Likewise.
7142 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
7143 emit p2align format with last argument if it's not needed.
7144 * doc/invoke.texi: Document extended format of -falign-*.
7145 * final.c: Use align_labels alignment.
7146 * flags.h (struct target_flag_state): Change type to use
7148 (struct align_flags_tuple): New.
7149 (struct align_flags): Likewise.
7150 (align_loops_log): Redefine macro to use new types.
7151 (align_loops_max_skip): Redefine macro to use new types.
7152 (align_jumps_log): Redefine macro to use new types.
7153 (align_jumps_max_skip): Redefine macro to use new types.
7154 (align_labels_log): Redefine macro to use new types.
7155 (align_labels_max_skip): Redefine macro to use new types.
7156 (align_functions_log): Redefine macro to use new types.
7157 (align_loops): Redefine macro to use new types.
7158 (align_jumps): Redefine macro to use new types.
7159 (align_labels): Redefine macro to use new types.
7160 (align_functions): Redefine macro to use new types.
7161 (align_functions_max_skip): Redefine macro to use new types.
7162 (align_loops_value): New macro.
7163 (align_jumps_value): New macro.
7164 (align_labels_value): New macro.
7165 (align_functions_value): New macro.
7166 * function.c (invoke_set_current_function_hook): Propagate
7167 alignment values from flags to global variables default in
7169 * ipa-icf.c (sem_function::equals_wpa): Use
7170 cl_optimization_option_eq instead of memcmp.
7171 * lto-streamer.h (cl_optimization_stream_out): Support streaming
7173 (cl_optimization_stream_in): Likewise.
7174 * optc-save-gen.awk: Support strings in cl_optimization.
7175 * opth-gen.awk: Likewise.
7176 * opts.c (finish_options): Remove error checking of invalid
7178 (MAX_CODE_ALIGN): Remove.
7179 (MAX_CODE_ALIGN_VALUE): Likewise.
7180 (parse_and_check_align_values): New function.
7181 (check_alignment_argument): Likewise.
7182 (common_handle_option): Use check_alignment_argument.
7183 * opts.h (parse_and_check_align_values): Declare.
7184 * toplev.c (init_alignments): Remove.
7185 (read_log_maxskip): New.
7186 (parse_N_M): Likewise.
7187 (parse_alignment_opts): Likewise.
7188 (backend_init_target): Remove usage of init_alignments.
7189 * toplev.h (parse_alignment_opts): Declare.
7190 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
7192 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
7193 * tree.c (cl_option_hasher::equal): New.
7194 * varasm.c: Use new global macros.
7196 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
7198 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
7199 Use a simpler align directive also if MAXSKIP = ALIGN-1.
7200 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7201 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7202 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7203 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7204 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
7205 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
7206 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
7207 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7208 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7209 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7210 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7212 2018-07-04 Martin Liska <mliska@suse.cz>
7213 Jonathan Wakely <jwakely@redhat.com>
7215 * coverage.c: Use correct type.
7216 * doc/invoke.texi: Language correction.
7218 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
7221 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
7222 ENDBRANCH for non-tail call which may return via indirect branch.
7223 * doc/extend.texi: Document indirect_return attribute.
7225 2018-07-03 Martin Sebor <msebor@redhat.com>
7227 PR tree-optimization/86274
7228 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
7230 (format_floating): Correct handling of infinities and NaNs.
7232 2018-07-03 Martin Sebor <msebor@redhat.com>
7234 * print-tree.c (print_real_cst): New function.
7235 (print_node_brief): Call it.
7236 (print_node): Ditto.
7238 2018-07-03 Jeff Law <law@redhat.com>
7240 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
7241 into a single pattern.
7243 * config/h8300/h8300.md (ors code_iterator): New.
7244 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
7245 a single pattern and single splitter.
7246 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
7247 (iorqi3_1, xorqi3_1): Likewise.
7248 (iorqi3, xorqi3 expanders): Similarly.
7250 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
7251 (movmd_internal) into a single pattern using the P mode iterator.
7252 (movmd splitters): Similarly.
7253 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
7254 (movsd splitters): Similarly.
7256 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
7257 ADDB, ADDW and ADDL into a single ADD attribute which selects the
7258 right table based on the size of the operand.
7259 * config/h8300/h8300.md (length_table): Corresponding changes. All
7260 references to "addb", "addw" and "addl" changed to "add".
7261 (btst patterns): Merge two variants into a single pattern.
7262 (tstqi, tsthi): Likewise.
7263 (addhi3_incdec, addsi3_incdec): Likewise.
7264 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
7265 (mulhi3, mulsi3): Likewise.
7266 (udivhi3, udivsi3): Likewise.
7267 (divhi3, divsi3): Likewise.
7268 (andorqi3, andorhi3, andorsi3): Likewise.
7270 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
7273 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
7274 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
7276 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7278 PR tree-optimization/85694
7279 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
7280 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
7282 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
7283 (<u>avg<mode>3_ceil): New patterns.
7285 2018-07-03 David Malcolm <dmalcolm@redhat.com>
7287 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
7288 scan-tree-dump directive.
7289 * gcc.dg/vect/slp-perm-2.c: Likewise.
7290 * gcc.dg/vect/slp-perm-3.c: Likewise.
7291 * gcc.dg/vect/slp-perm-5.c: Likewise.
7292 * gcc.dg/vect/slp-perm-6.c: Likewise.
7293 * gcc.dg/vect/slp-perm-7.c: Likewise.
7294 * gcc.dg/vect/slp-perm-8.c: Likewise.
7296 2018-07-03 Marek Polacek <polacek@redhat.com>
7299 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
7301 2018-07-03 Richard Biener <rguenther@suse.de>
7304 * tree-ssa-structalias.c (find_func_clobbers): Properly
7305 handle indirect calls.
7307 2018-07-03 Jeff Law <law@redhat.com>
7309 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
7310 (shifts): New code iterator.
7311 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
7312 expander. Fix HImode handling on H8/SX.
7313 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
7314 (subqi3, subhi3, subsi3 expanders): Likewise.
7315 (andqi3, andhi3, andsi3 expanders): Likewise.
7316 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
7317 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
7318 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
7319 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
7320 (zero_extendqihi2, zero_extendqisi2): Likewise.
7321 (extendqihi2, extendqisi2): Likewise.
7322 (rotlqi3, rotlhi3, rotlsi3): Likewise.
7323 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
7324 (rotlqi3_1, rotlhi3_1): Likewise.
7325 (logicalhi3_sn, logicalsi3_sn): Likewise.
7326 (logicalhi3, logicalsi3): Likewise.
7328 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7330 * tree-vect-patterns.c (vect_recog_rotate_pattern)
7331 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7332 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
7333 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
7334 type to append_pattern_def_seq instead of creating a stmt_vec_info
7336 (build_mask_conversion): Likewise. Remove vinfo argument.
7337 (vect_add_conversion_to_patterm): Likewise, renaming to...
7338 (vect_add_conversion_to_pattern): ...this.
7339 (vect_recog_mask_conversion_pattern): Update call to
7340 build_mask_conversion. Pass the vector type to
7341 append_pattern_def_seq here too.
7342 (vect_recog_gather_scatter_pattern): Update call to
7343 vect_add_conversion_to_pattern.
7345 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7347 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
7348 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7349 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
7350 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
7351 STMT_VINFO_PATTERN_DEF_SEQ to null here.
7352 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
7353 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
7354 append_pattern_def_seq instead of new_pattern_def_seq.
7355 (vect_recog_divmod_pattern): Do both of the above.
7356 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
7359 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7361 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
7362 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
7363 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
7364 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
7365 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
7366 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7367 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
7368 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
7369 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
7370 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
7371 parameter with a single stmt_vec_info.
7372 (vect_recog_func_ptr): Likewise.
7373 (vect_recog_gather_scatter_pattern): Likewise, folding in...
7374 (vect_try_gather_scatter_pattern): ...this.
7375 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
7376 the stmt_vec_info of the statement to be matched. Don't clear
7377 STMT_VINFO_RELATED_STMT.
7378 (vect_pattern_recog): Update call accordingly.
7380 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7382 PR tree-optimization/85694
7383 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
7384 (uavgM3_ceil): Document new optabs.
7385 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
7386 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
7388 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
7389 (savg_ceil_optab): New optabs.
7390 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
7391 (vect_vect_recog_func_ptrs): Add it.
7392 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
7393 constant directly from the associated lhs.
7395 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7397 * tree-vect-patterns.c (vect_split_statement): New function.
7398 (vect_convert_input): Use it to try to split an existing cast.
7400 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7402 * poly-int.h (print_hex): New function.
7403 * dumpfile.h (dump_dec, dump_hex): Declare.
7404 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
7405 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
7406 min_input_precision, operation_precision and operation_sign.
7407 * tree-vect-patterns.c (vect_get_range_info): New function.
7408 (vect_same_loop_or_bb_p, vect_single_imm_use)
7409 (vect_operation_fits_smaller_type): Delete.
7410 (vect_look_through_possible_promotion): Add an optional
7411 single_use_p parameter.
7412 (vect_recog_over_widening_pattern): Rewrite to use new
7413 stmt_vec_info infomration. Handle one operation at a time.
7414 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
7415 (vect_truncatable_operation_p, vect_set_operation_type)
7416 (vect_set_min_input_precision): New functions.
7417 (vect_determine_min_output_precision_1): Likewise.
7418 (vect_determine_min_output_precision): Likewise.
7419 (vect_determine_precisions_from_range): Likewise.
7420 (vect_determine_precisions_from_users): Likewise.
7421 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
7422 (vect_vect_recog_func_ptrs): Put over_widening first.
7424 (vect_pattern_recog): Call vect_determine_precisions.
7426 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7428 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
7429 statements that have been replaced by further pattern statements.
7430 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
7432 2018-07-03 Richard Biener <rguenther@suse.de>
7434 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
7435 always set *dt. Dump vectype in vectype overload.
7436 * dumpfile.h (dump_gimple_expr): New function.
7437 (dump_gimple_expr_loc): Likewise.
7438 * dumpfile.c (dump_gimple_expr): New function.
7439 (dump_gimple_expr_loc): Likewise.
7441 2018-07-02 Jeff Law <law@redhat.com>
7443 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
7444 the H8/300, H8/300H and H8/S variants into a single pattern.
7445 (movhi_h8300, movqi_h8300hs): Similarly.
7446 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
7447 (QHI mode iterator): New.
7449 * config/h8300/h8300.md: Remove trailing whitespace.
7451 2018-07-02 Jim Wilson <jimw@sifive.com>
7453 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
7454 instead of emit_insn for interrupt returns.
7455 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
7456 (riscv_sret, riscv_uret): Likewise.
7458 2018-07-02 David Malcolm <dmalcolm@redhat.com>
7460 * pretty-print.c (selftest::test_pp_format): Move save and restore
7461 of quotes to class auto_fix_quotes, and add an instance.
7462 * selftest.c: Include "intl.h".
7463 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
7464 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
7465 * selftest.h (selftest::auto_fix_quotes): New class.
7467 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
7469 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
7470 (aarch64_sve_prepare_conditional_op): Remove.
7471 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
7472 Allow aarch64_simd_reg_or_zero as select operand; remove
7473 the aarch64_sve_prepare_conditional_op call.
7474 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
7475 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7476 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
7477 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
7478 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
7479 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
7480 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
7481 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
7482 and a splitters to match all of the *_any patterns.
7483 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
7485 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
7486 (SVE_COND_FP_BINARY_REV): Remove.
7487 (sve_int_op_rev, sve_fp_op_rev): New.
7488 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
7489 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
7490 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
7491 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
7492 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
7493 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
7494 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
7495 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
7496 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
7498 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
7499 Remove match_dup 1 from the inner unspec.
7500 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7502 * config/aarch64/aarch64.md (movprfx): New attr.
7503 (length): Default movprfx to 8.
7504 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
7505 (*madd<SVE_I>, *msub<SVE_I): Likewise.
7506 (*<su>mul<SVE_I>3_highpart): Likewise.
7507 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
7508 (*v<ASHIFT><SVE_I>3): Likewise.
7509 (*<su><MAXMIN><SVE_I>3): Likewise.
7510 (*<su><MAXMIN><SVE_F>3): Likewise.
7511 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
7512 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
7513 (*div<SVE_F>4): Likewise.
7515 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
7517 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
7520 2018-07-02 Richard Biener <rguenther@suse.de>
7522 PR tree-optimization/86363
7523 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
7524 memset argument refers to a non-variable address.
7526 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
7528 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
7529 {PLUS,MINUS}_EXPR code to...
7530 (adjust_symbolic_bound): ...here,
7531 (combine_bound): ...here,
7532 (set_value_range_with_overflow): ...and here.
7534 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
7536 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
7538 (extract_range_from_abs_expr): ...here.
7540 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
7542 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
7543 -fno-omit-frame-pointer when not optimizing.
7545 2018-07-02 Martin Liska <mliska@suse.cz>
7548 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
7549 (propagate_nothrow): Likewise.
7551 2018-07-02 Martin Liska <mliska@suse.cz>
7554 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
7556 2018-07-02 David Malcolm <dmalcolm@redhat.com>
7558 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
7559 function in r262149, changing "loc" param from source_location to
7560 const dump_location_t &.
7561 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
7562 declaration, as above.
7564 2018-07-01 Paul Koning <ni1d@arrl.net>
7566 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
7567 -munit-asm, -mgnu-asm, -mdec-asm.
7568 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
7569 (pdp11_output_labelref): New.
7570 (pdp11_output_def): New.
7571 (pdp11_output_addr_vec_elt): New.
7572 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
7573 %# and %@ format codes.
7574 (pdp11_option_override): New.
7575 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
7576 (pdp11_output_ident): New.
7577 (pdp11_asm_named_section): New.
7578 (pdp11_asm_init_sections): New.
7579 (pdp11_file_start): New.
7580 (pdp11_file_end): New.
7581 (output_ascii): Use .ascii/.asciz for -mdec-asm.
7582 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
7583 %o, like %c but octal.
7584 (pdp11_option_override): New.
7585 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
7587 (DATA_SECTION_ASM_OP): Ditto.
7588 (READONLY_DATA_SECTION_ASM_OP): New.
7589 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
7590 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
7591 (ASM_OUTPUT_LABELREF): Ditto.
7592 (ASM_OUTPUT_DEF): Ditto.
7593 (ASM_OUTPUT_EXTERNAL): New.
7594 (ASM_OUTPUT_SOURCE_FILENAME): New.
7595 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
7596 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
7597 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
7598 %# and %@ format codes.
7599 * config/pdp11/pdp11.opt (mgnu-asm): New.
7600 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
7601 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
7602 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
7604 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
7606 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
7607 dereferencing path[] beyond its length.
7609 (debug_all_paths): New.
7610 (rewire_first_differing_edge): New.
7611 (adjust_paths_after_duplication): New.
7612 (duplicate_thread_path): Call adjust_paths_after_duplication.
7614 (thread_through_all_blocks): Add new argument to
7615 duplicate_thread_path.
7617 2018-06-30 Jim Wilson <jimw@sifive.com>
7619 * config/riscv/predicates.md (p2m1_shift_operand): New.
7620 (high_mask_shift_operand): New.
7621 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
7622 pattern using p2m1_shift_operand.
7623 (lshsi3_zero_extend_3+2): New combiner pattern using
7624 high_mask_shift_operand.
7626 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7628 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
7630 (vect_recog_rotate_pattern): ...here.
7631 (vect_convert_input): Try to insert casts of invariants in the
7633 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
7634 preheader to be empty.
7636 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7638 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
7639 vector type. If given, install it in the new statement's
7641 (vect_element_precision): New function.
7642 (vect_unpromoted_value): New struct.
7643 (vect_unpromoted_value::vect_unpromoted_value): New function.
7644 (vect_unpromoted_value::set_op): Likewise.
7645 (vect_look_through_possible_promotion): Likewise.
7646 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
7647 (vect_widened_op_tree, vect_convert_input): Likewise.
7648 (vect_convert_inputs, vect_convert_output): Likewise.
7649 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
7650 to handle the optional cast of the multiplication result and
7651 vect_widened_op_tree to detect the widened multiplication itself.
7652 Do not require the input and output of promotion casts to have
7653 the same sign, but base the signedness of the operation on the
7654 input rather than the result. If the pattern includes two
7655 promotions, check that those promotions have the same sign.
7656 Do not restrict the MULT_EXPR handling to a double-width result;
7657 handle quadruple-width results and wider. Use vect_convert_inputs
7658 to convert the inputs to the common type.
7659 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
7660 to handle the optional cast of the ABS result. Also allow a sign
7661 change or a sign extension between the ABS and MINUS.
7662 Use vect_widened_op_tree to detect the widened subtraction and use
7663 vect_convert_inputs to convert the inputs to the common type.
7664 (vect_handle_widen_op_by_const): Delete.
7665 (vect_recog_widen_op_pattern): New function.
7666 (vect_recog_widen_mult_pattern): Use it.
7667 (vect_recog_widen_shift_pattern): Likewise.
7668 (vect_recog_widen_sum_pattern): Use
7669 vect_look_through_possible_promotion to handle the promoted
7672 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7674 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
7675 the containing gimple_seq *.
7676 * gimple-iterator.h (gsi_for_stmt): Declare it.
7677 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
7678 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
7679 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7680 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7681 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
7683 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
7685 (vect_mark_pattern_stmts): ...here. Handle cases in which the
7686 statement being replaced is part of an existing pattern
7687 definition sequence, inserting the new pattern statements before
7689 (vect_pattern_recog_1): Don't return a bool. If the statement
7690 is already part of a pattern, instead apply pattern matching
7691 to the pattern definition statements. Don't clear the
7692 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
7693 (vect_pattern_recog): Don't break after the first match;
7694 continue processing the pattern definition statements instead.
7695 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
7697 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7699 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7700 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7701 (vect_recog_widen_sum_pattern): Use it.
7703 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7705 * tree-vect-loop.c (vectorizable_reduction): Assert that the
7706 phi is not a pattern statement and has not been replaced by
7707 a pattern statement.
7708 * tree-vect-patterns.c (type_conversion_p): Don't check
7709 STMT_VINFO_IN_PATTERN_P.
7710 (vect_recog_vector_vector_shift_pattern): Likewise.
7711 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
7712 the pattern statement rather than the original statement; check
7713 directly for a WIDEN_MULT_EXPR here.
7714 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
7715 vect_is_simple_use to return the pattern statement rather
7716 than the original statement; use is_pattern_stmt_p to check
7717 for such a pattern statement.
7718 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
7719 to return the pattern statement rather than the original statement;
7720 don't do the same transformation here.
7721 (vect_is_simple_use): If the defining statement has been replaced
7722 by a pattern statement, return the pattern statement instead.
7723 Remove the corresponding (local) transformation from the vectype
7726 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7728 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
7729 end and default to null.
7730 * tree-vect-loop.c (vect_create_epilog_for_reduction)
7731 (vectorizable_reduction): Update calls accordingly, dropping the
7732 gimple ** argument if the passed-back statement isn't needed.
7733 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
7734 (vect_recog_rotate_pattern): Likewise.
7735 (vect_recog_mask_conversion_pattern): Likewise.
7736 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7737 (vect_mask_constant_operand_p): Likewise.
7738 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
7739 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
7740 (get_group_load_store_type, get_load_store_type): Likewise.
7741 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
7742 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
7743 (vectorizable_conversion, vectorizable_assignment): Likewise.
7744 (vectorizable_shift, vectorizable_operation): Likewise.
7745 (vectorizable_store, vect_is_simple_cond): Likewise.
7746 (vectorizable_condition, vectorizable_comparison): Likewise.
7747 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
7748 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
7749 and move it to the end. Cope with null def_stmt_outs.
7751 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
7753 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
7755 2018-06-29 Jeff Law <law@redhat.com>
7757 * config/v850/v850.c (v850_legitimate_address_p): Handle large
7758 displacements for TARGET_V850E2V3 and newer.
7759 (TARGET_LRA_P): Remove. Defaults to LRA now.
7760 * config/v850/v850.md (sign23byte_load): Remove.
7761 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
7762 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
7764 2018-06-29 Martin Liska <mliska@suse.cz>
7767 * coverage.c (coverage_init): Mangle full path name.
7768 * doc/invoke.texi: Document the change.
7769 * gcov-io.c (mangle_path): New.
7770 * gcov-io.h (mangle_path): Likewise.
7771 * gcov.c (mangle_name): Use mangle_path for path mangling.
7773 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7775 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
7776 if starting source register is not even.
7778 2018-06-29 Martin Liska <mliska@suse.cz>
7780 PR tree-optimization/86263
7781 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
7782 Make edge redirection.
7784 2018-06-29 David Malcolm <dmalcolm@redhat.com>
7786 * dumpfile.c (dump_loc): Add indentation based on scope depth.
7787 (dump_scope_depth): New variable.
7788 (get_dump_scope_depth): New function.
7789 (dump_begin_scope): New function.
7790 (dump_end_scope): New function.
7791 * dumpfile.h (get_dump_scope_depth): New declaration.
7792 (dump_begin_scope): New declaration.
7793 (dump_end_scope): New declaration.
7794 (class auto_dump_scope): New class.
7795 (AUTO_DUMP_SCOPE): New macro.
7796 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
7799 2018-06-29 Richard Biener <rguenther@suse.de>
7801 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
7802 compute_all_dependences succeeds.
7803 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
7804 exceed --param loop-max-datarefs-for-datadeps.
7806 2018-06-29 Jakub Jelinek <jakub@redhat.com>
7808 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
7810 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
7813 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
7814 alternative 0 in preferred_for_speed attribute.
7816 2018-06-28 Paul Koning <ni1d@arrl.net>
7818 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
7819 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
7820 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
7821 * config/pdp11/pdp11.md: Correct "length" attribute calculation
7822 for shift insn patterns.
7824 2018-06-28 David Malcolm <dmalcolm@redhat.com>
7826 * cgraph.c (cgraph_node::get_body): Replace assignments to
7827 "dump_file" with calls to set_dump_file.
7828 * dumpfile.c (alt_dump_file): Make static, and group with...
7829 (alt_flags): ...this definition.
7830 (dumps_are_enabled): New variable.
7831 (refresh_dumps_are_enabled): New function.
7832 (set_dump_file): New function.
7833 (set_alt_dump_file): New function.
7834 (gcc::dump_manager::dump_start): Replace assignments to
7835 "dump_file" and "alt_dump_file" with calls to set_dump_file and
7837 (gcc::dump_manager::dump_finish): Likewise.
7838 * dumpfile.h (alt_dump_file): Delete decl.
7839 (dumps_are_enabled): New variable decl.
7840 (set_dump_file): New function decl.
7841 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
7843 * tree-nested.c (lower_nested_functions): Replace assignments to
7844 "dump_file" with calls to set_dump_file.
7846 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
7848 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
7849 goto_locus of each outgoing edge of each basic block.
7851 2018-06-28 Richard Biener <rguenther@suse.de>
7853 * dwarf2out.c (decl_scope_table): Remove.
7854 (push_decl_scope): Likewise.
7855 (pop_decl_scope): Likewise.
7856 (gen_type_die_for_member): Do not call push/pop_decl_scope.
7857 (gen_struct_or_union_type_die): Likewise.
7858 (gen_tagged_type_die): Likewise.
7859 (dwarf2out_init): Do not initialize decl_scope_table.
7860 (dwarf2out_c_finalize): Do not free it.
7862 2018-06-28 Richard Biener <rguenther@suse.de>
7864 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
7865 deciding whether to not re-use a DIE.
7867 2018-06-28 Richard Biener <rguenther@suse.de>
7869 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
7870 DW_AT_abstract_origin attribute.
7872 2018-06-28 Martin Liska <mliska@suse.cz>
7874 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
7875 Use newly introduced constants.
7876 * tree-switch-conversion.h (struct jump_table_cluster):
7877 Define max_ratio_for_size and max_ratio_for_speed.
7879 2018-06-28 Martin Liska <mliska@suse.cz>
7881 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7882 Add new checking assert to catch invalid state.
7883 (jump_table_cluster::can_be_handled): Handle single case
7885 (jump_table_cluster::is_beneficial): Bail out for such case.
7886 (bit_test_cluster::find_bit_tests):
7887 Add new checking assert to catch invalid state.
7888 (bit_test_cluster::can_be_handled): Handle single case
7890 (bit_test_cluster::is_beneficial): Bail out for such case.
7891 (switch_decision_tree::analyze_switch_statement):
7894 2018-06-28 Martin Liska <mliska@suse.cz>
7896 * common.opt: Introduce -completion option.
7897 * gcc.c (driver_handle_option): Handle it.
7898 (driver::main): Print completions if completion
7900 * opt-suggestions.c (option_proposer::get_completions):
7902 (option_proposer::suggest_completion): Likewise.
7903 (option_proposer::find_param_completions): Likewise.
7904 (verify_autocompletions): Likewise.
7905 (test_completion_valid_options): Likewise.
7906 (test_completion_valid_params): Likewise.
7907 (in_completion_p): Likewise.
7908 (empty_completion_p): Likewise.
7909 (test_completion_partial_match): Likewise.
7910 (test_completion_garbage): Likewise.
7911 (opt_proposer_c_tests): Likewise.
7912 * opt-suggestions.h: Declare new functions.
7913 * opts.c (common_handle_option): Handle OPT__completion_.
7914 * selftest-run-tests.c (selftest::run_tests): Add
7915 opt_proposer_c_tests.
7916 * selftest.c (assert_str_startswith): New.
7917 * selftest.h (assert_str_startswith): Likewise.
7918 (opt_proposer_c_tests): New.
7919 (ASSERT_STR_STARTSWITH): Likewise.
7921 2018-06-28 Martin Liska <mliska@suse.cz>
7923 * Makefile.in: Add opt-suggestions.o.
7924 * gcc-main.c: Include opt-suggestions.h.
7925 * gcc.c (driver::driver): Likewise.
7926 (driver::~driver): Remove m_option_suggestions.
7927 (driver::build_option_suggestions): Moved to option_proposer.
7928 (driver::suggest_option): Likewise.
7929 (driver::handle_unrecognized_options): Use option_proposer.
7930 * gcc.h (class driver): Add new memver m_option_proposer.
7931 * opt-suggestions.c: New file.
7932 * opt-suggestions.h: New file.
7934 2018-06-28 Martin Liska <mliska@suse.cz>
7936 * vec.h (class auto_string_vec): New (moved from auto_argvec).
7937 (auto_string_vec::~auto_string_vec): Likewise.
7939 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
7941 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
7942 prevent_decl_creation_for_types fields up and add reset_location field.
7943 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
7944 statement if id->reset_location is true.
7945 (copy_edges_for_bb): Do not set goto_locus on the new edges if
7946 id->reset_location is true.
7947 (copy_phis_for_bb): Force input_location on the arguments if
7948 id->reset_location is true.
7949 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
7950 is set on the function to be inlined.
7952 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
7954 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
7956 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
7958 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
7959 registers for Pmode.
7960 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
7961 hard registers for the clobbered pseudo.
7963 2018-06-27 Paul Koning <ni1d@arrl.net>
7965 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
7966 mutually exclusive options.
7967 * config/pdp11/constraints.md (h): New constraint.
7968 (O): Update definition to match shift code generation.
7969 (D): New constraint.
7970 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
7972 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
7974 (output_jump): Change arguments.
7975 (pdp11_fixed_cc_regs): New function.
7976 (pdp11_cc_mode): Ditto.
7977 (pdp11_expand_shift): Ditto.
7978 (pdp11_assemble_shift): Ditto.
7979 (pdp11_small_shift): Ditto.
7980 (pdp11_branch_cost): Remove.
7981 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
7983 (pdp11_register_move_cost): Update for CC registers.
7984 (pdp11_rtx_costs): Add case for LSHIFTRT.
7985 (pdp11_output_jump): Add CCNZ mode conditional branches.
7986 (notice_update_cc_on_set): Remove.
7987 (pdp11_cc_mode): New function.
7988 (simple_memory_operand): Correct pre/post decrement case.
7989 (no_side_effect_operand): New function.
7990 (pdp11_regno_reg_class): Add CC_REGS class.
7991 (pdp11_fixed_cc_regs): New function.
7992 (pdp11_small_shift): New function.
7993 (pdp11_expand_shift): New function to expand shift insns.
7994 (pdp11_assemble_shift): New function to output shifts.
7995 (pdp11_branch_cost): Remove.
7996 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
7997 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
7998 (WCHAR_TYPE): Ditto.
7999 (PTRDIFF_TYPE): Ditto.
8000 (ADJUST_INSN_LENGTH): New macro.
8001 (FIXED_REGISTERS): Add CC registers.
8002 (CALL_USED_REGISTERS): Ditto.
8004 (REG_CLASS_NAMES): Ditto.
8005 (REG_CLASS_CONTENTS): Ditto.
8006 (SELECT_CC_MODE): Use new function.
8007 (TARGET_FLAGS_REGNUM): New macro.
8008 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
8009 (cc0_reg_rtx): Remove.
8010 (CC_STATUS_MDEP): Remove.
8011 (CC_STATUS_MDEFP_INIT): Remove.
8012 (CC_IN_FPU): Remove.
8013 (NOTICE_UPDATE_CC): Remove.
8014 (REGISTER_NAMES): Add CC registers.
8015 (BRANCH_COST): Change to constant 1.
8016 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
8018 * config/pdp11/pdp11.opt (mbcopy): Remove.
8019 (mbcopy-builtin): Remove.
8020 (mbranch-cheap): Remove.
8021 (mbranch-expensive): Remove.
8022 * config/pdp11/predicates.md (expand_shift_operand): Update to
8023 match shift code generation.
8024 (ccnz_operator): New predicate.
8025 * doc/invoke.texi (PDP-11 Options): Remove deleted options
8026 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
8027 Remove non-existent option -mabshi, -mno-abshi. Document mutually
8029 * doc/md.texi (PDP-11): Document new D and h constraints. Update
8030 description of O constraint.
8032 2018-06-27 Jeff Law <law@redhat.com>
8033 Austin Law <austinklaw@gmail.com>
8035 * config/v850/v850.md (addsi3_set_flags): New pattern.
8036 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
8037 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
8038 (zero_extendhisi2_v850_set_flags): Likewise.
8039 (zero_extendqisi2_v850_set_flags): Likewise.
8040 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
8041 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
8042 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
8044 * config/v850/v850-protos.h (notice_update_cc): Remove.
8045 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
8046 (v850_print_operand): Handle 'D' and "d".
8047 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
8048 Add handling of arithmetic/logical operations compared against zero.
8049 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
8050 Do not look at v850_compare_op, instead get mode from last argument.
8051 (v850_gen_compare): Remove
8052 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
8053 after reload for prologue insns.
8054 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
8056 (construct_save_jarl): Likewise.
8057 (TARGET_FLAGS_REGNUM): Define.
8058 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
8059 (NOTICE_UPDATE_CC): Remove.
8060 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
8061 than cc0. Conditionalize on reload_completed.
8062 (cmpsi_insn, setfcc_insn): Likewise.
8063 (tst1 splitter): Turn into define_and_split which sets the flags
8065 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
8066 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
8067 (cstoresf4, cstoredf4): Clobber the flags.
8068 (cmpsi, cmpsf, cmpdf): Remove expanders.
8069 (setf_insn): Remove pattern.
8070 (addsi3): Turn into define_and_split which clobbers the flags after
8071 reload and a suitable pattern (addsi3_clobber_flags) for use after
8073 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
8074 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
8075 (ashrsi3, ashrsi3_v850e2): Likewise.
8076 (bins): Clobber the flags.
8077 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
8078 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
8079 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
8080 (call_value_internal_short, call_value_internal_long): Likewise.
8081 (callt_save_interrupt, callt_return_interrupt): Likewise.
8082 (save_interrupt, return_interrupt): Likewise.
8083 (callt_save_all_interrupt, save_all_interrupt): Likewise.
8084 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
8085 (restore_all_interrupt, _restore_all_interrupt): Likewise.
8086 (All FP comparisons): Only allow after reload has completed.
8088 (divh, divhu): Tweak output template.
8089 (branch_z_normal, branch_z_invert): Remove
8090 (branch_nz_normal, branch_nz_invert): Likewise.
8091 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
8093 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
8094 * config/v850/v850.c (notice_update_cc): Remove.
8095 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
8096 (CC_NO_CARRY): Likewise.
8097 (NOTICE_UPDATE_CC): Define to nothing.
8098 * config/v850/v850.md: Remove block comment on cc0 handling
8099 Remove "cc" attribute from all patterns. Remove cc_status handling
8100 from all patterns. Minor formatting fixes.
8102 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8104 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
8105 (cortex-a76.cortex-a55): Likewise.
8106 * config/aarch64/aarch64-tune.md: Regenerate.
8107 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
8108 cortex-a76.cortex-a55.
8110 2018-06-27 Jeff Law <law@redhat.com>
8112 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
8113 (MULTILIB_DIRNAMES): Similarly.
8115 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
8117 * gimple.h (gimple_return_retbnd): Delete.
8118 (gimple_return_set_retbnd): Likewise.
8119 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
8120 gimple_return_set_retbnd.
8121 * gimple-pretty-print.c (dump_gimple_return): Remove call to
8122 gimple_return_retbnd and adjust.
8123 * tree-inline.h (struct copy_body_data): Remove retbnd field.
8124 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
8125 Explicitly return NULL in a couple more cases. Move assertion
8126 on debug statements and remove unreachable code.
8127 (reset_debug_binding): Do not test id->retbnd.
8128 (expand_call_inline): Do not set it.
8130 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
8132 * configure.ac: Add --disable-gcov option.
8133 * configure: Regenerate.
8134 * Makefile.in: Honour @enable_gcov@.
8135 * doc/install.texi: Document --disable-gcov.
8137 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8139 * config/arm/arm-cpus.in (cortex-a76): New entry.
8140 (cortex-a76.cortex-a55): Likewise.
8141 * config/arm/arm-tables.opt: Regenerate.
8142 * config/arm/arm-tune.md: Likewise.
8143 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
8144 * doc/invoke.texi (ARM Options): Document cortex-a76 and
8145 cortex-a76.cortex-a55.
8147 2018-06-27 Tamar Christina <tamar.christina@arm.com>
8150 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
8152 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
8154 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
8156 (EPILOGUE_USES): Likewise.
8158 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
8160 * tree-inline.c (remap_location): New function extracted from...
8161 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
8162 (copy_phis_for_bb): ...here. Call remap_location.
8163 (copy_cfg_body): Adjust call to copy_edges_for_bb.
8165 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
8167 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
8168 unaligned vsx for 16B memset.
8170 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
8173 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
8174 ieee128_float_type_node to long_double_type_node unless
8175 TARGET_LONG_DOUBLE_128 is set.
8177 2018-06-26 David Malcolm <dmalcolm@redhat.com>
8179 * cfgloop.c (get_loop_location): Convert return type from
8180 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
8181 by implicit construction from rtx_insn *, and using
8182 dump_user_location_t::from_function_decl for the fallback case.
8183 * cfgloop.h (get_loop_location): Convert return type from
8184 location_t to dump_user_location_t.
8185 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
8186 dump_printf_loc to pass in a dump_location_t rather than a
8187 location_t, via the gimple stmt.
8188 * coverage.c (get_coverage_counts): Update calls to
8189 dump_printf_loc to pass in dump_location_t rather than a
8191 * doc/optinfo.texi (Dump types): Convert example of
8192 dump_printf_loc from taking "locus" to taking "insn". Update
8193 description of the "_loc" calls to cover dump_location_t.
8194 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
8196 (dump_user_location_t::dump_user_location_t): New constructors,
8197 from gimple *stmt and rtx_insn *.
8198 (dump_user_location_t::from_function_decl): New function.
8199 (dump_loc): Make static.
8200 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
8201 const dump_location_t &.
8202 (dump_generic_expr_loc): Delete.
8203 (dump_printf_loc): Convert param "loc" from location_t to
8204 const dump_location_t &.
8205 (selftest::test_impl_location): New function.
8206 (selftest::dumpfile_c_tests): New function.
8207 * dumpfile.h: Include "profile-count.h".
8208 (class dump_user_location_t): New class.
8209 (struct dump_impl_location_t): New struct.
8210 (class dump_location_t): New class.
8211 (dump_printf_loc): Convert 2nd param from source_location to
8212 const dump_location_t &.
8213 (dump_generic_expr_loc): Delete.
8214 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
8215 const dump_location_t &.
8216 * gimple-fold.c (fold_gimple_assign): Update call to
8217 dump_printf_loc to pass in a dump_location_t rather than a
8218 location_t, via the gimple stmt.
8219 (gimple_fold_call): Likewise.
8220 * gimple-loop-interchange.cc
8221 (loop_cand::analyze_iloop_reduction_var): Update for change to
8222 check_reduction_path.
8223 (tree_loop_interchange::interchange): Update for change to
8225 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
8226 change in return-type of find_loop_location.
8227 (graphite_regenerate_ast_isl): Likewise.
8228 * graphite-optimize-isl.c (optimize_isl): Likewise.
8229 * graphite.c (graphite_transform_loops): Likewise.
8230 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
8231 pass in a dump_location_t rather than a location_t, via the
8233 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
8234 * ipa.c (walk_polymorphic_call_targets): Likewise.
8235 * loop-unroll.c (report_unroll): Convert "locus" param from
8236 location_t to dump_location_t.
8237 (decide_unrolling): Update for change to get_loop_location's
8239 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
8240 location_t to dump_user_location_t.
8241 (grid_find_single_omp_among_assignments_1): Updates calls to
8242 dump_printf_loc to pass in a dump_location_t rather than a
8243 location_t, via the gimple stmt.
8244 (grid_parallel_clauses_gridifiable): Convert "tloc" from
8245 location_t to dump_location_t. Updates calls to dump_printf_loc
8246 to pass in a dump_location_t rather than a location_t, via the
8248 (grid_inner_loop_gridifiable_p): Likewise.
8249 (grid_dist_follows_simple_pattern): Likewise.
8250 (grid_gfor_follows_tiling_pattern): Likewise.
8251 (grid_target_follows_gridifiable_pattern): Likewise.
8252 (grid_attempt_target_gridification): Convert initialization
8253 of local "grid" from memset to zero-initialization; FIXME: does
8254 this require C++11? Update call to dump_printf_loc to pass in a
8255 optinfo_location rather than a location_t, via the gimple stmt.
8256 * profile.c (read_profile_edge_counts): Updates call to
8257 dump_printf_loc to pass in a dump_location_t rather than a
8259 (compute_branch_probabilities): Likewise.
8260 * selftest-run-tests.c (selftest::run_tests): Call
8262 * selftest.h (dumpfile_c_tests): New decl.
8263 * tree-loop-distribution.c (pass_loop_distribution::execute):
8264 Update for change in return type of find_loop_location.
8265 * tree-parloops.c (parallelize_loops): Likewise.
8266 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
8267 "locus" from location_t to dump_user_location_t.
8268 (canonicalize_loop_induction_variables): Likewise.
8269 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
8270 for change in return type of find_loop_location.
8271 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
8272 to dump_printf_loc to pass in a dump_location_t rather than a
8273 location_t, via the stmt.
8274 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
8276 * tree-vect-loop-manip.c (find_loop_location): Convert return
8277 type from source_location to dump_user_location_t.
8278 (vect_do_peeling): Update for above change.
8279 (vect_loop_versioning): Update for change in type of
8281 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
8282 from location_t to dump_user_location_t.
8283 (vect_estimate_min_profitable_iters): Update for change in type
8285 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
8286 location_t to dump_location_t.
8287 (vect_slp_bb): Update for change in type of vect_location.
8288 * tree-vectorizer.c (vect_location): Convert from source_location
8289 to dump_user_location_t.
8290 (try_vectorize_loop_1): Update for change in vect_location's type.
8291 (vectorize_loops): Likewise.
8292 (increase_alignment): Likewise.
8293 * tree-vectorizer.h (vect_location): Convert from source_location
8294 to dump_user_location_t.
8295 (find_loop_location): Convert return type from source_location to
8296 dump_user_location_t.
8297 (check_reduction_path): Convert 1st param from location_t to
8298 dump_user_location_t.
8299 * value-prof.c (check_counter): Update call to dump_printf_loc to
8300 pass in a dump_user_location_t rather than a location_t; update
8301 call to error_at for change in type of "locus".
8302 (check_ic_target): Update call to dump_printf_loc to
8303 pass in a dump_user_location_t rather than a location_t, via the
8306 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
8308 * config/s390/s390.h (enum processor_flags): Do not use
8309 default tune parameter when -march was specified.
8311 2018-06-26 Jakub Jelinek <jakub@redhat.com>
8314 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
8315 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
8318 2018-06-26 Richard Biener <rguenther@suse.de>
8320 PR tree-optimization/86287
8322 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
8323 (vect_analyze_loop): Initialize n_stmts.
8325 2018-06-26 Richard Biener <rguenther@suse.de>
8328 * fold-const.c (fold_convertible_p): Pointer extension
8331 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
8334 * dwarf2out.c (loc_list_has_views): Adjust comments.
8335 (dw_loc_list): Split single cross-partition range with
8338 2018-06-25 Jeff Law <law@redhat.com>
8340 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
8341 on -mbig-switch by default.
8343 * config/v850/predicates.md (const_float_1_operand): Fix match_code
8345 (const_float_0_operand): Remove unused predicate.
8346 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
8347 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
8348 (recipsf2): New expander. Original pattern now called
8350 (recipdf2, recipdf2_insn): Similarly.
8351 (rsqrtsf2, rsqrtsf2_insn): Similarly
8352 (rsqrtdf2, rsqrtdf2_insn): Similarly
8354 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
8356 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
8357 Simplify logic for FreeBSD (twice).
8359 2018-06-25 Martin Sebor <msebor@redhat.com>
8361 PR tree-optimization/86204
8362 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
8363 a strnlen result if it's less than the length of the string.
8365 2018-06-25 Martin Sebor <msebor@redhat.com>
8367 PR tree-optimization/85700
8368 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
8369 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
8370 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
8372 2018-06-25 Martin Sebor <msebor@redhat.com>
8374 * doc/extend.texi (Zero-length arrays): Update and clarify.
8376 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
8378 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
8379 added IEEE/IBM long double multilib support on PowerPC little
8380 endian Linux systems.
8381 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
8382 (MULTILIB_DEFAULTS): Likewise.
8383 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8385 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
8386 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
8387 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
8389 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
8392 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
8393 (REORDER_45): Likewise.
8395 2018-06-25 Jeff Law <law@redhat.com>
8397 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
8398 dividend to 32 bits. Adjust length.
8399 (udivmodhi4): Cleanup output template. Fix length.
8401 2018-06-25 Carl Love <cel@us.ibm.com>
8403 * config/rs6000/vsx.md: Change word selector to prefered location.
8405 2018-06-25 Richard Biener <rguenther@suse.de>
8407 PR tree-optimization/86304
8408 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
8409 epilogue-if-converted loops as well.
8411 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
8413 * lto-section-out.c (lto_begin_section): Do not print section
8414 name for noaddr and unnumbered dumps.
8416 2018-06-25 Richard Biener <rguenther@suse.de>
8418 * tree-vectorizer.h (struct vec_info_shared): New structure
8419 with parts split out from struct vec_info and loop_nest from
8420 struct _loop_vec_info.
8421 (struct vec_info): Adjust accordingly.
8422 (struct _loop_vec_info): Likewise.
8423 (LOOP_VINFO_LOOP_NEST): Adjust.
8424 (LOOP_VINFO_DATAREFS): Likewise.
8425 (LOOP_VINFO_DDRS): Likewise.
8426 (struct _bb_vec_info): Likewise.
8427 (BB_VINFO_DATAREFS): Likewise.
8428 (BB_VINFO_DDRS): Likewise.
8429 (struct _stmt_vec_info): Add dr_aux member.
8430 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
8431 (DR_MISALIGNMENT_UNINITIALIZED): New.
8432 (set_dr_misalignment): Adjust.
8433 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
8434 (vect_analyze_loop): Adjust prototype.
8435 (vect_analyze_loop_form): Likewise.
8436 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
8437 Compute dependences lazily.
8438 (vect_record_base_alignments): Use shared datarefs/ddrs.
8439 (vect_verify_datarefs_alignment): Likewise.
8440 (vect_analyze_data_refs_alignment): Likewise.
8441 (vect_analyze_data_ref_accesses): Likewise.
8442 (vect_analyze_data_refs): Likewise.
8443 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
8444 constructor parameter for shared part.
8445 (vect_analyze_loop_form): Pass in shared part and adjust.
8446 (vect_analyze_loop_2): Pass in storage for the number of
8447 stmts. Move loop nest finding to the caller. Compute
8449 (vect_analyze_loop): Pass in shared part.
8450 (vect_transform_loop): Verify shared datarefs are unchanged.
8451 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
8452 constructor parameter for shared part.
8453 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
8454 (vect_slp_bb): Verify shared datarefs are unchanged before
8456 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
8458 (new_stmt_vec_info): Initialize DR_AUX misalignment to
8459 DR_MISALIGNMENT_UNINITIALIZED.
8460 * tree-vectorizer.c (vec_info::vec_info): Add constructor
8461 parameter for shared part.
8462 (vec_info::~vec_info): Adjust.
8463 (vec_info_shared::vec_info_shared): New.
8464 (vec_info_shared::~vec_info_shared): Likewise.
8465 (vec_info_shared::save_datarefs): Likewise.
8466 (vec_info_shared::check_datarefs): Likewise.
8467 (try_vectorize_loop_1): Construct shared part live for analyses
8468 of a single loop for multiple vector sizes.
8469 * tree-parloops.c (gather_scalar_reductions): Adjust.
8471 2018-06-25 Richard Biener <rguenther@suse.de>
8473 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
8474 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
8475 (vect_analyze_data_refs): Remove similar code from here and
8476 simplify accordingly.
8478 2018-06-25 Richard Biener <rguenther@suse.de>
8480 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
8481 for reverse storage order accesses rather than asserting
8482 they cannot happen here.
8484 2018-06-25 Tom de Vries <tdevries@suse.de>
8487 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
8488 Use data16 instead of .byte for insn prefix.
8490 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
8493 * parser.c (make_char_string_pack): Pass this literal chars
8494 through cpp_interpret_string.
8495 (cp_parser_userdef_numeric_literal): Check the result of
8496 make_char_string_pack.
8498 2018-06-24 Maya Rashish <coypu@sdf.org>
8500 * ginclude/stddef.h: Simplify conditions around avoiding
8501 re-definition of __size_t.
8503 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8505 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
8506 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
8508 2018-06-22 Maya Rashish <coypu@sdf.org>
8510 * doc/invoke.texi (mno-fancy-math-387): Update for changes
8511 made to OpenBSD and NetBSD through the years.
8513 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8515 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8516 behavior of vec_pack (vector double, vector double) to match
8517 behavior of vec_float2 (vector double, vector double).
8519 2018-06-22 Olivier Hainque <hainque@adacore.com>
8521 * gimplify.c (gimplify_function_tree): Prevent creation
8522 of a trampoline for the address of the current function
8523 passed to entry/exit instrumentation hooks.
8525 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
8528 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
8531 2018-06-22 Martin Liska <mliska@suse.cz>
8533 PR tree-optimization/86263
8534 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8535 Bail out if is_enabled is false.
8536 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
8538 (jump_table_cluster::is_enabled): New function.
8540 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8542 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8543 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
8544 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
8545 (lto_input_ts_binfo_tree_pointers): Likewise.
8546 * tree-streamer-out.c (streamer_write_tree_bitfields,
8547 write_ts_binfo_tree_pointers): Likewise.
8548 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
8550 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8552 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
8554 2018-06-22 Martin Liska <mliska@suse.cz>
8556 * symbol-summary.h (get): Make it pure and inline move
8557 functionality from ::get function.
8558 (get): Remove and inline into ::get and ::get_create.
8559 (get_create): Move code from ::get function.
8561 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8564 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
8565 -x assembler-with-cpp.
8567 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8569 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
8570 _FILE_OFFSET_BITS=64 for C++.
8572 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
8574 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
8575 conversion insn that shows up when pr85657-3.c is compiled using
8576 IEEE 128-bit long double.
8577 (neg<mode>2_internal): Use the correct mode to check whether the
8578 mode is IBM extended.
8579 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
8580 multiply and divide external functions from being created more
8583 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8585 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
8587 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
8588 the edge can be forwarded.
8589 (cfg_layout_merge_blocks): Likewise.
8591 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8593 * except.c (finish_eh_generation): Commit edge insertions only after
8594 the EH edges have been redirected from post-landing to landing pads.
8596 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8598 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
8599 create_tmp_var_for to create the FRAME decl.
8600 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
8602 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8604 * tree-inline.c (copy_edges_for_bb): Minor tweak.
8605 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
8606 debug statement when resetting its value.
8607 (expand_call_inline): Copy the locus of the call onto the assignment
8608 of the return value, if any. Use local variable in more cases.
8610 2018-06-21 Martin Liska <mliska@suse.cz>
8612 * ipa-pure-const.c (propagate_nothrow): Use
8613 funct_state_summaries->get.
8614 (dump_malloc_lattice): Likewise.
8615 (propagate_malloc): Likewise.
8617 2018-06-21 Richard Biener <rguenther@suse.de>
8619 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
8620 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
8621 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
8623 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
8624 BLOCK_ABSTRACT_ORIGIN unconditionally.
8626 2018-06-21 David Malcolm <dmalcolm@redhat.com>
8628 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
8630 * ipa-reference.c (ipa_reference_c_finalize): Delete
8631 ipa_ref_opt_sum_summaries and set it to NULL.
8633 2018-06-21 Tom de Vries <tdevries@suse.de>
8635 PR tree-optimization/85859
8636 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
8637 test with comment from bb_no_side_effects_p.
8639 2018-06-21 Richard Biener <rguenther@suse.de>
8641 PR tree-optimization/86232
8642 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
8643 max for constant niter.
8645 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
8647 * config/aarch64/aarch64-simd.md
8648 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
8650 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
8652 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
8653 Make opernads of the unspec commutative.
8655 2018-06-21 Richard Biener <rguenther@suse.de>
8657 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
8658 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
8659 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
8660 (vect_analyze_data_ref_dependence): Re-order checks to deal with
8662 (vect_record_base_alignments): Do not record base alignment
8663 for gathers or scatters.
8664 (vect_compute_data_ref_alignment): Drop return value that is always
8665 true. Bail out early for gathers or scatters.
8666 (vect_enhance_data_refs_alignment): Bail out early for gathers
8668 (vect_find_same_alignment_drs): Likewise.
8669 (vect_analyze_data_refs_alignment): Remove dead code.
8670 (vect_slp_analyze_and_verify_node_alignment): Likewise.
8671 (vect_analyze_data_refs): For possible gathers or scatters do
8672 not create an alternate DR, just check their possible validity
8673 and mark them. Adjust DECL_NONALIASED handling to not rely
8675 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
8676 update inits of gathers or scatters.
8677 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
8678 Also copy gather/scatter flag to pattern vinfo.
8680 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
8682 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8683 behavior of vec_packsu (vector unsigned long long, vector unsigned
8684 long long) to match behavior of vec_packs with same signature.
8686 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
8687 Thomas Schwinge <thomas@codesourcery.com>
8688 Cesar Philippidis <cesar@codesourcery.com>
8690 * gimplify.c (gimplify_scan_omp_clauses): Add support for
8691 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8692 (gimplify_adjust_omp_clauses): Likewise.
8693 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
8694 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
8695 (gimplify_omp_target_update): Update handling of acc update and
8697 * omp-low.c (install_var_field): Remove unused parameter
8698 base_pointers_restrict.
8699 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
8700 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
8702 (omp_target_base_pointers_restrict_p): Delete.
8703 (scan_omp_target): Update call to scan_sharing_clauses.
8704 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
8706 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
8707 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8708 (convert_local_omp_clauses): Likewise.
8709 * tree-pretty-print.c (dump_omp_clause): Likewise.
8710 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
8712 (omp_clause_code_name): Likewise.
8714 2018-06-20 Jakub Jelinek <jakub@redhat.com>
8717 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
8720 PR tree-optimization/86231
8721 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
8722 anti-range don't overwrite *vr0min before using it to compute *vr0max.
8724 2018-06-20 Tom de Vries <tdevries@suse.de>
8726 PR tree-optimization/86097
8727 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
8728 iv type if signedness of iv type is not the same as that of *nit.
8730 2018-06-20 Jakub Jelinek <jakub@redhat.com>
8732 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
8733 EDGE_EH edges, verify they are all EDGE_EH.
8735 2018-06-20 Maya Rashish <coypu@sdf.org>
8737 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
8739 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8741 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
8742 * config/aarch64/aarch64.c (xgene1_tunings): Add
8743 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
8744 (aarch64_mode_valid_for_sched_fusion_p):
8745 Allow 16-byte modes.
8746 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
8747 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
8749 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
8751 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
8752 * config/aarch64/iterators.md (VQ2): New mode iterator.
8754 2018-06-20 Martin Liska <mliska@suse.cz>
8756 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
8757 Change default ratio from 10 to 8.
8759 2018-06-20 Martin Liska <mliska@suse.cz>
8761 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8763 (bit_test_cluster::find_bit_tests): Likewise.
8764 (switch_decision_tree::analyze_switch_statement): Find clusters.
8765 * tree-switch-conversion.h (struct jump_table_cluster): Document
8768 2018-06-20 Martin Liska <mliska@suse.cz>
8770 * tree-switch-conversion.c (switch_conversion::collect):
8771 Record m_uniq property.
8772 (switch_conversion::expand): Bail out for special conditions.
8773 (group_cluster::~group_cluster): New.
8774 (group_cluster::group_cluster): Likewise.
8775 (group_cluster::dump): Likewise.
8776 (jump_table_cluster::emit): New.
8777 (switch_decision_tree::fix_phi_operands_for_edges): New.
8778 (struct case_node): Remove struct.
8779 (jump_table_cluster::can_be_handled): New.
8780 (case_values_threshold): Moved to header.
8781 (reset_out_edges_aux): Likewise.
8782 (jump_table_cluster::is_beneficial): New.
8783 (bit_test_cluster::can_be_handled): Likewise.
8784 (add_case_node): Remove.
8785 (bit_test_cluster::is_beneficial): New.
8786 (case_bit_test::cmp): New.
8787 (bit_test_cluster::emit): New.
8788 (expand_switch_as_decision_tree_p): Remove.
8789 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
8790 (fix_phi_operands_for_edge): Likewise.
8791 (switch_decision_tree::analyze_switch_statement): New.
8792 (compute_cases_per_edge): Move ...
8793 (switch_decision_tree::compute_cases_per_edge): ... here.
8794 (try_switch_expansion): Likewise.
8795 (switch_decision_tree::try_switch_expansion): Likewise.
8796 (record_phi_operand_mapping): Likewise.
8797 (switch_decision_tree::record_phi_operand_mapping): Likewise.
8798 (emit_case_decision_tree): Likewise.
8799 (switch_decision_tree::emit): Likewise.
8800 (balance_case_nodes): Likewise.
8801 (switch_decision_tree::balance_case_nodes): Likewise.
8802 (dump_case_nodes): Likewise.
8803 (switch_decision_tree::dump_case_nodes): Likewise.
8804 (emit_jump): Likewise.
8805 (switch_decision_tree::emit_jump): Likewise.
8806 (emit_cmp_and_jump_insns): Likewise.
8807 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
8808 (emit_case_nodes): Likewise.
8809 (switch_decision_tree::emit_case_nodes): Likewise.
8810 (conditional_probability): Remove.
8811 * tree-switch-conversion.h (enum cluster_type): New.
8813 (struct cluster): Likewise.
8814 (cluster::cluster): Likewise.
8815 (struct simple_cluster): Likewise.
8816 (simple_cluster::simple_cluster): Likewise.
8817 (struct group_cluster): Likewise.
8818 (struct jump_table_cluster): Likewise.
8819 (struct bit_test_cluster): Likewise.
8820 (struct min_cluster_item): Likewise.
8821 (struct case_tree_node): Likewise.
8822 (case_tree_node::case_tree_node): Likewise.
8823 (jump_table_cluster::case_values_threshold): Likewise.
8824 (struct case_bit_test): Likewise.
8825 (struct switch_decision_tree): Likewise.
8826 (struct switch_conversion): Likewise.
8827 (switch_decision_tree::reset_out_edges_aux): Likewise.
8829 2018-06-20 Martin Liska <mliska@suse.cz>
8831 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
8832 (hoist_edge_and_branch_if_true): Likewise.
8833 (expand_switch_using_bit_tests_p): Likewise.
8834 (struct case_bit_test): Likewise.
8835 (case_bit_test_cmp): Likewise.
8836 (emit_case_bit_tests): Likewise.
8837 (switch_conversion::switch_conversion): New class.
8838 (struct switch_conv_info): Remove old struct.
8839 (collect_switch_conv_info): More to ...
8840 (switch_conversion::collect): ... this.
8841 (check_range): Likewise.
8842 (switch_conversion::check_range): Likewise.
8843 (check_all_empty_except_final): Likewise.
8844 (switch_conversion::check_all_empty_except_final): Likewise.
8845 (check_final_bb): Likewise.
8846 (switch_conversion::check_final_bb): Likewise.
8847 (create_temp_arrays): Likewise.
8848 (switch_conversion::create_temp_arrays): Likewise.
8849 (free_temp_arrays): Likewise.
8850 (gather_default_values): Likewise.
8851 (switch_conversion::gather_default_values): Likewise.
8852 (build_constructors): Likewise.
8853 (switch_conversion::build_constructors): Likewise.
8854 (constructor_contains_same_values_p): Likewise.
8855 (switch_conversion::contains_same_values_p): Likewise.
8856 (array_value_type): Likewise.
8857 (switch_conversion::array_value_type): Likewise.
8858 (build_one_array): Likewise.
8859 (switch_conversion::build_one_array): Likewise.
8860 (build_arrays): Likewise.
8861 (switch_conversion::build_arrays): Likewise.
8862 (gen_def_assigns): Likewise.
8863 (switch_conversion::gen_def_assigns): Likewise.
8864 (prune_bbs): Likewise.
8865 (switch_conversion::prune_bbs): Likewise.
8866 (fix_phi_nodes): Likewise.
8867 (switch_conversion::fix_phi_nodes): Likewise.
8868 (gen_inbound_check): Likewise.
8869 (switch_conversion::gen_inbound_check): Likewise.
8870 (process_switch): Use the newly created class.
8871 (switch_conversion::expand): New.
8872 (switch_conversion::~switch_conversion): New.
8873 * tree-switch-conversion.h: New file.
8875 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8877 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
8878 tree-vect-patterns.c.
8879 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
8880 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
8881 (vect_recog_sad_pattern): Likewise.
8882 (vect_recog_widen_sum_pattern): Likewise.
8883 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
8884 (vect_recog_widen_shift_pattern): Remove the type_in argument.
8885 (vect_recog_rotate_pattern): Likewise.
8886 (vect_recog_mult_pattern): Likewise.
8887 (vect_recog_vector_vector_shift_pattern): Likewise.
8888 (vect_recog_divmod_pattern): Likewise.
8889 (vect_recog_mixed_size_cond_pattern): Likewise.
8890 (vect_recog_bool_pattern): Likewise.
8891 (vect_recog_mask_conversion_pattern): Likewise.
8892 (vect_try_gather_scatter_pattern): Likewise.
8893 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
8894 (vect_recog_over_widening_pattern): Likewise.
8895 (vect_recog_gather_scatter_pattern): Likewise.
8896 (vect_recog_func_ptr): Move from tree-vectorizer.h
8897 (vect_vect_recog_func_ptrs): Move further down the file.
8898 (vect_recog_func): Likewise. Remove the third argument.
8899 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
8900 (vect_pattern_recog_1): Expect the pattern function to do any
8901 necessary target tests. Also expect it to provide a vector type.
8902 Remove the type_in handling.
8904 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8906 * tree-vect-patterns.c (vect_pattern_detected): New function.
8907 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
8908 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
8909 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
8910 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
8911 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
8912 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
8913 (vect_recog_mask_conversion_pattern)
8914 (vect_try_gather_scatter_pattern): Likewise.
8916 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8918 * tree-vect-patterns.c (vect_get_internal_def): New function.
8919 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8920 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
8921 (search_type_for_mask_1): Use it.
8923 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8925 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8926 redundant WIDEN_SUM_EXPR handling.
8927 (vect_recog_sad_pattern): Likewise.
8929 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8931 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8932 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
8933 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
8934 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
8935 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
8937 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8939 * tree-vect-stmts.c (vectorizable_call): Make sure that we
8940 use the stmt_vec_info of the original bb statement for the
8941 new zero assignment, even if the call is part of a pattern.
8943 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8945 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
8946 that the sequence is attached to the original statement rather
8947 than the pattern statement.
8948 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
8949 PATTERN_DEF_SEQ from the original statement rather than
8950 the main pattern statement.
8951 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
8952 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
8953 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
8955 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8957 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
8958 definition statements before the early exit for statements that aren't
8960 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
8962 (vect_transform_loop): ...here. Process pattern definition
8963 statements without first checking whether the main pattern
8964 statement is live or relevant.
8966 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
8968 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
8969 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
8971 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
8973 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
8974 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
8975 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
8976 (expand_block_compare): Change select_block_compare_mode call.
8977 (expand_strncmp_align_check): Use new functions, fix comment.
8978 (emit_final_str_compare_gpr): New function.
8979 (expand_strn_compare): Refactor and clean up code.
8980 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
8982 2018-06-19 Tony Reix <tony.reix@atos.com>
8983 Damien Bergamini <damien.bergamini@atos.com>
8984 David Edelsohn <dje.gcc@gmail.com>
8986 * collect2.c (static_obj): New variable.
8987 (static_libs): New variable.
8988 (is_in_list): Uncomment declaration.
8989 (main): Track AIX libraries linked statically.
8990 (is_in_list): Uncomment definition.
8991 (scan_prog_file): Don't add AIX shared libraries initializer
8992 to constructor list if linking statically.
8994 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
8996 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
8998 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
9000 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
9002 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
9005 2018-06-19 Martin Liska <mliska@suse.cz>
9007 * config/i386/i386.c (ix86_can_inline_p): Do not use
9008 ipa_fn_summaries::get_create.
9009 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
9011 (devirtualization_time_bonus): Likewise.
9012 (ipcp_propagate_stage): Likewise.
9013 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
9014 (edge_set_predicate): Likewise.
9015 (evaluate_conditions_for_known_args): Likewise.
9016 (evaluate_properties_for_edge): Likewise.
9017 (ipa_call_summary::reset): Tranform to ...
9018 (ipa_call_summary::~ipa_call_summary): ... this.
9019 (ipa_fn_summary::reset): Transform to ...
9020 (ipa_fn_summary::~ipa_fn_summary): ... this.
9021 (ipa_fn_summary_t::remove): Rename to ...
9022 (ipa_fn_summary_t::remove_callees): ... this.
9023 (ipa_fn_summary_t::duplicate): Use placement new
9024 instead of memory copy.
9025 (ipa_call_summary_t::duplicate): Likewise.
9026 (ipa_call_summary_t::remove): Remove.
9027 (dump_ipa_call_summary): Change get_create to get.
9028 (ipa_dump_fn_summary): Dump only when summary exists.
9029 (analyze_function_body): Use symbol_summary::get instead
9031 (compute_fn_summary): Likewise.
9032 (estimate_edge_devirt_benefit): Likewise.
9033 (estimate_edge_size_and_time): Likewise.
9034 (inline_update_callee_summaries): Likewise.
9035 (remap_edge_change_prob): Likewise.
9036 (remap_edge_summaries): Likewise.
9037 (ipa_merge_fn_summary_after_inlining): Likewise.
9038 (write_ipa_call_summary): Likewise.
9039 (ipa_fn_summary_write): Likewise.
9040 (ipa_free_fn_summary): Likewise.
9041 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
9042 (struct ipa_call_summary): Likewise.
9043 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
9045 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
9046 (estimate_size_after_inlining): Likewise.
9047 (estimate_growth): Likewise.
9048 (growth_likely_positive): Likewise.
9049 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
9050 (inline_call): Likewise.
9051 * ipa-inline.c (caller_growth_limits): Likewise.
9052 (can_inline_edge_p): Likewise.
9053 (can_inline_edge_by_limits_p): Likewise.
9054 (compute_uninlined_call_time): Likewise.
9055 (compute_inlined_call_time): Likewise.
9056 (want_inline_small_function_p): Likewise.
9057 (edge_badness): Likewise.
9058 (update_caller_keys): Likewise.
9059 (update_callee_keys): Likewise.
9060 (inline_small_functions): Likewise.
9061 (inline_to_all_callers_1): Likewise.
9062 (dump_overall_stats): Likewise.
9063 (early_inline_small_functions): Likewise.
9064 (early_inliner): Likewise.
9065 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
9066 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9067 * ipa-pure-const.c (malloc_candidate_p): Likewise.
9068 * ipa-split.c (execute_split_functions): Likewise.
9069 * symbol-summary.h: Likewise.
9070 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
9072 2018-06-19 Richard Biener <rguenther@suse.de>
9074 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
9075 (vectorize_loops): ... here. Fix dbgcnt handling.
9076 (try_vectorize_loop): Wrap try_vectorize_loop_1.
9078 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
9081 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
9082 ieee128 argument takes up only one (vector) register, not two (floating
9085 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
9087 * gimplify.c (gimplify_init_constructor): Really never clear for an
9088 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
9090 2018-06-19 Richard Biener <rguenther@suse.de>
9092 PR tree-optimization/86179
9093 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
9094 after failed recognition.
9096 2018-06-18 Martin Sebor <msebor@redhat.com>
9099 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
9100 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
9101 Handle integer subtraction.
9102 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
9103 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
9105 2018-06-18 David Malcolm <dmalcolm@redhat.com>
9107 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
9108 param from rtx to rtx_insn *.
9109 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
9111 (frv_ifcvt_modify_insn): Likwise.
9112 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
9113 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
9114 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
9115 as_a <rtx_insn *> cast to local "unprotected_region" once
9116 it's been established that it's not NULL or pc_rtx.
9117 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
9118 param "sethi" from rtx to rtx_insn *.
9119 (nds32_group_float_insns): Likewise for param "insn".
9120 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
9122 (vax_output_int_subtract): Likewise.
9123 * config/vax/vax.c (vax_output_int_add): Likewise for param
9125 (vax_output_int_subtract): Likewise.
9126 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
9127 (emit_pattern_after): Likewise for param "after".
9128 (emit_insn_after): Likewise.
9129 (emit_jump_insn_after): Likewise.
9130 (emit_call_insn_after): Likewise.
9131 (emit_debug_insn_after): Likewise.
9132 (emit_pattern_before): Likewise for param "before".
9133 (emit_insn_before): Likewise.
9134 (emit_jump_insn_before): Likewise.
9135 * final.c (get_insn_template): Likewise for param "insn", removing
9137 * output.h (get_insn_template): Likewise for 2nd param.
9138 * rtl.h (emit_insn_before): Likewise.
9139 (emit_jump_insn_before): Likewise.
9140 (emit_debug_insn_before_noloc): Likewise.
9141 (emit_insn_after): Likewise.
9142 (emit_jump_insn_after): Likewise.
9143 (emit_call_insn_after): Likewise.
9144 (emit_debug_insn_after): Likewise.
9145 (set_insn_deleted): Likewise for param.
9147 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
9150 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
9151 floating point modes, so that IFmode is numerically greater than
9152 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
9153 to declare the ordering. This prevents IFmode from being
9154 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
9155 machine. Include rs6000-modes.h to share the fractional values
9156 between genmodes* and the rest of the compiler.
9160 * config/rs6000/rs6000-modes.h: New file.
9161 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
9162 meaning of rs6000_long_double_size so that 126..128 selects an
9163 appropriate 128-bit floating point type.
9164 (rs6000_option_override_internal): Likewise.
9165 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
9166 (TARGET_LONG_DOUBLE_128): Change the meaning of
9167 rs6000_long_double_size so that 126..128 selects an appropriate
9168 128-bit floating point type.
9169 (LONG_DOUBLE_TYPE_SIZE): Update comment.
9170 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
9171 source and destination to match the standard usage.
9172 (truncifkf2): Likewise.
9173 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
9174 ISA 2.07 to use an explicit clobber, instead of passing in a
9176 (copysign<mode>3_soft): Likewise.
9178 2018-06-18 David Malcolm <dmalcolm@redhat.com>
9180 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
9181 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
9182 (vect_slp_analyze_instance_dependence): Likewise.
9183 (vect_enhance_data_refs_alignment): Likewise.
9184 (vect_analyze_data_refs_alignment): Likewise.
9185 (vect_slp_analyze_and_verify_instance_alignment
9186 (vect_analyze_data_ref_accesses): Likewise.
9187 (vect_prune_runtime_alias_test_list): Likewise.
9188 (vect_analyze_data_refs): Likewise.
9189 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
9190 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9191 (vect_analyze_scalar_cycles_1): Likewise.
9192 (vect_get_loop_niters): Likewise.
9193 (vect_analyze_loop_form_1): Likewise.
9194 (vect_update_vf_for_slp): Likewise.
9195 (vect_analyze_loop_operations): Likewise.
9196 (vect_analyze_loop): Likewise.
9197 (vectorizable_induction): Likewise.
9198 (vect_transform_loop): Likewise.
9199 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
9200 * tree-vect-slp.c (vect_analyze_slp): Likewise.
9201 (vect_make_slp_decision): Likewise.
9202 (vect_detect_hybrid_slp): Likewise.
9203 (vect_slp_analyze_operations): Likewise.
9204 (vect_slp_bb): Likewise.
9205 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
9206 (vectorizable_bswap): Likewise.
9207 (vectorizable_call): Likewise.
9208 (vectorizable_simd_clone_call): Likewise.
9209 (vectorizable_conversion): Likewise.
9210 (vectorizable_assignment): Likewise.
9211 (vectorizable_shift): Likewise.
9212 (vectorizable_operation): Likewise.
9213 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
9215 2018-06-18 Martin Sebor <msebor@redhat.com>
9217 PR tree-optimization/81384
9218 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
9219 * builtins.c (expand_builtin_strnlen): New function.
9220 (expand_builtin): Call it.
9221 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
9222 * builtins.def (BUILT_IN_STRNLEN): New.
9223 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
9224 Warn for bounds in excess of maximum object size.
9225 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
9226 single-value ranges. Handle strnlen.
9227 (handle_builtin_strlen): Handle strnlen.
9228 (strlen_check_and_optimize_stmt): Same.
9229 * doc/extend.texi (Other Builtins): Document strnlen.
9231 2018-06-18 Maya Rashish <coypu@sdf.org>
9233 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
9234 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
9235 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
9237 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
9239 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
9241 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9243 * tree.c (escaped_string::escape): Replace cast to char * by
9244 const_cast<char *> (unescaped).
9246 2018-06-18 Nick Clifton <nickc@redhat.com>
9249 * tree.c (escaped_string): New class. Converts an unescaped
9250 string into its escaped equivalent.
9251 (warn_deprecated_use): Use the new class to convert the
9252 deprecation message, if present.
9253 (test_escaped_strings): New self test.
9254 (test_c_tests): Add test_escaped_strings.
9255 * doc/extend.texi (deprecated): Add a note that the
9256 deprecation message is affected by the -fmessage-length
9257 option, and that control characters will be escaped.
9258 (#pragma GCC error): Document this pragma.
9259 (#pragma GCC warning): Likewise.
9260 * doc/invoke.texi (-fmessage-length): Document this option's
9261 effect on the #warning and #error preprocessor directives and
9262 the deprecated attribute.
9264 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
9266 * tree.c (decl_value_expr_lookup): Revert latest change.
9267 (decl_value_expr_insert): Likewise.
9269 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
9271 * gimplify.c (nonlocal_vlas): Delete.
9272 (nonlocal_vla_vars): Likewise.
9273 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
9275 (gimplify_body): Do not create and destroy nonlocal_vlas.
9276 * tree-nested.c: Include diagnostic.h.
9277 (use_pointer_in_frame): Tweak.
9278 (lookup_field_for_decl): Add assertion and declare the transformation.
9279 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
9280 internal error when the reference is in a wrong context. Do not
9281 create a debug decl by default.
9282 (note_nonlocal_block_vlas): Delete.
9283 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
9284 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
9285 create a debug decl by default.
9286 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
9287 call is in a wrong context.
9288 (fixup_vla_decls): New function.
9289 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
9290 debug variables were created.
9291 * tree.c (decl_value_expr_lookup): Add checking assertion.
9292 (decl_value_expr_insert): Likewise.
9294 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
9297 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
9298 * tree-scalar-evolution.c (interpret_expr): Likewise.
9299 (expression_expensive_p): Likewise.
9300 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
9301 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
9302 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
9303 (ssa_defined_by_minus_one_stmt_p): New.
9305 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
9308 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
9309 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
9310 * dojump.c (do_jump): Likewise.
9311 * expr.c (expand_expr_real_2): Check operand type's sign.
9312 * fold-const.c (const_unop): Handle ABSU_EXPR.
9313 (fold_abs_const): Likewise.
9314 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
9315 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
9316 (strip_sign_op_1): Likesise.
9317 * match.pd: Add new pattern to generate ABSU_EXPR.
9318 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
9319 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
9320 * tree-eh.c (operation_could_trap_helper_p): Likewise.
9321 * tree-inline.c (estimate_operator_cost): Likewise.
9322 * tree-pretty-print.c (dump_generic_node): Likewise.
9323 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
9324 * tree.def (ABSU_EXPR): New.
9326 2018-06-16 Jakub Jelinek <jakub@redhat.com>
9329 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
9330 documented as preserved for backward compatibility only.
9331 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
9333 PR rtl-optimization/86108
9334 * bb-reorder.c (create_forwarder_block): Renamed to ...
9335 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
9336 jump from new landing pad to the second part.
9337 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
9340 2018-06-15 Jakub Jelinek <jakub@redhat.com>
9343 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
9344 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
9345 Only call store_expr for halves if the mode is the same.
9348 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
9349 Fix up comment formatting.
9351 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
9353 * typed-splay-tree.h (typed_splay_tree::remove): New function.
9354 (typed_splay_tree::closure,
9355 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
9356 (typed_splay_tree::typed_splay_tree,
9357 typed_splay_tree::operator =): Declared private.
9358 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
9359 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
9360 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
9361 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
9362 typed_splay_tree::splay_tree_splay,
9363 typed_splay_tree::splay_tree_foreach_helper,
9364 typed_splay_tree::splay_tree_insert,
9365 typed_splay_tree::splay_tree_remove,
9366 typed_splay_tree::splay_tree_lookup,
9367 typed_splay_tree::splay_tree_predecessor,
9368 typed_splay_tree::splay_tree_successor,
9369 typed_splay_tree::splay_tree_min,
9370 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
9371 (typed_splay_tree::root, typed_splay_tree::comp,
9372 typed_splay_tree::delete_key,
9373 typed_splay_tree::delete_value): New data members.
9374 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
9375 typed_splay_tree::remove.
9377 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
9379 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
9380 -mginv and -mno-ginv to the assembler.
9381 * config/mips/mips.opt (-mcrc): New option.
9383 * doc/invoke.text (-mcrc): Document.
9386 2018-06-15 Nick Clifton <nickc@redhat.com>
9389 * tree.c (escaped_string): New class. Converts an unescaped
9390 string into its escaped equivalent.
9391 (warn_deprecated_use): Use the new class to convert the
9392 deprecation message, if present.
9393 (test_escaped_strings): New self test.
9394 (test_c_tests): Add test_escaped_strings.
9395 * doc/extend.texi (deprecated): Add a note that the
9396 deprecation message is affected by the -fmessage-length
9397 option, and that control characters will be escaped.
9398 (#pragma GCC error): Document this pragma.
9399 (#pragma GCC warning): Likewise.
9400 * doc/invoke.texi (-fmessage-length): Document this option's
9401 effect on the #warning and #error preprocessor directives and
9402 the deprecated attribute.
9404 2018-06-15 Richard Biener <rguenther@suse.de>
9406 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
9407 here, also noting vector size used.
9408 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
9409 size used in MSG_OPTIMIZED_LOCATIONS dump.
9410 (pass_slp_vectorize::execute): Adjust.
9412 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
9415 * config/arc/arc.c (arc_return_address_register): Fix
9418 2018-06-15 Richard Biener <rguenther@suse.de>
9421 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
9422 leave useless conversion stripping to force_gimple_operand_gsi.
9423 (gimplify_build2): Likewise.
9424 (gimplify_build1): Likewise.
9426 2018-06-15 Richard Biener <rguenther@suse.de>
9429 * tree-cfg.c (move_stmt_op): unshare invariant addresses
9430 before adjusting their block.
9432 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
9434 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
9435 multilibs for *-*-rtems*.
9436 * config/riscv/t-rtems: New file.
9438 2018-06-14 Jakub Jelinek <jakub@redhat.com>
9441 * match.pd ((A +- CST1) +- CST2): Punt if last resort
9442 unsigned_type_for returns NULL.
9445 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
9446 subregs of multi-word pseudos unless the float mode has word size.
9448 2018-06-14 Richard Biener <rguenther@suse.de>
9451 * tree-vect-generic.c (build_word_mode_vector_type): Remove
9452 duplicate and harmful type_hash_canon.
9453 * tree.c (type_hash_canon): Assert we didn't find ourselves.
9455 2018-06-14 Richard Biener <rguenther@suse.de>
9458 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
9461 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
9463 * config/rtems.h (STDINT_LONG32): Define.
9465 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
9466 Prachi Godbole <prachi.godbole@imgtec.com>
9468 * config/mips/mips-cpus.def: Define P6600.
9469 * config/mips/mips-tables.opt: Regenerate.
9470 * config/mips/mips.c (mips_ucbranch_type): New enum.
9471 (mips_rtx_cost_data): Add support for P6600.
9472 (mips_issue_rate): Likewise.
9473 (mips_multipass_dfa_lookahead): Likewise.
9474 (mips_avoid_hazard): Likewise.
9475 (mips_reorg_process_insns): Likewise.
9476 (mips_classify_branch_p6600): New function.
9477 * config/mips/mips.h (TUNE_P6600): New define.
9478 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
9479 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
9480 * config/mips/mips.md: Include p6600.md.
9481 (processor): Add p6600.
9482 * config/mips/p6600.md: New file.
9483 * doc/invoke.texi: Add p6600 to supported architectures.
9485 2018-06-13 Martin Sebor <msebor@redhat.com>
9487 PR tree-optimization/86114
9488 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
9490 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
9492 2018-06-13 Richard Biener <rguenther@suse.de>
9494 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
9495 Properly set vector type of the intermediate stmt.
9496 * tree-vect-stmts.c (vectorizable_operation): The destination
9497 var always has vectype_out type.
9499 2018-06-13 Jeff Law <law@redhat.com>
9501 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
9502 integer 0 for argument to print_rtl_with_bb.
9503 (rl78_reorg): Likewise.
9505 2018-06-13 David Malcolm <dmalcolm@redhat.com>
9507 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
9508 from rtx to rtx_insn *.
9509 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
9511 (add_sched_insns_for_speculation): Likewise for local "target",
9512 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9513 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
9514 from rtx_insn ** to rtx_code_label **.
9515 (reorg_emit_nops): Likewise.
9516 (c6x_reorg): Likewise for local "call_labels".
9517 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
9519 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
9520 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
9521 the loops over LABEL_REFS.
9522 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
9524 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9525 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
9526 (split_branches): Strengthen local "olabel" from rtx to
9527 rtx_insn *, adding a safe_as_a cast.
9528 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
9530 (add_insn_after): Likewise for first two params.
9531 (add_insn_before): Likewise.
9532 (remove_insn): Likewise for param.
9533 (emit_pattern_before_noloc): Likewise for second and third params.
9534 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
9535 (emit_call_insn_before_noloc): Likewise.
9536 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
9538 (emit_barrier_before): Likewise.
9539 (emit_label_before): Strengthen "label" param from "rtx" to
9540 "rtx_code_label *". Strengthen "before" param from "rtx" to
9542 (emit_insn_after_1): Strengthen "after" param from "rtx" to
9544 (emit_pattern_after_noloc): Likewise.
9545 (emit_insn_after_noloc): Likewise.
9546 (emit_jump_insn_after_noloc): Likewise.
9547 (emit_call_insn_after_noloc): Likewise.
9548 (emit_debug_insn_after_noloc): Likewise.
9549 (emit_barrier_after): Likewise.
9550 (emit_label_after): Likewise for both params.
9551 (emit_pattern_after_setloc): Likewise for "after" param. Convert
9552 "loc" param from "int" to "location_t".
9553 (emit_insn_after_setloc): Likewise.
9554 (emit_jump_insn_after_setloc): Likewise.
9555 (emit_call_insn_after_setloc): Likewise.
9556 (emit_debug_insn_after_setloc): Likewise.
9557 (emit_pattern_before_setloc): Likewise for "before" param. Convert
9558 "loc" param from "int" to "location_t".
9559 (emit_pattern_before): Convert NULL_RTX to NULL.
9560 (emit_insn_before_setloc): Convert "loc" param from "int" to
9562 (emit_jump_insn_before_setloc): Likewise.
9563 (emit_call_insn_before_setloc): Likewise.
9564 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
9565 rtx_insn *. Convert "loc" param from "int" to "location_t".
9566 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
9567 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
9568 Convert 3rd param from "int" to "location_t".
9569 (emit_barrier_before, emit_barrier_after, next_real_insn):
9570 Strengthen param from rtx to rtx_insn *.
9571 (emit_label_before): Strengthen 1st param from "rtx" to
9572 "rtx_code_label *". Strengthen 2nd param from "rtx" to
9574 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
9575 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
9576 Strengthen 2nd param from "rtx" to "rtx_insn *".
9577 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
9578 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
9579 Likewise. Convert 3rd param from "int" to "location_t".
9580 (emit_label_after): Strengthen 1st param from "rtx" to
9582 (next_real_insn, remove_insn): Strengthen param from "rtx" to
9584 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
9585 from "rtx" to "rtx_insn *".
9587 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
9589 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
9590 bodies streamed in with -Q.
9591 * dumpfile.c (dump_files): Add lto-stream-out dump file.
9592 * dumpfile.h (tree_dump_index): Add lto_stream_out.
9593 * gimple-streamer-out.c: Include gimple-pretty-print.h
9594 (output_bb): Dump stmts streamed.
9595 * lto-section-out.c: Include print-tree.h
9596 (lto_begin_section): Dump sections created.
9597 (lto_output_decl_index): Dump decl encoded.
9598 * lto-streamer-out.c: Include print-tree.h
9599 (create_output_block): Dump output block created.
9600 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
9601 (output_function): Dump function output.
9602 (output_constructor): Dump constructor streamed.
9603 (write_global_stream): Output indexes encoded.
9604 (produce_asm_for_decls): Dump streams encoded.
9605 * lto-streamer.c (streamer_dump_file): New global var.
9606 * lto-streamer.h (streamer_dump_file): Declare.
9607 * passes.c (ipa_write_summaries): Initialize streamer dump.
9608 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
9611 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
9614 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
9615 offsets for register save directives. Emit a second batch of save
9616 directives, if need be, when the function accesses prior frames.
9618 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9620 * config/arc/fpu.md (fmasf4): Force operand to register.
9621 (fnmasf4): Likewise.
9623 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9625 * config/arc/arc-protos.h (arc_pad_return): Remove.
9626 * config/arc/arc.c (machine_function): Remove force_short_suffix
9628 (arc_print_operand): Adjust printing of '&'.
9629 (arc_verify_short): Remove conditional printing of short suffix.
9630 (arc_final_prescan_insn): Remove reference to size_reason.
9631 (pad_return): New function.
9632 (arc_reorg): Call pad_return.
9633 (arc_pad_return): Remove.
9634 (arc_init_machine_status): Remove reference to force_short_suffix.
9635 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
9636 (attr length): When attribute iscompact is true force to 2
9637 regardless; in the case of maybe check if we want to force the
9638 instruction to have 4 bytes length.
9639 (nopv): Change it to generate 4 byte long nop as well.
9640 (blockage): New pattern.
9641 (simple_return): Remove call to arc_pad_return.
9642 (p_return_i): Likewise.
9644 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9646 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
9648 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9650 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
9653 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9655 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
9656 for ARC700 and ARCv2.
9658 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
9661 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
9662 operands[2] instead of operands[1].
9665 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
9667 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
9668 case, check whether the outer register overlaps an unallocatable
9669 register, not just whether it fits the required class.
9671 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
9673 * poly-int.h (can_div_trunc_p): Add new overload in which all values
9675 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
9676 (memrefs_conflict_p): Likewise.
9677 (init_alias_analysis): Likewise.
9678 * cfgexpand.c (expand_debug_expr): Likewise.
9679 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
9680 * cse.c (fold_rtx): Likewise.
9681 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
9682 * expr.c (emit_block_move_hints): Likewise.
9683 (clear_storage_hints, push_block, emit_push_insn): Likewise.
9684 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
9685 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
9686 (emit_group_store): Likewise.
9687 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
9688 to read the PRE/POST_MODIFY increment.
9689 * calls.c (store_one_arg): Use strip_offset.
9690 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
9692 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
9694 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
9695 (simplify_binary_operation_1): Extend CONST_INT handling to
9697 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
9698 than a HOST_WIDE_INT.
9699 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
9701 (adjust_mems, add_stores): Update accodingly.
9702 (vt_canonicalize_addr): Track polynomial offsets.
9703 (emit_note_insn_var_location): Likewise.
9704 (vt_add_function_parameter): Likewise.
9705 (vt_initialize): Likewise.
9707 2018-06-12 Jeff Law <law@redhat.com>
9709 * config.gcc (alpha*-*-freebsd*): Remove.
9710 * config/alpha/freebsd.h: Remove.
9712 2018-06-12 David Malcolm <dmalcolm@redhat.com>
9715 * spellcheck-tree.c (levenshtein_distance): Rename to...
9716 (get_edit_distance): ...this, and update for underlying renaming.
9717 * spellcheck-tree.h (levenshtein_distance): Rename to...
9718 (get_edit_distance): ...this.
9719 * spellcheck.c (levenshtein_distance): Rename to...
9720 (get_edit_distance): ...this. Convert from Levenshtein distance
9721 to Damerau-Levenshtein distance by supporting transpositions of
9722 adjacent characters. Rename "v1" to "v_next" and "v0" to
9724 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
9725 (selftest::test_edit_distance_unit_test_oneway): ...this, and
9726 update for underlying renaming.
9727 (selftest::levenshtein_distance_unit_test): Rename to...
9728 (selftest::test_get_edit_distance_unit): ...this, and update for
9729 underlying renaming.
9730 (selftest::test_find_closest_string): Add example from PR 69968
9731 where transposition helps
9732 (selftest::test_metric_conditions): Update for renaming.
9733 (selftest::test_metric_conditions): Likewise.
9734 (selftest::spellcheck_c_tests): Likewise.
9735 * spellcheck.h (levenshtein_distance): Rename both overloads to...
9736 (get_edit_distance): ...this.
9737 (best_match::consider): Update for renaming.
9739 2018-06-12 Martin Sebor <msebor@redhat.com>
9741 PR tree-optimization/85259
9742 * builtins.c (compute_objsize): Handle constant offsets.
9743 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
9744 true iff a warning has been issued.
9745 * gimple.h (gimple_nonartificial_location): New function.
9746 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
9747 gimple_nonartificial_location and handle -Wno-system-headers.
9748 (handle_builtin_stxncpy): Same.
9750 2018-06-12 Martin Sebor <msebor@redhat.com>
9753 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
9755 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
9757 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
9758 BUILTIN_VEC_XST entries for pointer to double and long long.
9760 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
9763 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9765 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9768 2018-06-12 Martin Liska <mliska@suse.cz>
9770 * doc/options.texi: Document IntegerRange.
9772 2018-06-12 Martin Liska <mliska@suse.cz>
9774 * config/i386/i386.opt: Make MPX-related options as Deprecated.
9775 * opt-functions.awk: Handle Deprecated flag.
9776 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
9778 (read_cmdline_option): Report warning for a deprecated option.
9779 * opts.h (struct cl_option): Add new field cl_deprecated.
9780 (CL_ERR_DEPRECATED): New.
9782 2018-06-12 Martin Liska <mliska@suse.cz>
9784 * doc/options.texi: Document Deprecated option flag.
9786 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9788 * config/arc/arc-arch.h (arc_extras): New enum.
9789 (arc_cpu_t):Add field extra.
9790 (arc_cpu_types): Consider the extras.
9791 * config/arc/arc-cpus.def: Add extras info.
9792 * config/arc/arc-opts.h (processor_type): Consider extra field.
9793 * config/arc/arc.c (arc_override_options): Handle extra field.
9795 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9797 * config/arc/arc-arch.h: Update ARC_OPTX macro.
9798 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
9800 * config/arc/arc.c (arc_init): Update pic warning.
9801 (irq_range): Update irq range parsing warnings.
9802 (arc_override_options): Update various warning messages.
9803 (arc_handle_aux_attribute): Likewise.
9805 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
9807 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
9809 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9811 * doc/sourcebuild.texi: Document usage of line number 0 in verify
9812 compiler messages directives.
9814 2018-06-12 Matthew Fortune <mfortune@gmail.com>
9816 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
9817 * config/mips/mips-tables.opt: Regenerate.
9818 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
9820 * doc/invoke.texi: Document -march=i6500.
9822 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
9824 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
9825 (i6400_gpmul): Add cpu_unit.
9826 (i6400_gpdiv): Likewise.
9827 (i6400_msa_add_d): Update reservations.
9828 (i6400_msa_int_add) Likewise.
9829 (i6400_msa_short_logic3) Likewise.
9830 (i6400_msa_short_logic2) Likewise.
9831 (i6400_msa_short_logic) Likewise.
9832 (i6400_msa_move) Likewise.
9833 (i6400_msa_cmp) Likewise.
9834 (i6400_msa_short_float2) Likewise.
9835 (i6400_msa_div_d) Likewise.
9836 (i6400_msa_long_logic1) Likewise.
9837 (i6400_msa_long_logic2) Likewise.
9838 (i6400_msa_mult) Likewise.
9839 (i6400_msa_long_float2) Likewise.
9840 (i6400_msa_long_float4) Likewise.
9841 (i6400_msa_long_float5) Likewise.
9842 (i6400_msa_long_float8) Likewise.
9843 (i6400_fpu_fadd): Include frint type.
9844 (i6400_fpu_store): New define_insn_reservation.
9845 (i6400_fpu_load): Likewise.
9846 (i6400_fpu_move): Likewise.
9847 (i6400_fpu_fcmp): Likewise.
9848 (i6400_fpu_fmadd): Likewise.
9849 (i6400_int_mult): Include imul3nc type and update reservation.
9850 (i6400_int_div): Include idiv3 type and update reservation.
9851 (i6400_int_load): Update to check type not move_type.
9852 (i6400_int_store): Likewise.
9853 (i6400_int_prefetch): Set zero latency.
9855 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
9857 * gcc.c: Document new %@{...} sequence.
9858 (LINK_COMMAND_SPEC): Use it for the -L switches.
9859 (cpp_unique_options): Use it for the -I switches.
9860 (at_file_argbuf): New global variable.
9861 (in_at_file): Likewise.
9862 (alloc_args): Create at_file_argbuf.
9863 (clear_args): Truncate at_file_argbuf.
9864 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
9865 (open_at_file): New function.
9866 (close_at_file): Likewise.
9867 (create_at_file): Delete.
9868 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
9871 (validate_switches_from_spec): Deal with %@{...} sequence.
9872 (validate_switches): Likewise.
9873 (driver::finalize): Call clear_args.
9875 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
9877 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
9879 2018-06-11 Martin Sebor <msebor@redhat.com>
9881 * doc/invoke.texi (-Wall): List -Wc++17-compat.
9882 (Wno-class-memaccess): Add @opindex.
9883 (Wno-templates, Wno-multiple-inheritance): Same.
9884 (Wno-virtual-inheritance, Wno-namespaces): Same.
9885 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
9886 (Wno-format-overflow, Wno-format-truncation): Same.
9887 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
9888 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
9889 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
9890 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
9891 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
9892 (Wno-misspelled-isr): Same.
9894 2018-06-11 Martin Sebor <msebor@redhat.com>
9896 * PR tree-optimization/86083
9897 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
9899 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
9901 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
9903 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
9906 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
9907 on the correct operand.
9908 (*movdi_internal64): Ditto.
9910 2018-06-11 Martin Liska <mliska@suse.cz>
9912 PR tree-optimization/86089
9913 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
9915 2018-06-11 Julia Koval <julia.koval@intel.com>
9917 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
9918 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
9919 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
9921 2018-06-11 Olivier Hainque <hainque@adacore.com>
9923 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
9924 for Ada with strict dwarf2.
9926 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
9929 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
9932 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
9934 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
9936 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
9938 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
9941 2018-06-08 Martin Liska <mliska@suse.cz>
9943 * tree-cfg.h (debug_function): Fix argument type to match
9946 2018-06-08 Martin Liska <mliska@suse.cz>
9948 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
9949 Remove usage of MPX-related (and removed) fields.
9950 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
9952 2018-06-08 David Malcolm <dmalcolm@redhat.com>
9954 * cfg.c (debug): Use TDF_NONE rather than 0.
9955 * cfghooks.c (debug): Likewise.
9956 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
9957 (struct dump_option_value_info): Convert to...
9958 (struct kv_pair): ...this template type.
9959 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
9961 (optinfo_verbosity_options): Likewise.
9962 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
9964 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
9965 than int for "optgroup_flags" param.
9966 (dump_generic_expr_loc): Use dump_flags_t rather than int for
9968 (dump_dec): Likewise.
9969 (dump_finish): Use TDF_NONE rather than 0.
9970 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
9971 rather than int for "optgroup_flags" param. Use TDF_NONE rather
9972 than 0. Update for change to option_ptr.
9973 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
9974 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
9975 0. Update for changes to optinfo_verbosity_options and
9977 (opt_info_switch_p): Convert optgroup_flags from int to
9979 (dump_basic_block): Use dump_flags_t rather than int
9980 for "dump_kind" param.
9981 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
9982 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
9983 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
9984 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
9985 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
9986 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
9987 TDF_NONE): Convert from macros to...
9988 (enum dump_flag): ...this new enum.
9989 (dump_flags_t): Update to use enum.
9990 (operator|, operator&, operator~, operator|=, operator&=):
9991 Implement for dump_flags_t.
9992 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
9993 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
9994 Convert from macros to...
9995 (enum optgroup_flag): ...this new enum.
9996 (optgroup_flags_t): New typedef.
9997 (operator|, operator|=): Implement for optgroup_flags_t.
9998 (struct dump_file_info): Convert field "alt_flags" to
9999 dump_flags_t. Convert field "optgroup_flags" to
10001 (dump_basic_block): Use dump_flags_t rather than int for param.
10002 (dump_generic_expr_loc): Likewise.
10003 (dump_dec): Likewise.
10004 (dump_register): Convert param "optgroup_flags" to
10006 (opt_info_enable_passes): Likewise.
10007 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
10009 * gimple-pretty-print.c (debug): Likewise.
10010 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
10011 (merged_store_group::apply_stores): Likewise.
10012 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
10013 * gimple.c (verify_gimple_pp): Likewise.
10014 * graphite-poly.c (print_pbb_body): Likewise.
10015 * passes.c (pass_manager::register_one_dump_file): Convert
10016 local "optgroup_flags" to optgroup_flags_t.
10017 * print-tree.c (print_node): Use TDF_NONE rather than 0.
10019 (debug_body): Likewise.
10020 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
10021 to optgroup_flags_t.
10022 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
10024 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
10025 (convert_mult_to_fma): Likewise.
10026 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
10027 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
10028 * tree-vect-data-refs.c (dump_lower_bound): Convert param
10029 "dump_kind" to dump_flags_t.
10031 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
10033 * config/rs6000/rs6000.c (min, max): Delete.
10035 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
10037 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
10038 -mabi=spe and -mabi=no-spe.
10040 2018-06-08 Martin Liska <mliska@suse.cz>
10042 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
10043 where we expect an existing summary.
10045 2018-06-08 Martin Liska <mliska@suse.cz>
10047 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
10048 * ipa-inline.h (estimate_edge_growth): Likewise.
10050 2018-06-08 Martin Liska <mliska@suse.cz>
10052 * cgraph.c (function_version_hasher::hash): Use
10053 cgraph_node::get_uid ().
10054 (function_version_hasher::equal):
10055 * cgraph.h (cgraph_node::get_uid): New method.
10056 * ipa-inline.c (update_caller_keys): Use
10057 cgraph_node::get_uid ().
10058 (update_callee_keys): Likewise.
10059 * ipa-utils.c (searchc): Likewise.
10060 (ipa_reduced_postorder): Likewise.
10061 * lto-cgraph.c (input_node): Likewise.
10062 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
10063 * symbol-summary.h (symtab_insertion): Likewise.
10064 (symtab_removal): Likewise.
10065 (symtab_duplication): Likewise.
10066 * tree-pretty-print.c (dump_function_header): Likewise.
10067 * tree-sra.c (convert_callers_for_node): Likewise.
10069 2018-06-08 Martin Liska <mliska@suse.cz>
10071 * cgraph.c (symbol_table::create_edge): Always assign a new
10073 (symbol_table::free_edge): Do not recycle numbers.
10074 * cgraph.h (cgraph_edge::get): New method.
10075 * symbol-summary.h (symtab_removal): Use it.
10076 (symtab_duplication): Likewise.
10077 (call_summary::hashable_uid): Remove.
10079 2018-06-08 Martin Liska <mliska@suse.cz>
10081 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
10082 (initialize_growth_caches): Remove.
10083 (free_growth_caches): Likewise.
10084 (do_estimate_edge_time): Use edge_growth_cache.
10085 (do_estimate_edge_size): Likewise.
10086 (do_estimate_edge_hints): Likewise.
10087 * ipa-inline.c (reset_edge_caches): Likewise.
10088 (recursive_inlining): Likewise.
10089 (inline_small_functions): Likewise.
10090 * ipa-inline.h (initialize_growth_caches): Remove.
10091 (estimate_edge_size): Likewise.
10092 (estimate_edge_time): Likewise.
10093 (estimate_edge_hints): Likewise.
10094 (reset_edge_growth_cache): Likewise.
10095 * symbol-summary.h (call_summary::remove): New method.
10097 2018-06-08 Martin Liska <mliska@suse.cz>
10099 * ipa-cp.c (class edge_clone_summary): New summary.
10100 (grow_edge_clone_vectors): Remove.
10101 (ipcp_edge_duplication_hook): Remove.
10102 (class edge_clone_summary_t): New call_summary class.
10103 (ipcp_edge_removal_hook): Remove.
10104 (edge_clone_summary_t::duplicate): New function.
10105 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
10106 (create_specialized_node): Likewise.
10107 (ipcp_driver): Initialize edge_clone_summaries and do not
10110 2018-06-08 Martin Liska <mliska@suse.cz>
10112 * symbol-summary.h (get): New function.
10113 (call_summary::m_initialize_when_cloning): New class member.
10115 2018-06-08 Martin Liska <mliska@suse.cz>
10117 * cgraph.c (cgraph_node::remove): Do not recycle uid.
10118 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
10119 (symbol_table::allocate_cgraph_symbol): Do not set uid.
10120 * passes.c (uid_hash_t): Record removed_nodes by their uids.
10121 (remove_cgraph_node_from_order): Use the removed_nodes set.
10122 (do_per_function_toporder): Likwise.
10123 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
10124 instead of summary_uid.
10125 (symtab_removal): Likewise.
10126 (symtab_duplication): Likewise.
10128 2018-06-08 Martin Liska <mliska@suse.cz>
10130 * ipa-cp.c (ipcp_store_bits_results): Use
10131 ipcp_transformation_sum.
10132 (ipcp_store_vr_results): Likewise.
10133 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
10135 (ipcp_transformation_initialize): ... this.
10136 (ipa_set_node_agg_value_chain):
10137 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
10138 (write_ipcp_transformation_info): Likewise.
10139 (read_ipcp_transformation_info): Likewise.
10140 (ipcp_update_bits): Likewise.
10141 (ipcp_update_vr): Likewise.
10142 (ipcp_transform_function): Likewise.
10143 * ipa-prop.h: Rename ipcp_transformation_summary to
10144 ipcp_transformation.
10145 (class ipcp_transformation_t): New function summary.
10146 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
10147 (ipa_get_agg_replacements_for_node): Likewise.
10149 2018-06-08 Martin Liska <mliska@suse.cz>
10151 * ipa-pure-const.c (struct funct_state_d): Do it class instead
10153 (class funct_state_summary_t): New function_summary class.
10154 (has_function_state): Remove.
10155 (get_function_state): Likewise.
10156 (set_function_state): Likewise.
10157 (add_new_function): Likewise.
10158 (funct_state_summary_t::insert): New function.
10159 (duplicate_node_data): Remove.
10160 (remove_node_data): Remove.
10161 (funct_state_summary_t::duplicate): New function.
10162 (register_hooks): Create new funct_state_summaries.
10163 (pure_const_generate_summary): Use it.
10164 (pure_const_write_summary): Likewise.
10165 (pure_const_read_summary): Likewise.
10166 (propagate_pure_const): Likewise.
10167 (propagate_nothrow): Likewise.
10168 (dump_malloc_lattice): Likewise.
10169 (propagate_malloc): Likewise.
10170 (execute): Do not register hooks, just remove summary
10172 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
10175 2018-06-08 Martin Liska <mliska@suse.cz>
10177 * ipa-reference.c (remove_node_data): Remove.
10178 (duplicate_node_data): Likewise.
10179 (class ipa_ref_var_info_summary_t): New class.
10180 (class ipa_ref_opt_summary_t): Likewise.
10181 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
10182 (get_reference_optimization_summary): Use
10183 ipa_ref_opt_sum_summaries.
10184 (set_reference_vars_info): Remove.
10185 (set_reference_optimization_summary): Likewise.
10186 (ipa_init): Create summaries.
10187 (init_function_info): Use function summary.
10188 (ipa_ref_opt_summary_t::duplicate): New function.
10189 (ipa_ref_opt_summary_t::remove): New function.
10190 (get_read_write_all_from_node): Fix GNU coding style.
10191 (propagate): Use function summary.
10192 (write_node_summary_p): Fix GNU coding style.
10193 (stream_out_bitmap): Likewise.
10194 (ipa_reference_read_optimization_summary): Use function summary.
10195 (ipa_reference_c_finalize): Do not release hooks.
10197 2018-06-08 Martin Liska <mliska@suse.cz>
10199 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
10200 (analyze_function_body): Extract multiple calls of get_create.
10201 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
10202 * ipa-inline.c (recursive_inlining): Use ::get method.
10203 * ipa-inline.h (estimate_edge_growth): Likewise.
10205 2018-06-08 Martin Liska <mliska@suse.cz>
10207 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
10209 (hsa_function_summary::hsa_function_summary): Use the new enum
10211 (hsa_gpu_implementation_p): Use hsa_summaries::get.
10212 * hsa-gen.c (hsa_get_host_function): Likewise.
10213 (get_brig_function_name): Likewise.
10214 * ipa-hsa.c (process_hsa_functions): Likewise.
10215 (ipa_hsa_write_summary): Likewise.
10216 * symbol-summary.h (symtab_duplication): Use ::get function/
10217 (get): New function.
10219 2018-06-08 Martin Liska <mliska@suse.cz>
10221 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
10223 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
10224 (hsa_register_kernel): Likewise.
10225 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
10226 * hsa-gen.c (hsa_get_host_function): Likewise.
10227 (get_brig_function_name): Likewise.
10228 (generate_hsa): Likewise.
10229 (pass_gen_hsail::execute): Likewise.
10230 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
10231 (devirtualization_time_bonus): Likewise.
10232 (ipcp_propagate_stage): Likewise.
10233 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
10234 (edge_set_predicate): Likewise.
10235 (evaluate_conditions_for_known_args): Likewise.
10236 (evaluate_properties_for_edge): Likewise.
10237 (ipa_fn_summary::reset): Likewise.
10238 (ipa_fn_summary_t::duplicate): Likewise.
10239 (dump_ipa_call_summary): Likewise.
10240 (ipa_dump_fn_summary): Likewise.
10241 (analyze_function_body): Likewise.
10242 (compute_fn_summary): Likewise.
10243 (estimate_edge_devirt_benefit): Likewise.
10244 (estimate_edge_size_and_time): Likewise.
10245 (estimate_calls_size_and_time): Likewise.
10246 (estimate_node_size_and_time): Likewise.
10247 (inline_update_callee_summaries): Likewise.
10248 (remap_edge_change_prob): Likewise.
10249 (remap_edge_summaries): Likewise.
10250 (ipa_merge_fn_summary_after_inlining): Likewise.
10251 (ipa_update_overall_fn_summary): Likewise.
10252 (read_ipa_call_summary): Likewise.
10253 (inline_read_section): Likewise.
10254 (write_ipa_call_summary): Likewise.
10255 (ipa_fn_summary_write): Likewise.
10256 (ipa_free_fn_summary): Likewise.
10257 * ipa-hsa.c (process_hsa_functions): Likewise.
10258 (ipa_hsa_write_summary): Likewise.
10259 (ipa_hsa_read_section): Likewise.
10260 * ipa-icf.c (sem_function::merge): Likewise.
10261 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
10262 (do_estimate_edge_time): Likewise.
10263 (estimate_size_after_inlining): Likewise.
10264 (estimate_growth): Likewise.
10265 (growth_likely_positive): Likewise.
10266 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
10267 (inline_call): Likewise.
10268 * ipa-inline.c (caller_growth_limits): Likewise.
10269 (can_inline_edge_p): Likewise.
10270 (can_inline_edge_by_limits_p): Likewise.
10271 (compute_uninlined_call_time): Likewise.
10272 (compute_inlined_call_time): Likewise.
10273 (want_inline_small_function_p): Likewise.
10274 (edge_badness): Likewise.
10275 (update_caller_keys): Likewise.
10276 (update_callee_keys): Likewise.
10277 (recursive_inlining): Likewise.
10278 (inline_small_functions): Likewise.
10279 (inline_to_all_callers_1): Likewise.
10280 (dump_overall_stats): Likewise.
10281 (early_inline_small_functions): Likewise.
10282 (early_inliner): Likewise.
10283 * ipa-inline.h (estimate_edge_growth): Likewise.
10284 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
10285 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10286 * ipa-prop.h (IPA_NODE_REF): Likewise.
10287 (IPA_EDGE_REF): Likewise.
10288 * ipa-pure-const.c (malloc_candidate_p): Likewise.
10289 (propagate_malloc): Likewise.
10290 * ipa-split.c (execute_split_functions): Likewise.
10291 * symbol-summary.h: Rename get to get_create.
10293 (get_create): Likewise.
10294 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
10296 2018-06-08 Martin Liska <mliska@suse.cz>
10298 * symbol-summary.h (release): Move definition out of class
10300 (symtab_removal): Likewise.
10301 (symtab_duplication): Likewise.
10303 2018-06-08 Martin Liska <mliska@suse.cz>
10305 * symbol-summary.h (function_summary): Move constructor
10306 implementation out of class declaration.
10307 (release): Likewise.
10308 (symtab_insertion): Likewise.
10309 (symtab_removal): Likewise.
10310 (symtab_duplication): Likewise.
10313 2018-06-08 Martin Liska <mliska@suse.cz>
10315 * Makefile.in: Remove support for MPX (macros, related functions,
10316 fields in cgraph_node, ...).
10317 * builtin-types.def (BT_BND): Likewise.
10318 (BT_FN_BND_CONST_PTR): Likewise.
10319 (BT_FN_CONST_PTR_BND): Likewise.
10320 (BT_FN_VOID_PTR_BND): Likewise.
10321 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
10322 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
10323 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
10324 (expand_builtin_mempcpy_with_bounds): Likewise.
10325 (expand_builtin_memset_with_bounds): Likewise.
10326 (expand_builtin_memset_args): Likewise.
10327 (std_expand_builtin_va_start): Likewise.
10328 (expand_builtin): Likewise.
10329 (expand_builtin_with_bounds): Likewise.
10330 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
10331 (DEF_LIB_BUILTIN_CHKP): Likewise.
10332 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
10333 (DEF_CHKP_BUILTIN): Likewise.
10334 (BUILT_IN_MEMCPY): Likewise.
10335 (BUILT_IN_MEMMOVE): Likewise.
10336 (BUILT_IN_MEMPCPY): Likewise.
10337 (BUILT_IN_MEMSET): Likewise.
10338 (BUILT_IN_STPCPY): Likewise.
10339 (BUILT_IN_STRCAT): Likewise.
10340 (BUILT_IN_STRCHR): Likewise.
10341 (BUILT_IN_STRCPY): Likewise.
10342 (BUILT_IN_STRLEN): Likewise.
10343 (BUILT_IN_MEMCPY_CHK): Likewise.
10344 (BUILT_IN_MEMMOVE_CHK): Likewise.
10345 (BUILT_IN_MEMPCPY_CHK): Likewise.
10346 (BUILT_IN_MEMSET_CHK): Likewise.
10347 (BUILT_IN_STPCPY_CHK): Likewise.
10348 (BUILT_IN_STRCAT_CHK): Likewise.
10349 (BUILT_IN_STRCPY_CHK): Likewise.
10350 * calls.c (store_bounds): Likewise.
10351 (emit_call_1): Likewise.
10352 (special_function_p): Likewise.
10353 (maybe_warn_nonstring_arg): Likewise.
10354 (initialize_argument_information): Likewise.
10355 (finalize_must_preallocate): Likewise.
10356 (compute_argument_addresses): Likewise.
10357 (expand_call): Likewise.
10358 * cfgexpand.c (expand_call_stmt): Likewise.
10359 (expand_return): Likewise.
10360 (expand_gimple_stmt_1): Likewise.
10361 (pass_expand::execute): Likewise.
10362 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
10363 (cgraph_node::remove): Likewise.
10364 (cgraph_node::dump): Likewise.
10365 (cgraph_node::verify_node): Likewise.
10366 * cgraph.h (chkp_function_instrumented_p): Likewise.
10367 (symtab_node::get_alias_target): Likewise.
10368 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
10369 (cgraph_local_p): Likewise.
10370 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
10371 (cgraph_edge::rebuild_references): Likewise.
10372 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
10373 (walk_polymorphic_call_targets): Likewise.
10374 (cgraph_node::expand_thunk): Likewise.
10375 (symbol_table::output_weakrefs): Likewise.
10376 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
10377 (ix86_handle_option): Likewise.
10378 * config/i386/constraints.md: Likewise.
10379 * config/i386/i386-builtin-types.def (BND): Likewise.
10383 * config/i386/i386-builtin.def (BDESC_END): Likewise.
10384 (BDESC_FIRST): Likewise.
10386 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
10387 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
10388 * config/i386/i386.c (enum reg_class): Likewise.
10389 (ix86_target_string): Likewise.
10390 (ix86_option_override_internal): Likewise.
10391 (ix86_conditional_register_usage): Likewise.
10392 (ix86_valid_target_attribute_inner_p): Likewise.
10393 (ix86_set_indirect_branch_type): Likewise.
10394 (ix86_set_current_function): Likewise.
10395 (ix86_function_arg_regno_p): Likewise.
10396 (init_cumulative_args): Likewise.
10397 (ix86_function_arg_advance): Likewise.
10398 (ix86_function_arg): Likewise.
10399 (ix86_pass_by_reference): Likewise.
10400 (ix86_function_value_regno_p): Likewise.
10401 (ix86_function_value_1): Likewise.
10402 (ix86_function_value_bounds): Likewise.
10403 (ix86_return_in_memory): Likewise.
10404 (ix86_setup_incoming_vararg_bounds): Likewise.
10405 (ix86_va_start): Likewise.
10406 (indirect_thunk_need_prefix): Likewise.
10407 (print_reg): Likewise.
10408 (ix86_print_operand): Likewise.
10409 (ix86_expand_call): Likewise.
10410 (ix86_output_function_return): Likewise.
10411 (reg_encoded_number): Likewise.
10412 (BDESC_VERIFYS): Likewise.
10413 (ix86_init_mpx_builtins): Likewise.
10414 (ix86_init_builtins): Likewise.
10415 (ix86_emit_cmove): Likewise.
10416 (ix86_emit_move_max): Likewise.
10417 (ix86_expand_builtin): Likewise.
10418 (ix86_builtin_mpx_function): Likewise.
10419 (ix86_get_arg_address_for_bt): Likewise.
10420 (ix86_load_bounds): Likewise.
10421 (ix86_store_bounds): Likewise.
10422 (ix86_load_returned_bounds): Likewise.
10423 (ix86_store_returned_bounds): Likewise.
10424 (ix86_class_likely_spilled_p): Likewise.
10425 (ix86_hard_regno_mode_ok): Likewise.
10426 (x86_order_regs_for_local_alloc): Likewise.
10427 (ix86_mitigate_rop): Likewise.
10428 (ix86_bnd_prefixed_insn_p): Likewise.
10429 (ix86_mpx_bound_mode): Likewise.
10430 (ix86_make_bounds_constant): Likewise.
10431 (ix86_initialize_bounds): Likewise.
10432 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
10433 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
10434 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
10435 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
10436 (TARGET_CHKP_BOUND_MODE): Likewise.
10437 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
10438 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
10439 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
10440 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
10441 * config/i386/i386.h (TARGET_MPX): Likewise.
10442 (TARGET_MPX_P): Likewise.
10443 (VALID_BND_REG_MODE): Likewise.
10444 (FIRST_BND_REG): Likewise.
10445 (LAST_BND_REG): Likewise.
10446 (enum reg_class): Likewise.
10447 (BND_REG_P): Likewise.
10448 (BND_REGNO_P): Likewise.
10449 (BNDmode): Likewise.
10450 (ADJUST_INSN_LENGTH): Likewise.
10451 * config/i386/i386.md: Likewise.
10452 * config/i386/i386.opt: Likewise.
10453 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
10454 (defined): Likewise.
10455 (LINK_MPX): Likewise.
10456 (MPX_SPEC): Likewise.
10457 (LIBMPX_SPEC): Likewise.
10458 (LIBMPXWRAPPERS_SPEC): Likewise.
10459 (CHKP_SPEC): Likewise.
10460 * config/i386/predicates.md: Likewise.
10461 * dbxout.c (dbxout_type): Likewise.
10462 * doc/extend.texi: Likewise.
10463 * doc/invoke.texi: Likewise.
10464 * doc/md.texi: Likewise.
10465 * doc/tm.texi: Likewise.
10466 * doc/tm.texi.in: Likewise.
10467 * dwarf2out.c (is_base_type): Likewise.
10468 (gen_formal_types_die): Likewise.
10469 (gen_subprogram_die): Likewise.
10470 (gen_type_die_with_usage): Likewise.
10471 (gen_decl_die): Likewise.
10472 (dwarf2out_late_global_decl): Likewise.
10473 * expr.c (expand_assignment): Likewise.
10474 (emit_storent_insn): Likewise.
10475 (store_expr_with_bounds): Likewise.
10476 (store_expr): Likewise.
10477 (expand_expr_real_1): Likewise.
10478 * expr.h (store_expr_with_bounds): Likewise.
10479 * function.c (use_register_for_decl): Likewise.
10480 (struct bounds_parm_data): Likewise.
10481 (assign_parms_augmented_arg_list): Likewise.
10482 (assign_parm_find_entry_rtl): Likewise.
10483 (assign_parm_is_stack_parm): Likewise.
10484 (assign_parm_load_bounds): Likewise.
10485 (assign_bounds): Likewise.
10486 (assign_parms): Likewise.
10487 (expand_function_start): Likewise.
10488 * gcc.c (CHKP_SPEC): Likewise.
10489 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
10490 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
10491 (wrestrict_dom_walker::check_call): Likewise.
10492 * gimple.c (gimple_build_call_from_tree): Likewise.
10493 * gimple.h (enum gf_mask): Likewise.
10494 (gimple_call_with_bounds_p): Likewise.
10495 (gimple_call_set_with_bounds): Likewise.
10496 * gimplify.c (gimplify_init_constructor): Likewise.
10497 * ipa-cp.c (initialize_node_lattices): Likewise.
10498 (propagate_constants_across_call): Likewise.
10499 (find_more_scalar_values_for_callers_subset): Likewise.
10500 * ipa-hsa.c (process_hsa_functions): Likewise.
10501 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
10502 * ipa-icf.c (sem_function::merge): Likewise.
10503 * ipa-inline.c (early_inliner): Likewise.
10504 * ipa-pure-const.c (warn_function_noreturn): Likewise.
10505 (warn_function_cold): Likewise.
10506 (propagate_pure_const): Likewise.
10507 * ipa-ref.h (enum GTY): Likewise.
10508 * ipa-split.c (find_retbnd): Likewise.
10509 (consider_split): Likewise.
10510 (split_function): Likewise.
10511 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
10512 * ipa.c (walk_polymorphic_call_targets): Likewise.
10513 (symbol_table::remove_unreachable_nodes): Likewise.
10514 (process_references): Likewise.
10515 (cgraph_build_static_cdtor_1): Likewise.
10516 * lto-cgraph.c (lto_output_node): Likewise.
10517 (output_refs): Likewise.
10518 (compute_ltrans_boundary): Likewise.
10519 (input_overwrite_node): Likewise.
10520 (input_node): Likewise.
10521 (input_cgraph_1): Likewise.
10522 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
10523 * passes.c (pass_manager::execute_early_local_passes): Likewise.
10524 (class pass_chkp_instrumentation_passes): Likewise.
10525 (make_pass_chkp_instrumentation_passes): Likewise.
10526 * passes.def: Likewise.
10527 * rtl.h (struct GTY): Likewise.
10528 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
10529 * stor-layout.c (layout_type): Likewise.
10530 * symtab.c: Likewise.
10531 * target.def: Likewise.
10532 * targhooks.c (default_chkp_bound_type): Likewise.
10533 (default_chkp_bound_mode): Likewise.
10534 (default_builtin_chkp_function): Likewise.
10535 (default_chkp_function_value_bounds): Likewise.
10536 (default_chkp_make_bounds_constant): Likewise.
10537 (default_chkp_initialize_bounds): Likewise.
10538 * targhooks.h (default_chkp_bound_type): Likewise.
10539 (default_chkp_bound_mode): Likewise.
10540 (default_builtin_chkp_function): Likewise.
10541 (default_chkp_function_value_bounds): Likewise.
10542 (default_chkp_make_bounds_constant): Likewise.
10543 (default_chkp_initialize_bounds): Likewise.
10544 * toplev.c (compile_file): Likewise.
10545 (process_options): Likewise.
10546 * tree-core.h (DEF_BUILTIN): Likewise.
10547 (DEF_BUILTIN_CHKP): Likewise.
10548 * tree-inline.c (declare_return_variable): Likewise.
10549 (remap_gimple_stmt): Likewise.
10550 (copy_bb): Likewise.
10551 (initialize_inlined_parameters): Likewise.
10552 (expand_call_inline): Likewise.
10553 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
10554 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
10555 (make_pass_ipa_chkp_produce_thunks): Likewise.
10556 (make_pass_chkp): Likewise.
10557 (make_pass_chkp_opt): Likewise.
10558 (make_pass_chkp_instrumentation_passes): Likewise.
10559 * tree-pretty-print.c (dump_generic_node): Likewise.
10560 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
10561 * tree-ssa-dce.c (propagate_necessity): Likewise.
10562 (eliminate_unnecessary_stmts): Likewise.
10563 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
10564 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
10565 * tree-ssa-sccvn.h: Likewise.
10566 * tree-ssa-strlen.c (get_string_length): Likewise.
10567 (valid_builtin_call): Likewise.
10568 (adjust_last_stmt): Likewise.
10569 (handle_builtin_strchr): Likewise.
10570 (handle_builtin_strcpy): Likewise.
10571 (handle_builtin_stxncpy): Likewise.
10572 (handle_builtin_memcpy): Likewise.
10573 (handle_builtin_strcat): Likewise.
10574 (strlen_check_and_optimize_stmt): Likewise.
10575 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
10576 * tree-streamer-in.c: Likewise.
10577 * tree-streamer.c (record_common_node): Likewise.
10578 * tree.c (tree_code_size): Likewise.
10579 (wide_int_to_tree_1): Likewise.
10580 (type_contains_placeholder_1): Likewise.
10581 (build_common_tree_nodes): Likewise.
10582 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
10583 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
10584 (POINTER_BOUNDS_P): Likewise.
10585 (BOUNDED_TYPE_P): Likewise.
10586 (BOUNDED_P): Likewise.
10587 (CALL_WITH_BOUNDS_P): Likewise.
10588 (pointer_bounds_type_node): Likewise.
10589 * value-prof.c (gimple_ic): Likewise.
10590 * var-tracking.c (vt_add_function_parameters): Likewise.
10591 * varasm.c (make_decl_rtl): Likewise.
10592 (assemble_start_function): Likewise.
10593 (output_constant): Likewise.
10594 (maybe_assemble_visibility): Likewise.
10595 * varpool.c (ctor_for_folding): Likewise.
10596 * chkp-builtins.def: Remove.
10597 * ipa-chkp.c: Remove.
10598 * ipa-chkp.h: Remove.
10599 * rtl-chkp.c: Remove.
10600 * rtl-chkp.h: Remove.
10601 * tree-chkp-opt.c: Remove.
10602 * tree-chkp.c: Remove.
10603 * tree-chkp.h: Remove.
10605 2018-06-07 Carl Love <cel@us.ibm.com>
10607 * config/rs6000/vsx.md (vextract_fp_from_shorth,
10608 vextract_fp_from_shortl): Add BE support.
10610 2018-06-07 Paul Koning <ni1d@arrl.net>
10612 * compare-elim.c (try_merge_compare): Don't merge compare if
10613 address contains a side effect.
10614 (try_eliminate_compare): Likewise.
10616 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
10618 * config.gcc: Support "tremont".
10619 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
10620 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10622 * config/i386/i386.c (m_TREMONT): Define.
10623 (processor_target_table): Add "tremont".
10624 (PTA_TREMONT): Define.
10625 (ix86_lea_outperforms): Add TARGET_TREMONT.
10626 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
10627 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
10628 and M_INTEL_GOLDMONT_PLUS.
10629 (fold_builtin_cpu): Add "tremont".
10630 (ix86_add_stmt_cost): Add TARGET_TREMONT.
10631 (ix86_option_override_internal): Add "tremont".
10632 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
10633 (processor_type): Add PROCESSOR_TREMONT.
10634 * config/i386/x86-tune.def: Add m_TREMONT.
10635 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
10637 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10639 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
10640 symbol defined for msp430i* devices to be lower case.
10642 2018-06-07 Richard Biener <rguenther@suse.de>
10644 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
10645 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
10646 Properly wrap signed arithmetic if overflow wraps.
10648 2018-06-07 Jakub Jelinek <jakub@redhat.com>
10650 PR tree-optimization/69615
10651 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
10652 of a cast from a same precision integral SSA_NAME in a bb dominated
10653 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
10654 cast to utype if rhs2 has already a compatible type.
10656 2018-06-07 Richard Biener <rguenther@suse.de>
10658 PR tree-optimization/85935
10659 * graphite-scop-detection.c (find_params_in_bb): Analyze
10660 condition operands with respect to the correct loop. Assert
10661 the analysis doesn't fail.
10663 2018-06-04 Carl Love <cel@us.ibm.com>
10665 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
10666 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
10667 as it is slightly cheaper.
10668 (first_match_or_eos_index_<mode>):
10669 Calculate index using natural element order.
10670 (first_match_index_<mode>):
10671 Calculate index using natural element order.
10672 (first_match_or_eos_index_<mode>):
10673 Calculate index using natural order.
10674 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
10675 for BE and LE modes.
10676 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
10677 P9V_BUILTIN_VCLZLSBB_V16QI.
10678 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
10681 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
10683 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
10684 indentation and line wrap for many prototypes. Add missing
10685 @smallexample directives around block of prototypes for vec_xl and
10688 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
10690 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
10691 track if we pass or return IEEE 128-bit floating point.
10692 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
10693 C++ mangling that is compatible with GCC 8.1.
10694 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
10695 (init_cumulative_args): Note if we pass or return IEEE 128-bit
10696 floating point types.
10697 (rs6000_function_arg_advance_1): Likewise.
10698 (rs6000_mangle_type): Optionally generate mangled names that match
10699 what GCC 8.1 generated for IEEE 128-bit floating point types.
10700 (rs6000_globalize_decl_name): If we have an external function that
10701 passes or returns IEEE 128-bit types, generate a weak reference
10702 from the mangled name used in GCC 8.1 to the current mangled
10704 (rs6000_init_builtins): Make __ibm128 use the long double type if
10705 long double is IBM extended double. Make __float128 use the long
10706 double type if long double is IEEE 128-bit.
10709 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
10710 macro for __ibm128 built-in functions.
10711 (PACK_IF): Add __ibm128 pack/unpack functions.
10712 (UNPACK_IF): Likewise.
10713 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
10714 enable long double built-in functions if long double is IEEE
10715 128-bit floating point.
10716 (rs6000_invalid_builtin): Update long double built-in function
10718 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
10719 functions, adjust the built-in function to use the long double
10720 built-in function if __ibm128 and long double are the same type.
10721 * doc/extend.texi (PowerPC builtins): Update documention for
10722 __builtin_{,un}pack_longdouble. Add documentation for
10723 __builtin_{,un}pack_ibm128.
10725 2018-06-06 Jim Wilson <jimw@sifive.com>
10727 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
10728 (struct machine_function): New field interrupt_mode.
10729 (riscv_handle_type_attribute): New function. Add forward declaration.
10730 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
10731 (riscv_expand_epilogue): Check interrupt_mode field.
10732 (riscv_set_current_function): Check interrupt attribute args and
10733 set interrupt_mode field.
10734 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
10735 (riscv_sret, riscv_uret): New.
10736 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
10737 new arguments to interrupt attribute.
10739 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
10742 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
10743 Don't handle -mcpu=power8 if -mpower9-vector is also used.
10745 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
10747 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
10748 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
10749 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
10750 several redundant entries.
10752 2018-06-06 David Malcolm <dmalcolm@redhat.com>
10754 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
10755 type from "rtx" to "rtx_insn *".
10756 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
10757 for local "call_insn", removing cast.
10758 (ix86_expand_call): Likewise, introducing a "call_insn" local.
10760 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
10762 PR tree-optimization/86066
10763 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
10764 for BIT_INSERT_EXPR stores.
10766 2018-06-06 Richard Biener <rguenther@suse.de>
10768 PR tree-optimization/86062
10769 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
10770 component refs ontop
10771 of to be offsetted base.
10773 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10775 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
10776 to be static and remove check on interrupt attribute name.
10778 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
10780 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
10781 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
10783 2018-06-05 Steve Ellcey <sellcey@cavium.com>
10786 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
10788 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
10789 Remove second argument, change how error is called.
10790 (aarch64_layout_arg): Remove second argument from
10791 aarch64_err_no_fpadvsimd call.
10792 (aarch64_init_cumulative_args): Ditto.
10793 (aarch64_gimplify_va_arg_expr): Ditto.
10794 * config/aarch64/aarch64.md (mov<mode>): Ditto.
10796 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
10798 * config/i386/i386.md (simple_return_indirect_internal): New expander.
10799 (*simple_return_indirect_internal<mode>): Rename from
10800 simple_return_indirect_internal. Use W mode iterator.
10801 (rstorssp): New expander.
10802 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
10803 (clrssbsy): New expander.
10804 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
10806 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
10808 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
10810 (cmse_check_pointed_object): Likewise.
10812 2018-06-05 Martin Liska <mliska@suse.cz>
10814 PR gcov-profile/47618
10815 * doc/invoke.texi: Document how -fprofile-dir format
10818 2018-06-05 Richard Biener <rguenther@suse.de>
10820 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
10821 removal pretend DOM info isn't available so we do not update
10822 it and only remove edges, not dominated blocks. Actually free
10823 DOM info in case we removed something. Remove unreachable blocks.
10824 (mfb_keep_latches): Work with either DOM info or marked backedges.
10825 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
10826 first. Mark backedges if DOM info isn't available.
10827 (Re-)compute DOM info after cleanup_control_flow_pre.
10829 2018-06-05 Richard Biener <rguenther@suse.de>
10831 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
10832 (locus_discrim_hasher::hash): Adjust.
10833 (locus_discrim_hasher::equal): Likewise.
10834 (next_discriminator_for_locus): Work on line directly.
10835 (same_line_p): Pass in expanded locus1 as well.
10836 (assign_discriminators): Avoid redundant location expansions.
10838 2018-06-05 Richard Biener <rguenther@suse.de>
10840 PR tree-optimization/86046
10841 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
10842 if required after clearing TREE_ADDRESSABLE.
10844 2018-06-05 Richard Biener <rguenther@suse.de>
10846 PR tree-optimization/86047
10847 * tree-ssa-loop.c (for_each_index): Glob handling of all
10848 decls and constants and really handle all of them.
10850 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10853 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
10854 qualifier_void_pointer and qualifier_const_void_pointer.
10855 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
10856 (arm_init_builtins): Handle the above.
10857 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
10858 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
10859 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
10862 2018-06-05 Martin Liska <mliska@suse.cz>
10864 * auto-profile.c (read_autofdo_file): Do not use
10865 gcov_ctr_summary struct.
10866 (afdo_callsite_hot_enough_for_early_inline): Likewise.
10867 * coverage.c (struct counts_entry): Likewise.
10868 (read_counts_file): Read just single summary entry.
10869 (get_coverage_counts): Use gcov_summary struct.
10870 * coverage.h (get_coverage_counts): Likewise.
10871 * gcov-dump.c (dump_working_sets): Likewise.
10872 (tag_summary): Dump just single summary.
10873 * gcov-io.c (gcov_write_summary): Write just histogram
10875 (gcov_read_summary): Read just single summary.
10876 (compute_working_sets): Use gcov_summary struct.
10877 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
10878 of GCOV_COUNTERS_SUMMABLE.
10879 (GCOV_COUNTERS_SUMMABLE): Remove.
10880 (GCOV_FIRST_VALUE_COUNTER): Replace with
10881 GCOV_COUNTER_V_INTERVAL.
10882 (struct gcov_ctr_summary): Remove.
10883 (struct gcov_summary): Directly use fields of former
10885 (compute_working_sets): Use gcov_summary struct.
10886 * gcov.c (read_count_file): Do not use ctrs fields.
10887 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
10889 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
10891 * profile.c: Likewise.
10892 * profile.h: Likewise.
10894 2018-06-05 Martin Liska <mliska@suse.cz>
10896 PR gcov-profile/84846
10897 * gcov.c (output_lines): Print working directory only
10898 in intermediate format.
10900 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
10902 * config/s390/s390-builtin-types.def: Add void function type.
10903 * config/s390/s390-builtins.def: Use the function type for the
10906 2018-06-04 Jim Wilson <jimw@sifive.com>
10908 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
10910 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
10911 and maybe_eh_return. Change regno to unsigned int. Use new args to
10912 handle EH_RETURN_DATA_REGNO registers properly.
10913 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
10914 (riscv_expand_epilogue): Update comment. Change argument name and
10915 type. Update code to use new name and type. Pass new args to
10916 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
10918 * config/riscv/riscv.md (NORMAL_RETURN): New.
10919 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
10920 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
10921 (eh_return): Call gen_eh_return_internal and emit barrier.
10922 (eh_return_internal): Call riscv_expand_epilogue.
10924 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
10926 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
10927 bit_insertion field and declare can_be_merged_into method.
10928 (merged_store_group::can_be_merged_into): New method.
10929 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
10930 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
10931 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
10933 2018-06-04 Richard Biener <rguenther@suse.de>
10935 PR tree-optimization/85955
10936 * builtins.c (fold_builtin_sincos): Convert pointers to
10937 destination to appropriate type before dereferencing.
10939 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
10941 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
10943 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
10945 * expr.c (expand_expr_real_1): Force the operand into memory if
10946 its TYPE_MODE is BLKmode and if there is no integer mode for
10947 the number of bits being extracted.
10949 2018-06-04 Jakub Jelinek <jakub@redhat.com>
10953 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
10954 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
10956 2018-06-04 Richard Biener <rguenther@suse.de>
10958 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
10959 (cleanup_tree_cfg_noloop): ... single caller. Do
10960 start_recording_case_labels later.
10962 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
10964 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
10965 to _IMMINTRIN_H_INCLUDED.
10966 * config/i386/pconfigintrin.h: Ditto.
10967 * config/i386/waitpkgintrin.h: Ditto.
10968 * config/i386/immintrin.h: Add includes for sgxintrin.h,
10969 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
10970 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
10971 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
10972 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
10973 waitpkgintrin.h and cldemoteintrin.h.
10975 2018-06-04 Richard Biener <rguenther@suse.de>
10977 PR tree-optimization/86038
10978 * tracer.c (find_best_successor): Check probability for
10979 being initialized, bail out if not.
10981 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
10984 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
10985 of bits to ignore when comparing architectures.
10987 2018-06-04 Jakub Jelinek <jakub@redhat.com>
10989 PR tree-optimization/69615
10990 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
10991 maximum or minimum of the type, try to merge it also as if
10992 range1 is + [-, x - 1] or + [x + 1, -].
10995 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
10997 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
10999 PR tree-optimization/86034
11000 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
11001 the unsigned bitfield type in a bit insertion sequence if it does not
11002 have a larger precision than the bitfield size.
11003 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
11005 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
11007 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
11009 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
11011 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
11012 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
11013 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
11014 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
11016 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
11018 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
11019 Disable -fdelete-null-pointer-checks for ELF toolchain.
11021 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
11022 Kito Cheng <kito.cheng@gmail.com>
11024 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
11025 (nds32le-*-*, nds32be-*-*): Integrate checking process.
11026 (nds32*-*-*): Add glibc and uclibc conditions.
11027 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
11028 (TARGET_EXCEPT_UNWIND_INFO): Define.
11029 * config/nds32/elf.h: New file.
11030 * config/nds32/linux.h: New file.
11031 * config/nds32/nds32-elf.opt: New file.
11032 * config/nds32/nds32-linux.opt: New file.
11033 * config/nds32/nds32-fp-as-gp.c
11034 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
11035 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
11037 (nds32_asm_file_end): Ditto.
11038 (nds32_print_operand): Ditto.
11039 (nds32_insert_attributes): Ditto.
11040 (nds32_init_libfuncs): New function.
11041 (TARGET_HAVE_TLS): Define.
11042 (TARGET_INIT_LIBFUNCS): Define.
11043 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
11045 (TARGET_ELF): Apply different mcmodel setting.
11046 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
11047 been migrated into elf.h and linux.h files.
11048 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
11049 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
11050 (mcmodel): The content has been migrated into nds32-elf.opt and
11051 nds32-linux.opt files.
11052 * config/nds32/t-elf: New file.
11053 * config/nds32/t-linux: New file.
11055 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
11056 Shiva Chen <shiva0217@gmail.com>
11058 * config/nds32/constants.md (unspec_volatile_element): Add
11059 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
11060 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
11062 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
11063 (make_pass_nds32_fp_as_gp): Declare.
11064 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
11066 (nds32_asm_function_end_prologue): Remove unused asm output.
11067 (nds32_asm_function_begin_epilogue): Remove unused asm output.
11068 (nds32_asm_file_start): Output necessary fp_as_gp information.
11069 (nds32_option_override): Adjust register usage.
11070 (nds32_expand_prologue): Consider fp_as_gp situation.
11071 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
11072 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
11075 (simple_return): Ditto.
11076 (omit_fp_begin): Output special directive for fp_as_gp.
11077 (omit_fp_end): Output special directive for fp_as_gp.
11078 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
11079 mforbid-fp-as-gp): New options.
11081 2018-06-01 Mark Wielaard <mark@klomp.org>
11083 * dwarf2out.c (dwarf2out_finish): Remove generation of
11084 DW_AT_loclists_base.
11086 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
11088 * gimple-ssa-store-merging.c: Include gimple-fold.h.
11089 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
11090 (struct merged_store_group): Add bit_insertion field.
11091 (dump_char_array): Use standard hexadecimal format.
11092 (merged_store_group::merged_store_group): Set bit_insertion to false.
11093 (merged_store_group::apply_stores): Use optimal buffer size. Deal
11094 with BIT_INSERT_EXPR stores. Move up code updating the mask and
11095 also print the mask in the dump file.
11096 (pass_store_merging::gate): Minor tweak.
11097 (imm_store_chain_info::coalesce_immediate): Fix wrong association
11098 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
11099 stores with INTEGER_CST stores.
11100 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
11101 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
11102 and use it throughout. Generate bit insertion sequences if need be.
11103 (pass_store_merging::process_store): Remove redundant condition.
11104 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
11106 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
11108 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
11109 the 128-bit floating point types. Fix function comment.
11111 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11113 * config/aarch64/aarch64-simd.md
11114 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
11116 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
11119 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
11121 PR tree-optimization/85989
11122 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
11124 (backprop::intersect_uses): Check it when deciding whether this
11125 is a backedge reference.
11126 (backprop::process_block): Add each phi to m_visited_phis
11127 after visiting it, then clear it at the end.
11129 2018-06-01 Richard Biener <rguenther@suse.de>
11131 * tree-vectorizer.h (vect_dr_stmt): New function.
11132 (vect_get_load_cost): Adjust.
11133 (vect_get_store_cost): Likewise.
11134 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
11135 Use vect_dr_stmt instead of DR_SMTT.
11136 (vect_record_base_alignments): Likewise.
11137 (vect_calculate_target_alignment): Likewise.
11138 (vect_compute_data_ref_alignment): Likewise and make static.
11139 (vect_update_misalignment_for_peel): Likewise.
11140 (vect_verify_datarefs_alignment): Likewise.
11141 (vector_alignment_reachable_p): Likewise.
11142 (vect_get_data_access_cost): Likewise. Pass down
11143 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
11144 (vect_get_peeling_costs_all_drs): Likewise.
11145 (vect_peeling_hash_get_lowest_cost): Likewise.
11146 (vect_enhance_data_refs_alignment): Likewise.
11147 (vect_find_same_alignment_drs): Likewise.
11148 (vect_analyze_data_refs_alignment): Likewise.
11149 (vect_analyze_group_access_1): Likewise.
11150 (vect_analyze_group_access): Likewise.
11151 (vect_analyze_data_ref_access): Likewise.
11152 (vect_analyze_data_ref_accesses): Likewise.
11153 (vect_vfa_segment_size): Likewise.
11154 (vect_small_gap_p): Likewise.
11155 (vectorizable_with_step_bound_p): Likewise.
11156 (vect_prune_runtime_alias_test_list): Likewise.
11157 (vect_analyze_data_refs): Likewise.
11158 (vect_supportable_dr_alignment): Likewise.
11159 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
11160 (vect_gen_prolog_loop_niters): Likewise.
11161 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11162 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
11164 (vect_recog_mask_conversion_pattern): Likewise.
11165 (vect_try_gather_scatter_pattern): Likewise.
11166 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
11167 to vect_get_store_cost.
11168 (vect_get_store_cost): Get stmt_info instead of DR.
11169 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
11170 (vect_get_load_cost): Get stmt_info instead of DR.
11172 2018-06-01 Richard Biener <rguenther@suse.de>
11174 PR middle-end/86017
11175 * gimple-fold.c (var_decl_component_p): Also allow offsetted
11176 vars wrapped in MEM_REFs.
11178 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
11180 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
11181 Fix subreg tests so that we only return a choice between
11182 GENERAL_REGS and FP_REGS if the original classes included both.
11184 2018-06-01 Richard Biener <rguenther@suse.de>
11187 * tree-ssa-structalias.c (get_function_part_constraint):
11188 Handle NULL fi->decl.
11189 (find_func_aliases_for_call): Properly handle indirect
11190 fi from direct call.
11191 (find_func_clobbers): Likewise.
11192 (ipa_pta_execute): Likewise.
11193 (create_variable_info_for): For functions that are ifunc_resolver
11194 resolve to a varinfo that contains the result of the resolver call.
11195 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
11198 2018-05-31 Michael Collison <michael.collison@arm.com>
11200 * config/aarch64/aarch64.md:
11201 (*fix_to_zero_extenddfdi2): New pattern.
11202 * gcc.target/aarch64/fix_extend1.c: New testcase.
11204 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
11206 PR middle-end/78809
11207 PR middle-end/83026
11208 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
11209 and BUILT_IN_STRNCMP_EQ.
11210 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
11211 BUILT_IN_STRNCMP_EQ.
11212 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
11213 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
11214 (gimple_fold_builtin): Likewise.
11215 * tree-ssa-strlen.c (compute_string_length): New function.
11216 (determine_min_obsize): New function.
11217 (handle_builtin_string_cmp): New function to handle calls to
11218 string compare functions.
11219 (strlen_optimize_stmt): Add handling to builtin string compare
11221 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
11222 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
11223 * tree.c (build_common_builtin_nodes): Add new defines of
11224 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
11226 2018-05-31 Jakub Jelinek <jakub@redhat.com>
11229 * bb-reorder.c (pass_partition_blocks::gate): Return false for
11230 functions with naked attribute.
11232 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
11234 * config/i386/sse.md (avx_vec_concat<mode>):
11235 Substitute concat_tg_mode mode attribute with xtg_mode.
11236 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
11237 (concat_tg_mode): Remove mode attribute.
11239 2018-05-31 Martin Sebor <msebor@redhat.com>
11242 * calls.c (alloc_max_size): Correct a logic error/typo.
11243 Treat excessive arguments as infinite. Warn for invalid arguments.
11244 * doc/invoke.texi (-Walloc-size-larger-than): Update.
11246 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
11249 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
11250 and movx for Haswell.
11252 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
11253 Cesar Philippidis <cesar@codesourcery.com>
11255 PR middle-end/85879
11256 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
11257 when emitting error on private/firstprivate reductions.
11258 * omp-low.c (lower_omp_target): Avoid reference-type processing
11259 on pointers for firstprivate clause.
11261 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
11263 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
11266 * config/aarch64/aarch64-simd.md
11267 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
11268 (aarch64_ld1_x3_<mode>): Likewise
11269 (aarch64_st1x2<VALLDIF:mode>): Likewise
11270 (aarch64_st1_x2_<mode>): Likewise
11271 (aarch64_st1x3<VALLDIF:mode>): Likewise
11272 (aarch64_st1_x3_<mode>): Likewise
11273 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
11274 (vld1_s8_x3): Likewise.
11275 (vld1_u16_x3): Likewise.
11276 (vld1_s16_x3): Likewise.
11277 (vld1_u32_x3): Likewise.
11278 (vld1_s32_x3): Likewise.
11279 (vld1_u64_x3): Likewise.
11280 (vld1_s64_x3): Likewise.
11281 (vld1_f16_x3): Likewise.
11282 (vld1_f32_x3): Likewise.
11283 (vld1_f64_x3): Likewise.
11284 (vld1_p8_x3): Likewise.
11285 (vld1_p16_x3): Likewise.
11286 (vld1_p64_x3): Likewise.
11287 (vld1q_u8_x3): Likewise.
11288 (vld1q_s8_x3): Likewise.
11289 (vld1q_u16_x3): Likewise.
11290 (vld1q_s16_x3): Likewise.
11291 (vld1q_u32_x3): Likewise.
11292 (vld1q_s32_x3): Likewise.
11293 (vld1q_u64_x3): Likewise.
11294 (vld1q_s64_x3): Likewise.
11295 (vld1q_f16_x3): Likewise.
11296 (vld1q_f32_x3): Likewise.
11297 (vld1q_f64_x3): Likewise.
11298 (vld1q_p8_x3): Likewise.
11299 (vld1q_p16_x3): Likewise.
11300 (vld1q_p64_x3): Likewise.
11301 (vst1_s64_x2): Likewise.
11302 (vst1_u64_x2): Likewise.
11303 (vst1_f64_x2): Likewise.
11304 (vst1_s8_x2): Likewise.
11305 (vst1_p8_x2): Likewise.
11306 (vst1_s16_x2): Likewise.
11307 (vst1_p16_x2): Likewise.
11308 (vst1_s32_x2): Likewise.
11309 (vst1_u8_x2): Likewise.
11310 (vst1_u16_x2): Likewise.
11311 (vst1_u32_x2): Likewise.
11312 (vst1_f16_x2): Likewise.
11313 (vst1_f32_x2): Likewise.
11314 (vst1_p64_x2): Likewise.
11315 (vst1q_s8_x2): Likewise.
11316 (vst1q_p8_x2): Likewise.
11317 (vst1q_s16_x2): Likewise.
11318 (vst1q_p16_x2): Likewise.
11319 (vst1q_s32_x2): Likewise.
11320 (vst1q_s64_x2): Likewise.
11321 (vst1q_u8_x2): Likewise.
11322 (vst1q_u16_x2): Likewise.
11323 (vst1q_u32_x2): Likewise.
11324 (vst1q_u64_x2): Likewise.
11325 (vst1q_f16_x2): Likewise.
11326 (vst1q_f32_x2): Likewise.
11327 (vst1q_f64_x2): Likewise.
11328 (vst1q_p64_x2): Likewise.
11329 (vst1_s64_x3): Likewise.
11330 (vst1_u64_x3): Likewise.
11331 (vst1_f64_x3): Likewise.
11332 (vst1_s8_x3): Likewise.
11333 (vst1_p8_x3): Likewise.
11334 (vst1_s16_x3): Likewise.
11335 (vst1_p16_x3): Likewise.
11336 (vst1_s32_x3): Likewise.
11337 (vst1_u8_x3): Likewise.
11338 (vst1_u16_x3): Likewise.
11339 (vst1_u32_x3): Likewise.
11340 (vst1_f16_x3): Likewise.
11341 (vst1_f32_x3): Likewise.
11342 (vst1_p64_x3): Likewise.
11343 (vst1q_s8_x3): Likewise.
11344 (vst1q_p8_x3): Likewise.
11345 (vst1q_s16_x3): Likewise.
11346 (vst1q_p16_x3): Likewise.
11347 (vst1q_s32_x3): Likewise.
11348 (vst1q_s64_x3): Likewise.
11349 (vst1q_u8_x3): Likewise.
11350 (vst1q_u16_x3): Likewise.
11351 (vst1q_u32_x3): Likewise.
11352 (vst1q_u64_x3): Likewise.
11353 (vst1q_f16_x3): Likewise.
11354 (vst1q_f32_x3): Likewise.
11355 (vst1q_f64_x3): Likewise.
11356 (vst1q_p64_x3): Likewise.
11358 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11360 * config/msp430/msp430.c (msp430_output_labelref): Prepend
11361 user_label_prefix to name.
11363 * tree-core.h: Update comment about the format of NAME string
11364 passed to handler in attribute_spec.
11366 * config/msp430/msp430.md: Remove erroneous subreg expression from
11367 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
11368 zero_extend{q,h}isi2.
11370 2018-05-30 Borislav Petkov <bp@suse.de>
11372 * doc/extend.texi: Document some architecture specific
11373 constraints and sort entries.
11375 2018-05-30 Martin Sebor <msebor@redhat.com>
11377 PR middle-end/85369
11378 * builtins.c (expand_builtin_stpcpy_1): New function.
11379 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
11380 only if the former succeeds.
11382 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
11384 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
11387 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11389 * doc/invoke.texi (-flinker-output): Document
11391 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11393 * passes.c (ipa_write_summaries): Only modify statements if body
11395 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
11396 incrementally linking.
11397 (ipa_passes): Likewise.
11398 * lto-cgraph.c (lto_output_node): When incrementally linking do not
11399 pass down resolution info.
11400 * common.opt (flag_incremental_link): Update info.
11401 * gcc.c (plugin specs): Turn flinker-output=* to
11402 -plugin-opt=-linker-output-known
11403 * toplev.c (compile_file): Also cut compilation when doing incremental
11405 * flag-types. (enum lto_partition_model): Add
11406 LTO_LINKER_OUTPUT_NOLTOREL.
11407 (invoke.texi): Add -flinker-output docs.
11408 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
11409 link same way as WPA; do not stream in dead initializers.
11411 * dwarf2out.c (dwarf2out_die_ref_for_decl,
11412 darf2out_register_external_decl): Support incremental link.
11414 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11416 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
11418 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11420 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
11421 it down to simple_object_copy_lto_debug_sections.
11422 (run_gcc): Determine incremental LTO link time and configure
11423 lto1 into non-wpa mode, disable renaming of debug sections.
11425 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
11427 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
11428 descriptions of various incorrectly documented functions.
11430 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
11433 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11434 address check not strict.
11436 2018-05-30 Richard Biener <rguenther@suse.de>
11438 PR tree-optimization/85964
11439 * tracer.c (better_p): Drop initialized count check, we only
11440 call the function with initialized counts now.
11441 (find_best_successor): Do find a best edge if one
11442 has uninitialized count.
11443 (find_best_predecessor): Likewise. Do BB frequency check only
11444 if count is initialized.
11446 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
11448 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
11449 (aarch64_ldrstr_offset_compare): New.
11450 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
11451 load/store orderings.
11452 (aarch64_gen_adjusted_ldpstp): Likewise.
11454 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
11456 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
11457 Check for subset of GENERAL_REGS and FP_REGS.
11458 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
11461 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
11463 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
11464 and wi::to_poly_offset. Add the current offset and then check
11465 whether the sum fits, rather than using an unchecked addition of
11466 a checked term. Check for a shwi rather than a uhwi.
11467 * expr.c (get_bit_range): Use tree_to_poly_uint64.
11468 (store_constructor): Use poly_int_tree_p.
11469 (expand_expr_real_1): Likewise.
11470 * function.c (assign_temp): Likewise.
11471 * fold-const.c (const_binop): Use poly_int_tree_p and
11472 wi::to_poly_offset.
11473 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
11475 * ipa-icf-gimple.c (func_checker::compare_operand): Use
11476 to_poly_offset for MEM offsets.
11477 * ipa-icf.c (sem_variable::equals): Likewise.
11478 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
11479 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
11480 wi::to_poly_offset for BIT_FIELD_REF offsets.
11481 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
11482 wi::to_poly_offset.
11483 * var-tracking.c (emit_note_insn_var_location): Use
11484 tree_to_poly_uint64.
11486 2018-05-29 Jim Wilson <jimw@sifive.com>
11488 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
11490 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
11493 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
11494 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
11496 (sse4_1_round<mode>2): Use nonimmediate_operand
11497 for operand 1 predicate.
11499 2018-05-29 Martin Sebor <msebor@redhat.com>
11500 Richard Biener <rguenther@suse.de>
11503 * calls.c (get_size_range): Call determine_value_range instead
11504 of get_value_range..
11505 * tree-vrp.h (determine_value_range): Declared new function.
11506 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
11508 2018-05-29 Richard Biener <rguenther@suse.de>
11510 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
11511 sure to use non-pattern stmts for get_earlier_stmt arguments.
11512 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
11513 called on pattern stmts.
11514 (get_later_stmt): Likewise.
11516 2018-05-29 Martin Liska <mliska@suse.cz>
11518 PR gcov-profile/85759
11519 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
11522 2018-05-29 Jakub Jelinek <jakub@redhat.com>
11524 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
11526 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
11530 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11531 VEC_PACK_FLOAT_EXPR): New tree codes.
11532 * tree-pretty-print.c (op_code_prio): Handle
11533 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11534 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11535 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11536 * tree-inline.c (estimate_operator_cost): Likewise.
11537 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
11538 * fold-const.c (const_binop): Likewise.
11539 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
11540 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11541 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11542 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
11543 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11544 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11545 * expr.c (expand_expr_real_2): Likewise.
11546 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
11547 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
11548 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
11550 * optabs.c (expand_widen_pattern_expr): For
11551 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
11552 sign from result type rather than operand's type.
11553 (expand_binop_directly): For vec_packu_float_optab and
11554 vec_packs_float_optab allow result type to be different from operand's
11556 * optabs-tree.c (optab_for_tree_code): Handle
11557 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11558 VEC_PACK_FLOAT_EXPR. Formatting fixes.
11559 * tree-vect-generic.c (expand_vector_operations_1): Handle
11560 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11561 VEC_PACK_FLOAT_EXPR.
11562 * tree-vect-stmts.c (supportable_widening_operation): Handle
11564 (supportable_narrowing_operation): Handle FLOAT_EXPR.
11565 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
11566 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
11567 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
11568 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
11570 (vec_pack<floatprefix>_float_<mode>): New expander.
11571 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
11573 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
11574 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
11575 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
11576 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
11577 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
11579 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
11580 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
11581 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11582 VEC_PACK_FLOAT_EXPR): Document.
11584 2018-05-29 Richard Biener <rguenther@suse.de>
11586 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
11588 (stmt_vec_info_vec): Make pointer.
11589 (init_stmt_vec_info_vec): Remove.
11590 (free_stmt_vec_info_vec): Likewise.
11591 (set_stmt_vec_info_vec): New function.
11592 (free_stmt_vec_infos): Likewise.
11593 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
11594 (set_vinfo_for_stmt): Likewise.
11595 (get_earlier_stmt): Likewise.
11596 (get_later_stmt): Likewise.
11597 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
11598 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
11599 (vec_info::~vec_info): Free stmt_vec_infos.
11600 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
11601 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
11602 (pass_slp_vectorize::execute): Likewise.
11603 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
11604 (free_stmt_vec_info_vec): Likewise.
11605 (set_stmt_vec_info_vec): New function.
11606 (free_stmt_vec_infos): Likewise.
11607 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
11608 the global stmt_vec_info_vec.
11609 * tree-parloops.c (gather_scalar_reductions): Use
11610 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
11613 2018-05-29 Richard Biener <rguenther@suse.de>
11615 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
11617 2018-05-29 Martin Liska <mliska@suse.cz>
11618 David Malcolm <dmalcolm@redhat.com>
11620 * vec.c (test_reverse): New.
11621 (vec_c_tests): Add new test.
11622 * vec.h (vl_ptr>::reverse): New function.
11624 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
11626 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
11628 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
11631 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
11633 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
11635 2018-05-28 Richard Biener <rguenther@suse.de>
11637 PR tree-optimization/85933
11638 * tree-vect-data-refs.c (vect_record_base_alignments): Only
11639 look at stmts marked as vectorizable.
11641 2018-05-28 Richard Biener <rguenther@suse.de>
11643 PR tree-optimization/85934
11644 * tree-vect-generic.c (expand_vector_operations_1): Hoist
11645 vector boolean check before scalar optimization.
11647 2018-05-28 Jakub Jelinek <jakub@redhat.com>
11649 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
11652 2018-05-28 Mark Wielaard <mark@klomp.org>
11654 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
11655 if it is an expression containing a minus sign.
11657 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
11659 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
11661 2018-05-27 Paul Koning <ni1d@arrl.net>
11663 * config/pdp11/pdp11.md (truncsihi2): Remove.
11665 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
11666 Chung-Ju Wu <jasonwucj@gmail.com>
11668 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
11670 (unaligned_store_dw): Ditto.
11671 * config/nds32/nds32-memory-manipulation.c
11672 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
11673 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
11674 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
11675 (emit_setmem_word_loop): Rename to ...
11676 (emit_setmem_doubleword_loop): ... this.
11677 (nds32_gen_dup_4_byte_to_word_value): New function.
11678 (nds32_gen_dup_8_byte_to_double_word_value): New function.
11679 (nds32_expand_setmem_loop): Refine implementation.
11680 (nds32_expand_setmem_loop_v3m): Ditto.
11681 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
11684 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
11686 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
11688 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
11690 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
11691 (nds32_init_machine_status): Initialize machine->attr_naked_p and
11692 machine->attr_no_prologue_p.
11693 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
11694 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
11695 (nds32_expand_epilogue): Consider attr_naked_p.
11696 (nds32_expand_epilogue_v3pop): Likewise.
11697 (nds32_can_use_return_insn): Likewise.
11698 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
11699 attr_no_prologue_p fields.
11700 * config/nds32/nds32.opt (mret-in-naked-func): New option.
11702 2018-05-27 Jakub Jelinek <jakub@redhat.com>
11705 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
11707 * config/i386/sse.md
11708 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
11710 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
11712 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
11714 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
11716 (*<floatsuffix>floatv2div2sf2): Rename to ...
11717 (*float<floatunssuffix>v2div2sf2): ... this.
11718 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
11719 (float<floatunssuffix>v2div2sf2_mask): ... this.
11720 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
11721 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
11722 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
11724 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
11726 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11728 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11730 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11732 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11734 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
11735 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
11736 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
11737 gen_ufix_truncv8dfv8si2.
11738 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
11739 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
11740 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
11741 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
11742 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
11743 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
11744 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
11745 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
11747 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
11751 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
11753 2018-05-25 Jim Wilson <jimw@sifive.com>
11755 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
11756 * config/riscv/riscv.c (struct machine_function): Add
11757 interrupt_handler_p and attribute_checked_p fields.
11758 (riscv_attribute_table): Add interrupt.
11759 (riscv_interrupt_type_p): New.
11760 (riscv_save_reg_p): Save extra regs for interrupt handler.
11761 (riscv_use_save_libcall): Return false for interrupt handler.
11762 (riscv_first_stack_step): Add forward declaration.
11763 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
11764 for interrupt handler with large frame. Use it for saved reg list.
11765 (riscv_expand_prologue): Move flag_stack_usage_info support to
11766 eliminate duplication.
11767 (riscv_expand_epilogue): Generate mret for interrupt handler.
11768 (riscv_epilogue_uses): New.
11769 (riscv_can_use_return_insn): Return false for interrupt handler.
11770 (riscv_function_ok_for_sibcall): Likewise.
11771 (riscv_set_current_function): Add interrupt handler support.
11772 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
11773 * config/riscv/riscv.md (UNSPECV_MRET): New.
11775 (riscv_frflags, riscv_fsflags): Use tab after opcode.
11777 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
11779 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
11781 PR tree-optimization/85712
11782 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
11783 this candidate has already been replaced in-situ by a copy.
11785 2018-05-25 Jason Merrill <jason@redhat.com>
11787 PR c++/80485 - inline function non-zero address.
11788 * symtab.c (nonzero_address): Check DECL_COMDAT.
11790 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
11793 * config/alpha/alpha.md (ashlsi3): New insn pattern.
11794 (*ashlsi_se): Rename from *ashldi_se. Define as sign
11795 extension of SImode operation. Use const123_operand predicate.
11796 (*saddsi_1): Remove.
11797 (*saddl_se_1): Ditto.
11798 (*ssubsi_1): Ditto.
11799 (*ssubl_se_1): Ditto.
11800 * config/alpha/predicates.md (const123_operand): New predicate.
11801 * config/alpha/constraints.md (P): Use IN_RANGE.
11803 2018-05-25 Richard Biener <rguenther@suse.de>
11805 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
11807 (ref_maybe_used_by_stmt_p): Likewise.
11808 (stmt_may_clobber_ref_p): Likewise.
11809 (stmt_may_clobber_ref_p_1): Likewise.
11810 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
11812 (ref_maybe_used_by_stmt_p): Likewise.
11813 (stmt_may_clobber_ref_p): Likewise.
11814 (stmt_may_clobber_ref_p_1): Likewise.
11815 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
11816 the alias oracle to disambiguate DRs with stmts DR analysis
11818 (vect_analyze_data_refs): Do not give up on not analyzable
11819 DRs for BB vectorization. Remove code truncating the dataref
11822 2018-05-25 Jakub Jelinek <jakub@redhat.com>
11825 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
11826 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
11827 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
11829 2018-05-25 Richard Biener <rguenther@suse.de>
11831 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
11832 function, combining stmt data ref gathering and fatal analysis
11834 (vect_analyze_data_refs): Remove now redudnant code and simplify.
11835 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
11836 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
11837 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
11838 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
11840 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11842 PR tree-optimization/85720
11843 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
11844 SCC if all partitions are builtins.
11845 (version_loop_by_alias_check): New parameter. Generate cancelable
11846 runtime alias check if all partitions are builtins.
11847 (distribute_loop): Update call to above function.
11849 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11851 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
11852 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
11853 (parm_default_def_partition_arg): Ditto.
11854 (set_parm_default_def_partition): Ditto.
11855 (get_parm_default_def_partitions): Ditto and make it static.
11856 (get_undefined_value_partitions): Ditto and make it static.
11857 (remove_ssa_form): Refactor call to init_var_map here.
11858 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
11859 computation for loop region.
11860 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
11861 (register_default_def): Delete.
11862 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
11863 (parm_default_def_partition_arg): Ditto.
11864 (set_parm_default_def_partition): Ditto.
11865 (get_parm_default_def_partitions): Ditto and make it static.
11866 (get_undefined_value_partitions): Ditto and make it static.
11867 (coalesce_with_default, coalesce_with_default): Update comment.
11868 (create_coalesce_list_for_region): New func factored out from
11869 create_outofssa_var_map.
11870 (populate_coalesce_list_for_outofssa): New func factored out from
11871 create_outofssa_var_map and coalesce_ssa_name.
11872 (create_outofssa_var_map): Delete.
11873 (coalesce_ssa_name): Refactor to support live range computation.
11874 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
11875 (get_parm_default_def_partitions): Delete.
11876 (get_undefined_value_partitions): Ditto.
11877 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
11878 computation for loop region.
11879 (new_tree_live_info, loe_visit_block): Ditto.
11880 (live_worklist, set_var_live_on_entry): Ditto.
11881 (calculate_live_on_exit, verify_live_on_entry): Ditto.
11882 * tree-ssa-live.h (struct _var_map): New fields.
11883 (init_var_map): Change decl.
11884 (region_contains_p): New.
11886 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11888 * tree-ssa-live.h (live_merge_and_clear): Delete.
11890 2018-05-25 Richard Biener <rguenther@suse.de>
11893 * tree-dump.c (dequeue_and_dump): Remove access to removed
11894 operand 2 of a SWITCH_EXPR.
11896 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11898 * doc/sourcebuild.texi (vect_double_cond_arith): Include
11899 multiplication and division.
11900 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
11901 (cond_udiv@var{m}, cond_umod@var{m}): Document.
11902 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
11903 (cond_udiv_optab, cond_umod_optab): New optabs.
11904 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
11905 (IFN_COND_RDIV): New internal functions.
11906 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
11907 TRUNC_MOD_EXPR and RDIV_EXPR.
11908 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
11909 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
11911 (SVE_INT_BINARY): Include mult.
11912 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
11913 (optab, sve_int_op): Handle mult.
11914 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
11916 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
11917 for SVE_INT_BINARY_SD.
11919 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11921 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
11922 (optab, sve_int_op): Handle div and udiv.
11923 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
11924 for SVE_INT_BINARY_SD.
11925 (*<optab><mode>3): New insn for the same.
11927 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11929 * tree-vect-patterns.c: Include predict.h.
11930 (vect_recog_divmod_pattern): Restrict check for division support
11931 to when optimizing for size.
11933 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11935 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
11936 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
11937 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
11938 (gimple_match_op::set_op): Likewise.
11939 (gimple_resimplify4): Declare.
11940 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
11941 (expr::gen_transform): Likewise.
11942 (decision_tree::gen): Generate a simplification routine for 4 operands.
11943 * gimple-match-head.c (gimple_simplify): Add an overload for
11944 4 operands. In the top-level function, handle up to 4 call
11945 arguments and call gimple_resimplify4.
11946 (gimple_resimplify4): New function.
11947 (build_call_internal): Pass a fourth operand.
11948 (maybe_push_to_seq): Likewise.
11949 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
11950 Fold VEC_COND_EXPRs of an operation and a default value into
11951 an IFN_COND_* function if possible.
11952 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
11954 (SVE_COND_FP_BINARY): Include them.
11955 (optab, sve_fp_op): Handle them.
11956 (SVE_INT_BINARY_REV): New code iterator.
11957 (SVE_COND_FP_BINARY_REV): New int iterator.
11958 (commutative): New int attribute.
11959 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
11961 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
11963 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
11964 (*cond_<optab><mode>): New patterns for reversed operands.
11966 2018-05-25 Richard Biener <rguenther@suse.de>
11968 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
11969 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
11970 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
11971 (STMT_VINFO_GROUPED_ACCESS): Adjust.
11972 * tree-vect-data-refs.c (everywhere): Adjust users.
11973 * tree-vect-loop.c (everywhere): Likewise.
11974 * tree-vect-slp.c (everywhere): Likewise.
11975 * tree-vect-stmts.c (everywhere): Likewise.
11976 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
11978 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11980 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
11982 (gcc_cv_as_section_exclude): ... this.
11983 Try Solaris as #exclude syntax.
11984 * configure: Regenerate.
11985 * config.in: Regenerate.
11986 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
11988 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
11989 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
11991 * varasm.c (default_elf_asm_named_section): Don't check if
11992 HAVE_GAS_SECTION_EXCLUDE is defined.
11994 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11996 * doc/md.texi: Update the documentation of the cond_* optabs
11997 to mention the new final operand. Fix GET_MODE_NUNITS call.
11998 Describe the scalar case too.
11999 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
12000 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
12002 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
12003 (get_conditional_internal_fn): Update comment.
12004 * tree-vect-loop.c (vectorizable_reduction): Pass the original
12005 accumulator value as a final argument to conditional functions.
12006 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
12007 a define_expand and add an "else" operand. Assert for now that
12008 the else operand is equal to operand 2. Use SVE_INT_BINARY and
12009 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
12010 (*cond_<optab><mode>): New patterns.
12011 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
12012 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12013 (UNSPEC_COND_EOR): Delete.
12014 (optab): Remove associated mappings.
12015 (SVE_INT_BINARY): New code iterator.
12016 (sve_int_op): Remove int attribute and add "minus" to the code
12018 (SVE_COND_INT_OP): Delete.
12019 (SVE_COND_FP_OP): Rename to...
12020 (SVE_COND_FP_BINARY): ...this.
12022 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12024 * optabs.c (can_reuse_operands_p): New function.
12025 (maybe_legitimize_operands): Try to reuse the results for
12028 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
12030 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
12031 Add {q} suffix to insn mnemonic.
12033 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12035 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
12036 (msp430_warn_func_return): New.
12038 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
12040 * fold-const.c (tree_nonzero_bits): New function.
12041 * fold-const.h (tree_nonzero_bits): Likewise.
12042 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
12043 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
12045 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
12049 * varasm.c (assemble_alias): Check ifunc_resolver only on
12052 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
12055 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
12056 when memory input operand is handled.
12058 2018-05-24 Luis Machado <luis.machado@linaro.org>
12060 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
12062 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
12064 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
12066 * match.pd: Delay FMA folds until after vectorization.
12068 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
12071 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12072 address check not strict.
12074 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
12076 * gimple-match.h (gimple_match_op): New class.
12077 (mprts_hook): Replace parameters with a gimple_match_op *.
12078 (maybe_build_generic_op): Likewise.
12079 (gimple_simplified_result_is_gimple_val): Replace parameters with
12080 a const gimple_match_op *.
12081 (gimple_simplify): Replace code_helper * and tree * parameters with
12082 a gimple_match_op * parameter.
12083 (gimple_resimplify1): Replace code_helper *, tree and tree *
12084 parameters with a gimple_match_op * parameter.
12085 (gimple_resimplify2): Likewise.
12086 (gimple_resimplify3): Likewise.
12087 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
12088 parameters with a gimple_match_op * parameter.
12089 * gimple-match-head.c (gimple_simplify): Change prototypes of
12090 auto-generated functions to take a gimple_match_op * instead of
12091 separate code_helper * and tree * parameters. Make the same
12092 change in the top-level overload and update calls to the
12093 gimple_resimplify routines. Update calls to the auto-generated
12094 functions and to maybe_push_res_to_seq in the publicly-facing
12095 operation-specific gimple_simplify overloads.
12096 (gimple_match_op::MAX_NUM_OPS): Define.
12097 (gimple_resimplify1): Replace rcode and ops with a single res_op
12098 parameter. Update call to gimple_simplify.
12099 (gimple_resimplify2): Likewise.
12100 (gimple_resimplify3): Likewise.
12101 (mprts_hook): Replace parameters with a gimple_match_op *.
12102 (maybe_build_generic_op): Likewise.
12103 (build_call_internal): Replace type, nargs and ops with
12104 a gimple_match_op *.
12105 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
12106 with a single gimple_match_op *. Update calls to mprts_hook,
12107 build_call_internal and gimple_simplified_result_is_gimple_val.
12108 Factor out code that is common to the tree_code and combined_fn cases.
12109 * genmatch.c (expr::gen_transform): Replace tem_code and
12110 tem_ops with a gimple_match_op called tem_op. Update calls
12111 to the gimple_resimplify functions and maybe_push_res_to_seq.
12112 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
12113 res_ops. Update call to the gimple_resimplify functions.
12114 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
12115 (decision_tree::gen): Make the functions take a gimple_match_op *
12116 called res_op instead of separate res_code and res_ops parameters.
12117 Update call accordingly.
12118 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
12119 and ops with a single res_op parameter. Update calls to
12120 maybe_build_generic_op and maybe_push_res_to_seq.
12121 (fold_stmt_1): Update calls to gimple_simplify and
12122 replace_stmt_with_simplification.
12123 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
12124 and gimple_simplified_result_is_gimple_val.
12125 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
12127 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
12128 with a gimple_match_op *.
12129 (vn_nary_build_or_lookup): Likewise. Update call to
12130 vn_nary_build_or_lookup_1.
12131 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
12132 gimple_match_op *. Update calls to the gimple_resimplify routines
12133 and to gimple_simplified_result_is_gimple_val.
12134 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
12135 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
12136 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
12137 (visit_nary_op): Likewise.
12138 (visit_reference_op_load): Likewise.
12140 2018-05-23 Luis Machado <luis.machado@linaro.org>
12142 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
12143 modifier for printing the step amount.
12145 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
12148 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
12151 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
12153 * doc/sourcebuild.texi (Endianness): New subsubsection.
12155 2018-05-23 Luis Machado <luis.machado@linaro.org>
12157 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12158 <prefetch_dynamic_strides>: New const bool field.
12159 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12160 prefetch_dynamic_strides.
12161 (exynosm1_prefetch_tune): Likewise.
12162 (thunderxt88_prefetch_tune): Likewise.
12163 (thunderx_prefetch_tune): Likewise.
12164 (thunderx2t99_prefetch_tune): Likewise.
12165 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
12167 (aarch64_override_options_internal): Update to set
12168 PARAM_PREFETCH_DYNAMIC_STRIDES.
12169 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12170 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12171 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12172 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12173 prefetch-dynamic-strides setting.
12175 2018-05-23 Luis Machado <luis.machado@linaro.org>
12177 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12178 <minimum_stride>: New const int field.
12179 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12180 minimum_stride field defaulting to -1.
12181 (exynosm1_prefetch_tune): Likewise.
12182 (thunderxt88_prefetch_tune): Likewise.
12183 (thunderx_prefetch_tune): Likewise.
12184 (thunderx2t99_prefetch_tune): Likewise.
12185 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
12186 <default_opt_level>: Set to 3.
12187 (aarch64_override_options_internal): Update to set
12188 PARAM_PREFETCH_MINIMUM_STRIDE.
12189 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12190 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12191 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12192 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12193 stride is constant and is below the minimum stride threshold.
12195 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12197 * config/arm/arm-cpus.in (mode26): Delete.
12198 (armv4): Delete mode26 reference.
12199 * config/arm/arm.c (arm_configure_build_target): Delete use of
12202 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
12204 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
12206 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
12207 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
12209 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
12210 (floatunsdidf2): Ditto.
12212 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
12214 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
12215 (fixuns_trunc<mode>si2_avx512f): Ditto.
12216 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
12217 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
12218 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
12220 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
12222 PR rtl-optimization/79985
12223 * df-scan.c (df_insn_refs_collect): Remove special case for
12224 global registers and asm statements.
12226 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
12228 * extend.texi (Global Register Variables): Rewrite the bullet list.
12229 Note that the register is available for allocation. Note that access
12230 via inline asm must use constraints. Add note about async-signal
12231 handlers. Remove paragraph about automagic register selection.
12233 2018-05-23 Richard Biener <rguenther@suse.de>
12235 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
12236 of fixed offset from memset VN.
12238 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
12240 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
12241 first_interp field.
12242 (alloc_cand_and_find_basis): Initialize first_interp field.
12243 (slsr_process_mul): Modify first_interp field.
12244 (slsr_process_add): Likewise.
12245 (slsr_process_cast): Modify first_interp field for each new
12247 (slsr_process_copy): Likewise.
12248 (dump_candidate): Dump first_interp field.
12249 (replace_mult_candidate): Process all interpretations, not just
12251 (replace_rhs_if_not_dup): Likewise.
12252 (replace_one_candidate): Likewise.
12254 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
12256 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
12258 (aarch64_needs_frame_chain): New function.
12259 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
12261 2018-05-23 Sudakshina Das <sudi.das@arm.com>
12264 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
12265 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
12266 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
12267 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
12268 as true for strict-align.
12269 (aarch64_can_inline_p): Perform checks even when callee has no
12270 attributes to check for strict alignment.
12271 * doc/extend.texi (AArch64 Function Attributes): Document
12273 * doc/invoke.texi: (AArch64 Options): Likewise.
12275 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
12277 PR tree-optimization/85853
12278 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
12279 the handling of the root of the node to...
12280 (vect_slp_analyze_node_operations_1): ...this new function,
12281 and run the whole thing with the child nodes' def types
12282 set according to their SLP node's def type.
12284 2018-05-23 Richard Biener <rguenther@suse.de>
12286 PR middle-end/85874
12287 * tree-data-ref.c (create_runtime_alias_checks): Defer
12288 and ignore overflow warnings.
12290 2018-05-23 Yury Gribov <tetra2005@gmail.com>
12292 PR tree-optimization/85822
12293 * tree-vrp.c (is_masked_range_test): Fix handling of negative
12296 2018-05-23 Richard Biener <rguenther@suse.de>
12298 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
12299 memset constants via native_interpret_expr.
12301 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
12304 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
12306 (cgraph_node::create_alias): Likewise.
12307 (cgraph_node::get_availability): Check ifunc_resolver instead
12308 of looking up ifunc attribute.
12309 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
12310 * varasm.c (do_assemble_alias): Likewise.
12311 (assemble_alias): Likewise.
12312 (default_binds_local_p_3): Likewise.
12313 * cgraph.h (cgraph_node): Add ifunc_resolver.
12314 (cgraph_node::only_called_directly_or_aliased_p): Return false
12315 for IFUNC resolver.
12316 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
12318 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
12319 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
12320 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
12321 instead of looking up ifunc attribute.
12323 2018-05-22 Luis Machado <luis.machado@linaro.org>
12325 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
12327 2018-05-22 Martin Sebor <msebor@redhat.com>
12329 PR middle-end/85359
12330 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
12331 only when expasion succeeds.
12332 (expand_builtin_strcmp): Same.
12333 (expand_builtin_strncmp): Same.
12335 2018-05-22 Martin Sebor <msebor@redhat.com>
12337 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
12339 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
12340 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12342 * config/aarch64/aarch64-ldpstp.md: Replace uses of
12343 aarch64_mem_pair_operand with memory_operand and delete operand swapping
12345 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12346 Add check for legitimate_address.
12347 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
12348 (aarch64_swap_ldrstr_operands): New.
12349 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
12352 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
12353 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12355 * config/aarch64/aarch64.md: New patterns to generate stp
12357 (store_pair_sw, store_pair_dw): New patterns to generate stp for
12358 single words and double words.
12359 (load_pair_sw, load_pair_dw): Likewise.
12360 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
12362 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
12364 * config/aarch64/aarch64-ldpstp.md: Modify peephole
12365 for different mode ldpstp and add peephole for merged zero stores.
12366 Likewise for loads.
12367 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12369 (aarch64_gen_store_pair): Rename calls to match new patterns.
12370 (aarch64_gen_load_pair): Rename calls to match new patterns.
12371 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
12372 (load_pair<DREG:mode><DREG2:mode>): ... This.
12373 (store_pair<mode>): Rename to...
12374 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
12375 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
12376 New mode iterators.
12377 (V_INT_EQUIV): Handle SImode.
12378 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
12381 2018-05-22 Martin Sebor <msebor@redhat.com>
12384 * calls.c (maybe_warn_nonstring_arg): Use string length to set
12385 or ajust the presumed bound on an operation to avoid unnecessary
12388 2018-05-22 Martin Sebor <msebor@redhat.com>
12390 PR tree-optimization/85826
12391 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
12392 assuming that a DECL necesarily has a constant size.
12394 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
12396 PR middle-end/85862
12397 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
12399 2018-05-22 Richard Biener <rguenther@suse.de>
12401 PR tree-optimization/85834
12402 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
12403 non-constant and non-zero memset arguments.
12405 2018-05-22 Martin Liska <mliska@suse.cz>
12408 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
12410 2018-05-22 Richard Biener <rguenther@suse.de>
12412 PR tree-optimization/85863
12413 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
12414 comparisons when vectype is specified.
12415 (vectorizable_condition): Do not specify vectype for
12416 vect_is_simple_cond when SLP vectorizing.
12418 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
12421 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
12422 define __ibm128 as long double.
12423 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
12424 as a distinct type when IEEE 128-bit support is enabled.
12425 (init_float128_ieee): Fix up conversions between IFmode and IEEE
12426 128-bit types to use the correct functions.
12427 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
12428 convert between 128-bit floating point types that have different
12429 modes but the same representation, instead of using gen_lowpart to
12431 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
12433 (IFKF_reg): New attributes to give the register constraints for
12435 (extend<mode>tf2_internal): New insns to mark an explicit
12436 conversion between 128-bit floating point types that have a
12437 different mode but share the same representation.
12439 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
12441 PR tree-optimization/85814
12442 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
12443 a null return from get_strinfo when unsharing the next
12444 strinfo in the chain.
12446 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
12449 * varasm.c (weak_finish): Clean up weak_decls.
12451 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12453 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
12454 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
12455 UNSPEC_UADALP values.
12456 * config/aarch64/iterators.md (ABAL): New int iterator.
12459 (sur): Add mappings for the above.
12460 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
12462 (aarch64_<sur>abal<mode>_4): Likewise.
12463 (aarch64_<sur>adalp<mode>_3): Likewise.
12464 (<sur>sadv16qi): New define_expand.
12466 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
12468 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
12469 (*movdf_internal): Ditto.
12470 (*rcpsf2_sse): Ditto.
12471 (*rsqrtsf2_sse): Ditto.
12472 (*sqrt<mode>2_sse): Ditto.
12474 2018-05-21 Tamar Christina <tamar.christina@arm.com>
12476 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
12478 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
12479 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
12480 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
12481 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
12483 * config/aarch64/arm_neon.h: Likewise.
12484 * config/aarch64/iterators.md (VQ_I): New.
12486 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
12488 * config.gcc: Add arc/t-multilib-linux to tmake_file for
12490 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
12493 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
12495 * config/nds32/constraints.md (S): New constraint.
12496 * config/nds32/nds32.md (call_internal): Use constraint S.
12497 (call_value_internal): Likewise.
12498 (sibcall_internal): Likewise.
12499 (sibcall_value_internal): Likewise.
12501 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
12502 Chung-Ju Wu <jasonwucj@gmail.com>
12504 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
12505 into consideration.
12507 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
12508 Chung-Ju Wu <jasonwucj@gmail.com>
12510 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
12511 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
12512 (nds32_rtx_costs_impl): Simplify.
12513 (nds32_address_cost_impl): Simplify.
12514 (nds32_init_rtx_costs): New function.
12515 (nds32_rtx_costs_speed_prefer): Likewise.
12516 (nds32_rtx_costs_size_prefer): Likewise.
12517 (nds32_address_cost_speed_prefer): Likewise.
12518 (nds32_address_cost_speed_fwprop): Likewise.
12519 (nds32_address_cost_size_prefer): Likewise.
12520 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
12521 * config/nds32/nds32.c (nds32_option_override): Use
12522 nds32_init_rtx_costs function.
12524 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
12526 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
12527 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
12528 (TARGET_PIPELINE_N8): Likewise.
12529 (TARGET_PIPELINE_N10): Likewise.
12530 (TARGET_PIPELINE_N13): Likewise.
12531 (TARGET_PIPELINE_GRAYWOLF): Likewise.
12533 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
12535 * config/nds32/nds32-fpu.md: Update copyright year.
12537 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12539 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
12541 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12543 * config/nds32/nds32.c
12544 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
12545 * config/nds32/nds32.opt (minline-asm-r15): New option.
12547 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12549 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
12551 * config/nds32/nds32.md (abssi2): New pattern.
12553 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
12555 * config/i386/i386.md (rex64namesuffix): New mode attribute.
12556 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
12557 Merge insn pattern from sse_cvtsi2ss<round_name> and
12558 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
12559 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
12560 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
12561 using SWI48 mode iterator.
12562 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
12563 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
12564 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
12565 pattern from sse_cvttss2si<round_saeonly_name>
12566 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
12567 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
12568 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
12569 using SWI48 mode iterator.
12570 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
12571 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
12572 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
12573 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
12574 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
12575 using SWI48 mode iterator.
12576 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
12577 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
12578 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
12579 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
12580 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
12581 SWI48 mode iterator.
12582 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
12583 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
12584 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
12585 pattern from sse_cvttsd2si<round_saeonly_name>
12586 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
12588 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12590 * config/nds32/nds32-md-auxiliary.c
12591 (nds32_valid_smw_lwm_base_p): Refine.
12592 (nds32_output_smw_single_word): Refine.
12593 (nds32_output_smw_double_word): New.
12594 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
12596 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12598 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
12599 (nds32_output_stack_pop): Refine.
12600 (nds32_expand_unaligned_load): Refine.
12601 (nds32_expand_unaligned_store): Refine.
12603 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12604 Chung-Ju Wu <jasonwucj@gmail.com>
12606 * config/nds32/constants.md: Add TP_REGNUM constant.
12607 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
12608 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
12610 * config/nds32/nds32-doubleword.md: Consider flag_pic.
12611 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
12612 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
12613 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
12614 and PIC code generation.
12615 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
12617 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
12619 * config/nds32/nds32.md: Support TLS and PIC.
12620 * config/nds32/nds32.c: Support TLS and PIC.
12621 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
12622 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
12625 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12627 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
12628 mode with E_ prefix.
12630 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12631 Chung-Ju Wu <jasonwucj@gmail.com>
12633 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
12634 * config/nds32/nds32-md-auxiliary.c
12635 (symbolic_reference_mentioned_p): New.
12636 (nds32_legitimize_ict_address): New.
12637 (nds32_expand_ict_move): New.
12638 (nds32_indirect_call_referenced_p): New.
12639 (nds32_symbol_binds_local_p): Delete.
12640 (nds32_long_call_p): Modify.
12641 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
12642 * config/nds32/nds32-protos.h
12643 (symbolic_reference_mentioned_p): Declare.
12644 (nds32_legitimize_ict_address): Declare.
12645 (nds32_expand_ict_move): Declare.
12646 (nds32_indirect_call_referenced_p): Declare.
12647 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
12648 (nds32_relax_group): Use nds32_ict_const_p as condition.
12649 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
12650 (nds32_asm_file_start): Output ict_model directive in asm code.
12651 (nds32_legitimate_address_p): Consider indirect call.
12652 (nds32_print_operand): Consider indirect call.
12653 (nds32_print_operand_address): Consider indirect call.
12654 (nds32_insert_attributes): Handle "indirect_call" attribute.
12655 (TARGET_LEGITIMATE_ADDRESS_P): Define.
12656 (TARGET_LEGITIMATE_CONSTANT_P): Define.
12657 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
12658 (TARGET_DELEGITIMIZE_ADDRESS): Define.
12659 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
12660 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
12661 (TARGET_ICT_MODEL_SMALL): Define.
12662 (TARGET_ICT_MODEL_LARGE): Define.
12663 * config/nds32/nds32.md (movsi): Consider ict model.
12664 (call, call_value): Consider ict model.
12665 (sibcall, sibcall_value): Consider ict model.
12666 * config/nds32/nds32.opt (mict-model): New option.
12667 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
12670 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
12671 Monk Chiang <sh.chiang04@gmail.com>
12672 Jim Wilson <jimw@sifive.com>
12674 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
12675 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
12676 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
12677 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
12678 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
12679 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
12680 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
12681 compute save_libcall_adjustment properly.
12682 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
12683 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
12684 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
12685 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
12686 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
12687 (ABI_SPEC): Handle mabi=ilp32e.
12688 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
12689 (RVE): Add RVE mask.
12690 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
12691 <-march>: Add rv32e as an example.
12693 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
12696 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
12697 (intra_create_variable_infos): Handle C++ constructors.
12699 2018-05-18 Martin Liska <mliska@suse.cz>
12701 * passes.def: Remove a redundant pass.
12703 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
12706 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
12708 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12710 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
12712 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
12714 (armv2, armv2a, armv3, armv3m): Delete architectures.
12715 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
12716 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
12717 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
12719 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
12720 (*mulsidi3adddi): Likewise.
12721 (mulsidi3): Likewise.
12722 (*mulsidi3_nov6): Likewise.
12723 (umulsidi3): Likewise.
12724 (umulsidi3_nov6): Likewise.
12725 (umaddsidi4): Likewise.
12726 (*umulsidi3adddi): Likewise.
12727 (smulsi3_highpart): Likewise.
12728 (*smulsi3_highpart_nov6): Likewise.
12729 (umulsi3_highpart): Likewise.
12730 (*umulsi3_highpart_nov6): Likewise.
12731 * config/arm/arm.h (arm_arch3m): Delete.
12732 * config/arm/arm.c (arm_arch3m): Delete.
12733 (arm_option_override_internal): Update armv3-related comment.
12734 (arm_configure_build_target): Delete use of isa_bit_mode32.
12735 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
12736 (arm_rtx_costs_internal): Delete check of arm_arch3m.
12737 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
12738 (mulsa3): Likewise.
12739 (mulusa3): Likewise.
12740 * config/arm/arm-protos.h (arm_arch3m): Delete.
12741 * config/arm/arm-tables.opt: Regenerate.
12742 * config/arm/arm-tune.md: Likewise.
12743 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
12744 deleted architectures.
12746 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12748 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
12749 (armv5t, armv5te): New features.
12750 (ARMv5, ARMv5e): Delete fgroups.
12751 (ARMv5t, ARMv5te): Adjust for above changes.
12752 (ARMv6m): Likewise.
12753 (armv5, armv5e): Delete arches.
12754 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
12756 (*call_reg_arm): Likewise.
12757 (*call_value_reg_armv5): Likewise.
12758 (*call_value_reg_arm): Likewise.
12759 (*call_symbol): Likewise.
12760 (*call_value_symbol): Likewise.
12761 (*sibcall_insn): Likewise.
12762 (*sibcall_value_insn): Likewise.
12763 (clzsi2): Likewise.
12764 (prefetch): Likewise.
12765 (define_split and define_peephole2 dependent on arm_arch5):
12767 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
12769 (TARGET_ARM_QBIT): Likewise.
12770 (TARGET_DSP_MULTIPLY): Likewise.
12771 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
12772 (arm_arch5, arm_arch5e): Delete.
12773 (arm_arch5t, arm_arch5te): Declare.
12774 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
12775 (arm_arch5t): Declare.
12776 (arm_option_reconfigure_globals): Update for the above.
12777 (arm_options_perform_arch_sanity_checks): Update comment, replace
12778 use of arm_arch5 with arm_arch5t.
12779 (use_return_insn): Likewise.
12780 (arm_emit_call_insn): Likewise.
12781 (output_return_instruction): Likewise.
12782 (arm_final_prescan_insn): Likewise.
12783 (arm_coproc_builtin_available): Likewise.
12784 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
12785 arm_arch5e with arm_arch5t and arm_arch5te.
12786 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
12787 (arm_arch5t, arm_arch5te): Declare.
12788 * config/arm/arm-tables.opt: Regenerate.
12789 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
12790 * config/arm/t-multilib: Likewise.
12791 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
12792 instead of arm_arch5.
12793 (*call_reg_thumb1): Likewise.
12794 (*call_value_reg_thumb1_v5): Likewise.
12795 (*call_value_reg_thumb1): Likewise.
12796 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
12798 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
12800 2018-05-18 Martin Liska <mliska@suse.cz>
12802 PR gcov-profile/84846
12803 * doc/gcov.texi: Document -t option of gcov tool.
12805 2018-05-18 Martin Liska <mliska@suse.cz>
12807 PR gcov-profile/84846
12808 * gcov.c (print_usage): Add new -t option.
12809 (process_args): Handle the option.
12810 (generate_results): Use stdout as output when requested by
12813 2018-05-18 Martin Liska <mliska@suse.cz>
12815 PR gcov-profile/84846
12816 * coverage.c (coverage_init): Write PWD to .gcno file.
12817 * doc/gcov.texi: Document how working directory is printed.
12818 * gcov-dump.c (dump_gcov_file): Print PWD.
12819 * gcov.c (output_intermediate_file): Likewise.
12820 (read_graph_file): Read PWD string.
12821 (output_lines): Print PWD.
12823 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12825 PR middle-end/85817
12826 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
12827 for retval and return false if all args to phi are zero.
12829 2018-05-18 Richard Biener <rguenther@suse.de>
12831 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
12833 (evrp_dom_walker::before_dom_children): Call it.
12835 2018-05-18 Richard Biener <rguenther@suse.de>
12837 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
12838 results when processing array refs with variable index.
12840 2018-05-18 Toon Moene <toon@moene.org>
12842 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
12843 directly after that of -floop-interchange. Indicate that both
12844 options are enabled by default when specifying -O3.
12846 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12848 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
12849 iterator. Delete separate integer-mode vec_set<mode> expander.
12850 (aarch64_simd_vec_setv2di): Delete.
12851 (vec_setv2di): Delete.
12852 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
12853 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
12854 the "w, r" alternative.
12856 2018-05-18 Martin Liska <mliska@suse.cz>
12858 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
12859 * tree-pass.h (make_pass_lower_switch_O0): New function.
12860 * tree-switch-conversion.c (node_has_low_bound): Remove.
12861 (node_has_high_bound): Likewise.
12862 (node_is_bounded): Likewise.
12863 (class pass_lower_switch): Make it a template type and create
12865 (pass_lower_switch::execute): Add template argument.
12866 (make_pass_lower_switch): New function.
12867 (make_pass_lower_switch_O0): New function.
12868 (do_jump_if_equal): Remove.
12869 (emit_case_nodes): Simplify to just handle all 3 cases and leave
12870 all the hard work to tree optimization passes.
12872 2018-05-18 Martin Liska <mliska@suse.cz>
12874 * dbgcnt.c (limit_low): Renamed from limit.
12875 (limit_high): New variable.
12876 (dbg_cnt_is_enabled): Check for upper limit.
12877 (dbg_cnt): Adjust dumping.
12878 (dbg_cnt_set_limit_by_index): Add new argument for high
12880 (dbg_cnt_set_limit_by_name): Likewise.
12881 (dbg_cnt_process_single_pair): Parse new format.
12882 (dbg_cnt_process_opt): Use strtok.
12883 (dbg_cnt_list_all_counters): Remove 'value' and add
12885 * doc/invoke.texi: Document changes.
12887 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
12889 * doc/sourcebuild.texi (scalar_all_fma): Document.
12890 * tree.def (FMA_EXPR): Delete.
12891 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
12892 * internal-fn.c (ternary_direct): New macro.
12893 (expand_ternary_optab_fn): Likewise.
12894 (direct_ternary_optab_supported_p): Likewise.
12895 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
12896 * builtins.c (fold_builtin_fma): Delete.
12897 (fold_builtin_3): Don't call it.
12898 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
12899 * expr.c (expand_expr_real_2): Likewise.
12900 * fold-const.c (operand_equal_p): Likewise.
12901 (fold_ternary_loc): Likewise.
12902 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
12903 * gimple.c (DEFTREECODE): Likewise.
12904 * gimplify.c (gimplify_expr): Likewise.
12905 * optabs-tree.c (optab_for_tree_code): Likewise.
12906 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
12907 * tree-eh.c (operation_could_trap_p): Likewise.
12908 (stmt_could_throw_1_p): Likewise.
12909 * tree-inline.c (estimate_operator_cost): Likewise.
12910 * tree-pretty-print.c (dump_generic_node): Likewise.
12911 (op_code_prio): Likewise.
12912 * tree-ssa-loop-im.c (stmt_cost): Likewise.
12913 * tree-ssa-operands.c (get_expr_operands): Likewise.
12914 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
12915 * fold-const-call.h (fold_fma): Delete.
12916 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
12917 CFN_FNMA and CFN_FNMS.
12918 (fold_fma): Delete.
12919 * genmatch.c (combined_fn): New enum.
12920 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
12921 (commutative_op): New function.
12922 (commutate): Use it. Handle more than 2 operands.
12923 (dt_operand::gen_gimple_expr): Use commutative_op.
12924 (parser::parse_expr): Allow :c to be used with non-binary
12925 operators if the commutative operand is known.
12926 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
12927 CFN_FMS, CFN_FNMA and CFN_FNMS.
12928 (backprop::process_assign_use): Remove FMA_EXPR handling.
12929 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
12930 (gen_hsa_fma): New function.
12931 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
12932 IFN_FNMA and IFN_FNMS.
12933 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
12934 * gimple-fold.h (follow_all_ssa_edges): Declare.
12935 * gimple-fold.c (follow_all_ssa_edges): New function.
12936 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
12937 gimple_build interface and use follow_all_ssa_edges to fold the result.
12938 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
12939 instead of checking for optabs directly.
12940 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
12941 rather than FMA_EXPRs.
12942 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
12943 call to IFN_FMA instead of an FMA_EXPR.
12945 2018-05-17 Jim Wilson <jimw@sifive.com>
12947 * expr.c (do_tablejump): When converting index to Pmode, if we have a
12948 sign extended promoted subreg, and the range does not have the sign bit
12949 set, then do a sign extend.
12951 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
12952 test, check for sign extended subreg and/or constant operands, and
12953 do a sign extend in that case.
12955 2018-05-17 Steve Ellcey <sellcey@cavium.com>
12957 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
12958 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
12960 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
12961 Change logics_shift_reg to logics_shift_imm.
12962 (thunderx2t99_fp_loadpair_basic): Delete.
12963 (thunderx2t99_fp_storepair_basic): Delete.
12964 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
12965 (thunderx2t99_asimd_polynomial): Delete.
12966 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
12967 and neon_fp_mul_d_scalar_q.
12968 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
12969 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
12970 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
12971 (thunderx2t99_asimd_lut): Add missing tbl types.
12972 (thunderx2t99_asimd_ext): Delete.
12973 (thunderx2t99_asimd_load1_1_mult): Delete.
12974 (thunderx2t99_asimd_load1_2_mult): Delete.
12975 (thunderx2t99_asimd_load1_ldp): New.
12976 (thunderx2t99_asimd_load1): New.
12977 (thunderx2t99_asimd_load2): Add missing *load2* types.
12978 (thunderx2t99_asimd_load3): New.
12979 (thunderx2t99_asimd_load4): New.
12980 (thunderx2t99_asimd_store1_1_mult): Delete.
12981 (thunderx2t99_asimd_store1_2_mult): Delete.
12982 (thunderx2t99_asimd_store2_mult): Delete.
12983 (thunderx2t99_asimd_store2_onelane): Delete.
12984 (thunderx2t99_asimd_store_stp): New.
12985 (thunderx2t99_asimd_store1): New.
12986 (thunderx2t99_asimd_store2): New.
12987 (thunderx2t99_asimd_store3): New.
12988 (thunderx2t99_asimd_store4): New.
12990 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
12992 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
12993 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
12995 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
12996 Segher Boessenkool <segher@kernel.crashing.org>
12999 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
13002 2018-05-17 Richard Biener <rguenther@suse.de>
13004 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
13005 for pruning loop and prune defs feeding only already visited PHIs.
13007 2018-05-17 Richard Biener <rguenther@suse.de>
13009 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
13011 2018-05-17 Bin Cheng <bin.cheng@arm.com>
13012 Richard Biener <rguenther@suse.de>
13014 PR tree-optimization/85793
13015 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
13016 for VMAT_ELEMENTWISE.
13018 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
13020 * internal-fn.h (lookup_internal_fn): Declare
13021 * internal-fn.c (lookup_internal_fn): New function.
13022 * gimple.c (gimple_build_call_from_tree): Handle calls to
13023 internal functions.
13024 * gimple-pretty-print.c (dump_gimple_call): Print "." before
13025 internal function names.
13026 * tree-pretty-print.c (dump_generic_node): Likewise.
13027 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
13029 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
13031 * gimple-fold.h (gimple_build): Make the function forms take
13032 combined_fn rather than built_in_function.
13033 (gimple_simplify): Likewise.
13034 * gimple-match-head.c (gimple_simplify): Likewise.
13035 * gimple-fold.c (gimple_build): Likewise.
13036 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
13037 rather than gimple_build_call_internal.
13038 (get_initial_defs_for_reduction): Likewise.
13039 (vect_create_epilog_for_reduction): Likewise.
13040 (vectorizable_live_operation): Likewise.
13042 2018-05-17 Martin Liska <mliska@suse.cz>
13044 * gimple-ssa-sprintf.c (format_directive): Do not use
13045 space in between 'G_' and '('.
13047 2018-05-17 Jakub Jelinek <jakub@redhat.com>
13050 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
13051 even if the mask is not all ones.
13054 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
13056 (ix86_gimple_fold_builtin): Likewise.
13059 * config/i386/i386.c: Include tree-vector-builder.h.
13060 (ix86_vector_shift_count): New function.
13061 (ix86_fold_builtin): Fold shift builtins by scalar count.
13062 (ix86_gimple_fold_builtin): Likewise.
13064 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
13065 _mm512_setzero): New intrinsics.
13067 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
13068 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13070 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
13071 code generation for cases where splatting a value is not useful.
13072 * simplify-rtx.c (simplify_ternary_operation): Simplify
13073 vec_merge across a vec_duplicate and a paradoxical subreg forming
13074 a vector mode to a vec_concat.
13076 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
13078 * config.gcc: Support "goldmont-plus".
13079 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
13081 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13082 PROCESSOR_GOLDMONT_PLUS.
13083 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
13084 (processor_target_table): Add "goldmont-plus".
13085 (PTA_GOLDMONT_PLUS): Define.
13086 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
13087 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
13088 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
13089 (fold_builtin_cpu): Add "goldmont-plus".
13090 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
13091 (ix86_option_override_internal): Add "goldmont-plus".
13092 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
13093 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
13094 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
13095 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
13097 2018-05-17 Richard Biener <rguenther@suse.de>
13099 PR tree-optimization/85757
13100 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
13101 remove defs that only feed that PHI from further processing.
13103 2018-05-16 Jim Wilson <jimw@sifive.com>
13105 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
13107 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
13109 2018-05-16 Mark Wielaard <mark@klomp.org>
13111 * dwarf2out.c (count_index_strings): New function.
13112 (output_indirect_strings): Call count_index_strings and generate
13113 header for dwarf_version >= 5.
13115 2018-05-16 Mark Wielaard <mark@klomp.org>
13117 * dwarf2out.c (dwarf_FORM): New function.
13118 (set_indirect_string): Use dwarf_FORM.
13119 (reset_indirect_string): Likewise.
13120 (size_of_die): Likewise.
13121 (value_format): Likewise.
13122 (output_die): Likewise.
13123 (add_skeleton_AT_string): Likewise.
13124 (output_macinfo_op): Likewise.
13125 (index_string): Likewise.
13126 (output_index_string_offset): Likewise.
13127 (output_index_string): Likewise.
13128 (count_index_strings): Likewise.
13130 2018-05-16 Carl Love <cel@us.ibm.com>
13132 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13133 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
13135 2018-05-16 Martin Jambor <mjambor@suse.cz>
13137 * ipa-prop.c (ipa_free_all_edge_args): Remove.
13138 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
13140 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
13142 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
13143 (fnma<mode>4): Likewise.
13144 (fms<mode>4): Likewise.
13145 (fnms<mode>4): Likewise.
13146 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
13147 (aarch64_fnma<mode>4): Likewise.
13148 (aarch64_fms<mode>4): Likewise.
13149 (aarch64_fnms<mode>4): Likewise.
13150 (aarch64_fnmadd<mode>4): Likewise.
13152 2018-05-16 Jason Merrill <jason@redhat.com>
13154 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
13156 2018-05-16 Richard Biener <rguenther@suse.de>
13158 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
13159 (dump_stmt_cost): Declare.
13160 (add_stmt_cost): Dump cost we add.
13161 (add_stmt_costs): New function.
13162 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
13163 No longer exported.
13164 (vect_analyze_stmt): Adjust prototype.
13165 (vectorizable_condition): Likewise.
13166 (vectorizable_live_operation): Likewise.
13167 (vectorizable_reduction): Likewise.
13168 (vectorizable_induction): Likewise.
13169 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
13170 cost vector to pass to vectorizable_ and record afterwards.
13171 (vect_model_reduction_cost): Take cost vector argument and adjust.
13172 (vect_model_induction_cost): Likewise.
13173 (vectorizable_reduction): Likewise.
13174 (vectorizable_induction): Likewise.
13175 (vectorizable_live_operation): Likewise.
13176 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
13177 SLP_TREE_NUMBER_OF_VEC_STMTS.
13178 (vect_analyze_slp_cost_1): Remove.
13179 (vect_analyze_slp_cost): Likewise.
13180 (vect_slp_analyze_node_operations): Take visited args and
13181 a target cost vector. Avoid processing already visited stmt sets.
13182 (vect_slp_analyze_operations): Use a local cost vector to gather
13183 costs and register those of non-discarded instances.
13184 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
13185 (vect_schedule_slp_instance): Remove copying of
13186 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
13188 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
13189 adding cost. Record cost entry location.
13190 (vect_prologue_cost_for_slp_op): Function to compute cost of
13191 a constant or invariant generated for SLP vect in the prologue,
13192 split out from vect_analyze_slp_cost_1.
13193 (vect_model_simple_cost): Make static. Adjust for SLP costing.
13194 (vect_model_promotion_demotion_cost): Likewise.
13195 (vect_model_store_cost): Likewise, make static.
13196 (vect_model_load_cost): Likewise.
13197 (vectorizable_bswap): Add cost vector arg and adjust.
13198 (vectorizable_call): Likewise.
13199 (vectorizable_simd_clone_call): Likewise.
13200 (vectorizable_conversion): Likewise.
13201 (vectorizable_assignment): Likewise.
13202 (vectorizable_shift): Likewise.
13203 (vectorizable_operation): Likewise.
13204 (vectorizable_store): Likewise.
13205 (vectorizable_load): Likewise.
13206 (vectorizable_condition): Likewise.
13207 (vectorizable_comparison): Likewise.
13208 (can_vectorize_live_stmts): Likewise.
13209 (vect_analyze_stmt): Likewise.
13210 (vect_transform_stmt): Adjust calls to vectorizable_*.
13211 * tree-vectorizer.c: Include gimple-pretty-print.h.
13212 (dump_stmt_cost): New function.
13214 2018-05-16 Richard Biener <rguenther@suse.de>
13216 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
13217 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
13218 * tree-ssa-dse.c: Include tree-ssa-loop.h.
13219 (check_name): New callback.
13220 (dse_classify_store): Track cycles via a visited bitmap of PHI
13221 defs and simplify handling of in-loop and across loop dead stores
13222 and properly fail for loop-variant refs. Handle byte-tracking with
13223 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
13226 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
13228 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
13229 (vect_get_mask_type_for_stmt): Likewise.
13230 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
13232 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
13233 to determine the statement's vector type and the vector type that
13234 should be used for calculating nunits. Deal with cases in which
13235 the type has to be deferred.
13236 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
13237 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
13238 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
13239 (vect_determine_vf_for_stmt): New functions, split out from...
13240 (vect_determine_vectorization_factor): ...here.
13241 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
13242 (vect_get_mask_type_for_stmt): New functions, split out from
13243 vect_determine_vectorization_factor.
13245 2018-05-16 Richard Biener <rguenther@suse.de>
13247 * tree-cfg.c (verify_gimple_assign_ternary): Properly
13248 verify the [VEC_]COND_EXPR embedded comparison.
13250 2018-05-15 Martin Sebor <msebor@redhat.com>
13252 PR tree-optimization/85753
13253 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
13254 RECORD_TYPE in addition to ARRAY_TYPE.
13256 2018-05-15 Martin Sebor <msebor@redhat.com>
13258 PR middle-end/85643
13259 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
13261 2018-05-15 Richard Biener <rguenther@suse.de>
13263 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
13264 add by_clobber_p one. Change algorithm to collect all defs
13265 representing uses we need to walk and try reducing them to
13266 a single one before failing.
13267 (dse_dom_walker::dse_optimize_stmt): Adjust.
13269 2018-05-13 Mark Wielaard <mark@klomp.org>
13271 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
13272 (size_of_loc_descr): Likewise.
13273 (output_loc_operands): Likewise.
13274 (output_loc_operands_raw): Likewise.
13275 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
13276 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
13277 (hash_loc_operands): Likewise.
13278 (compare_loc_operands): Likewise.
13280 2018-05-14 Mark Wielaard <mark@klomp.org>
13282 * dwarf2out.c (count_index_addrs): New function.
13283 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
13285 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13287 PR tree-optimization/83648
13288 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
13289 return value as malloc candidate.
13291 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13294 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
13295 param as true in call to suggest_attribute.
13297 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
13299 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
13300 -mreadonly-in-sdata.
13302 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13304 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
13306 (aarch64_crypto_aesd_fused): Likewise.
13308 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
13310 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
13311 (movsi_aarch64): Likewise.
13312 (load_pairsi): Likewise.
13313 (load_pairdi): Likewise.
13314 (store_pairsi): Likewise.
13315 (store_pairdi): Likewise.
13316 (load_pairsf): Likewise.
13317 (load_pairdf): Likewise.
13318 (store_pairsf): Likewise.
13319 (store_pairdf): Likewise.
13320 (zero_extend): Likewise.
13321 (trunc): Swap alternatives.
13322 (fcvt_target): Add '?' to prefer w over r.
13324 2018-05-14 Jakub Jelinek <jakub@redhat.com>
13327 * config/i386/i386.md: Disallow non-commutative arithmetics in
13328 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
13329 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
13330 in the peephole2 before it.
13332 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
13334 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
13335 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
13336 (ix86_handle_option): Handle -mcldemote.
13337 * config.gcc: New header.
13338 * config/i386/cldemoteintrin.h: New file.
13339 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
13340 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
13342 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13343 OPTION_MASK_ISA_CLDEMOTE.
13344 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
13345 (ix86_valid_target_attribute_inner_p): Ditto.
13346 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
13347 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
13348 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
13349 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
13350 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
13352 * config/i386/i386.opt: Add -mcldemote.
13353 * config/i386/x86intrin.h: New header.
13354 * doc/invoke.texi: Add -mcldemote.
13356 2018-05-14 Richard Biener <rguenther@suse.de>
13358 * doc/match-and-simplify.texi: Adjust :s documentation.
13360 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
13362 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
13363 intended memcpy size.
13364 (REORDER_45): Likewise.
13366 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
13368 * sort.cc: New file.
13369 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
13370 * vec.c (qsort_chk): Use gcc_qsort.
13371 * Makefile.in (OBJS-libcommon): Add sort.o.
13372 (build/sort.o): New target. Use it...
13373 (BUILD_RTL): ... here, and...
13374 (build/gencfn-macros): ... here, and...
13375 (build/genmatch): ... here.
13377 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13378 Chung-Ju Wu <jasonwucj@gmail.com>
13380 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
13381 * config/nds32/nds32-graywolf.md: New file.
13382 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
13383 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
13385 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
13386 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
13387 * config/nds32/nds32.md (pipeline_model): Add graywolf.
13388 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
13389 * config/nds32/pipelines.md: Include n15 settings.
13391 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13392 Chung-Ju Wu <jasonwucj@gmail.com>
13394 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
13395 * config/nds32/nds32-n13.md: New file.
13396 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
13397 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
13399 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
13400 * config/nds32/nds32.md (pipeline_model): Add n13.
13401 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
13402 * config/nds32/pipelines.md: Include n13 settings.
13404 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13405 Chung-Ju Wu <jasonwucj@gmail.com>
13407 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
13408 * config/nds32/nds32-n10.md: New file.
13409 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
13410 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
13412 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
13413 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
13414 * config/nds32/nds32.md (pipeline_model): Add n10.
13415 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
13416 * config/nds32/pipelines.md: Include n10 settings.
13418 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
13419 Kito Cheng <kito.cheng@gmail.com>
13420 Chung-Ju Wu <jasonwucj@gmail.com>
13422 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
13423 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13424 Add enum values for DSP extension instructions.
13425 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
13427 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
13428 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
13429 New code iterators.
13430 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
13431 * config/nds32/nds32-dspext.md: New file for DSP implementation.
13432 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
13433 * config/nds32/nds32-intrinsic.md: Likewise.
13434 * config/nds32/nds32_intrinsic.h: Likewise.
13435 * config/nds32/nds32-md-auxiliary.c: Likewise.
13436 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
13437 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
13438 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
13439 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
13440 * config/nds32/nds32-protos.h: New declarations for DSP extension.
13441 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
13442 TYPE_DMAC in switch statement.
13443 * config/nds32/nds32.c: New checking and implementation for DSP
13444 extension instructions.
13445 * config/nds32/nds32.h: Likewise.
13446 * config/nds32/nds32.md: Likewise.
13447 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
13448 * config/nds32/predicates.md: Implement new predicates for DSP
13451 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13453 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
13454 Reformat alternatives and attributes so it is easier to identify
13455 which constraints/attributes go with which instruction.
13456 (mov<mode>_hardfloat32, FMOVE64): Likewise.
13457 (mov<mode>_softfloat32, FMOVE64): Likewise.
13458 (mov<mode>_hardfloat64, FMOVE64): Likewise.
13459 (mov<mode>_softfloat64, FMOVE64): Likewise.
13461 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
13463 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13465 (Basic PowerPC Built-in Functions): The new name of the
13466 subsection previously known as "PowerPC Built-in Functions".
13467 (Basic PowerPC Built-in Functions Available on all Configurations):
13469 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
13470 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
13471 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
13472 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
13474 2018-05-11 Martin Jambor <mjambor@suse.cz>
13477 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
13480 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
13483 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
13485 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
13487 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
13488 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
13489 (ix86_handle_option): Handle -mwaitpkg.
13490 * config.gcc: New header.
13491 * config/i386/cpuid.h (bit_WAITPKG): New bit.
13492 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
13493 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
13495 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13496 OPTION_MASK_ISA_WAITPKG.
13497 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
13498 (ix86_option_override_internal): Add PTA_WAITPKG.
13499 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
13500 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
13501 IX86_BUILTIN_TPAUSE.
13502 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
13503 __builtin_ia32_umwait and __builtin_ia32_tpause.
13504 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
13505 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
13506 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
13507 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
13508 UNSPECV_TPAUSE): New.
13509 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
13510 * config/i386/i386.opt: Add -mwaitpkg.
13511 * config/i386/waitpkgintrin.h: New file.
13512 * config/i386/x86intrin.h: New header.
13513 * doc/invoke.texi: Add -mwaitpkg.
13515 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
13518 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
13520 (cortex-m0): Use armv6s-m isa.
13521 (cortex-m0plus): Likewise.
13522 (cortex-m1): Likewise.
13523 (cortex-m0.small-multiply): Likewise.
13524 (cortex-m0plus.small-multiply): Likewise.
13525 (cortex-m1.small-multiply): Likewise.
13527 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
13528 Jakub Jelinek <jakub@redhat.com>
13530 PR tree-optimization/85692
13531 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
13532 source permute as well.
13534 2018-05-11 Martin Liska <mliska@suse.cz>
13537 * doc/extend.texi: Document LLVM style format for no_sanitize
13540 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13542 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
13543 mode_supports_vsx_dform_quad to mode_supports_dq_form.
13544 (mode_supports_vsx_dform_quad): Likewise.
13545 (mode_supports_vmx_dform): Move these functions to be next to the
13546 other mode_supports functions.
13547 (mode_supports_dq_form): Likewise.
13548 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
13549 mode_supports_dq_form.
13550 (reg_offset_addressing_ok_p): Likewise.
13551 (offsettable_ok_by_alignment): Likewise.
13552 (rs6000_legitimate_offset_address_p): Likewise.
13553 (legitimate_lo_sum_address_p): Likewise.
13554 (rs6000_legitimize_address): Likewise.
13555 (rs6000_legitimize_reload_address): Likewise.
13556 (rs6000_secondary_reload_inner): Likewise.
13557 (rs6000_preferred_reload_class): Likewise.
13558 (rs6000_output_move_128bit): Likewise.
13560 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
13562 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
13563 Generate SImode target register for null target.
13564 <case IX86_BUILTIN_XGETBV>: Ditto.
13565 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
13566 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
13568 2018-05-10 Carl Love <cel@us.ibm.com>
13570 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13571 dcbtt and dcbtstt if operands[2] is 0.
13573 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
13576 * config/i386/sse.md (usadv64qi): New expander.
13578 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
13580 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
13581 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
13582 -maltivec=be support.
13583 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
13584 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
13585 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
13586 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
13587 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
13588 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
13589 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
13590 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
13591 altivec_vsumsws): Adjust.
13592 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
13593 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
13594 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
13596 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
13597 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
13598 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
13599 (altivec_lve<VI_char>x): Delete expand.
13600 (*altivec_lve<VI_char>x_internal): Rename to...
13601 (altivec_lve<VI_char>x): ... this.
13602 (altivec_lvxl_<mode>): Delete expand.
13603 (*altivec_lvxl_<mode>_internal): Rename to ...
13604 (altivec_lvxl_<mode>): ... this.
13605 (altivec_stvxl_<mode>): Delete expand.
13606 (*altivec_stvxl_<mode>_internal): Rename to ...
13607 (altivec_stvxl_<mode>): ... this.
13608 (altivec_stve<VI_char>x): Delete expand.
13609 (*altivec_stve<VI_char>x_internal): Rename to ...
13610 (altivec_stve<VI_char>x): ... this.
13611 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
13612 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
13613 reduc_plus_scal_<mode>): Adjust.
13614 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
13616 (rs6000_cpu_cpp_builtins): Adjust.
13617 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
13618 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
13619 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
13620 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
13621 -maltivec=be support.
13622 (rs6000_split_vec_extract_var): Adjust.
13623 (rs6000_split_v4si_init): Adjust.
13624 (swap_selector_for_mode): Delete.
13625 (altivec_expand_lvx_be, altivec_expand_stvx_be,
13626 altivec_expand_stvex_be): Delete.
13627 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
13628 -maltivec=be support.
13629 (rs6000_gimple_fold_builtin): Ditto.
13630 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
13632 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
13633 (TARGET_DIRECT_MOVE_64BIT): Adjust.
13634 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
13635 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
13636 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
13637 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
13638 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
13639 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
13640 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
13641 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
13642 anonymous split): Adjust.
13643 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
13644 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
13646 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
13648 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
13649 when --with-gxx-include-dir is also specified.
13650 * configure: Regenerate.
13652 2018-05-09 Jim Wilson <jimw@sifive.com>
13655 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
13656 * config/riscv/t-withmultilib: New.
13657 * config/riscv/withmultilib.h: New.
13658 * doc/install.texi: Document RISC-V --with-multilib-list support.
13660 2018-05-09 Richard Biener <rguenther@suse.de>
13662 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
13664 (vect_bb_vectorization_profitable_p): Adjust. Compute
13665 actual scalar cost using the cost vector and the add_stmt_cost
13668 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13670 PR rtl-optimization/85645
13671 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
13672 in the REG_CFA_REGISTER note for LR, don't leave it empty.
13674 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13676 PR rtl-optimization/85645
13677 * shrink-wrap.c (spread_components): Return a boolean saying if
13678 anything was changed.
13679 (try_shrink_wrapping_separate): Iterate spread_components until
13680 nothing changes anymore.
13682 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13684 PR rtl-optimization/85645
13685 * regrename.c (build_def_use): Also kill the chains that include the
13686 destination of a REG_CFA_REGISTER note.
13688 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13690 PR rtl-optimization/85645
13691 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
13692 insn that has a REG_CFA_REGISTER note.
13694 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
13696 * cfgexpand.c (expand_clobber): New function.
13697 (expand_gimple_stmt_1): Use it.
13698 * tree-vect-stmts.c (vect_clobber_variable): New function,
13700 (vectorizable_simd_clone_call): ...here.
13701 (vectorizable_store): Emit a clobber either side of an
13702 IFN_STORE_LANES sequence.
13703 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
13705 2018-05-09 Tom de Vries <tom@codesourcery.com>
13708 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
13709 (define_insn "trap_if_false"): Add exit after trap.
13711 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
13713 PR rtl-optimization/85638
13714 * bb-reorder.c: Include common/common-target.h.
13715 (create_forwarder_block): New function extracted from...
13716 (fix_up_crossing_landing_pad): ...here. Rename into...
13717 (dw2_fix_up_crossing_landing_pad): ...this.
13718 (sjlj_fix_up_crossing_landing_pad): New function.
13719 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
13720 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
13721 from both partitions and exit the loop after one iteration.
13723 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
13726 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13728 (Basic PowerPC Built-in Functions): The new name of the
13729 subsection previously known as "PowerPC Built-in Functions".
13730 (Basic PowerPC Built-in Functions Available on all Configurations):
13732 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13734 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13736 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13738 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13741 2018-05-08 Jim Wilson <jimw@sifive.com>
13743 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
13744 (LD_EMUL_SUFFIX): New.
13745 (LINK_SPEC): Use it.
13747 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
13749 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13751 (Basic PowerPC Built-in Functions): The new name of the
13752 subsection previously known as "PowerPC Built-in Functions".
13753 (Basic PowerPC Built-in Functions Available on all Configurations):
13755 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13757 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13759 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13761 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13764 2018-05-08 Jakub Jelinek <jakub@redhat.com>
13767 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
13768 after cmpelim optimization.
13770 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
13772 * config.gcc: Support "goldmont".
13773 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
13774 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13775 PROCESSOR_GOLDMONT.
13776 * config/i386/i386.c (m_GOLDMONT): Define.
13777 (processor_target_table): Add "goldmont".
13778 (PTA_GOLDMONT): Define.
13779 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
13780 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
13781 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
13782 (fold_builtin_cpu): Add "goldmont".
13783 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
13784 (ix86_option_override_internal): Add "goldmont".
13785 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
13786 (processor_type): Add PROCESSOR_GOLDMONT.
13787 * config/i386/i386.md: Add CPU "glm".
13788 * config/i386/glm.md: New file.
13789 * config/i386/x86-tune.def: Add m_GOLDMONT.
13790 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
13792 2018-05-08 Jakub Jelinek <jakub@redhat.com>
13795 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
13797 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
13798 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
13799 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
13802 * config/i386/i386.c (ix86_fold_builtin): Handle
13803 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
13806 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
13807 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
13809 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
13812 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
13813 (check_arch): Likewise.
13814 (check_fpu): Return the result rather than printing it.
13815 (end arch): Fix operator precedence.
13816 (end cpu): Likewise.
13817 (END): Print the result from check_fpu.
13819 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
13820 Alan Hayward <alan.hayward@arm.com>
13821 David Sherwood <david.sherwood@arm.com>
13823 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
13824 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
13825 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
13826 (*fcmuo<mode>_and): New patterns.
13828 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
13830 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
13831 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
13832 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
13833 (cmp_op, sve_imm_con): New code attributes.
13834 (SVE_COND_INT_CMP, imm_con): Delete.
13835 (cmp_op): Remove above unspecs from int attribute.
13836 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
13838 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
13839 comparison-specific unspecs.
13840 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
13841 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
13842 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
13843 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
13844 (*vec_fcm<cmp_op><mode>): Rename to...
13845 (*fcm<cmp_op><mode>): ...this and adjust likewise.
13846 (*vec_fcmuo<mode>): Rename to...
13847 (*fcmuo<mode>): ...this and adjust likewise.
13848 (*pred_fcm<cmp_op><mode>): New pattern.
13849 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
13850 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
13852 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
13854 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
13855 (aarch64_emit_sve_predicated_cond): New function.
13856 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
13857 (aarch64_emit_unspec_cond_or): Replace with...
13858 (aarch64_emit_sve_or_conds): ...this new function. Use
13859 aarch64_emit_sve_ptrue_op for the individual comparisons and
13860 aarch64_emit_binop to OR them together.
13861 (aarch64_emit_inverted_unspec_cond): Replace with...
13862 (aarch64_emit_sve_inverted_cond): ...this new function. Use
13863 aarch64_emit_sve_ptrue_op for the comparison and
13864 aarch64_emit_unop to invert the result.
13865 (aarch64_expand_sve_vec_cmp_float): Update after the above
13866 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
13868 2018-05-07 Nathan Sidwell <nathan@acm.org>
13870 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
13871 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
13872 (Backwards Compatibility): Likewise.
13874 2018-05-07 Luis Machado <luis.machado@linaro.org>
13878 2018-05-07 Luis Machado <luis.machado@linaro.org>
13880 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13881 <prefetch_dynamic_strides>: New const bool field.
13882 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13883 prefetch_dynamic_strides.
13884 (exynosm1_prefetch_tune): Likewise.
13885 (thunderxt88_prefetch_tune): Likewise.
13886 (thunderx_prefetch_tune): Likewise.
13887 (thunderx2t99_prefetch_tune): Likewise.
13888 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13890 (aarch64_override_options_internal): Update to set
13891 PARAM_PREFETCH_DYNAMIC_STRIDES.
13892 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13893 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13894 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13895 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13896 prefetch-dynamic-strides setting.
13898 2018-05-07 Luis Machado <luis.machado@linaro.org>
13900 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13901 <minimum_stride>: New const int field.
13902 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13903 minimum_stride field.
13904 (exynosm1_prefetch_tune): Likewise.
13905 (thunderxt88_prefetch_tune): Likewise.
13906 (thunderx_prefetch_tune): Likewise.
13907 (thunderx2t99_prefetch_tune): Likewise.
13908 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13909 (aarch64_override_options_internal): Update to set
13910 PARAM_PREFETCH_MINIMUM_STRIDE.
13911 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13912 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13913 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13914 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13915 stride is constant and is below the minimum stride threshold.
13917 2018-05-07 Luis Machado <luis.machado@linaro.org>
13919 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
13922 2018-05-07 Luis Machado <luis.machado@linaro.org>
13924 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13925 <prefetch_dynamic_strides>: New const bool field.
13926 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13927 prefetch_dynamic_strides.
13928 (exynosm1_prefetch_tune): Likewise.
13929 (thunderxt88_prefetch_tune): Likewise.
13930 (thunderx_prefetch_tune): Likewise.
13931 (thunderx2t99_prefetch_tune): Likewise.
13932 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13934 (aarch64_override_options_internal): Update to set
13935 PARAM_PREFETCH_DYNAMIC_STRIDES.
13936 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13937 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13938 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13939 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13940 prefetch-dynamic-strides setting.
13942 2018-05-07 Luis Machado <luis.machado@linaro.org>
13944 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13945 <minimum_stride>: New const int field.
13946 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13947 minimum_stride field.
13948 (exynosm1_prefetch_tune): Likewise.
13949 (thunderxt88_prefetch_tune): Likewise.
13950 (thunderx_prefetch_tune): Likewise.
13951 (thunderx2t99_prefetch_tune): Likewise.
13952 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13953 (aarch64_override_options_internal): Update to set
13954 PARAM_PREFETCH_MINIMUM_STRIDE.
13955 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13956 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13957 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13958 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13959 stride is constant and is below the minimum stride threshold.
13961 2018-05-06 Jakub Jelinek <jakub@redhat.com>
13964 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
13965 the type is addressable. Don't force op into register if it has
13968 2018-05-05 Roland McGrath <mcgrathr@google.com>
13971 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
13972 any section for which we don't know a specific type it should have,
13973 regardless of name. Previously this was done only for the exact
13974 names ".init_array", ".fini_array", and ".preinit_array".
13975 (default_elf_asm_named_section): Add comment about
13976 relationship with default_section_type_flags and SECTION_NOTYPE.
13977 (get_section): Don't consider it a type conflict if one side has
13978 SECTION_NOTYPE and the other doesn't, as long as neither has the
13979 SECTION_BSS et al used in the default_section_type_flags logic.
13981 2018-05-05 Tom de Vries <tom@codesourcery.com>
13984 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
13985 (workaround_barsyncs): New function.
13986 (nvptx_reorg): Use workaround_barsyncs.
13987 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
13988 (define_expand "nvptx_membar_cta"): New define_expand.
13989 (define_insn "*nvptx_membar_cta"): New insn.
13991 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
13993 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
13994 To improve optimization opportunities.
13995 * builtin-types.def: The new needed builtin types for the above.
13997 2018-05-04 Richard Biener <rguenther@suse.de>
13999 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
14000 * gimple-ssa-store-merging.c
14001 (imm_store_chain_info::output_merged_store): Remove redundant create,
14002 release split_store vector contents on failure.
14003 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
14004 scalar stmt vector on cache hit.
14006 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
14008 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
14010 * config.gcc (powerpc-xilinx-eabi*): Remove.
14011 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
14013 (fusion_addis_mem_combo_load): Ditto.
14014 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
14016 (rs6000_cpu_cpp_builtins): Ditto.
14017 * config/rs6000/rs6000-linux.c
14018 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
14019 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
14020 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
14022 (rs6000_setup_reg_addr_masks): Ditto.
14023 (rs6000_init_hard_regno_mode_ok): Ditto.
14024 (rs6000_option_override_internal): Ditto.
14025 (legitimate_lo_sum_address_p): Ditto.
14026 (rs6000_legitimize_address): Ditto.
14027 (rs6000_legitimize_reload_address): Ditto.
14028 (rs6000_legitimate_address_p): Ditto.
14029 (abi_v4_pass_in_fpr): Ditto.
14030 (setup_incoming_varargs): Ditto.
14031 (rs6000_gimplify_va_arg): Ditto.
14032 (rs6000_split_multireg_move): Ditto.
14033 (rs6000_savres_strategy): Ditto.
14034 (rs6000_emit_prologue_components): Ditto.
14035 (rs6000_emit_epilogue_components): Ditto.
14036 (rs6000_emit_prologue): Ditto.
14037 (rs6000_emit_epilogue): Ditto.
14038 (rs6000_elf_file_end): Ditto.
14039 (rs6000_function_value): Ditto.
14040 (rs6000_libcall_value): Ditto.
14041 * config/rs6000/rs6000.h: Ditto.
14042 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
14043 (TARGET_MINMAX): ... this. New.
14044 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
14045 * config/rs6000/rs6000.md: Remove Xilinx FP support.
14046 (*movsi_internal1_single): Delete.
14047 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
14048 mfpu=, mxilinx-fpu): Delete.
14049 * config/rs6000/singlefp.h: Delete.
14050 * config/rs6000/sysv4.h: Remove Xilinx FP support.
14051 * config/rs6000/t-rs6000: Ditto.
14052 * config/rs6000/t-xilinx: Delete.
14053 * config/rs6000/titan.md: Adjust for fp_type removal.
14054 * config/rs6000/vsx.md: Remove Xilinx FP support.
14055 (VStype_simple): Delete.
14056 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
14057 * config/rs6000/xfpu.h: Delete.
14058 * config/rs6000/xfpu.md: Delete.
14059 * config/rs6000/xilinx.h: Delete.
14060 * config/rs6000/xilinx.opt: Delete.
14061 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
14062 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
14064 2018-05-04 Tom de Vries <tom@codesourcery.com>
14067 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
14070 2018-05-04 Richard Biener <rguenther@suse.de>
14072 PR middle-end/85627
14073 * tree-complex.c (update_complex_assignment): We are always in SSA form.
14074 (expand_complex_div_wide): Likewise.
14075 (expand_complex_operations_1): Likewise.
14076 (expand_complex_libcall): Preserve EH info of the original stmt.
14077 (tree_lower_complex): Handle removed blocks.
14078 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
14079 on complex multiplication and division libcall builtins.
14081 2018-05-04 Richard Biener <rguenther@suse.de>
14083 PR middle-end/85574
14084 * fold-const.c (negate_expr_p): Restrict negation of operand
14085 zero of a division to when we know that can happen without
14087 (fold_negate_expr_1): Likewise.
14089 2018-05-04 Jakub Jelinek <jakub@redhat.com>
14092 * real.h (real_nextafter): Declare.
14093 * real.c (real_nextafter): New function.
14094 * fold-const-call.c (fold_const_nextafter): New function.
14095 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
14096 CASE_CFN_NEXTTOWARD.
14097 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
14098 even when arg1_mode is different from arg0_mode.
14100 2018-05-03 Nathan Sidwell <nathan@acm.org>
14102 * doc/extend.texi (Deprecated Features): Remove
14103 -ffriend-injection.
14104 (Backwards Compatibility): Likewise.
14105 * doc/invoke.texi (C++ Language Options): Likewise.
14106 (C++ Dialect Options): Likewise.
14108 2018-05-03 Jakub Jelinek <jakub@redhat.com>
14111 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
14112 _mm512_mask_mullox_epi64): New intrinsics.
14114 2018-05-03 Tom de Vries <tom@codesourcery.com>
14117 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14118 dump files): Add offload-tree.
14120 2018-05-03 Richard Biener <rguenther@suse.de>
14122 PR tree-optimization/85615
14123 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
14124 to loops not nested in BBs loop father to avoid creating multi-entry
14127 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14129 PR tree-optimization/70291
14130 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
14131 arguments. Change return type to tree. Emit libcall as a new
14132 statement rather than replacing existing one when inplace_p is true.
14133 (expand_complex_multiplication_components): New function.
14134 (expand_complex_multiplication): Expand floating-point complex
14135 multiplication using the above.
14136 (expand_complex_division): Rename inner_type parameter to type.
14137 Update expand_complex_libcall call-site.
14138 (expand_complex_operations_1): Update expand_complex_multiplication
14139 and expand_complex_division call-sites.
14141 2018-05-02 Jakub Jelinek <jakub@redhat.com>
14144 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
14145 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
14146 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
14147 the highest significant bit of the shift count mask is clear. In
14148 check whether and[sq]i3 is needed verify that all significant bits
14149 of the shift count other than the highest are set.
14151 2018-05-02 Tom de Vries <tom@codesourcery.com>
14154 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
14155 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
14156 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
14157 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
14158 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
14159 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
14160 __builtin_goacc_parlevel_size.
14162 2018-05-02 Richard Biener <rguenther@suse.de>
14164 PR tree-optimization/85597
14165 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
14166 do not use split vect_get_vec_defs call but call vect_get_slp_defs
14169 2018-05-02 Tom de Vries <tom@codesourcery.com>
14172 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14173 dump files): Add ltrans-tree.
14175 2018-05-02 Tom de Vries <tom@codesourcery.com>
14178 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14179 dump files): Add wpa-ipa.
14181 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
14183 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
14184 powerpc*-*-linux*paired* target.
14185 * config/rs6000/750cl.h: Delete.
14186 * config/rs6000/paired.h: Delete.
14187 * config/rs6000/paired.md: Delete.
14188 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
14190 * config/rs6000/rs6000-builtin.def: Remove paired float support.
14191 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
14192 comment. Remove paired float support.
14193 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
14194 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
14196 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
14197 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
14199 * config/rs6000/rs6000.c: Remove paired float support.
14200 (paired_expand_vector_init, paired_expand_vector_move,
14201 paired_emit_vector_compare, paired_emit_vector_cond_expr,
14202 (paired_expand_lv_builtin, paired_expand_stv_builtin,
14203 paired_expand_builtin, paired_expand_predicate_builtin,
14204 paired_init_builtins): Delete.
14205 * config/rs6000/rs6000.h: Remove paired float support.
14206 * config/rs6000/rs6000.md: Remove paired float support.
14207 (move_from_CR_ov_bit): Delete.
14208 * config/rs6000/rs6000.opt (mpaired): Delete.
14209 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
14210 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
14212 2018-05-02 Richard Biener <rguenther@suse.de>
14214 PR middle-end/85567
14215 * gimplify.c (gimplify_save_expr): When in SSA form allow
14216 SAVE_EXPRs to compute to SSA vars.
14218 2018-05-02 Jakub Jelinek <jakub@redhat.com>
14221 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
14222 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
14223 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
14224 clobber operands[2], instead use a new pseudo. Formatting fixes.
14226 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
14228 PR tree-optimization/85586
14229 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
14230 exit early for statements in the same group if the accesses are
14233 2018-05-02 Tom de Vries <tom@codesourcery.com>
14236 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
14239 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
14241 PR tree-optimization/85143
14242 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
14244 2018-05-01 Tom de Vries <tom@codesourcery.com>
14247 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
14248 not found" error message.
14250 2018-05-01 Tom de Vries <tom@codesourcery.com>
14253 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
14254 * vec.c (test_ordered_remove_if): New function.
14255 (vec_c_tests): Call test_ordered_remove_if.
14256 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
14257 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
14258 * tree-vect-patterns.c (vect_pattern_recog_1): Use
14259 VEC_ORDERED_REMOVE_IF.
14261 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14263 PR tree-optimization/82665
14264 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
14265 pointer subtraction where arguments come from a memchr call.
14267 2018-05-01 Jakub Jelinek <jakub@redhat.com>
14269 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
14270 --push-state --as-needed and --pop-state instead of --as-needed and
14271 --no-as-needed if ld supports it.
14272 * configure: Regenerated.
14275 * doc/install.texi2html: Replace _002d with - and _002a with * in
14276 generated html files using sed.
14278 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14281 * gcc-rich-location.c (blank_line_before_p): New function.
14282 (use_new_line): New function.
14283 (gcc_rich_location::add_fixit_insert_formatted): New function.
14284 * gcc-rich-location.h
14285 (gcc_rich_location::add_fixit_insert_formatted): New function.
14287 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14289 * selftest.c (assert_streq): Rename "expected" and "actual" to
14290 "val1" and "val2". Extend NULL-handling to cover both inputs
14291 symmetrically, while still requiring both to be non-NULL for a pass.
14292 * selftest.h (assert_streq): Rename "expected" and "actual" to
14294 (ASSERT_EQ): Likewise.
14295 (ASSERT_EQ_AT): Likewise.
14296 (ASSERT_KNOWN_EQ): Likewise.
14297 (ASSERT_KNOWN_EQ_AT): Likewise.
14298 (ASSERT_NE): Likewise.
14299 (ASSERT_MAYBE_NE): Likewise.
14300 (ASSERT_MAYBE_NE_AT): Likewise.
14301 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
14302 the assertion to pass.
14303 (ASSERT_STREQ_AT): Likewise.
14305 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
14307 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
14308 interaction with -pie.
14310 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14312 * selftest.h: Fix alphabetization of per-source-file selftest
14315 2018-04-30 Jason Merrill <jason@redhat.com>
14317 PR c++/61982 - dead stores to destroyed objects.
14318 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
14321 2018-04-30 Jason Merrill <jason@redhat.com>
14323 * tree.c (build_clobber): New.
14324 * tree.h: Declare it.
14325 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
14327 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14329 * diagnostic-show-locus.c (layout::layout): Update for
14330 location_get_source_line returning a char_span.
14331 (struct char_span): Move to input.h.
14332 (struct correction): Update for fields in char_span becoming
14334 (struct source_line): Update for location_get_source_line
14335 returning a char_span.
14336 (layout::print_line): Likewise.
14337 * edit-context.c (edited_file::print_content): Likewise.
14338 (edited_file::print_diff_hunk): Likewise.
14339 (edited_file::print_run_of_changed_lines): Likewise.
14340 (edited_file::get_num_lines): Likewise.
14341 (edited_line::edited_line): Likewise.
14342 * final.c (asm_show_source): Likewise.
14343 * input.c (location_get_source_line): Convert return type
14344 from const char * to char_span, losing the final "line_len"
14346 (dump_location_info): Update for the above.
14347 (get_substring_ranges_for_loc): Likewise. Use a char_span
14348 when handling the literal within the line.
14349 (test_reading_source_line): Update for location_get_source_line
14350 returning a char_span.
14351 * input.h (class char_span): Move here from
14352 diagnostic-show-locus.c, converting from a struct to a class.
14353 Make data members private.
14354 (char_span::operator bool): New.
14355 (char_span::length): New.
14356 (char_span::get_buffer): New.
14357 (char_span::operator[]): New.
14358 (char_span::subspan): Make const.
14359 (char_span::xstrdup): New.
14360 (location_get_source_line): Convert return type from const char *
14361 to char_span, losing the final "line_size" param.
14363 2018-04-30 Jan Hubicka <jh@suse.cz>
14365 * lto-wrapper.c (ltrans_priorities): New static var.
14366 (cmp_priority): New.
14367 (run_gcc): Read priorities and if doing parallel build order
14368 the Makefile by them.
14370 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14372 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
14374 2018-04-30 Richard Biener <rguenther@suse.de>
14376 * tree-cfg.c (verify_address): Remove base argument, add
14377 flag whether to check TREE_ADDRESSABLE and do that.
14378 (verify_expr): Remove.
14379 (verify_types_in_gimple_reference): Add pieces from verify_expr.
14380 (verify_gimple_assign_single): Likewise.
14381 (verify_gimple_switch): Likewise.
14382 (verify_expr_location_1): Dereference tp once. Add (disabled)
14383 piece from verify_expr.
14384 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
14386 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14388 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
14390 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14392 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
14393 (small_data_pattern): Likewise.
14394 (arc_rewrite_small_data): Likewise.
14395 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
14396 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
14397 (get_symbol_alignment): New function.
14398 (legitimate_small_data_address_p): Likewise.
14399 (legitimate_scaled_address): Update, call
14400 legitimate_small_data_address_p.
14401 (output_sdata): New static variable.
14402 (arc_print_operand): Update how we handle small data operands.
14403 (arc_print_operand_address): Likewise.
14404 (arc_legitimate_address_p): Update, use
14405 legitimate_small_data_address_p.
14406 (arc_rewrite_small_data_p): Remove.
14407 (arc_rewrite_small_data_1): Likewise.
14408 (arc_rewrite_small_data): Likewise.
14409 (small_data_pattern): Likewise.
14410 (compact_sda_memory_operand): Update to use
14411 legitimate_small_data_address_p and get_symbol_alignment.
14412 (prepare_move_operands): Don't rewite sdata pattern.
14413 (prepare_extend_operands): Remove.
14414 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
14416 (zero_extendqisi2): Likewise.
14417 (zero_extendhisi2): Likewise.
14418 (extendqihi2): Likewise.
14419 (extendqisi2): Likewise.
14420 (extendhisi2): Likewise.
14421 (addsi3): Likewise.
14422 (subsi3): Likewise.
14423 (andsi3): Likewise.
14424 * config/arc/constraints.md (Usd): Change it to memory constraint.
14426 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14428 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
14429 as source of std instructions.
14430 * config/arc/arc.md (movsi_insn): Update pattern predicate to
14431 allow 6-bit constants as source for store instructions.
14432 (movdi_insn): Update instruction pattern to allow 6-bit constants
14433 as source for store instructions.
14435 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
14437 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
14439 2018-04-30 Nathan Sidwell <nathan@acm.org>
14440 Sandra Loosemore <sandra@codesourcery.com>
14442 * dumpfile.c (dump_open): Allow '-' for stdout.
14443 * doc/invoke.texi (Developer Options): Document dump filename
14444 determination early. Document stdin/stdout selection.
14446 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
14448 Microblaze Target: PIC data text relative
14450 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
14451 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
14453 * config/microblaze/microblaze.h (microblaze_constant_address_p):
14454 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
14455 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
14456 New addressing mode for data-text relative position indepenedent code.
14457 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
14458 'ADDRESS_SYMBOLIC_TXT_REL'.
14459 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
14460 (microblaze_legitimate_pic_operand): Exclude function calls from
14461 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
14462 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
14464 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14465 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14466 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
14467 for 'address + offset'.
14468 (microblaze_expand_prologue): Add new function prologue call for
14470 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
14471 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
14472 table in case of TARGET_PIC_DATA_TEXT_REL.
14473 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
14474 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
14475 Add new macros 'UNSPEC_TEXT',
14476 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
14477 + exclude function calls from 'UNSPEC_PLT' in case of data text
14479 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14480 new target hook for generating address diff vector tables in case of
14482 * doc/tm.texi : Regenerate.
14483 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
14484 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
14485 of addr diff vector generation.
14486 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14487 target hook definition.
14488 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
14489 Add default function for generate_pic_addr_diff_vec -> flag_pic.
14490 * doc/invoke.texi (Add new pic option): Add new microblaze pic
14491 option for data text relative.
14493 2018-04-30 Richard Biener <rguenther@suse.de>
14495 * tree-chrec.h (evolution_function_is_constant_p): Remove
14497 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
14499 2018-04-30 Richard Biener <rguenther@suse.de>
14502 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
14504 2018-04-30 Richard Biener <rguenther@suse.de>
14506 PR tree-optimization/28364
14507 PR tree-optimization/85275
14508 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
14509 copying first exit test.
14511 2018-04-28 Mark Wielaard <mark@klomp.org>
14513 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
14514 dwarf_version >= 5.
14515 (dwarf_AT): Handle DW_AT_addr_base.
14516 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
14518 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
14521 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
14522 (*ashl<dwi>3_doubleword_mask_1): Ditto.
14523 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
14524 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
14526 2018-04-28 Richard Biener <rguenther@suse.de>
14528 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
14529 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
14530 to reflect use. Only add interesting stmts.
14532 2018-04-27 Martin Jambor <mjambor@suse.cz>
14535 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
14536 the jump function allows for passing through aggregate values.
14538 2018-04-27 David Malcolm <dmalcolm@redhat.com>
14540 * input.h (in_system_header_at): Convert from macro to inline
14542 (from_macro_expansion_at): Likewise.
14543 (from_macro_definition_at): Likewise.
14545 2018-04-27 Jeff Law <law@redhat.com>
14547 * config.gcc: Mark tile* targets as deprecated/obsolete.
14549 2018-04-27 Richard Biener <rguenther@suse.de>
14551 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
14554 2018-04-27 Richard Biener <rguenther@suse.de>
14556 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
14558 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
14560 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
14561 with Yd constraint. Set "preferred_for_speed" attribute from
14562 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
14563 with Yd constraint.
14564 (*movdi_internal): Ditto.
14565 (movti_interunit splitters): Remove
14566 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
14567 (movdi_interunit splitters): Ditto.
14568 * config/i386/constraints.md (Ye): Remove.
14569 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
14571 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14574 * config/aarch64/constraints.md (Usg): Limit to 31.
14575 (Usj): Limit to 63.
14577 2018-04-27 Jakub Jelinek <jakub@redhat.com>
14579 PR tree-optimization/85529
14580 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
14581 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
14582 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
14583 zero extension or masking of the MSB bit.
14584 (optimize_range_tests): Add FIRST_BB argument, pass it through
14585 to optimize_range_tests_var_bound.
14586 (maybe_optimize_range_tests, reassociate_bb): Adjust
14587 optimize_range_tests callers.
14589 2018-04-26 Richard Biener <rguenther@suse.de>
14590 Jakub Jelinek <jakub@redhat.com>
14592 * cgraph.h (symbol_table): Just declare debug method here.
14593 * symtab.c (symbol_table::debug): Define.
14595 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
14597 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
14599 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
14601 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
14602 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
14603 (*movdi_internal): Substitute Yi and Yj constraint with x
14604 and Ym and Yn constraint with y constraint. Update "isa"
14605 attribute and set "preferred_for_speed" attribute from
14606 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
14607 (*movsi_internal): Ditto.
14608 (*movdf_internal): Ditto.
14609 (*movsf_internal): Ditto.
14610 (*zero_extendsidi2): Ditto.
14611 * config/i386/sse.md (vec_set<mode>_0): Ditto.
14612 (sse2_loadld): Ditto.
14613 (*vec_extract<ssevecmodelower>_0): Ditto.
14614 (*vec_extractv4si_0_zext_sse4): Ditto.
14615 (vec_concatv2di): Ditto.
14616 (*vec_dup<mode>): Ditto.
14617 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
14618 * config/i386/constraints.md (Yi): Remove.
14623 2018-04-26 Nathan Sidwell <nathan@acm.org>
14625 * dumpfile.c (dump_open): New.
14626 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
14627 (dump_finish): Detect stdio/stderr by value not name.
14629 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
14631 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
14633 2018-04-26 Tom de Vries <tom@codesourcery.com>
14636 * config/nvptx/nvptx.c (verify_neutering_jumps)
14637 (verify_neutering_labels): New function
14638 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
14640 2018-04-26 Tom de Vries <tom@codesourcery.com>
14643 * config/nvptx/nvptx.c (needs_neutering_p): New function.
14644 (nvptx_single): Use needs_neutering_p to skip over insns that do not
14647 2018-04-26 Richard Biener <rguenther@suse.de>
14648 Tom de Vries <tom@codesourcery.com>
14651 * lto-streamer-out.c (output_function): Fixup loops if required to match
14652 discovery done in the reader.
14654 2018-04-26 Richard Biener <rguenther@suse.de>
14656 PR tree-optimization/85116
14657 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
14658 have a loop exit from the single latch predecessor. Remove
14659 case of header with just condition.
14660 (ch_base::copy_headers): Exclude infinite loops from any
14662 (pass_ch::execute): Record exits.
14664 2018-04-26 Richard Biener <rguenther@suse.de>
14666 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
14667 prologue cost vector and pass it to vect_get_load_cost.
14668 (vect_get_peeling_costs_all_drs): Likewise.
14669 (vect_peeling_hash_get_lowest_cost): Likewise.
14670 (vect_enhance_data_refs_alignment): Likewise.
14672 2018-04-26 Richard Biener <rguenther@suse.de>
14674 PR middle-end/85450
14675 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
14676 checking of integer<->pointer conversions.
14677 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
14678 sign-/zero-extending pointer types.
14679 (expand_omp_for_static_chunk): Likewise.
14681 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
14682 Jean Lee <xiaoyur347@gmail.com>
14684 * config/mips/mips.c (mips_asan_shadow_offset): New function.
14685 (TARGET_ASAN_SHADOW_OFFSET): Define.
14686 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
14687 true for -fsanitize=address.
14689 2018-04-25 Mark Wielaard <mark@klomp.org>
14691 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
14694 2018-04-25 Jakub Jelinek <jakub@redhat.com>
14696 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
14697 than "alu", remove explicit "memory" and "imm_disp" attributes.
14698 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
14700 PR middle-end/85414
14701 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
14702 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
14703 gen_lowpart_no_emit.
14705 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
14708 * config/i386/i386.c (ix86_expand_builtin): Change memory
14709 operand to XI, extend p0 to Pmode.
14710 * config/i386/i386.md: Change unspec volatile and operand
14711 1 mode to XI, change operand 0 mode to P.
14713 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14715 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
14716 GET_MODE_MASK before any checking.
14717 (nds32_can_use_bset_p): Likewise.
14718 (nds32_can_use_btgl_p): Likewise.
14720 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14722 * config/nds32/nds32-doubleword.md: New define_split pattern for
14723 illegal register number.
14725 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14727 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
14729 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14731 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
14733 2018-04-25 Richard Biener <rguenther@suse.de>
14735 * lto-streamer.h (LTO_major_version): Bump to 8.
14737 2018-04-25 Jakub Jelinek <jakub@redhat.com>
14739 * BASE-VER: Set to 9.0.0.
14741 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
14743 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
14744 in __abskf2 and __powikf2.
14746 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14749 * config/aarch64/constraints.md (Usg, Usj): New constraints.
14750 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
14751 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
14752 Use the above on operand 2. Reindent.
14753 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
14755 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
14758 * common/config/i386/i386-common.c (ix86_handle_option): Don't
14760 * config/i386/i386.opt (mcet): Removed.
14761 * doc/install.texi: Remove -mcet documentation.
14762 * doc/invoke.texi: Likewise.
14764 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
14767 * doc/install.texi: Remove -mcet from bootstrap-cet.
14769 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14772 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
14773 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
14777 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
14778 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
14779 containing a CONST_VECTOR.
14781 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
14783 * doc/install.texi: Update newlib dependency for nvptx.
14785 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14788 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
14789 instead of INTVAL when shifting x left.
14791 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
14793 PR tree-optimization/85478
14794 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
14795 vect_grouped_store_supported for single element vectors.
14797 2018-04-24 Richard Biener <rguenther@suse.de>
14800 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
14801 load cost increase to the case of non-constant step.
14803 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14806 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
14807 destination if any_malformed_asm.
14809 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
14811 PR middle-end/85496
14812 * expr.c (store_field): In the bitfield case, if the value comes from
14813 a function call and is returned in registers by means of a PARALLEL,
14814 do not change the mode of the temporary unless BLKmode and VOIDmode.
14816 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
14818 PR rtl-optimization/85423
14819 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
14820 dependencies to debug insns when the previous insn is non-debug.
14822 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
14824 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
14825 enums into a single definition.
14826 (fls): Fix predicates and printing.
14829 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
14831 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
14832 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
14833 and short u6 immediate.
14834 (check_if_valid_sleep_operand): Remove.
14835 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
14837 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14839 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
14840 flag_always_save_lp condition.
14841 * config/nds32/nds32.opt (malways-save-lp): New option.
14843 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14845 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
14846 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
14847 * config/nds32/nds32.h
14848 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
14849 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
14851 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14853 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
14854 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
14856 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14857 Chung-Ju Wu <jasonwucj@gmail.com>
14859 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
14861 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
14862 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
14864 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14866 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
14868 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14870 * config/nds32/nds32-protos.h (nds32_data_alignment,
14871 nds32_local_alignment): Declare.
14872 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
14873 nds32_local_alignment): New functions.
14874 (TARGET_CONSTANT_ALIGNMENT): Define.
14875 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
14877 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14879 * config/nds32/nds32.c
14880 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
14881 (TARGET_MODES_TIEABLE_P): Likewise.
14883 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14885 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
14886 level Ofast and Og.
14888 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
14889 Chung-Ju Wu <jasonwucj@gmail.com>
14891 * config/nds32/constants.md (unspec_volatile_element): Add enum values
14892 for unaligned access.
14893 * config/nds32/nds32-intrinsic.c: Implementation of expanding
14895 * config/nds32/nds32-intrinsic.md: Likewise.
14896 * config/nds32/nds32_intrinsic.h: Likewise.
14897 * config/nds32/nds32.h (nds32_builtins): Likewise.
14898 * config/nds32/nds32.opt (munaligned-access): New option.
14899 * config/nds32/nds32.c (nds32_asm_file_start): Display
14900 flag_unaligned_access status.
14902 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
14904 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
14905 -mno-relax is present.
14906 * config/riscv/linux.h (LINK_SPEC): Ditto.
14908 2018-04-20 Martin Sebor <msebor@redhat.com>
14911 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
14913 (gimple_fold_builtin_stxcpy_chk): Same.
14914 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
14916 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
14919 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
14920 __powikf2 when long double is IEEE 128-bit.
14922 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
14924 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
14925 step to make sure stack always aligned.
14927 2018-04-20 Carl Love <cel@us.ibm.com>
14930 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
14931 size check for arg0.
14933 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
14934 Tom de Vries <tom@codesourcery.com>
14937 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
14938 Emit insns for calls too.
14939 (nvptx_find_par): Always look for worker-level predecessor insn.
14940 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
14942 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
14943 (nvptx_process_pars): Propagate frames for calls.
14945 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
14948 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
14950 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
14951 (ix86_handle_option): Don't handle OPT_mibt.
14952 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
14954 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
14956 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
14957 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
14958 (ix86_target_macros): Define __CET__ with flag_cf_protection
14959 for -fcf-protection.
14960 * config/i386/i386.c (isa2_opts): Remove -mibt.
14961 * config/i386/i386.h (TARGET_IBT): Removed.
14962 (TARGET_IBT_P): Likewise.
14963 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
14964 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
14965 * config/i386/i386.opt (mcet): Update help message.
14966 (mshstk): Likewise.
14968 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
14969 -mcet as an alias for -mshstk.
14971 2018-04-20 Richard Biener <rguenther@suse.de>
14973 PR middle-end/85475
14974 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
14975 complexity by forcing a single use of the multiply operand.
14977 2018-04-20 Martin Jambor <mjambor@suse.cz>
14980 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
14981 recursion dependency to only apply to non-clones.
14983 2018-04-20 Martin Jambor <mjambor@suse.cz>
14986 * ipa-cp.c (create_specialized_node): Check that clones of
14987 self-recursive edges exist during IPA-CP.
14989 2018-04-19 Toon Moene <toon@moene.org>
14991 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
14994 2018-04-19 Jakub Jelinek <jakub@redhat.com>
14996 PR tree-optimization/85467
14997 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
14998 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
14999 VECTOR_CST element to type.
15001 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
15004 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
15005 * config/i386/i386.md (builtin_setjmp_setup): Removed.
15006 (builtin_longjmp): Likewise.
15007 (save_stack_nonlocal): New pattern.
15008 (restore_stack_nonlocal): Likewise.
15010 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
15013 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
15014 Replace ASM_OUTPUT_LABEL with fprintf.
15016 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
15019 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
15020 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
15021 * config/i386/i386-c.c (ix86_target_macros_internal): Also
15022 define __IBT__ and __SHSTK__ for -fcf-protection.
15023 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
15025 (ix86_trampoline_init): Likewise.
15026 (x86_output_mi_thunk): Likewise.
15027 (ix86_notrack_prefixed_insn_p): Likewise.
15028 (ix86_option_override_internal): Don't disallow -fcf-protection.
15029 * config/i386/i386.md (rdssp<mode>): Also enable for
15031 (incssp<mode>): Likewise.
15032 (nop_endbr): Likewise.
15033 * config/i386/i386.opt (mcet): Change help message to built-in
15036 (mshstk): Likewise.
15037 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
15038 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
15039 enable CET built-in functions.
15041 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
15043 * common/config/i386/i386-common.c
15044 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
15045 OPTION_MASK_ISA_MOVDIRI_UNSET,
15046 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
15047 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
15048 * config.gcc (movdirintrin.h): New header.
15049 * config/i386/cpuid.h (bit_MOVDIRI,
15050 bit_MOVDIR64B): New bits.
15051 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
15053 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
15054 (VOID, PVOID, PCVOID)): New function types.
15055 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
15056 __builtin_ia32_directstoreu_u64,
15057 __builtin_ia32_movdir64b): New builtins.
15058 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
15059 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
15061 (ix86_valid_target_attribute_inner_p): Ditto.
15062 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
15063 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
15064 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
15065 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
15066 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
15067 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
15068 (movdiri<mode>, movdir64b_<mode>): New.
15069 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
15070 * config/i386/immintrin.h: Include movdirintrin.h.
15071 * config/i386/movdirintrin.h: New file.
15072 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
15074 2018-04-19 Richard Biener <rguenther@suse.de>
15076 PR middle-end/85455
15077 * cfg.c (clear_bb_flags): When loop state says we have
15078 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
15080 2018-04-19 Richard Biener <rguenther@suse.de>
15082 PR tree-optimization/84737
15083 * tree-vect-data-refs.c (vect_copy_ref_info): New function
15084 copying restrict info.
15085 (vect_setup_realignment): Use it.
15086 * tree-vectorizer.h (vect_copy_ref_info): Declare.
15087 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
15088 the first DR to all generated stores.
15089 (vectorizable_load): Likewise for loads.
15091 2018-04-19 Jakub Jelinek <jakub@redhat.com>
15093 PR tree-optimization/85446
15094 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
15095 the integral and pointer types to have the same precision.
15097 * doc/install.texi: Document --disable-cet being the default and
15100 2018-04-18 Martin Liska <mliska@suse.cz>
15102 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
15105 2018-04-18 Martin Liska <mliska@suse.cz>
15108 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
15111 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
15112 arguments if they are comparable.
15114 2018-04-18 Martin Liska <mliska@suse.cz>
15117 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
15120 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
15123 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
15126 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
15127 ENDBR after calling __morestack.
15129 2018-04-18 David Malcolm <dmalcolm@redhat.com>
15132 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
15133 by using gcc_base_ver to generate a gcc_driver_version, and use
15134 it when generating GCC_DRIVER_NAME.
15135 * configure: Regenerate.
15137 2018-04-18 Jakub Jelinek <jakub@redhat.com>
15140 * config.gcc: Obsolete powerpc*-*-*spe*.
15142 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15145 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
15146 (stabstr_D): Change type of unum from unsigned int to
15147 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
15150 2018-04-17 Jim Wilson <jimw@sifive.com>
15153 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
15154 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
15155 Set arg_pointer_offset after using pretend_args_size.
15157 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15159 PR rtl-optimization/85431
15160 * dse.c (record_store): Ignore zero width stores.
15163 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
15164 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
15165 __builtin_stack_restore rather than after it.
15166 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
15167 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
15168 argument instead of virtual_dynamic_stack_rtx.
15170 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
15172 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
15174 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15175 Add note to error message to explain internal mapping of overloaded
15176 built-in function name to non-overloaded built-in function name.
15177 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
15180 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
15183 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
15184 where the inputs overlap with the output.
15186 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15188 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
15189 (=v, v) alternative and explicit "memory" attribute.
15190 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
15191 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15193 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
15194 "sselog1" type instead of "sselog".
15195 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
15196 "sselog". Remove explicit "memory" attribute.
15197 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
15198 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15200 (vec_extract_hi_v32hi): Merge all alternatives into one, use
15201 "sselog1" type instead of "sselog". Remove explicit "memory"
15203 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
15204 use "sselog1" type instead of "sselog". Remove explicit "memory"
15206 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
15207 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15209 (vec_extract_hi_v64qi): Merge all alternatives into one, use
15210 "sselog1" type instead of "sselog". Remove explicit "memory"
15212 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
15213 use "sselog1" type instead of "sselog". Remove explicit "memory"
15217 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
15219 PR middle-end/85414
15220 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
15223 2018-04-17 Martin Jambor <mjambor@suse.cz>
15226 * ipa-cp.c (create_specialized_node): Call
15227 expand_all_artificial_thunks if necessary.
15229 2018-04-17 Martin Liska <mliska@suse.cz>
15232 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
15233 in message, remote space in between '_G' and '('.
15235 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15238 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
15239 avx512f_vmcmp<mode>3<round_saeonly_name>,
15240 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
15241 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
15242 avx512f_rndscale<mode><round_saeonly_name>,
15243 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
15244 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
15245 Use %<iptr>2 instead of %2 for -masm=intel.
15246 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
15247 avx512f_vcvttss2usi<round_saeonly_name>,
15248 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
15250 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
15251 avx512f_vcvttsd2usi<round_saeonly_name>,
15252 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
15253 Use %q1 instead of %1 for -masm=intel.
15254 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
15255 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
15256 of %3 for -masm=intel.
15257 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
15259 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
15261 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
15263 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
15264 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
15266 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
15268 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
15269 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
15270 %g1 and one with %0 and %1.
15271 (avx512er_vmrcp28<mode><round_saeonly_name>,
15272 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
15273 %1 for -masm=intel.
15274 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
15275 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
15276 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
15277 of %0 and %{%4%} for -masm=intel.
15278 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
15279 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
15280 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
15281 order of %0 and %{%5%}%{z%} for -masm=intel.
15283 2018-04-17 Jan Hubicka <jh@suse.cz>
15286 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
15288 2018-04-17 Martin Liska <mliska@suse.cz>
15291 * multiple_target.c (create_dispatcher_calls): Set apostrophes
15292 for target_clone error message. Make default implementation
15293 clone to be a local declaration.
15294 (separate_attrs): Add new argument and check for an empty
15296 (expand_target_clones): Handle it.
15297 (ipa_target_clone): Make redirection just for target_clones
15300 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
15301 Tom de Vries <tom@codesourcery.com>
15303 PR middle-end/84955
15304 * omp-expand.c (expand_oacc_for): Add dummy false branch for
15305 tiled basic blocks without omp continue statements.
15307 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
15310 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
15311 vec_extract expression as having side effects to make sure it gets
15314 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
15317 * config/i386/i386.c (get_builtin_code_for_version): Check
15320 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
15323 * config.gcc: Support "skylake".
15324 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15326 * config/i386/i386.c (m_SKYLAKE): Define.
15327 (processor_target_table): Add "skylake".
15328 (ix86_option_override_internal): Add "skylake".
15329 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
15330 PROCESSOR_CANNONLAKE.
15331 (get_builtin_code_for_version): Fix priority for
15332 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
15333 PROCESSOR_SKYLAKE-AVX512.
15334 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
15335 (processor_type): Add PROCESSOR_SKYLAKE.
15337 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
15338 Jason Merrill <jason@redhat.com>
15341 * convert.c (convert_to_integer_1): Use direct recursion for
15342 enumeral types and types with a precision less than the number
15343 of bits in their mode.
15345 2018-04-16 Julia Koval <julia.koval@intel.com>
15348 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
15349 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
15351 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
15354 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
15355 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
15356 and -mno-direct-move.
15358 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
15361 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
15362 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
15363 Ensure negative shifts result in {0}.
15365 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
15367 PR rtl-optimization/79916
15368 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
15369 regs (if any) to define how to gnerate SD moves when LRA is in
15372 2018-04-13 Jakub Jelinek <jakub@redhat.com>
15374 PR rtl-optimization/85393
15375 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
15376 * except.c (expand_dw2_landing_pad_for_region): Make static.
15377 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
15378 a label and unconditional jump to old_bb, rather than
15379 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
15382 PR rtl-optimization/85376
15383 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
15384 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
15385 instead of a specific value.
15387 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
15388 Bin Cheng <bin.cheng@arm.com>
15390 PR tree-optimization/82965
15391 PR tree-optimization/83991
15392 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
15393 by_profile_only parameter.
15394 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
15395 information if the loop was predicted to iterate too many times.
15396 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
15398 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
15401 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
15404 2018-04-13 Martin Liska <mliska@suse.cz>
15405 Jakub Jelinek <jakub@redhat.com>
15407 PR middle-end/81657
15408 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
15409 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
15410 * builtins.c (expand_builtin_memory_copy_args): Use
15411 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
15412 handle dest_addr == pc_rtx.
15414 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
15417 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
15418 asked to not generate direct moves.
15419 (fix_trunc<mode>si2_stfiwx): Similar.
15420 (fix_trunc<mode>si2_internal): Similar.
15422 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15425 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
15426 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
15427 lookup if dest in some wider mode is known to be const0_rtx and
15428 if so, record permanent equivalence for it to be ZERO_EXTEND of
15429 the narrower mode destination.
15431 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
15433 * lto-streamer-out.c (output_function): Revert 259346.
15434 * omp-expand.c (expand_oacc_for): Likewise.
15436 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
15438 PR rtl-optimization/85354
15439 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
15440 * sel-sched.c (sel_global_init): ... here.
15442 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
15445 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
15446 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
15447 mode for PE-COFF targets.
15448 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
15449 (i386_pe_asm_lto_end): Likewise.
15450 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
15451 (TARGET_ASM_LTO_END): Likewise.
15452 * config/i386/winnt.c (saved_debug_info_level): New static variable.
15453 (i386_pe_asm_lto_start): New function.
15454 (i386_pe_asm_lto_end): Likewise.
15456 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
15457 Richard Biener <rguenther@suse.de>
15459 PR middle-end/84955
15460 * lto-streamer-out.c (output_function): Fix CFG loop state before
15462 * omp-expand.c (expand_oacc_for): Handle calls to internal
15463 functions like regular functions.
15465 2018-04-12 Richard Biener <rguenther@suse.de>
15468 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
15469 for the early LTO debug to properly generate references to it
15470 during DIE emission. Do not re-use that for the skeleton for
15472 (dwarf2out_early_finish): Likewise.
15474 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15477 * config/i386/sse.md
15478 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
15479 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
15480 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
15481 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
15482 and output is a reg, avoid creating invalid lowpart subreg, but
15483 instead split into a 512-bit move. Don't split if not AVX512VL,
15484 input is xmm16+ reg and output is a mem.
15485 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
15486 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
15487 xmm16+ reg and output is a mem.
15489 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15491 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
15492 also for flag_dwarf2_cfi_asm.
15494 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15496 PR rtl-optimization/85342
15497 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
15498 a bool scalar var inside of the loop instead. Don't try to update
15499 recog_data.operand after failed apply_change_group.
15501 2018-04-12 Tom de Vries <tom@codesourcery.com>
15504 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
15505 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
15506 array with flexible array member as array without given dimension.
15507 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
15508 argument for undefined param to true.
15510 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
15513 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15514 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
15515 from PowerPC section.
15516 * config/rs6000/sysv4.opt (mcall-): Improve help text.
15517 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
15518 help text that is too long.
15519 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
15520 help text that is too long.
15521 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
15522 help text that is too long.
15524 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
15526 * config/alpha/alpha.md (stack_probe_internal): Rename
15527 from "probe_stack". Update all callers.
15529 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15531 PR rtl-optimization/84566
15532 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
15533 sched_macro_fuse_insns.
15535 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15538 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
15539 (compute_block_dependences): ... from here.
15541 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15543 PR tree-optimization/85331
15544 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
15545 from int to HOST_WIDE_INT.
15547 2018-04-11 Martin Jambor <mjambor@suse.cz>
15550 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
15551 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
15552 not the same as the source val.
15553 (cgraph_edge_brings_value_p): New parameter.
15554 (gather_edges_for_value): Pass destination value to
15555 cgraph_edge_brings_value_p.
15556 (perhaps_add_new_callers): Likewise.
15557 (get_info_about_necessary_edges): Likewise and exclude values brought
15558 only by self-recursive edges.
15559 (create_specialized_node): Redirect only clones of self-calling edges.
15560 (+self_recursive_pass_through_p): New function.
15561 (find_more_scalar_values_for_callers_subset): Use it.
15562 (find_aggregate_values_for_callers_subset): Likewise.
15563 (known_aggs_to_agg_replacement_list): Removed.
15564 (decide_whether_version_node): Re-calculate known constants for all
15565 remaining context clones.
15567 2018-04-11 Richard Biener <rguenther@suse.de>
15570 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
15571 from early DWARF output.
15572 (dwarf2out_early_finish): Output line info unconditionally into
15573 early DWARF and add reference to it.
15575 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15578 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
15579 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
15580 other than V2DFmode using iptr mode attribute.
15581 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
15583 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15585 PR rtl-optimization/84659
15586 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
15588 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15591 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
15593 (output_loc_list): Pass address of a dummy size variable even in the
15594 locview handling loop.
15595 (index_location_lists): Add comment on why skip_loc_list_entry can't
15598 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
15601 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
15604 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
15607 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15608 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
15609 and -mstring-compare-inline-limit.
15611 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15614 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
15615 for stack clash protection in a register whenever we need it to be in
15618 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15620 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
15621 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
15623 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15626 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
15628 (mlong-double-): Ditto.
15629 * config/rs6000/sysv4.opt (msdata=): Ditto.
15630 (mtls-size=): Ditto.
15632 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15634 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
15635 erroneous entries for
15636 "vector int vec_ldl (int, long int *)", and
15637 "vector unsigned int vec_ldl (int, unsigned long int *)".
15638 Add comments and entries for
15639 "vector bool char vec_ldl (int, bool char *)",
15640 "vector bool short vec_ldl (int, bool short *)",
15641 "vector bool int vec_ldl (int, bool int *)",
15642 "vector bool long long vec_ldl (int, bool long long *)",
15643 "vector pixel vec_ldl (int, pixel *)",
15644 "vector long long vec_ldl (int, long long *)",
15645 "vector unsigned long long vec_ldl (int, unsigned long long *)".
15646 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
15647 type tree bool_long_long_type_node and correct definition of
15648 bool_V2DI_type_node to make reference to this new type tree.
15649 (rs6000_mangle_type): Replace erroneous reference to
15650 bool_long_type_node with bool_long_long_type_node.
15651 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
15652 comments to emphasize sign distinctions for char and int types and
15653 replace RS6000_BTI_bool_long constant with
15654 RS6000_BTI_bool_long_long constant. Also add comment to restrict
15655 use of RS6000_BTI_pixel.
15656 (bool_long_type_node): Remove this macro definition.
15657 (bool_long_long_type_node): New macro definition
15659 2018-04-10 Jakub Jelinek <jakub@redhat.com>
15661 PR rtl-optimization/85300
15662 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
15663 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
15664 simplify_unary_operation fails.
15666 2018-04-10 Martin Liska <mliska@suse.cz>
15668 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
15669 cgraph_edge and ipa_ref.
15671 2018-04-10 Jakub Jelinek <jakub@redhat.com>
15675 * config/i386/sse.md
15676 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
15677 computation of the VEC_MERGE selector from mask.
15678 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
15679 Fix decoding of the VEC_MERGE selector into mask.
15681 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
15683 PR tree-optimization/85286
15684 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
15686 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
15688 * final.c (final_1): Set insn_last_address as well as
15689 insn_current_address.
15691 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15694 * explow.c (emit_stack_probe): Call validize_mem on memory location
15695 before passing it to gen_probe_stack. Create address operand and
15696 legitimize it for the probe_stack_address case.
15698 2018-04-09 Jan Hubicka <jh@suse.cz>
15701 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
15702 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
15703 * tree.c (free_lang_data_in_type): Fix handling of binfos;
15705 (free_lang_data): Rebuild type inheritance graph.
15707 2018-04-09 Martin Sebor <msebor@redhat.com>
15709 * invoke.texi (-finline-small-functions): Mention other optimization
15711 (-findirect-inlining, -fpartial-inlining): Same.
15712 (-finline-functions-called-once): Same.
15713 (-freorder-blocks-and-partition): Same.
15715 2018-04-09 Jan Hubicka <jh@suse.cz>
15718 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
15719 jumps; choose last target that matches the criteria (i.e.
15720 no partition changes for non-crossing jumps).
15721 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
15722 support for redirecting crossing jumps to non-crossing.
15724 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
15726 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
15727 also for naked functions.
15729 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
15731 * config/arc/arc.md (add_shift): New pattern.
15732 (add_shift2): Likewise.
15733 (sub_shift): Likewise.
15734 (sub_shift_cmp0_noout): Likewise.
15735 (compare_si_ashiftsi): Likewise.
15736 (xbfu_cmp0_noout): New combine pattern.
15737 (xbfu_cmp0"): Likewise.
15738 (movsi_set_cc_insn): Place the predicable variant first.
15739 (commutative_binary_cmp0_noout): Remove clobber.
15740 (commutative_binary_cmp0): New pattern.
15741 (noncommutative_binary_cmp0): Likewise.
15742 (noncommutative_binary_cmp0_noout): Likewise.
15743 (noncommutative_binary_comparison_result_used): Removed.
15744 (rsub_cmp0): New pattern.
15745 (rsub_cmp0_noout): Likewise.
15746 (extzvsi): Changed, keep only meaningful variants.
15747 (SQH, SEZ): New iterators.
15748 (SQH_postfix): New mode attribute.
15749 (SEZ_prefix): New code attribute.
15750 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
15751 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
15752 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
15753 of numerical value.
15754 (noncommutative_operator): Check the availability of barrel
15757 2018-04-09 Richard Biener <rguenther@suse.de>
15759 PR tree-optimization/85284
15760 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
15761 Only use the niter constraining form of simple_iv when the exit
15762 is always executed.
15764 2018-04-09 Tom de Vries <tom@codesourcery.com>
15767 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15768 (define_expand "*memory_barrier"): New define_expand.
15769 (define_insn "memory_barrier"): New insn.
15771 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15773 PR rtl-optimization/80463
15774 PR rtl-optimization/83972
15775 PR rtl-optimization/83480
15777 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
15778 correct producer for the insn.
15779 (tidy_control_flow): Fixup seqnos in case of debug insns.
15781 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15783 PR rtl-optimization/83913
15785 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
15786 different sched-times when merging exprs.
15788 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15790 PR rtl-optimization/83962
15792 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
15793 tidy_fallthru_edge and tidy_control_flow.
15795 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15797 PR rtl-optimization/83530
15799 * sel-sched.c (force_next_insn): New global variable.
15800 (remove_insn_for_debug): When force_next_insn is true, also leave only
15801 next insn in the ready list.
15802 (sel_sched_region): When the region wasn't scheduled, make another pass
15803 over it with force_next_insn set to 1.
15805 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
15807 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
15809 * config/nds32/constants.md (unspec_volatile_element): Add enum values
15810 for interrupt control.
15811 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15812 functions for interrupt control.
15813 * config/nds32/nds32-intrinsic.md: Likewise.
15814 * config/nds32/nds32_intrinsic.h: Likewise.
15815 * config/nds32/nds32.h (nds32_builtins): Likewise.
15817 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
15819 * config/nds32/nds32.c (nds32_init_machine_status,
15820 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
15821 strict_aligned_p field.
15822 (nds32_expand_to_rtl_hook): New function.
15823 (TARGET_EXPAND_TO_RTL_HOOK): Define.
15824 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
15826 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15827 Chung-Ju Wu <jasonwucj@gmail.com>
15829 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
15830 * config/nds32/nds32-n7.md: New file.
15831 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
15832 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
15834 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
15835 * config/nds32/nds32.md (pipeline_model): Add n7.
15836 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
15837 * config/nds32/pipelines.md: Include n7 settings.
15839 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15840 Chung-Ju Wu <jasonwucj@gmail.com>
15842 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
15843 * config/nds32/nds32-e8.md: New file.
15844 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
15845 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
15847 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
15848 * config/nds32/nds32.md (pipeline_model): Add e8.
15849 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
15850 * config/nds32/pipelines.md: Include e8 settings.
15852 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15853 Chung-Ju Wu <jasonwucj@gmail.com>
15855 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
15856 * config/nds32/nds32-n8.md: New file.
15857 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
15858 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
15860 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
15861 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
15862 * config/nds32/nds32.md (pipeline_model): Add n8.
15863 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
15864 * config/nds32/pipelines.md: Include n8 settings.
15866 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15867 Chung-Ju Wu <jasonwucj@gmail.com>
15869 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
15870 * config/nds32/nds32-n9-2r1w.md: New file.
15871 * config/nds32/nds32-n9-3r2w.md: New file.
15872 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
15873 nds32_register_ports): New or modify for cpu n9.
15874 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
15876 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
15877 * config/nds32/nds32-utils.c: New file.
15878 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
15879 TARGET_MUL_SLOW): Define.
15880 * config/nds32/nds32.md (pipeline_model): New attribute.
15881 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
15882 New options that support cpu n9.
15883 * config/nds32/pipelines.md: Include n9 settings.
15884 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
15886 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
15888 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
15889 information if necessary.
15890 (output_cond_branch_compare_zero): Likewise.
15891 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
15892 (nds32_target_alignment): Refine for alignment.
15893 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
15894 (FUNCTION_BOUNDARY): Modify.
15895 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
15897 * config/nds32/nds32.opt (malways-align, malign-functions): New.
15899 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
15901 * config/nds32/constants.md (unspec_volatile_element): Add values for
15902 TLB operation and data prefetch.
15903 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15904 functions for TLB operation and data prefetch.
15905 * config/nds32/nds32-intrinsic.md: Likewise.
15906 * config/nds32/nds32_intrinsic.h: Likewise.
15907 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
15908 (nds32_print_operand): Likewise.
15909 * config/nds32/nds32.h (nds32_builtins): Likewise.
15911 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
15912 Andrew Pinski <pinsika@gcc.gnu.org>
15914 PR middle-end/82976
15915 * match.pd: Use constant_boolean_node of correct type instead of
15916 boolean_true_node or boolean_false_node for simplifying
15917 pointer comparisons to zero.
15919 2018-04-07 Jakub Jelinek <jakub@redhat.com>
15921 PR tree-optimization/80021
15922 * tree.c (verify_type_variant): Make error call in verify_variant_match
15923 translatable and remove final full stop.
15925 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15927 * config/nds32/constants.md (unspec_volatile_element): Add
15928 UNSPEC_VOLATILE_EH_RETURN.
15929 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
15930 nds32_output_stack_pop): Support dwarf exception handling process.
15931 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
15932 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
15933 exception handling process.
15934 (nds32_compute_stack_frame): Likewise.
15935 (nds32_return_addr_rtx): Likewise.
15936 (nds32_initial_elimination_offset): Likewise.
15937 (nds32_expand_prologue): Likewise.
15938 (nds32_expand_epilogue): Likewise.
15939 (nds32_dynamic_chain_address): New function.
15940 * config/nds32/nds32.h (machine_function): Add fields for dwarf
15941 exception handling.
15942 (DYNAMIC_CHAIN_ADDRESS): Define.
15943 (EH_RETURN_DATA_REGNO): Define.
15944 (EH_RETURN_STACKADJ_RTX): Define.
15945 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
15946 patterns for dwarf exception handling.
15948 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15950 * config/nds32/nds32.h: Clean up obsolete macros.
15952 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15954 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15955 Add enum values for particular instructions.
15956 * config/nds32/nds32-intrinsic.c: Implementation of expanding
15957 particular intrinsic functions.
15958 * config/nds32/nds32-intrinsic.md: Likewise.
15959 * config/nds32/nds32_intrinsic.h: Likewise.
15960 * config/nds32/nds32.h (nds32_builtins): Likewise.
15961 * config/nds32/nds32.md (type): Add pbsad and pbsada.
15962 (btst, ave): New patterns for particular instructions.
15964 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15966 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15967 Add enum values for atomic load/store and memory sync.
15968 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
15970 * config/nds32/nds32-intrinsic.md: Likewise.
15971 * config/nds32/nds32_intrinsic.h: Likewise.
15972 * config/nds32/nds32.h (nds32_builtins): Likewise.
15974 2018-04-07 Jakub Jelinek <jakub@redhat.com>
15976 PR tree-optimization/85257
15977 * fold-const.c (native_encode_vector): If not all elts could fit
15978 and off is -1, return 0 rather than offset.
15979 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
15980 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
15981 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
15982 adjust buffer in native_interpret_expr call.
15984 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15986 * config/nds32/constants.md (unspec_volatile_element): Add cache
15987 control enum values.
15988 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
15989 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
15990 * config/nds32/nds32.c (nds32_cctl_names): New.
15991 (nds32_print_operand): Handle cache control register names.
15992 * config/nds32/nds32.h (nds32_builtins): New enum values.
15993 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
15995 * config/nds32/nds32.md (type): Add mmu.
15996 * config/nds32/pipelines.md (simple_insn): Add mmu.
15998 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
16000 * config/nds32/nds32.md (type): Remove call.
16001 * config/nds32/pipelines.md (simple_insn): Likewise.
16003 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16005 * config/nds32/constants.md (unspec_volatile_element): Add
16006 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
16007 UNSPEC_VOLATILE_FMFCFG.
16008 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
16009 description for fmfcfg and fmfcsr.
16010 (bdesc_1arg): Add fmtcsr.
16011 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
16012 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
16013 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
16014 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
16015 unspec_fmfcfg): New patterns.
16016 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
16017 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
16018 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
16019 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
16020 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
16021 __nds32__fmfcfg): Define.
16023 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16025 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
16026 intrinsic register names.
16027 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
16028 intrinsic register enum values and macros.
16030 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
16032 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
16033 for load/store addressing form.
16034 (nds32_print_operand_address): Likewise.
16036 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
16039 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
16040 based on LABEL_REF. Remove useless assertion.
16041 (pic_address_needs_scratch): Fix formatting.
16042 (sparc_legitimize_pic_address): Minor tweaks.
16043 (sparc_delegitimize_address): Adjust assertion accordingly.
16044 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
16045 into symbolic_operand.
16046 (movsi_high_pic_label_ref): Likewise.
16047 (movsi_lo_sum_pic_label_ref): Likewise.
16048 (movdi_pic_label_ref): Likewise.
16049 (movdi_high_pic_label_ref): Likewise.
16050 (movdi_lo_sum_pic_label_ref): Likewise.
16052 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
16054 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
16055 custom LIB_SPEC setup.
16057 2018-04-06 Ruslan Bukin <br@bsdpad.com>
16058 Kito Cheng <kito.cheng@gmail.com>
16060 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
16061 * config/riscv/freebsd.h: New.
16063 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
16065 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
16066 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
16069 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
16070 Kito Cheng <kito.cheng@gmail.com>
16072 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
16073 nds32_output_call, nds32_symbol_binds_local_p): New functions.
16074 * config/nds32/nds32-protos.h (nds32_output_call,
16075 nds32_output_return): Declare.
16076 * config/nds32/nds32.md: Refine all the call and return patterns.
16078 2018-04-06 Jakub Jelinek <jakub@redhat.com>
16081 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
16082 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
16084 PR rtl-optimization/84872
16085 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
16086 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
16087 EDGE_CROSSING edge.
16089 2018-04-06 Tamar Christina <tamar.christina@arm.com>
16091 * expr.c (copy_blkmode_to_reg): Revert 254862.
16092 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
16094 2018-04-06 Richard Biener <rguenther@suse.de>
16096 PR middle-end/85244
16097 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
16098 after seeing a component reference with an adjacent field. Treat
16099 refs to arrays at struct end of external decls similar to
16100 refs to unconstrained commons.
16102 2018-04-06 Jakub Jelinek <jakub@redhat.com>
16105 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
16106 look through SAVE_EXPRs with non-side-effects argument. Adjust
16108 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
16111 2018-04-06 Richard Biener <rguenther@suse.de>
16113 PR middle-end/85180
16114 * alias.c (find_base_term): New wrapper around find_base_term
16115 unwinding CSELIB_VAL_PTR changes.
16116 (find_base_term): Do not restore CSELIB_VAL_PTR during the
16119 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16121 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
16123 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
16124 constant definitions.
16125 ("nop"): lr 0,0 -> nopr r0
16126 ("nop_lr0", "nop_lr1"): New insn definitions.
16128 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
16130 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
16131 NDS32_V3PUSH_AVAILABLE_P macro.
16133 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
16134 Chung-Ju Wu <jasonwucj@gmail.com>
16136 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
16137 (nds32*-*-*): Add float and fpu_config into supported_defaults.
16138 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
16139 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
16140 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
16141 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
16142 * config/nds32/constraints.md: New constraints and checking for hard
16143 float configuration.
16144 * config/nds32/iterators.md: New mode iterator and attribute for hard
16145 float configuration.
16146 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
16148 * config/nds32/nds32-fpu.md: New file.
16149 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
16150 deal with hard float code generation.
16151 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
16153 (abi_type, float_reg_number): New enum type.
16154 * config/nds32/nds32-predicates.c: New predicates for hard float.
16155 * config/nds32/nds32-protos.h: Declare functions for hard float.
16156 * config/nds32/nds32.c: Implementation for hard float configuration.
16157 * config/nds32/nds32.h: Definitions for hard float configuration.
16158 * config/nds32/nds32.md: Include hard float machine description and
16159 modify patterns for hard float configuration.
16160 * config/nds32/nds32.opt: New options for hard float configuration.
16161 * config/nds32/predicates.md: New predicates for hard float
16164 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
16166 * common/config/nds32/nds32-common.c
16167 (nds32_option_optimization_table): Enable -mreleax-hint by default.
16169 2018-04-05 Jakub Jelinek <jakub@redhat.com>
16171 PR middle-end/85195
16172 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
16173 CONSTRUCTOR_ELT (ctor, ...)->value.
16175 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
16178 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
16180 2018-04-05 Tom de Vries <tom@codesourcery.com>
16183 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
16186 2018-04-05 Shiva Chen <shiva0217@gmail.com>
16187 Kito Cheng <kito.cheng@gmail.com>
16189 * config/nds32/constraints.md (U33): Fine-tune checking condition.
16190 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
16191 * config/nds32/nds32.h (nds32_16bit_address_type): Add
16192 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
16194 2018-04-05 Shiva Chen <shiva0217@gmail.com>
16195 Kito Cheng <kito.cheng@gmail.com>
16197 * config/nds32/constraints.md (Ufe): New memory constraint.
16198 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
16199 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
16200 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
16202 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
16203 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
16205 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16207 * config/nds32/nds32.md: Use optimize_size in the condition for
16208 alu-shift instructions.
16210 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16212 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
16214 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16216 * config/nds32/nds32.md (negsi2): Refine pattern.
16218 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
16219 Chung-Ju Wu <jasonwucj@gmail.com>
16221 * config/nds32/iterators.md (shift_rotate): New code iterator.
16222 (shift): New code attribute.
16223 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
16224 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
16225 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
16226 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
16227 bit-wise operations.
16228 (andsi3, *andsi3): Ditto.
16229 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
16230 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
16231 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
16232 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
16233 nds32_ior_operand, nds32_xor_operand): New predicates.
16235 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16237 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
16238 (addsi3, subsi3): ... this.
16240 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16242 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
16244 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16246 * config/nds32/nds32.md: Adjust indention.
16248 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
16250 * config/nds32/nds32.md (feature): New attribute.
16252 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
16254 * config/nds32/nds32.md (subtype): New attribute.
16256 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
16259 * config/arm/arm-builtins.c (arm_expand_builtin): Change
16260 expansion to perform a bitwise AND of the argument followed by a
16261 boolean negation of the result.
16263 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
16265 PR rtl-optimization/84878
16266 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
16267 the basic block. Assert the use reference is not artificial and that
16268 it has an associated insn.
16270 2018-04-04 Michael Matz <matz@suse.de>
16272 * builtins.c (compute_objsize): Pass correct operand
16273 to array_at_struct_end_p.
16275 2018-04-04 Richard Biener <rguenther@suse.de>
16278 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
16279 from contexts for DINFO_LEVEL_TERSE and below.
16281 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
16283 * config/nds32/nds32-doubleword.md (move_<mode>): Require
16284 resiter_operand condition.
16285 * config/nds32/nds32.md (*move<mode>): Ditto.
16287 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
16288 Monk Chiang <sh.chiang04@gmail.com>
16290 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
16292 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16294 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
16296 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16297 Kito Cheng <kito.cheng@gmail.com>
16299 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
16300 nds32_cond_code_str, output_cond_branch,
16301 output_cond_branch_compare_zero, nds32_expand_cbranch,
16302 nds32_expand_cstore, nds32_expand_movcc,
16303 nds32_output_cbranchsi4_equality_zero,
16304 nds32_output_cbranchsi4_equality_reg,
16305 nds32_output_cbranchsi4_equality_reg_or_const_int,
16306 nds32_output_cbranchsi4_greater_less_zero: New functions.
16307 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
16308 nds32_expand_cstore, nds32_expand_movcc,
16309 nds32_output_cbranchsi4_equality_zero,
16310 nds32_output_cbranchsi4_equality_reg,
16311 nds32_output_cbranchsi4_equality_reg_or_const_int,
16312 nds32_output_cbranchsi4_greater_less_zero): Declare.
16313 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
16314 nds32_rimm11s_operand): New predicates.
16315 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
16316 * config/nds32/nds32.md: Rewrite all the branch and conditional move
16319 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
16321 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
16322 * config/nds32/nds32.md: Ditto.
16323 * config/nds32/pipelines.md: Ditto.
16325 2018-04-04 Richard Biener <rguenther@suse.de>
16327 PR tree-optimization/85168
16328 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
16329 propagating abnormals.
16331 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16333 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
16335 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16336 Kito Cheng <kito.cheng@gmail.com>
16338 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
16339 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
16340 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
16341 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
16342 * config/nds32/nds32.md (sibcall_internal): New.
16343 (sibcall_register): Remove.
16344 (sibcall_immediate): Remove.
16345 (sibcall_value_internal): New.
16346 (sibcall_value_register): Remove.
16347 (sibcall_value_immediate): Remove.
16348 * config/nds32/predicates.md (nds32_general_register_operand): New.
16349 (nds32_call_address_operand): New.
16351 2018-04-03 Jakub Jelinek <jakub@redhat.com>
16353 PR rtl-optimization/85167
16354 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
16355 bb_defs if *split_p, instead preinitialize it to NULL.
16357 PR tree-optimization/85156
16358 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
16359 evaluating the argument multiple times.
16361 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
16363 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
16365 (_mm_cvtpd_ps): Likewise.
16366 (_mm_cvttpd_epi32): Likewise.
16367 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
16368 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
16369 vector, pixel, and bool following altivec.h include.
16371 2018-04-03 Martin Sebor <msebor@redhat.com>
16373 * doc/extend.texi (Common Function Attributes): Clarify.
16374 (const attribute): Likewise.
16375 (pure attribute): Likewise.
16377 2018-04-03 Jakub Jelinek <jakub@redhat.com>
16380 * config/i386/i386.c (ix86_expand_vector_set): Use
16381 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
16383 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
16385 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
16386 instructions when changing rounding bits to preserve precision bits
16387 in the x87 control word.
16389 2018-04-03 Martin Liska <mliska@suse.cz>
16391 PR tree-optimization/82491
16392 * rtl.h (strip_offset_and_add): Replace += suboffset with
16393 poly_uint64 () + suboffset.
16395 2018-03-29 Martin Liska <mliska@suse.cz>
16396 Martin Jambor <mjambor@suse.cz>
16399 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
16400 param_type is not an integral or pointer type.
16402 2018-04-03 Richard Biener <rguenther@suse.de>
16404 * sese.h (recompute_all_dominators): Remove.
16406 2018-04-02 Martin Sebor <msebor@redhat.com>
16408 * doc/invoke.texi (-Wrestrict): Fix typos.
16410 2018-04-02 Jim Wilson <jimw@sifive.com>
16412 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
16413 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
16414 (<optab>di3, <optab>si3_extend): Likewise.
16415 (<optab>si3_mask, <optab>si3_mask_1): New.
16416 (<optab>di3_mask, <optab>di3_mask_1): New.
16417 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
16418 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
16419 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
16421 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
16423 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
16426 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
16428 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
16429 (nds32_canonicalize_comparison): New function.
16431 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16432 Kito Cheng <kito.cheng@gmail.com>
16433 Kuan-Lin Chen <kuanlinchentw@gmail.com>
16435 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
16436 * config/nds32/constants.md (unspec_volatile_element): Add
16437 UNSPEC_VOLATILE_RELAX_GROUP.
16438 * config/nds32/nds32-relax-opt.c: New file.
16439 * config/nds32/nds32-predicates.c
16440 (nds32_symbol_load_store_p): New function.
16441 * config/nds32/nds32-protos.h
16442 (nds32_symbol_load_store_p): Declare function.
16443 (make_pass_nds32_relax_opt): Declare new rtl pass function.
16444 * config/nds32/nds32.c
16445 (nds32_register_pass): New function to register pass.
16446 (nds32_register_passes): New function to register passes.
16447 * config/nds32/nds32.md (relax_group): New pattern.
16448 * config/nds32/nds32.opt (mrelax-hint): New option.
16449 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
16451 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
16453 * config/nds32/t-nds32: Modify files dependency.
16455 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16457 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
16458 (PROFILE_HOOK): Define its implementation.
16460 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16462 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
16463 type and 32-bit size.
16465 2018-04-01 Jakub Jelinek <jakub@redhat.com>
16467 PR middle-end/85090
16468 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
16469 (V_128_256): New mode iterator.
16470 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
16471 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
16472 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
16474 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
16475 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
16477 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
16480 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
16481 NaN inputs correctly.
16483 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
16486 * config/rs6000/vsx.md (??r): New mode attribute.
16487 (*vsx_mov<mode>_64bit): Use it.
16488 (*vsx_mov<mode>_32bit): Likewise.
16490 2018-03-30 Martin Sebor <msebor@redhat.com>
16492 PR tree-optimization/84818
16493 * builtins.c (check_access): Use warning_n.
16495 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16498 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
16500 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
16503 2018-03-30 Julia Koval <julia.koval@intel.com>
16506 * x86-tune.def (movx, partial_reg_dependency): Enable for
16509 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
16511 PR inline-asm/84985
16512 * lra-constraints.c (process_alt_operands): Move setting
16513 this_alternative_matches below.
16515 2018-03-29 Martin Liska <mliska@suse.cz>
16518 * doc/invoke.texi: Document how LTO works with debug info.
16519 Describe auto-load support of binutils. Mention 'x86-64'
16520 as valid option value of -march option.
16522 2018-03-29 Jakub Jelinek <jakub@redhat.com>
16524 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
16527 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
16528 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
16529 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
16532 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
16535 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
16536 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
16537 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
16538 for RS6000_BTM_POWERPC64.
16539 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
16540 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
16541 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
16544 (DIVDEU): Likewise.
16546 2018-03-28 Carl Love <cel@us.ibm.com>
16549 2017-09-27 Carl Love <cel@us.ibm.com>
16551 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
16552 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
16553 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
16556 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16558 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
16559 instead of __vector bool.
16560 (_mm_max_pu8): Likewise.
16561 (_mm_min_pi16): Likewise.
16563 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
16566 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
16567 (DIVWEUO): Likewise.
16568 (DIVDEO): Likewise.
16569 (DIVDEUO): Likewise.
16570 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
16571 DIVWEUO and DIVDEUO.
16572 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
16573 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
16574 (div_extend): Likewise.
16575 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
16577 (__builtin_divweuo): Likewise.
16578 (__builtin_divdeo): Likewise.
16579 (__builtin_divdeuo): Likewise.
16581 2018-03-28 Jakub Jelinek <jakub@redhat.com>
16584 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
16585 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
16587 PR tree-optimization/82004
16588 * gimple-match-head.c (optimize_pow_to_exp): New function.
16589 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
16590 Don't fold to exp if optimize_pow_to_exp is false.
16592 2018-03-28 Martin Liska <mliska@suse.cz>
16595 * calls.c (initialize_argument_information): Fix trailing space.
16596 * common.opt: Fix typo and provide better explanation for
16597 -fsanitize-coverage option.
16598 * config/i386/i386.opt: Fix typo.
16600 2018-03-28 Jakub Jelinek <jakub@redhat.com>
16601 Martin Liska <mliska@suse.cz>
16604 * gimplify.c (asan_poison_variable): Don't do the check for
16605 gimplify_omp_ctxp here.
16606 (gimplify_decl_expr): Do it here.
16607 (gimplify_target_expr): Likewise.
16609 2018-03-28 Martin Liska <mliska@suse.cz>
16612 * config/i386/i386.c (ix86_function_arg_advance): Do not call
16613 chkp_type_bounds_count if MPX is not enabled.
16615 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
16617 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
16619 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
16622 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
16623 function to create the function decl for complex long double
16624 multiply and divide for -mabi=ieeelongdouble.
16625 (init_float128_ieee): Call it.
16627 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
16630 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
16631 -fcf-protection=branch -mibt.
16632 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
16634 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16637 * config/arm/arm.c (arm_valid_symbolic_address): Handle
16638 arm_word_relocations.
16640 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
16643 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
16644 extern array declarations.
16646 2018-03-27 Richard Biener <rguenther@suse.de>
16648 PR middle-end/84067
16649 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
16650 explicit single_use checks.
16652 2018-03-27 Richard Biener <rguenther@suse.de>
16654 PR tree-optimization/85082
16655 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
16658 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16660 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
16661 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
16662 Turn on fasynchronous-unwind-tables and funwind-tables.
16664 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
16667 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
16668 (*bmi_blsr_<mode>_ccz): Ditto.
16670 2018-03-26 Tom de Vries <tom@codesourcery.com>
16672 PR tree-optimization/85063
16673 * omp-general.c (offloading_function_p): New function. Factor out
16675 * omp-offload.c (pass_omp_target_link::gate): ... here.
16676 * omp-general.h (offloading_function_p): Declare.
16677 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
16678 with attribute omp declare target for offloading functions.
16680 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
16682 PR tree-optimization/84005
16683 * tree-data-ref.h (get_base_for_alignment): Declare.
16684 * tree-data-ref.c (get_base_for_alignment_1): New function.
16685 (get_base_for_alignment): Likewise.
16686 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
16687 get_base_for_alignment to find a suitable base object, instead
16688 of always using drb->base_address.
16690 2018-03-23 Jakub Jelinek <jakub@redhat.com>
16692 PR inline-asm/85022
16693 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
16694 known size by default.
16696 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
16698 PR inline-asm/85030
16699 * lra-constraints.c (process_alt_operands): Don't match BLKmode
16700 and non BLKmode operands.
16702 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16705 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
16706 Clean up attributes.
16708 2018-03-23 Richard Biener <rguenther@suse.de>
16711 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
16712 we are going to emit early debug for LTO.
16714 2018-03-23 Jakub Jelinek <jakub@redhat.com>
16716 PR inline-asm/85034
16717 * function.c (match_asm_constraints_1): Don't optimize if input
16718 doesn't satisfy general_operand predicate for output's mode.
16720 PR inline-asm/85022
16721 * alias.c (write_dependence_p): Don't require for x_canonicalized
16722 non-VOIDmode if x has VOIDmode.
16725 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
16726 just don't try to optimize it rather than assert it never happens.
16728 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
16730 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
16731 macro expansions for definition of ST_INTERNAL_<mode> and
16732 LD_INTERNAL_<mode> builtins.
16733 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
16735 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
16737 (altivec_expand_st_builtin): Likewise.
16738 (altivec_expand_builtin): Remove calls to deleted functions.
16739 (rs6000_address_for_altivec): Delete this function.
16740 * config/rs6000/vector.md: Remove expands for
16741 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
16743 2018-03-22 Sudakshina Das <sudi.das@arm.com>
16746 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
16747 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
16748 re-computing once computed.
16749 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
16750 (arm_init_machine_status): Initialize
16751 machine->static_chain_stack_bytes.
16753 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
16756 * doc/extend.texi: Add four new prototypes for vec_ld.
16757 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
16758 definitions for more logical presentation.
16759 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
16760 entries for V1TI variants of __builtin_altivec_ld builtin.
16761 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
16762 handling of V1TI variant of LVX icode pattern.
16763 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
16764 (rs6000_gimple_fold_builtin): Likewise.
16765 (altivec_init_builtins): Add code to define
16766 __builtin_altivec_lvx_v1ti function.
16768 2018-03-22 Jakub Jelinek <jakub@redhat.com>
16770 PR inline-asm/84941
16771 * function.c (match_asm_constraints_1): Don't do the optimization
16772 if input isn't a REG, SUBREG, MEM or constant.
16774 2018-03-22 Tom de Vries <tom@codesourcery.com>
16776 PR tree-optimization/84956
16777 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
16778 bb_has_abnormal_pred.
16780 2018-03-22 Jakub Jelinek <jakub@redhat.com>
16783 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
16784 DECL_INITIAL (decl) to decl at the end.
16785 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
16786 adjust the comment.
16788 2018-03-21 Joseph Myers <joseph@codesourcery.com>
16790 * doc/extend.texi (__builtin_tgmath): Document when complex
16791 integer types are treated as _Complex _Float64.
16793 2018-03-21 Tom de Vries <tom@codesourcery.com>
16795 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
16797 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16799 PR tree-optimization/84960
16800 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
16801 if it is ENTRY block, move them into single succ of ENTRY in that case.
16803 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
16805 PR tree-optimization/84811
16806 * poly-int.h (poly_span_traits): Remove the T3 parameter and
16807 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
16808 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
16809 (known_subrange_p): Update accordingly. Cast each value involved
16810 in the size comparison, rather than casting the result of the
16813 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16815 PR tree-optimization/84982
16816 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
16817 by flipping the least significant bit rather than all bits from
16818 bitpos to bitpos + bitsize - 1.
16820 2018-03-21 Nathan Sidwell <nathan@acm.org>
16822 * doc/extend.texi (Deprecated Features): Remove mention of
16823 long-deleted deprecations.
16825 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16828 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
16829 * configure: Regenerate.
16831 2018-03-21 Tom de Vries <tom@codesourcery.com>
16833 PR tree-optimization/83126
16834 * tree-parloops.c (num_phis): New function.
16835 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
16837 2018-03-21 Nathan Sidwell <nathan@acm.org>
16839 * doc/extend.texi (Deprecated Features): Update deprecated flags,
16840 mention anon-struct/union members and trailing attributes.
16842 2018-03-21 Bin Cheng <bin.cheng@arm.com>
16844 PR tree-optimization/84969
16845 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
16846 builtin memset partitions if they set different rhs values.
16848 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16850 PR rtl-optimization/84989
16851 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
16852 VEC_DUPLICATE with scalar result mode.
16854 2018-03-21 Martin Liska <mliska@suse.cz>
16857 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
16858 not intended return statement.
16860 2018-03-21 Martin Liska <mliska@suse.cz>
16863 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
16864 (chkp_find_bound_slots_1): Limit number of iterations.
16866 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
16869 * Minor grammar fixes for x86 options.
16871 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16874 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
16875 holding REG_CFA_RESTORE notes, instead turn them into a USE.
16877 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
16880 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
16881 (altivec_lvx_<mode>_1op): Likewise.
16882 (altivec_stvx_<mode>_2op): Likewise.
16883 (altivec_stvx_<mode>_1op): Likewise.
16884 (altivec_lvx_<VM2:mode>): New define_expand.
16885 (altivec_stvx_<VM2:mode>): Likewise.
16886 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
16887 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16888 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
16889 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16890 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
16891 (rs6000_gen_lvx): Likewise.
16892 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
16893 (altivec_expand_stv_builtin): Likewise.
16894 (altivec_expand_builtin): Likewise.
16895 * config/rs6000/vector.md: Likewise.
16897 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16900 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
16903 2018-03-20 Richard Biener <rguenther@suse.de>
16906 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
16907 sign-conversions as zero, fall back to standard scalar_stmt
16910 2018-03-20 Martin Liska <mliska@suse.cz>
16913 * predict.c (rebuild_frequencies): Handle case when we have
16914 PROFILE_ABSENT, but flag_guess_branch_prob is false.
16916 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16919 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
16920 flag_section_anchors.
16921 * varasm.c (use_blocks_for_decl_p): Remove hack for
16922 dw2_force_const_mem.
16925 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
16927 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
16928 be created, use lowpart_subreg of operands[0] rather than operands[0]
16930 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
16931 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
16932 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
16933 and n constraint instead of aarch64_shift_imm_di and Usd.
16934 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
16935 (*aarch64_<optab>_reg_minus<mode>3): ... this.
16937 2018-03-20 Sudakshina Das <sudi.das@arm.com>
16940 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
16941 to favor GPR over NEON registers.
16942 (<shift>di3_neon): Likewise.
16944 2018-03-20 Tom de Vries <tom@codesourcery.com>
16947 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
16948 (nvptx_process_pars): Emit bar.sync asap and alap.
16950 2018-03-20 Tom de Vries <tom@codesourcery.com>
16953 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
16954 seen_label if seen_label is already set.
16956 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16959 * config/i386/i386.c (fold_builtin_cpu): For features above 31
16960 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
16961 Use 1U instead of 1. Formatting fixes.
16964 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
16965 instead of TREE_TYPE (s1) for the return value.
16967 2018-03-19 Jakub Jelinek <jakub@redhat.com>
16969 PR tree-optimization/84946
16970 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
16971 bitsize + bitsize in poly_uint64 rather than poly_int64.
16974 * dwarf2asm.c: Include fold-const.c.
16975 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
16976 of decl rather than decl itself.
16978 PR rtl-optimization/84643
16979 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
16981 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
16984 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
16985 calling assemble_variable.
16987 2018-03-19 Sudakshina Das <sudi.das@arm.com>
16990 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
16991 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
16993 2018-03-19 Jim Wilson <jimw@sifive.com>
16996 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
16997 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
16998 (riscv_first_stack_step): Likewise.
16999 (riscv_option_override): Use STACK_BOUNDARY instead of
17000 MIN_STACK_BOUNDARY.
17001 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
17002 MIN_STACK_BOUNDARY.
17003 (BIGGEST_ALIGNMENT): Set to 128.
17004 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
17005 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
17008 2018-03-19 Richard Biener <rguenther@suse.de>
17010 PR tree-optimization/84933
17011 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
17012 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
17014 2018-03-19 Richard Biener <rguenther@suse.de>
17016 PR tree-optimization/84859
17017 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
17018 (cond_if_else_store_replacement): Perform sinking operation on
17019 single-store BBs regardless of MAX_STORES_TO_SINK setting.
17020 Generalize what a BB with a single eligible store is.
17022 2018-03-19 Richard Biener <rguenther@suse.de>
17024 PR tree-optimization/84929
17025 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
17026 chrec_is_positive against non-chrec arg.
17028 2018-03-19 Tamar Christina <tamar.christina@arm.com>
17031 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
17033 2018-03-18 Martin Liska <mliska@suse.cz>
17035 PR rtl-optimization/84635
17036 * regrename.c (build_def_use): Use matches_mode only when
17039 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
17041 PR tree-optimization/84913
17042 * tree-vect-loop.c (vectorizable_reduction): Don't try to
17043 vectorize chains of COND_EXPRs.
17045 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
17047 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
17049 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
17051 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
17053 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
17055 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
17057 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
17058 Kito Cheng <kito.cheng@gmail.com>
17060 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
17061 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
17062 (nds32_adjust_reg_alloc_order): New function.
17063 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
17065 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
17067 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
17068 nds32_print_operand, nds32_print_operand_address): Use
17069 HOST_WIDE_INT_PRINT_DEC instead.
17071 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
17073 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
17075 2018-03-17 Jakub Jelinek <jakub@redhat.com>
17078 * config/i386/i386.c (initial_ix86_tune_features,
17079 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
17080 unsigned long long.
17081 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
17082 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
17083 rather than 1u << ix86_tune. Formatting fix.
17084 (ix86_option_override_internal): Change ix86_arch_mask from
17085 unsigned int to unsigned HOST_WIDE_INT, initialize to
17086 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
17087 (ix86_function_specific_restore): Likewise.
17089 2018-03-16 Jakub Jelinek <jakub@redhat.com>
17092 * postreload.c (reload_combine_recognize_pattern): Perform
17093 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
17094 truncate_int_for_mode the result for the destination's mode.
17097 * hsa-gen.c (mem_type_for_type): Fix comment typo.
17098 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
17100 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17103 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
17106 * lra-assigns.c (lra_split_hard_reg_for): Don't use
17107 regno_allocno_class_array and sorted_pseudos.
17108 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
17109 insns where regno is used.
17111 2018-03-16 Martin Liska <mliska@suse.cz>
17114 * multiple_target.c (create_dispatcher_calls): Redirect
17115 reference in the symbol table.
17117 2018-03-16 Martin Liska <mliska@suse.cz>
17120 * multiple_target.c (create_dispatcher_calls): Redirect also
17123 2018-03-16 Jakub Jelinek <jakub@redhat.com>
17127 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
17128 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
17129 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
17131 2018-03-16 Julia Koval <julia.koval@intel.com>
17133 * doc/invoke.texi (Skylake Server): Add CLWB.
17134 Cannonlake): Remove CLWB.
17136 2018-03-16 Jakub Jelinek <jakub@redhat.com>
17138 PR tree-optimization/84841
17139 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
17141 (FLOAT_ONE_CONST_TYPE): Define.
17142 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
17143 (sort_by_operand_rank): Put entries with higher constant_type last
17144 rather than first to match comments.
17146 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
17148 * config/nios2/nios2.md (movsi_internal): Fix thinko in
17151 2018-03-15 Jakub Jelinek <jakub@redhat.com>
17154 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
17155 check and use address of target always.
17157 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
17160 * config/i386/i386.c (indirect_thunk_needed): Update comments.
17161 (indirect_thunk_bnd_needed): Likewise.
17162 (indirect_thunks_used): Likewise.
17163 (indirect_thunks_bnd_used): Likewise.
17164 (indirect_return_needed): New.
17165 (indirect_return_bnd_needed): Likewise.
17166 (output_indirect_thunk_function): Add a bool argument for
17168 (output_indirect_thunk_function): Don't generate alias for
17169 function return thunk.
17170 (ix86_code_end): Call output_indirect_thunk_function to generate
17171 function return thunks.
17172 (ix86_output_function_return): Set indirect_return_bnd_needed
17173 and indirect_return_needed instead of indirect_thunk_bnd_needed
17174 and indirect_thunk_needed.
17176 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
17178 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
17179 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
17180 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
17182 2018-03-15 David Malcolm <dmalcolm@redhat.com>
17183 Paul Hua <paul.hua.gm@gmail.com>
17186 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
17188 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
17190 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
17191 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
17192 resp. SFmode cases.
17194 2018-03-15 Tamar Christina <tamar.christina@arm.com>
17197 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
17198 instead of GET_MODE_SIZE when comparing Units.
17200 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
17203 * varasm.c (hash_section): Return an unchangeble hash value
17204 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
17205 Return !aarch64_can_use_per_function_literal_pools_p ().
17207 2018-03-15 Jakub Jelinek <jakub@redhat.com>
17210 * optabs.c (emit_conditional_move): Pass address of cmode's copy
17211 rather than address of cmode as last argument to prepare_cmp_insn.
17213 2018-03-15 Julia Koval <julia.koval@intel.com>
17215 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
17216 F_AVX512VNNI, F_AVX512BITALG): New.
17218 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
17221 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
17222 insn for floating-point loads and stores.
17224 2018-03-14 Carl Love <cel@us.ibm.com>
17226 * config/rs6000/rs6000-c.c: Add macro definitions for
17227 ALTIVEC_BUILTIN_VEC_PERMXOR.
17228 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
17229 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
17230 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
17231 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
17233 * config/doc/extend.texi: Add prototypes for vec_permxor.
17235 2018-03-14 David Malcolm <dmalcolm@redhat.com>
17238 * diagnostic-show-locus.c (class layout_point): Convert m_line
17239 from int to linenum_type.
17240 (line_span::comparator): Use linenum "compare" function when
17241 comparing line numbers.
17242 (test_line_span): New function.
17243 (layout_range::contains_point): Convert param "row" from int to
17245 (layout_range::intersects_line_p): Likewise.
17246 (layout::will_show_line_p): Likewise.
17247 (layout::print_source_line): Likewise.
17248 (layout::should_print_annotation_line_p): Likewise.
17249 (layout::print_annotation_line): Likewise.
17250 (layout::print_leading_fixits): Likewise.
17251 (layout::annotation_line_showed_range_p): Likewise.
17252 (struct line_corrections): Likewise for field m_row.
17253 (line_corrections::line_corrections): Likewise for param "row".
17254 (layout::print_trailing_fixits): Likewise.
17255 (layout::get_state_at_point): Likewise.
17256 (layout::get_x_bound_for_row): Likewise.
17257 (layout::print_line): Likewise.
17258 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
17259 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
17260 * input.c (selftest::test_linenum_comparisons): New function.
17261 (selftest::input_c_tests): Call it.
17262 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
17263 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
17264 * selftest.h (ASSERT_GT): New macro.
17265 (ASSERT_GT_AT): New macro.
17266 (ASSERT_LT): New macro.
17267 (ASSERT_LT_AT): New macro.
17269 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
17271 PR rtl-optimization/84780
17272 * combine.c (distribute_links): Don't make a link based on pc_rtx.
17274 2018-03-14 Martin Liska <mliska@suse.cz>
17276 * tree.c (record_node_allocation_statistics): Use
17277 get_stats_node_kind.
17278 (get_stats_node_kind): New function extracted from
17279 record_node_allocation_statistics.
17280 (free_node): Use get_stats_node_kind.
17282 2018-03-14 Richard Biener <rguenther@suse.de>
17284 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
17285 that the value-set of ANTIC_IN doesn't grow.
17288 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
17290 (BB_VISITED_WITH_VISITED_SUCCS): New define.
17291 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
17293 2018-03-14 Julia Koval <julia.koval@intel.com>
17295 * config.gcc (icelake-client, icelake-server): New.
17297 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
17298 (initial_ix86_arch_features): Ditto.
17299 (PTA_SKYLAKE): Add SGX.
17300 (PTA_ICELAKE): Remove.
17301 (PTA_ICELAKE_CLIENT): New.
17302 (PTA_ICELAKE_SERVER): New.
17303 (ix86_option_override_internal): Split up icelake on icelake client and
17305 (get_builtin_code_for_version): Ditto.
17306 (fold_builtin_cpu): Ditto.
17307 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
17308 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
17309 * config/i386/i386.h (processor_type): Ditto.
17310 * doc/invoke.texi: Ditto.
17312 2018-03-14 Jakub Jelinek <jakub@redhat.com>
17315 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
17316 INTEGER_CST offset, add it together with bitpos / 8 and
17317 sign extend based on POINTER_SIZE.
17321 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
17324 * config/i386/constraints.md (Yc): New register constraint.
17325 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
17326 Use Yc constraint for alternative 2 of operand 0. Remove
17327 preferred_for_speed attribute.
17329 2018-03-14 Richard Biener <rguenther@suse.de>
17331 PR tree-optimization/84830
17332 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
17333 with the old one to avoid oscillations.
17335 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
17338 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
17340 (assign_by_spills): Return a flag of reload assignment failure.
17341 Do not process the reload assignment failures. Do not spill other
17342 reload pseudos if they has the same reg class. Update n if
17344 (lra_assign): Add a return arg. Set up from the result of
17345 assign_by_spills call.
17346 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17347 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17348 usage_insns if it is not NULL.
17349 (spill_hard_reg_in_range): New function.
17350 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17351 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17352 function prototypes.
17353 (lra_assign): Change prototype.
17354 * lra.c (lra): Add code to deal with fails by splitting hard reg
17357 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
17359 * config/riscv/riscv.opt (mrelax): New option.
17360 * config/riscv/riscv.c (riscv_file_start): Emit ".option
17361 "norelax" when riscv_mrelax is disabled.
17362 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
17364 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17367 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
17368 reassociation for int modes.
17370 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17372 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
17373 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
17375 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
17376 * config/aarch64/aarch64-sve.md
17377 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
17378 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
17379 (*extend<mode><Vwide>2): Rename to...
17380 (aarch64_sve_extend<mode><Vwide>2): ...this.
17381 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
17382 renaming the old pattern to...
17383 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
17385 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
17386 define_expand, renaming the old pattern to...
17387 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
17388 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
17389 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
17390 account when deciding which SVE instruction the optab should use.
17391 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
17393 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17395 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
17396 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
17397 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
17398 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
17399 (tlsdesc_small_<mode>): Turn a define_expand and use
17400 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
17401 (tlsdesc_small_advsimd_<mode>): ...this.
17402 (tlsdesc_small_sve_<mode>): New pattern.
17404 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17406 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
17407 (UNSPEC_UMUL_HIGHPART): New constants.
17408 (MUL_HIGHPART): New int iteraor.
17409 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
17410 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
17412 (*<su>mul<mode>3_highpart): New define_insn.
17414 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
17417 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
17420 2018-03-13 Martin Liska <mliska@suse.cz>
17423 * (sem_item_optimizer::sem_item_optimizer): Initialize new
17425 (sem_item_optimizer::~sem_item_optimizer): Release it.
17426 (sem_item_optimizer::merge_classes): Register variable aliases.
17427 (sem_item_optimizer::fixup_pt_set): New function.
17428 (sem_item_optimizer::fixup_points_to_sets): Likewise.
17429 * ipa-icf.h: Declare new variables and functions.
17431 2018-03-13 Jakub Jelinek <jakub@redhat.com>
17433 PR middle-end/84834
17434 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
17435 integer_pow2p@2 and test integer_pow2p in condition.
17436 (A < 0 ? C : 0): Similarly for @1.
17438 PR middle-end/84831
17439 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
17440 characters starting at p contain '\0' character, don't look beyond
17444 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
17445 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
17448 * reg-stack.c (change_stack): Change update_end var from int to
17449 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
17450 also call set_block_for_insn on the newly added insns and rescan.
17453 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
17454 on the last operand.
17457 * tree.c (stabilize_reference_1): Return save_expr (e) for
17458 STATEMENT_LIST even if it doesn't have side-effects.
17460 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
17462 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
17464 2018-03-12 Renlin Li <renlin.li@arm.com>
17466 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
17467 aarch64_output_scalar_simd_mov_immediate.
17469 2018-03-12 Martin Sebor <msebor@redhat.com>
17471 PR tree-optimization/83456
17472 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
17473 for perfectly overlapping calls to memcpy.
17474 (gimple_fold_builtin_memory_chk): Same.
17475 (gimple_fold_builtin_strcpy): Handle no-warning.
17476 (gimple_fold_builtin_stxcpy_chk): Same.
17477 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
17479 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
17481 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
17482 parameter. Use it for SFmode.
17483 (rs6000_function_arg_advance_1): Adjust.
17484 (rs6000_function_arg): Adjust.
17485 (rs6000_gimplify_va_arg): Pass false for that new parameter.
17487 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
17489 PR rtl-optimization/84169
17490 PR rtl-optimization/84780
17491 * combine.c (can_combine_p): Check for a 2-insn combination whether
17492 the destination register is used between the two insns, too.
17494 2018-03-12 Richard Biener <rguenther@suse.de>
17496 PR tree-optimization/84803
17497 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
17498 for refs DR analysis didn't process.
17500 2018-03-12 Richard Biener <rguenther@suse.de>
17502 PR tree-optimization/84777
17503 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
17504 force-vectorize loops ignore whether we are optimizing for size.
17506 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
17508 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
17509 (TARGET_MD_ASM_ADJUST): Define.
17511 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
17512 Kito Cheng <kito.cheng@gmail.com>
17513 Chung-Ju Wu <jasonwucj@gmail.com>
17515 * config/nds32/nds32.c (nds32_compute_stack_frame,
17516 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
17517 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
17518 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
17519 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
17520 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
17521 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
17522 * config/nds32/nds32.md (prologue, epilogue): Use macro
17523 NDS32_V3PUSH_AVAILABLE_P to do checking.
17525 2018-03-11 Jakub Jelinek <jakub@redhat.com>
17528 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
17529 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
17530 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
17531 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
17532 addition of most attributes on !orig_type_die or the attribute not
17533 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
17535 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17536 Chung-Ju Wu <jasonwucj@gmail.com>
17538 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
17539 __NDS32_VH__ macro.
17540 * config/nds32/nds32.opt (mvh): New option.
17542 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17543 Chung-Ju Wu <jasonwucj@gmail.com>
17545 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
17547 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
17548 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
17551 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17552 Chung-Ju Wu <jasonwucj@gmail.com>
17554 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
17556 * config/nds32/nds32-multiple.md (strlensi): New pattern.
17557 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
17559 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
17560 Kito Cheng <kito.cheng@gmail.com>
17561 Chung-Ju Wu <jasonwucj@gmail.com>
17563 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
17564 UNSPEC_FFMISM and UNSPEC_FLMISM.
17565 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
17566 for ffb, ffmism and flmism.
17567 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
17568 (unspec_ffmism): Ditto.
17569 (unspec_flmism): Ditto.
17570 (nds32_expand_builtin_impl): Check if string extension is available.
17571 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
17572 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
17574 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
17577 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
17580 * lra-assigns.c (assign_by_spills): Return a flag of reload
17581 assignment failure. Do not process the reload assignment
17582 failures. Do not spill other reload pseudos if they has the same
17584 (lra_assign): Add a return arg. Set up from the result of
17585 assign_by_spills call.
17586 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17587 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17588 usage_insns if it is not NULL.
17589 (spill_hard_reg_in_range): New function.
17590 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17591 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17592 function prototypes.
17593 (lra_assign): Change prototype.
17594 * lra.c (lra): Add code to deal with fails by splitting hard reg
17597 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
17600 * config/i386/i386.opt: Replace Enforcment with Enforcement.
17602 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
17605 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
17606 (dw_val_node): Add val_symbolic_view.
17607 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
17608 (symview_upper_bound): New.
17609 (new_line_info_table): Initialize symviews_since_reset.
17610 (dwarf2out_source_line): Count symviews_since_reset and set
17611 symview_upper_bound.
17612 (dw_val_equal_p): Handle symview.
17613 (add_AT_symview): New.
17614 (print_dw_val): Handle symview.
17615 (attr_checksum, attr_checksum_ordered): Likewise.
17616 (same_dw_val_p, size_of_die): Likewise.
17617 (value_format, output_die): Likewise.
17618 (add_high_low_attributes): Use add_AT_symview for entry_view.
17619 (dwarf2out_finish): Reset symview_upper_bound, clear
17622 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
17625 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
17626 Add strict argument and use it.
17627 (rs6000_split_multireg_move): Update for new strict argument.
17628 (mem_operand_gpr): Disallow all non-offsettable addresses.
17629 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
17631 2018-03-09 Jakub Jelinek <jakub@redhat.com>
17634 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
17635 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
17636 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
17639 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
17640 decl, use remap_type if we want to use the type.
17642 2018-03-09 Martin Sebor <msebor@redhat.com>
17644 PR tree-optimization/84526
17645 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17647 (builtin_access::generic_overlap): Be prepared to handle non-array
17650 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
17652 PR rtl-optimization/84682
17653 * lra-constraints.c (process_address_1): Check is_address flag
17654 for address constraints.
17655 (process_alt_operands): Likewise.
17656 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
17657 preprocess_constraints.
17658 * recog.h (preprocess_constraints): Add oploc parameter.
17660 * recog.c (preprocess_constraints): Test address_operand for
17661 CT_ADDRESS constraints.
17663 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
17666 * lra-assigns.c (assign_by_spills): Return a flag of reload
17667 assignment failure. Do not process the reload assignment
17668 failures. Do not spill other reload pseudos if they has the same
17670 (lra_assign): Add a return arg. Set up from the result of
17671 assign_by_spills call.
17672 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17673 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17674 usage_insns if it is not NULL.
17675 (spill_hard_reg_in_range): New function.
17676 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17677 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17678 function prototypes.
17679 (lra_assign): Change prototype.
17680 * lra.c (lra): Add code to deal with fails by splitting hard reg
17683 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17686 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
17687 Accept complain bool parameter. Only emit errors if it is true.
17688 (arm_parse_cpu_option_name): Likewise.
17689 (arm_target_thumb_only): Adjust callers of the above.
17690 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
17691 prototype to take a default true bool parameter.
17692 (arm_parse_arch_option_name): Likewise.
17694 2018-03-09 David Malcolm <dmalcolm@redhat.com>
17695 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
17699 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
17700 * configure: Regenerate.
17701 * configure.ac ("linker --version-script option"): New.
17702 ("linker soname option"): New.
17704 2018-03-09 Richard Biener <rguenther@suse.de>
17706 PR tree-optimization/84775
17707 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
17708 immediate uses of predicate stmts and mark them modified.
17711 PR tree-optimization/84178
17712 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17714 (version_loop_for_if_conversion): Delay update_ssa call.
17715 (tree_if_conversion): Delay update_ssa until after predicate
17718 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
17721 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
17722 when the function accesses prior frames.
17724 2018-03-08 Jakub Jelinek <jakub@redhat.com>
17727 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
17728 gen_llsym, otherwise call maybe_gen_llsym.
17730 PR inline-asm/84742
17731 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
17732 has ',' character inside of it.
17734 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17737 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
17738 as clobbering CC_REGNUM.
17740 2018-03-08 Richard Biener <rguenther@suse.de>
17742 PR middle-end/84552
17743 * tree-scalar-evolution.c: Include tree-into-ssa.h.
17744 (follow_copies_to_constant): Do not follow SSA names registered
17747 2018-03-08 Richard Biener <rguenther@suse.de>
17749 PR tree-optimization/84178
17750 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17752 (version_loop_for_if_conversion): Delay update_ssa call.
17753 (tree_if_conversion): Delay update_ssa until after predicate
17756 2018-03-08 David Malcolm <dmalcolm@redhat.com>
17758 PR tree-optimization/84178
17759 * tree-if-conv.c (release_bb_predicate): Remove the
17760 the assertion that the stmts have NULL use_ops.
17761 Discard the statements, asserting that they haven't
17762 yet been added to a BB.
17764 2018-03-08 Richard Biener <rguenther@suse.de>
17766 PR tree-optimization/84746
17767 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
17768 (phi_translate): Pass in destination ANTIC_OUT set.
17769 (phi_translate_1): Likewise. For a simplified result lookup
17770 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
17771 (phi_translate_set): Adjust.
17772 (do_pre_regular_insertion): Likewise.
17773 (do_pre_partial_partial_insertion): Likewise.
17775 2018-03-08 Martin Liska <mliska@suse.cz>
17777 PR gcov-profile/84735
17778 * doc/gcov.texi: Document usage of profile files.
17779 * gcov-io.h: Document changes in the format.
17781 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
17785 * dwarf2out.c (struct dw_line_info_table): Update comments for
17787 (FORCE_RESET_NEXT_VIEW): New.
17788 (FORCE_RESETTING_VIEW_P): New.
17789 (RESETTING_VIEW_P): Check for -1 too.
17790 (ZERO_VIEW_P): Likewise.
17791 (new_line_info_table): Force-reset next view.
17792 (dwarf2out_begin_function): Likewise.
17793 (dwarf2out_source_line): Simplify zero_view_p initialization.
17794 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
17795 view directly. Omit view when omitting .loc at line 0.
17797 2018-03-08 Jakub Jelinek <jakub@redhat.com>
17799 PR tree-optimization/84740
17800 * tree-switch-conversion.c (process_switch): Call build_constructors
17801 only if info.phi_count is non-zero.
17803 PR tree-optimization/84739
17804 * tree-tailcall.c (find_tail_calls): Check call arguments against
17805 DECL_ARGUMENTS (current_function_decl) rather than
17806 DECL_ARGUMENTS (func) when checking for tail recursion.
17808 2018-03-07 Jakub Jelinek <jakub@redhat.com>
17810 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
17811 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
17812 Volker Reichelt's entry and add entries for people that perform
17813 GCC fuzzy testing and report numerous bugs.
17815 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
17818 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
17819 readonly data in sdata, if that is disabled.
17820 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
17821 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
17822 -mreadonly-in-sdata option.
17824 2018-03-07 Martin Sebor <msebor@redhat.com>
17826 PR tree-optimization/84468
17827 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
17828 basic block when looking for nul assignment.
17830 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
17833 * except.h (output_function_exception_table): Adjust prototype.
17834 * except.c (output_function_exception_table): Remove FNNAME parameter
17835 and add SECTION parameter. Ouput one part of the table at a time.
17836 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
17837 the first part of the exception table and emit unwind directives.
17838 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
17839 (i386_pe_seh_cold_init): Likewise.
17840 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
17841 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
17842 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
17843 (ix86_output_call_insn): Emit a nop in one more case for SEH.
17844 * config/i386/winnt.c: Include except.h.
17845 (struct seh_frame_state): Add reg_offset, after_prologue and
17846 in_cold_section fields.
17847 (i386_pe_seh_end_prologue): Set seh->after_prologue.
17848 (i386_pe_seh_cold_init): New function.
17849 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
17850 to seh->in_cold_section.
17851 (seh_emit_push): Record the offset of the push.
17852 (seh_emit_save): Record the offet of the save.
17853 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
17854 Test seh->after_prologue to disregard the epilogue.
17855 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
17856 (i386_pe_end_cold_function): New function.
17858 2018-03-07 Jakub Jelinek <jakub@redhat.com>
17861 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
17862 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
17865 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
17867 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
17868 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
17870 PR middle-end/84723
17871 * multiple_target.c: Include tree-inline.h and intl.h.
17872 (expand_target_clones): Diagnose and fail if node->definition and
17873 !tree_versionable_function_p (node->decl).
17875 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
17877 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
17879 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
17880 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
17881 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
17883 2018-03-06 Jakub Jelinek <jakub@redhat.com>
17886 * combine.c (try_combine): Use reg_or_subregno instead of handling
17887 just paradoxical SUBREGs and REGs.
17889 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
17891 * config/arc/arc.c (arc_finalize_pic): Remove function.
17892 (arc_must_save_register): We use single base PIC register, remove
17893 checks to save/restore the PIC register.
17894 (arc_expand_prologue): Likewise.
17895 * config/arc/arc-protos.h (arc_set_default_type_attributes):
17897 (arc_verify_short): Likewise.
17898 (arc_attr_type): Likewise.
17899 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
17900 (walk_stores): Likewise.
17901 (arc_address_cost): Make it static.
17902 (arc_verify_short): Likewise.
17903 (branch_dest): Likewise.
17904 (arc_attr_type): Likewise.
17905 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
17906 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
17907 (arc_final_prescan_insn): Remove inserting the nops due to
17908 hardware hazards. It is done in reorg step.
17909 (insn_length_variant_t): Remove.
17910 (insn_length_parameters_t): Likewise.
17911 (arc_insn_length_parameters): Likewise.
17912 (arc_get_insn_variants): Likewise.
17913 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
17915 2018-03-06 Jakub Jelinek <jakub@redhat.com>
17917 PR inline-asm/84683
17918 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
17921 PR tree-optimization/84687
17922 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
17924 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
17926 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17928 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
17929 Rename to ppc_speculation_barrier.
17930 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
17931 __builtin_ppc_speculation_barrier.
17933 2018-03-05 Jakub Jelinek <jakub@redhat.com>
17936 * combine.c (combine_simplify_rtx): Don't try to simplify if
17937 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
17940 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
17942 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
17943 to 32 bytes when compiling for POWER9.
17945 2018-03-05 Jakub Jelinek <jakub@redhat.com>
17948 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
17949 regparm >= 3 with no arg reg available also for calls with
17950 flag_force_indirect_call. Pass decl to ix86_function_regparm.
17953 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
17955 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
17957 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
17960 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
17962 2018-03-05 Richard Biener <rguenther@suse.de>
17964 PR tree-optimization/84486
17965 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
17966 When inserting a __builtin_assume_aligned call set the LHS
17967 SSA name alignment info accordingly.
17969 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
17971 PR tree-optimization/84114
17972 * config/aarch64/aarch64.c (aarch64_reassociation_width)
17973 Avoid reassociation of FLOAT_MODE addition.
17975 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
17977 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
17978 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
17979 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
17980 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
17981 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
17982 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
17983 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
17985 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
17986 __builtin_ia32_wbinvd): New builtins.
17987 (SPECIAL_ARGS2): New.
17988 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
17989 (SPECIAL_ARGS2): New.
17990 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
17991 (ix86_valid_target_attribute_inner_p): Ditto.
17992 (ix86_init_mmx_sse_builtins): Add special_args2.
17993 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
17994 TARGET_WBNOINVD_P): New.
17995 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
17996 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
17997 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
17998 * config/i386/immintrin.h (_wbinvd): New intrinsic.
17999 * config/i386/pconfigintrin.h: New file.
18000 * config/i386/wbnoinvdintrin.h: Ditto.
18001 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
18003 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
18005 2018-03-05 Richard Biener <rguenther@suse.de>
18007 PR tree-optimization/84670
18008 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
18010 (BB_VISITED_WITH_VISITED_SUCCS): New define.
18011 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
18012 (compute_antic_aux): Only assert the number of values in ANTIC_IN
18013 doesn't grow if all successors (recursively) were visited at least
18016 2018-03-05 Richard Biener <rguenther@suse.de>
18018 PR tree-optimization/84650
18019 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
18020 if executed in the loop pipeline.
18022 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
18024 * doc/configfiles.texi (Configuration Files): Move info about
18025 conditionalizing $target-protos.h to...
18026 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
18027 differs from $target-protos.h.
18029 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
18030 Chung-Ju Wu <jasonwucj@gmail.com>
18032 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
18033 * config/nds32/nds32-multiple.md (setmemsi): Define.
18034 * config/nds32/nds32-memory-manipulation.c
18035 (nds32_gen_dup_4_byte_to_word_value): New.
18036 (emit_setmem_word_loop): New.
18037 (emit_setmem_byte_loop): New.
18038 (nds32_expand_setmem_loop): New.
18039 (nds32_expand_setmem_loop_v3m): New.
18040 (nds32_expand_setmem_unroll): New.
18041 (nds32_expand_setmem): New.
18043 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
18044 Chung-Ju Wu <jasonwucj@gmail.com>
18046 * config/nds32/nds32-memory-manipulation.c
18047 (nds32_emit_load_store): New.
18048 (nds32_emit_post_inc_load_store): New.
18049 (nds32_emit_mem_move): New.
18050 (nds32_emit_mem_move_block): New.
18051 (nds32_expand_movmemsi_loop_unknown_size): New.
18052 (nds32_expand_movmemsi_loop_known_size): New.
18053 (nds32_expand_movmemsi_loop): New.
18054 (nds32_expand_movmemsi_unroll): New.
18055 (nds32_expand_movmemqi): Rename ...
18056 (nds32_expand_movmemsi): ... to this.
18057 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
18058 (movmemsi): ... to this.
18059 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
18060 (nds32_expand_movmemsi): ... to this.
18062 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
18063 Monk Chiang <sh.chiang04@gmail.com>
18064 Chung-Ju Wu <jasonwucj@gmail.com>
18066 * config/nds32/nds32-protos.h
18067 (nds32_expand_load_multiple): New arguments.
18068 (nds32_expand_store_multiple): Ditto.
18069 (nds32_valid_multiple_load_store): Rename ...
18070 (nds32_valid_multiple_load_store_p): ... to this.
18071 * config/nds32/nds32-memory-manipulation.c
18072 (nds32_expand_load_multiple): Refine implementation.
18073 (nds32_expand_store_multiple): Ditto.
18074 * config/nds32/nds32-multiple.md
18075 (load_multiple): Update nds32_expand_load_multiple interface.
18076 (store_multiple): Update nds32_expand_store_multiple interface.
18077 * config/nds32/nds32-predicates.c
18078 (nds32_valid_multiple_load_store): Rename ...
18079 (nds32_valid_multiple_load_store_p): ... to this and refine
18081 * config/nds32/predicates.md
18082 (nds32_load_multiple_and_update_address_operation): New predicate.
18083 (nds32_store_multiple_and_update_address_operation): New predicate.
18085 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
18086 Chung-Ju Wu <jasonwucj@gmail.com>
18088 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
18089 (combo): New attribute.
18090 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
18092 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
18094 * config/nds32/nds32.opt: Change -mcmodel= default value.
18096 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
18097 Monk Chiang <sh.chiang04@gmail.com>
18098 Chung-Ju Wu <jasonwucj@gmail.com>
18100 * config/nds32/constants.md (unspec_element): New enum.
18101 * config/nds32/constraints.md (Umw): New constraint.
18102 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
18103 * config/nds32/nds32-intrinsic.md: Likewise.
18104 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
18105 (nds32_valid_smw_lwm_base_p): New.
18106 (nds32_output_smw_single_word): New.
18107 (nds32_output_lmw_single_word): New.
18108 (nds32_expand_unaligned_load): New.
18109 (nds32_expand_unaligned_store): New.
18110 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
18111 (nds32_output_smw_single_word): Declare.
18112 (nds32_output_lmw_single_word): Declare.
18113 (nds32_expand_unaligned_load): Declare.
18114 (nds32_expand_unaligned_store): Declare.
18115 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
18116 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
18117 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
18118 NDS32_BUILTIN_UASTORE_DW.
18119 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
18122 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
18123 Kito Cheng <kito.cheng@gmail.com>
18124 Chung-Ju Wu <jasonwucj@gmail.com>
18126 * config/nds32/nds32-intrinsic.c
18127 (nds32_expand_builtin_null_ftype_reg): Delete.
18128 (nds32_expand_builtin_reg_ftype_imm): Ditto.
18129 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
18130 (nds32_read_argument): New.
18131 (nds32_legitimize_target): Ditto.
18132 (nds32_legitimize_argument): Ditto.
18133 (nds32_check_constant_argument): Ditto.
18134 (nds32_expand_unop_builtin): Ditto.
18135 (nds32_expand_unopimm_builtin): Ditto.
18136 (nds32_expand_binop_builtin): Ditto.
18137 (nds32_builtin_decl_impl): Ditto.
18138 (builtin_description): Ditto.
18139 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
18140 (nds32_init_builtins_impl): Ditto.
18141 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
18142 (nds32_builtin_decl): New.
18143 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
18144 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
18146 2018-03-02 Jeff Law <law@redhat.com>
18148 * reorg.c (stop_search_p): Handle DEBUG_INSN.
18149 (redundant_insn, fill_simple_delay_slots): Likewise.
18150 (fill_slots_from_thread): Likewise.
18151 * resource.c (mark_referenced_resources): Likewise.
18152 (mark_set_resources, find_dead_or_set_registers): Likewise.
18154 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18156 * substring-locations.h (format_warning_va): Formatting fix for
18157 ATTRIBUTE_GCC_DIAG.
18158 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
18160 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
18161 * substring-locations.c: Include intl.h.
18162 (format_warning_va): Turned into small wrapper around
18163 format_warning_n_va, renamed to ...
18164 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
18165 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
18167 (format_warning_at_substring_n): New function.
18168 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
18169 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
18170 format_warning_at_substring with just a shorter name instead of
18171 const function pointer.
18172 (fmtwarn_n): New function.
18173 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
18174 appropriate, get rid of all the fmtstr temporaries, move conditionals
18175 with G_() wrapped string literals directly into fmtwarn arguments,
18176 cast dir.len to (int), formatting fixes.
18178 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
18180 * doc/invoke.texi: Remove "Cilk Plus" references.
18182 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18183 Richard Biener <rguenther@suse.de>
18186 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
18187 for error or warning attributes if CALL_FROM_THUNK_P is set.
18190 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18193 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
18194 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
18197 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
18198 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
18200 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
18201 instead of -1U in last predictors element's probability member.
18203 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
18206 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
18207 arguments if they are comparable.
18209 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
18211 PR tree-optimization/84634
18212 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
18213 masks and masked_loop_p with a single loop_masks, making sure it's
18214 null for bb vectorization.
18216 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
18218 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
18219 (vect_analyze_data_ref_access): Use loop->safe_len rather than
18220 loop->force_vectorize to check whether there is no alias.
18222 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18225 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
18227 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
18229 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
18230 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
18231 instead of a loop around prev_real_insn.
18232 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
18235 PR inline-asm/84625
18236 * config/i386/i386.c (ix86_print_operand): Use conditional
18237 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
18240 2018-03-02 Richard Biener <rguenther@suse.de>
18242 PR tree-optimization/84427
18243 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
18244 (bitmap_set_subtract_values): Rewrite to handle multiple
18247 (prune_clobbered_mems): Likewise.
18248 (phi_translate): Take edge instead of pred/phiblock.
18249 (phi_translate_1): Likewise.
18250 (phi_translate_set): Likewise. Insert all translated
18251 exprs for a value into the set, keeping possibly multiple
18252 expressions per value.
18253 (compute_antic_aux): Adjust for phi_translate changes.
18254 When intersecting union the expressions and prune those
18255 not in the final value set, keeping possibly multiple
18256 expressions per value. Do not use value-insertion
18257 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
18258 all expressions. Add verification that the value-sets
18259 only shrink during iteration.
18260 (compute_partial_antic_aux): Adjust for the phi_translate changes.
18261 (do_pre_regular_insertion): Likewise.
18262 (do_pre_partial_partial_insertion): Likewise.
18264 2018-03-02 Richard Biener <rguenther@suse.de>
18267 * config/darwin.c (saved_debug_info_level): New static global.
18268 (darwin_asm_lto_start): Disable debug info generation for LTO out.
18269 (darwin_asm_lto_end): Restore debug info generation settings.
18271 2018-03-01 Martin Liska <mliska@suse.cz>
18274 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
18275 volatile arguments.
18277 2018-03-01 Richard Biener <rguenther@suse.de>
18280 * dwarf2out.c (gen_variable_die): Properly handle late VLA
18281 type annotation with LTO when debug was disabled at compile-time.
18283 2018-03-01 Matthew Fortune <mfortune@gmail.com>
18285 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
18287 (mips_final_postscan_insn): Likewise.
18289 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
18291 PR rtl-optimization/84528
18292 * alias.c (init_alias_target): Add commentary.
18293 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
18294 a unique base value if the frame pointer is not eliminated
18295 to the stack pointer.
18297 2018-03-01 Tom de Vries <tom@codesourcery.com>
18299 PR rtl-optimization/83327
18300 * lra-int.h (hard_regs_spilled_into): Declare.
18301 * lra.c (hard_regs_spilled_into): Define.
18302 (init_reg_info): Init hard_regs_spilled_into.
18303 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
18304 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
18305 (process_bb_lives): Handle hard_regs_spilled_into.
18306 (lra_create_live_ranges_1): Before doing liveness propagation, clear
18307 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
18309 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
18311 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
18312 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
18313 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
18314 * config/rs6000/aix72.h: New file.
18316 2018-02-28 Jakub Jelinek <jakub@redhat.com>
18318 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
18319 instead of warning_at with conditional singular and plural messages
18323 * stor-layout.c (update_alignment_for_field): For
18324 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
18325 && !DECL_PACKED (field), do the alignment update, just use
18326 only desired_align instead of MAX (type_align, desired_align)
18328 (place_field): Don't do known_align < desired_align handling
18329 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
18330 is non-NULL, instead do it after rli->prev_field handling and
18331 only if not within a bitfield word. For DECL_PACKED (field)
18332 use type_align of BITS_PER_UNIT.
18334 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
18336 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
18337 superfluous parentheses and trailing spaces.
18339 2018-02-28 Richard Biener <rguenther@suse.de>
18341 PR tree-optimization/84584
18342 * graphite-scop-detection.c (scop_detection::add_scop): Discard
18343 SCoPs with fake exit edge.
18345 2018-02-28 Martin Liska <mliska@suse.cz>
18348 * timevar.c (timer::print): Fix format to properly print 100%
18351 2018-02-28 Richard Biener <rguenther@suse.de>
18353 PR middle-end/84607
18354 * genmatch.c (capture_info::walk_match): Do not mark
18355 captured expressions without operands as expr_p given
18356 they act more like predicates and should be subject to
18357 "lost tail" side-effect preserving.
18359 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
18361 PR rtl-optimization/81611
18362 * auto-inc-dec.c (attempt_change): Move dead note from
18363 mem_insn if it's the next use of regno
18364 (find_address): Take address use of reg holding
18365 non-incremented value. Add parm to limit search to the named
18367 (merge_in_block): Attempt to use a mem insn that is the next
18368 use of the original regno.
18370 2018-02-27 Martin Sebor <msebor@redhat.com>
18373 * doc/invoke.texi (-Wmissing-attributes): New option.
18374 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
18376 2018-02-27 Martin Sebor <msebor@redhat.com>
18378 PR translation/84207
18379 * diagnostic-core.h (warning_n, error_n, inform_n): Change
18380 n argument to unsigned HOST_WIDE_INT.
18381 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
18382 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
18383 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
18384 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
18386 2018-02-27 Richard Biener <rguenther@suse.de>
18388 PR tree-optimization/84512
18389 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
18390 Do not use the estimate returned from record_stmt_cost for
18391 the scalar iteration cost but sum properly using add_stmt_cost.
18393 2018-02-27 Richard Biener <rguenther@suse.de>
18395 PR tree-optimization/84466
18396 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
18397 Adjust last change to less strictly validate use operands.
18399 2018-02-27 Martin Liska <mliska@suse.cz>
18401 PR gcov-profile/84548
18402 * gcov.c (process_file): Allow partial overlap and consider it
18403 also as group functions.
18404 (output_lines): Properly calculate range of lines for a group.
18406 2018-02-27 Martin Liska <mliska@suse.cz>
18408 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
18409 'ggc' suffixes. Change first column width.
18410 (timer::print): Fix formatting of the column.
18412 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
18414 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
18415 preserve inline entry blocks for the sake of debug inline
18416 entry point markers alone.
18417 (remove_unused_locals): Suggest in comments a better place to
18418 force the preservation of inline entry blocks that are
18419 otherwise unused, but do not preserve them.
18421 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18423 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
18425 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18428 * config/i386/constraints.md (Bs): Replace
18429 ix86_indirect_branch_register with
18430 TARGET_INDIRECT_BRANCH_REGISTER.
18432 * config/i386/i386.md (indirect_jump): Likewise.
18433 (tablejump): Likewise.
18434 (*sibcall_memory): Likewise.
18435 (*sibcall_value_memory): Likewise.
18436 Peepholes of indirect call and jump via memory: Likewise.
18437 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
18438 (*sibcall_value_GOT_32): Likewise.
18439 * config/i386/predicates.md (indirect_branch_operand): Likewise.
18440 (GOT_memory_operand): Likewise.
18441 (call_insn_operand): Likewise.
18442 (sibcall_insn_operand): Likewise.
18443 (GOT32_symbol_operand): Likewise.
18444 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
18446 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
18448 PR rtl-optimization/83496
18449 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
18450 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
18451 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
18452 redundant insn, if any.
18453 (relax_delay_slots): Likewise.
18454 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
18456 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
18458 PR tree-optimization/83965
18459 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
18460 that grouped statements are part of a reduction chain. Return
18461 true if the statement is not marked as a reduction itself but
18462 is part of a group.
18463 (vect_recog_dot_prod_pattern): Don't check whether the statement
18464 is part of a group here.
18465 (vect_recog_sad_pattern): Likewise.
18466 (vect_recog_widen_sum_pattern): Likewise.
18468 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
18471 * final.c (rest_of_clean_state): Also look for calls inside sequences.
18473 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18476 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
18478 (ix86_output_indirect_function_return): New prototype.
18479 (ix86_split_simple_return_pop_internal): Likewise.
18480 * config/i386/i386.c (indirect_return_via_cx): New.
18481 (indirect_return_via_cx_bnd): Likewise.
18482 (indirect_thunk_name): Handle return va CX_REG.
18483 (output_indirect_thunk_function): Create alias for
18484 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
18485 (ix86_output_indirect_jmp): Remove the bool argument.
18486 (ix86_output_indirect_function_return): New function.
18487 (ix86_split_simple_return_pop_internal): Likewise.
18488 * config/i386/i386.md (*indirect_jump): Don't pass false
18489 to ix86_output_indirect_jmp.
18490 (*tablejump_1): Likewise.
18491 (simple_return_pop_internal): Change it to define_insn_and_split.
18492 Call ix86_split_simple_return_pop_internal to split it for
18493 -mfunction-return=.
18494 (simple_return_indirect_internal): Call
18495 ix86_output_indirect_function_return instead of
18496 ix86_output_indirect_jmp.
18498 2018-02-26 Jakub Jelinek <jakub@redhat.com>
18501 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
18502 memset and value initialization afterwards.
18504 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
18506 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
18508 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
18511 * common/config/aarch64/aarch64-common.c
18512 (aarch_option_optimization_table[]): Switch
18513 off fomit-frame-pointer
18515 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
18516 Chung-Ju Wu <jasonwucj@gmail.com>
18518 * config/nds32/nds32-multiple.md (load_multiple): Disallow
18520 (store_multiple): Ditto.
18522 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
18524 * config.gcc: Add --with-cpu support for nds32 target.
18525 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
18526 * config/nds32/nds32.opt: Add -mcpu= option.
18528 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
18530 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
18531 isel=yes): Warn for these deprecated options.
18533 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
18535 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
18536 ISA_2_5_MASKS_EMBEDDED.
18538 2018-02-23 Jakub Jelinek <jakub@redhat.com>
18540 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
18541 p->max as pointers rather than using iterative_hash_expr.
18543 2018-02-23 Carl Love <cel@us.ibm.com>
18545 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
18546 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
18548 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
18549 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
18550 P8V_BUILTIN_VEC_VUNSIGNED2.
18552 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
18555 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
18556 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
18558 * lra-constraints.c (curr_insn_transform): Set up
18559 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
18561 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
18562 * lra-eliminations.c (spill_pseudos): Ditto.
18563 (process_insn_for_elimination): Ditto.
18564 * lra-lives.c (reg_early_clobber_p): Use the new macros.
18565 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
18566 LRA_NON_CLOBBERED_ALT.
18568 2018-02-22 Martin Sebor <msebor@redhat.com>
18570 PR tree-optimization/84480
18571 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
18572 to maybe_diag_stxncpy_trunc. Call it.
18573 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
18574 from gimple_fold_builtin_strcpy. Print inlining stack.
18575 (handle_builtin_stxncpy): Print inlining stack.
18576 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
18578 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
18581 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
18582 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
18583 and -fcheck-pointer-bounds are used together.
18584 (indirect_thunk_prefix): New enum.
18585 (indirect_thunk_need_prefix): New function.
18586 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
18587 "_nt" instead of "_bnd" for NOTRACK prefix.
18588 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
18589 (output_indirect_thunk_function): Likewise.
18591 (ix86_code_end): Update output_indirect_thunk_function calls.
18592 (ix86_output_indirect_branch_via_reg): Replace
18593 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
18594 (ix86_output_indirect_branch_via_push): Likewise.
18595 (ix86_output_function_return): Likewise.
18596 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
18597 incompatible with -fcf-protection=branch and
18598 -fcheck-pointer-bounds.
18600 2018-02-22 Steve Ellcey <sellcey@cavium.com>
18603 * config/aarch64/aarch64.c (aarch64_print_address_internal):
18604 Change gcc_assert call to output_operand_lossage.
18606 2018-02-22 Steve Ellcey <sellcey@cavium.com>
18608 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
18610 2018-02-22 DJ Delorie <dj@redhat.com>
18611 Sebastian Perta <sebastian.perta@renesas.com>
18612 Oleg Endo <olegendo@gcc.gnu.org>
18614 * config/rx/rx.c (rx_rtx_costs): New function.
18615 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
18617 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
18619 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
18621 2018-02-22 Martin Liska <mliska@suse.cz>
18624 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
18625 Add "native" as a possible value.
18627 2018-02-22 Martin Liska <mliska@suse.cz>
18630 * config/i386/i386.c (ix86_option_override_internal):
18631 Add "native" as a possible value for -march and -mtune.
18633 2018-02-22 Jakub Jelinek <jakub@redhat.com>
18636 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
18637 to all type variants.
18639 PR tree-optimization/84503
18640 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
18641 width as info->bitpos + info->bitsize - start.
18642 (merged_store_group::merge_overlapping): Simplify width computation.
18643 (check_no_overlap): New function.
18644 (imm_store_chain_info::try_coalesce_bswap): Compute expected
18645 start + width and last_order of the group, fail if check_no_overlap
18647 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
18648 to group if check_no_overlap fails.
18650 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
18652 * config/rs6000/altivec.md: Delete contraint arguments to
18653 define_expand, define_split, and define_peephole2, and in
18654 define_insn_and_split if always unused.
18655 * config/rs6000/darwin.md: Ditto.
18656 * config/rs6000/dfp.md: Ditto.
18657 * config/rs6000/rs6000.md: Ditto.
18658 * config/rs6000/sync.md: Ditto.
18659 * config/rs6000/vector.md: Ditto.
18660 * config/rs6000/vsx.md: Ditto.
18662 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
18664 * config/rs6000/altivec.md: Write output control strings as braced
18665 blocks instead of double-quoted strings.
18666 * config/rs6000/darwin.md: Ditto.
18667 * config/rs6000/rs6000.md: Ditto.
18668 * config/rs6000/vector.md: Ditto.
18669 * config/rs6000/vsx.md: Ditto.
18671 2018-02-21 Jason Merrill <jason@redhat.com>
18673 PR c++/84314 - ICE with templates and fastcall attribute.
18674 * attribs.c (build_type_attribute_qual_variant): Remove assert.
18676 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
18678 * ipa-cp.c (determine_versionability): Fix comment typos.
18680 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
18683 * ipa-cp.c (determine_versionability): Do not version functions caling
18686 2018-02-21 Martin Liska <mliska@suse.cz>
18689 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
18690 Add "native" as a possible value.
18691 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
18692 the macro when native cpu detection is available.
18694 2018-02-21 Martin Liska <mliska@suse.cz>
18697 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
18698 Add "native" as a possible value.
18699 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
18700 when native cpu detection is available.
18702 2018-02-21 Jakub Jelinek <jakub@redhat.com>
18703 Martin Sebor <msebor@redhat.com>
18705 PR tree-optimization/84478
18706 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
18708 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
18709 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
18710 support which is conservatively correct, for 2 only stay conservative
18711 for maxlen. Formatting and comment capitalization fixes. Add STRICT
18712 argument to the 2 argument get_range_strlen, adjust 6 arg
18713 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
18715 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
18716 (gimple_fold_builtin_strlen): Pass true as last argument to
18719 2018-02-20 Martin Sebor <msebor@redhat.com>
18721 PR middle-end/84095
18722 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
18723 (builtin_memref::set_base_and_offset): Same. Handle inner references.
18724 (builtin_memref::builtin_memref): Factor out parts into
18725 set_base_and_offset and call it.
18727 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
18729 PR middle-end/84406
18730 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
18731 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
18732 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
18733 search at the associated MODE_INT.
18735 2018-02-20 Jeff Law <law@redhat.com>
18737 PR middle-end/82123
18738 PR tree-optimization/81592
18739 PR middle-end/79257
18740 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
18741 for range data rather than using global data.
18742 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
18743 range data rather than using global data.
18744 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
18745 pass it to children as needed.
18746 (struct directive::fmtresult): Similarly.
18747 (struct directive::set_width): Similarly.
18748 (struct directive::set_precision): Similarly.
18749 (format_integer, format_directive, parse_directive): Similarly.
18750 (format_none): Accept unnamed vr_values parameter.
18751 (format_percent, format_floating, format_character): Similarly.
18752 (format_string, format_plain): Similarly.
18753 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
18754 the EVRP range analyzer for range data rather than using global data.
18755 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
18756 gimple-ssa-evrp-analyze.h
18757 (class sprintf_dom_walker): Add after_dom_children member function.
18758 Add evrp_range_analyzer member.
18759 (sprintf_dom_walker::before_dom_children): Call into the EVRP
18760 range analyzer as needed.
18761 (sprintf_dom_walker::after_dom_children): New member function.
18762 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
18764 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
18765 (evrp_range_analyzer::pop_to_marker): Likewise.
18767 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
18769 PR tree-optimization/84419
18770 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
18771 with the required type if its current type is compatible but
18774 2018-02-20 Jakub Jelinek <jakub@redhat.com>
18776 PR middle-end/82004
18777 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
18778 after vectorization.
18780 2018-02-20 Martin Liska <mliska@suse.cz>
18783 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
18784 possible values if we don't have a hint.
18786 2018-02-20 Martin Liska <mliska@suse.cz>
18790 * final.c (shorten_branches): Build align_tab array with one
18792 * opts.c (finish_options): Add alignment option limit check.
18793 (MAX_CODE_ALIGN): Likewise.
18794 (MAX_CODE_ALIGN_VALUE): Likewise.
18795 * doc/invoke.texi: Document maximum allowed option value for
18796 all -falign-* options.
18798 2018-02-19 Jakub Jelinek <jakub@redhat.com>
18801 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
18802 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
18803 * var-tracking.c (emit_note_insn_var_location): Remove all references
18804 to NOTE_INSN_CALL_ARG_LOCATION.
18805 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
18806 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
18807 Use copy_rtx_if_shared.
18808 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
18809 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
18810 (dwarf2out_var_location): Remove handling of
18811 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
18813 * final.c (final_scan_insn): Remove all references to
18814 NOTE_INSN_CALL_ARG_LOCATION.
18815 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
18816 before dumping final insns.
18817 * except.c (emit_note_eh_region_end): Remove all references to
18818 NOTE_INSN_CALL_ARG_LOCATION.
18819 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
18820 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
18821 * config/arc/arc.c (hwloop_optimize): Likewise.
18822 * config/arm/arm.c (create_fix_barrier): Likewise.
18823 * config/s390/s390.c (s390_chunkify_start): Likewise.
18824 * config/sh/sh.c (find_barrier): Likewise.
18825 * config/i386/i386.c (rest_of_insert_endbranch,
18826 ix86_seh_fixup_eh_fallthru): Likewise.
18827 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
18828 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
18829 * config/frv/frv.c (frv_function_prologue): Likewise.
18830 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
18832 (note_outside_basic_block_p): Remove all references to
18833 NOTE_INSN_CALL_ARG_LOCATION.
18834 * gengtype.c (adjust_field_rtx_def): Likewise.
18835 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
18837 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
18838 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
18841 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
18844 PR tree-optimization/84452
18845 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
18846 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
18849 2018-02-19 Martin Liska <mliska@suse.cz>
18852 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
18854 2018-02-19 Martin Liska <mliska@suse.cz>
18855 Richard Sandiford <richard.sandiford@linaro.org>
18857 PR tree-optimization/82491
18858 * gimple-fold.c (get_base_constructor): Make earlier bail out
18861 2018-02-19 Carl Love <cel@us.ibm.com>
18863 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
18864 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
18866 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
18867 P8V_BUILTIN_VEC_NEG.
18869 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
18871 * config/rl78/rl78.md (movdf): New define expand.
18873 2018-02-19 Martin Liska <mliska@suse.cz>
18876 * doc/invoke.texi: Fix typo.
18877 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
18879 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
18881 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
18882 handle rs6000_single_float and rs6000_double_float specially for
18885 2018-02-16 Jeff Law <law@redhat.com>
18887 * config/rx/rx.c (add_pop_cfi_notes): New function.;
18888 (pop_regs): Use it.
18890 2018-02-16 Jakub Jelinek <jakub@redhat.com>
18893 * ipa-inline.c (inline_small_functions): Fix a typo.
18895 2018-02-16 Nathan Sidwell <nathan@acm.org>
18897 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
18899 2018-02-16 Carl Love <cel@us.ibm.com>
18901 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
18902 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
18903 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
18904 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
18905 expansion to P8V_BUILTIN_VEC_FLOAT2.
18907 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
18909 PR rtl-optimization/70023
18910 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
18911 src_regno into account.
18913 2018-02-16 Carl Love <cel@us.ibm.com>
18915 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
18916 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
18917 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
18918 * config/rs6000/rs6000.c: Remove case statements for
18919 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
18920 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
18921 and P9V_BUILTIN_VEC_VINSERT4B.
18922 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
18923 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
18924 * config/rs6000/vsx.md:
18925 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
18928 2018-02-16 Carl Love <cel@us.ibm.com>
18930 * config/rs6000/altivec.h: Add builtin names vec_extract4b
18932 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
18934 * config/rs6000/rs6000-c.c: Add the definitions for
18935 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
18936 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
18937 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
18938 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
18939 definition for insert4b and define insn *insert3b_internal.
18940 * doc/extend.texi: Add documentation for vec_extract4b.
18942 2018-02-16 Nathan Sidwell <nathan@acm.org>
18944 * doc/extend.texi (Backwards Compatibility): Mention friend
18945 injection. Note for-scope is deprecated.
18946 * doc/invoke.texi (-ffriend-injection): Deprecate.
18948 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
18950 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
18951 that moved to I2, also allow destinations that are a paradoxical
18952 subreg (instead of a normal reg).
18954 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
18957 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
18960 2018-02-16 Richard Biener <rguenther@suse.de>
18962 PR tree-optimization/84037
18963 PR tree-optimization/84016
18965 * config/i386/i386.c (ix86_builtin_vectorization_cost):
18966 Adjust vec_construct for the fact we need additional higher latency
18967 128bit inserts for AVX256 and AVX512 vector builds.
18968 (ix86_add_stmt_cost): Scale vector construction cost for
18971 2018-02-16 Richard Biener <rguenther@suse.de>
18973 PR tree-optimization/84417
18974 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
18975 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
18976 (non_rewritable_lvalue_p): Likewise, use poly-ints.
18978 2018-02-16 Martin Liska <mliska@suse.cz>
18981 * internal-fn.def (ASAN_CHECK): Set proper flags.
18982 (ASAN_MARK): Likewise.
18984 2018-02-16 Julia Koval <julia.koval@intel.com>
18986 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
18987 from PTA_CANNONLAKE.
18989 2018-02-16 Jakub Jelinek <jakub@redhat.com>
18992 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
18993 Use ++iter rather than iter++ for std::list iterators.
18994 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
18995 defer deleting them until all nodes in the forest are processed. Do
18996 free even leaf nodes. Change to_process into auto_vec.
18999 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
19000 * vec.h (vec_default_construct): Use memset instead of placement new
19001 if BROKEN_VALUE_INITIALIZATION is defined.
19002 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
19003 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
19006 PR rtl-optimization/83723
19007 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
19008 * lra.c (lra_substitute_pseudo): Likewise. If true, use
19009 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
19011 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
19013 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
19015 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
19017 PR rtl-optimization/81443
19018 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
19019 from inner REGs to paradoxical SUBREGs.
19021 2018-02-16 Richard Biener <rguenther@suse.de>
19023 PR tree-optimization/84399
19024 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
19025 For operands we can analyze at their definition make sure we can
19026 analyze them at each use as well.
19028 2018-02-16 Richard Biener <rguenther@suse.de>
19030 PR tree-optimization/84190
19031 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
19032 volatile accesses if the decl isn't volatile.
19034 2018-02-15 Jason Merrill <jason@redhat.com>
19036 PR c++/84314 - ICE with templates and fastcall attribute.
19037 * attribs.c (build_type_attribute_qual_variant): Don't clobber
19038 TYPE_CANONICAL on an existing type.
19040 2018-02-15 Jakub Jelinek <jakub@redhat.com>
19042 PR tree-optimization/84383
19043 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
19044 dstoff nor call operand_equal_p if dstbase is NULL.
19046 PR tree-optimization/84334
19047 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
19048 also a CONSTANT_CLASS_P, punt.
19050 2018-02-14 Jim Wilson <jimw@sifive.com>
19052 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
19053 first SMALL_OPERAND check. New local min_second_step. Move assert
19054 to where locals are set. Add TARGET_RVC support.
19055 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
19057 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
19059 * doc/invoke.texi: Correct -Wformat-overflow code sample.
19061 2018-02-14 Martin Sebor <msebor@redhat.com>
19063 PR tree-optimization/83698
19064 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
19065 arrays constrain the offset range to their bounds.
19066 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
19067 (builtin_access::overlap): Avoid setting the size of overlap if it's
19069 (maybe_diag_overlap): Also consider arrays when deciding what values
19070 of offsets to include in diagnostics.
19072 2018-02-14 Martin Sebor <msebor@redhat.com>
19075 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
19076 that correspond to the kind of a declaration.
19078 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
19081 * config/pa/pa.md: Load address of PIC label using the linkage table
19082 if the label is nonlocal.
19084 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
19086 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
19087 warning message if user requests -maltivec=be.
19088 * doc/invoke.texi: Document deprecation of -maltivec=be.
19090 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
19093 * config/rs6000/rs6000-c.c: Update definitions for
19094 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
19095 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
19097 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
19100 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
19101 add _get_ssp intrinsics. Remove argument from
19102 __builtin_ia32_rdssp[d|q].
19103 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
19104 * config/i386/i386-builtin.def: Remove argument from
19105 __builtin_ia32_rdssp[d|q].
19106 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
19107 ix86_expand_special_args_builtin for _rdssp[d|q].
19108 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
19109 Clear register before usage.
19110 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
19111 Add documentation for new _get_ssp and _inc_ssp intrinsics.
19113 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
19115 PR tree-optimization/84357
19116 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
19117 operand 1 of an ARRAY_REF too.
19119 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
19122 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
19123 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
19125 (set_of_reg): New struct.
19126 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
19127 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
19128 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
19130 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
19131 Split into bitclr, bitset, bitinvert patterns if appropriate.
19132 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
19133 use rx_fuse_in_memory_bitop.
19134 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
19135 to named insn, correct maximum insn length.
19137 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
19140 * machmode.def: Define a complex mode for PARTIAL_INT.
19141 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
19143 * doc/rtl.texi: Document CSPImode.
19144 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
19146 (msp430_hard_regno_nregs_with_padding): Likewise.
19148 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
19151 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
19153 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
19155 PR rtl-optimization/84169
19156 * combine.c (try_combine): New variable split_i2i3. Set it to true if
19157 we generated a parallel as new i3 and we split that to new i2 and i3
19158 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
19159 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
19160 those to i2, not i1. Partially rewrite this scan code.
19162 2018-02-13 Jakub Jelinek <jakub@redhat.com>
19165 * stor-layout.c (place_field): For variable length fields, adjust
19166 offset_align afterwards not just based on the field's alignment,
19167 but also on the size.
19169 PR middle-end/84309
19170 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
19171 of exps and logs in the use_exp2 case.
19173 2018-02-13 Jeff Law <law@redhat.com>
19175 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
19176 entry for "vector".
19178 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
19181 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
19185 * common.opt (gas-loc-support, gas-locview-support): New.
19186 (ginline-points, ginternal-reset-location-views): New.
19187 * doc/invoke.texi: Document them. Use @itemx where intended.
19188 (gvariable-location-views): Adjust.
19189 * target.def (reset_location_view): New.
19190 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
19191 (TARGET_RESET_LOCATION_VIEW): New.
19192 * doc/tm.texi: Rebuilt.
19193 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
19194 (dwarf2out_default_as_locview_support): New.
19195 (output_asm_line_debug_info): Use option variables.
19196 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
19197 (output_loc_list): Likewise.
19198 (add_high_low_attributes): Check option variables.
19199 Don't output entry view attribute in strict mode.
19200 (gen_inlined_subroutine_die): Check option variables.
19201 (dwarf2out_inline_entry): Likewise.
19202 (init_sections_and_labels): Likewise.
19203 (dwarf2out_early_finish): Likewise.
19204 (maybe_reset_location_view): New, from...
19205 (dwarf2out_var_location): ... here. Call it.
19206 * debug.h (dwarf2out_default_as_loc_support): Declare.
19207 (dwarf2out_default_as_locview_support): Declare.
19208 * hooks.c (hook_int_rtx_insn_0): New.
19209 * hooks.h (hook_int_rtx_insn_0): Declare.
19210 * toplev.c (process_options): Take -gas-loc-support and
19211 -gas-locview-support from dwarf2out. Enable
19212 -gvariable-location-views by default only with locview
19213 assembler support. Enable -ginternal-reset-location-views by
19214 default only if the target defines the corresponding hook.
19215 Enable -ginline-points by default if location views are
19216 enabled; force it disabled if statement frontiers are
19218 * tree-inline.c (expand_call_inline): Check option variables.
19219 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
19221 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
19223 PR tree-optimization/84321
19224 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
19225 handling. Also check whether the anti-range contains any values
19226 that satisfy the mask; switch to a VR_RANGE if not.
19228 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
19231 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
19233 2018-02-13 Martin Jambor <mjambor@suse.cz>
19236 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
19237 of call statements, also set location of a load to a temporary.
19239 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
19241 * config/rl78/rl78.c (add_vector_labels): New function.
19242 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
19243 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
19244 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
19245 which checks that no arguments are passed.
19246 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
19247 * doc/extend.texi: Documentation for the new attribute.
19249 2018-02-13 Andreas Schwab <schwab@suse.de>
19251 * config/riscv/linux.h (CPP_SPEC): Define.
19253 2018-02-13 Jakub Jelinek <jakub@redhat.com>
19256 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
19257 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
19258 OPTION_MASK_ISA_AES as first argument to def_builtin_const
19259 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
19260 instead of OPTION_MASK_ISA_PCLMUL as first argument to
19261 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
19262 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
19263 temporarily for AES and PCLMUL builtins.
19265 PR tree-optimization/84339
19266 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
19267 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
19270 PR middle-end/84309
19271 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
19272 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
19273 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
19275 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
19277 * omp-simd-clone.h: New file.
19278 * omp-simd-clone.c: Include omp-simd-clone.h.
19279 (expand_simd_clones): No longer static.
19280 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
19281 cgraph.h and omp-simd-clone.h.
19282 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
19283 (vect_recog_widen_shift_pattern): Formatting fix.
19284 (vect_pattern_recog_1): Don't check optab for calls.
19287 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
19288 operands[2] into a REG before using gen_lowpart on it.
19290 2018-02-12 Jeff Law <law@redhat.com>
19293 * config/sh/sh.c (find_barrier): Consider a sibling call
19296 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
19297 successfully back substituting a reg.
19299 2018-02-12 Richard Biener <rguenther@suse.de>
19301 PR tree-optimization/84037
19302 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
19303 parameter, move visited init to caller.
19304 (vect_slp_analyze_operations): Separate cost from validity
19305 check, initialize visited once for all instances.
19306 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
19308 * tree-vect-stmts.c (vect_model_simple_cost): Make early
19310 (vect_model_promotion_demotion_cost): Likewise.
19311 (vectorizable_bswap): Guard cost modeling with !slp_node
19312 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
19314 (vectorizable_call): Likewise.
19315 (vectorizable_conversion): Likewise.
19316 (vectorizable_assignment): Likewise.
19317 (vectorizable_shift): Likewise.
19318 (vectorizable_operation): Likewise.
19319 (vectorizable_store): Likewise.
19320 (vectorizable_load): Likewise.
19321 (vectorizable_condition): Likewise.
19322 (vectorizable_comparison): Likewise.
19324 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
19327 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
19328 (ASAN_MARK): Fix fnspec to account for return value, change pointer
19329 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
19331 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
19333 PR middle-end/83665
19334 * params.def (inline-min-speedup): Increase from 8 to 15.
19335 (max-inline-insns-auto): Decrease from 40 to 30.
19336 * ipa-split.c (consider_split): Add some buffer for function to
19337 be considered inlining candidate.
19338 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
19341 2018-02-12 Richard Biener <rguenther@suse.de>
19343 PR tree-optimization/84037
19344 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
19345 matched stmts if we cannot swap the non-matched ones.
19347 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
19349 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
19350 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
19351 _mm_maskz_scalef_round_ss): New intrinsics.
19352 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
19353 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
19354 __builtin_ia32_scalefss_round): Remove.
19355 (__builtin_ia32_scalefsd_mask_round,
19356 __builtin_ia32_scalefss_mask_round): New intrinsics.
19357 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
19358 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
19359 ((match_operand:VF_128 2 "<round_nimm_predicate>"
19360 "<round_constraint>")): Changed to ...
19361 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
19362 "<round_scalar_constraint>")): ... this.
19363 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
19364 %0, %1, %2<round_op3>}"): Changed to ...
19365 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
19366 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
19367 %2<round_scalar_mask_op3>}"): ... this.
19368 * config/i386/subst.md (round_scalar_nimm_predicate): New.
19370 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
19372 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
19373 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
19374 (_mm_maskz_sqrt_round_ss): New intrinsics.
19375 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
19376 (__builtin_ia32_sqrtsd_mask_round)
19377 (__builtin_ia32_sqrtss_mask_round): New builtins.
19378 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
19379 (__builtin_ia32_sqrtss_round): Remove.
19380 (__builtin_ia32_sqrtsd_mask_round)
19381 (__builtin_ia32_sqrtss_mask_round): New builtins.
19382 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
19383 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
19384 ((match_operand:VF_128 1 "vector_operand"
19385 "xBm,<round_constraint>")): Changed to ...
19386 ((match_operand:VF_128 1 "vector_operand"
19387 "xBm,<round_scalar_constraint>")): ... this.
19388 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
19389 %0, %2, %<iptr>1<round_op3>}): Changed to ...
19390 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
19391 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
19392 %<iptr>1<round_scalar_mask_op3>}): ... this.
19393 ((set_attr "prefix" "<round_prefix>")): Changed to ...
19394 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
19396 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
19399 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
19400 Cast vec_cmpeq result to correct type.
19401 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
19402 Cast vec_cmpgt result to correct type.
19404 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
19406 * final.c (final_scan_insn_1): Renamed from...
19407 (final_scan_insn): ... this. New wrapper, to recover
19408 seen from the outermost call in recursive ones.
19409 * config/sparc/sparc.c (output_return): Drop seen from call.
19410 (output_sibcall): Likewise.
19411 * config/visium/visium.c (output_branch): Likewise.
19413 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
19415 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
19418 2018-02-10 Alan Modra <amodra@gmail.com>
19421 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
19422 Specify LR as an input.
19424 2018-02-10 Jakub Jelinek <jakub@redhat.com>
19427 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
19428 remove_member_access_dummy_vars): New functions.
19429 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
19430 lower_omp_1, execute_lower_omp): Use them.
19432 PR rtl-optimization/84308
19433 * shrink-wrap.c (spread_components): Release todo vector.
19435 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
19437 PR rtl-optimization/57193
19438 * ira-color.c (struct allocno_color_data): Add member
19439 conflict_allocno_hard_prefs.
19440 (update_conflict_allocno_hard_prefs): New.
19441 (bucket_allocno_compare_func): Add a preference based on
19442 conflict_allocno_hard_prefs.
19443 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
19444 (color_allocnos): Remove a dead code. Initiate
19445 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
19447 2018-02-09 Jakub Jelinek <jakub@redhat.com>
19450 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
19451 constraint from =wa to wa. Avoid a subreg on the output operand,
19452 instead use a pseudo and subreg it in a move.
19453 (p9_xxbrd_<mode>): Changed to ...
19454 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
19455 (p9_xxbrd_v2df): New expander.
19456 (p9_xxbrw_<mode>): Changed to ...
19457 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
19458 (p9_xxbrw_v4sf): New expander.
19460 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
19462 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
19464 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
19467 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
19468 multiply in 32-bit mode.
19469 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
19470 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
19473 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
19475 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
19476 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
19477 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
19478 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
19480 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
19483 * dwarf2out.c (is_trivial_indirect_ref): New function.
19484 (dwarf2out_late_global_decl): Do not generate a location
19485 attribute for variables that have a non-trivial DECL_VALUE_EXPR
19486 and that are not defined in the current unit.
19488 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
19490 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
19491 instead of a libcall for UNORDERED.
19493 2018-02-09 Tamar Christina <tamar.christina@arm.com>
19496 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
19497 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
19499 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19502 * config/s390/s390.c (s390_set_current_function): Invoke
19503 s390_indirect_branch_settings also if fndecl didn't change.
19505 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
19507 * config/rs6000/rs6000.md (blockage): Set length to zero.
19509 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
19511 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
19513 2018-02-09 Jakub Jelinek <jakub@redhat.com>
19516 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
19517 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
19521 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
19522 PARALLEL incoming that failed vt_get_decl_and_offset check.
19524 PR middle-end/84237
19525 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
19526 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
19528 (get_variable_section): For decls in named .bss* sections pass true as
19529 second argument to bss_initializer_p.
19531 2018-02-09 Marek Polacek <polacek@redhat.com>
19532 Jakub Jelinek <jakub@redhat.com>
19535 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
19536 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
19537 Sync some changes from cxx_fold_indirect_ref.
19539 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
19541 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
19543 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
19544 (BLOCK_INLINE_ENTRY_LABEL): New.
19545 (dwarf2out_var_location): Disregard inline entry markers.
19546 (inline_entry_data): New struct.
19547 (inline_entry_data_hasher): New hashtable type.
19548 (inline_entry_data_hasher::hash): New.
19549 (inline_entry_data_hasher::equal): New.
19550 (inline_entry_data_table): New variable.
19551 (add_high_low_attributes): Add DW_AT_entry_pc and
19552 DW_AT_GNU_entry_view attributes if a pending entry is found
19553 in inline_entry_data_table. Add old entry_pc attribute only
19554 if debug nonbinding markers are disabled.
19555 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
19556 markers are enabled.
19557 (block_within_block_p, dwarf2out_inline_entry): New.
19558 (dwarf2out_finish): Check that no entries remained in
19559 inline_entry_data_table.
19560 * final.c (reemit_insn_block_notes): Handle inline entry notes.
19561 (final_scan_insn, notice_source_line): Likewise.
19562 (rest_of_clean_state): Skip inline entry markers.
19563 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
19565 * gimple.c (gimple_build_debug_inline_entry): New.
19566 * gimple.h (enum gimple_debug_subcode): Add
19567 GIMPLE_DEBUG_INLINE_ENTRY.
19568 (gimple_build_debug_inline_entry): Declare.
19569 (gimple_debug_inline_entry_p): New.
19570 (gimple_debug_nonbind_marker_p): Adjust.
19571 * insn-notes.def (INLINE_ENTRY): New.
19572 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
19573 inline entry marker notes.
19574 (print_insn): Likewise.
19575 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
19576 (INSN_DEBUG_MARKER_KIND): Likewise.
19577 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
19578 * tree-inline.c (expand_call_inline): Build and insert
19579 debug_inline_entry stmt.
19580 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
19581 inline entry blocks early, if nonbind markers are enabled.
19582 (dump_scope_block): Dump fragment info.
19583 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
19584 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
19585 (gimple_build_debug_inline_entry): New.
19586 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
19587 Enable/disable inline entry points too.
19588 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
19589 (DEBUG_INSN): Describe inline entry markers.
19591 * common.opt (gvariable-location-views): New.
19592 (gvariable-location-views=incompat5): New.
19593 * config.in: Rebuilt.
19594 * configure: Rebuilt.
19595 * configure.ac: Test assembler for view support.
19596 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
19597 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
19598 * dwarf2out.c (var_loc_view): New typedef.
19599 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
19600 (dwarf2out_locviews_in_attribute): New.
19601 (dwarf2out_locviews_in_loclist): New.
19602 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
19603 (enum dw_line_info_opcode): Add LI_adv_address.
19604 (struct dw_line_info_table): Add view.
19605 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
19606 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
19607 (zero_view_p): New variable.
19608 (ZERO_VIEW_P): New macro.
19609 (output_asm_line_debug_info): New.
19610 (struct var_loc_node): Add view.
19611 (add_AT_view_list, AT_loc_list): New.
19612 (add_var_loc_to_decl): Add view param. Test it against last.
19613 (new_loc_list): Add view params. Record them.
19614 (AT_loc_list_ptr): Handle loc and view lists.
19615 (view_list_to_loc_list_val_node): New.
19616 (print_dw_val): Handle dw_val_class_view_list.
19617 (size_of_die): Likewise.
19618 (value_format): Likewise.
19619 (loc_list_has_views): New.
19620 (gen_llsym): Set vl_symbol too.
19621 (maybe_gen_llsym, skip_loc_list_entry): New.
19622 (dwarf2out_maybe_output_loclist_view_pair): New.
19623 (output_loc_list): Output view list or entries too.
19624 (output_view_list_offset): New.
19625 (output_die): Handle dw_val_class_view_list.
19626 (output_dwarf_version): New.
19627 (output_compilation_unit_header): Use it.
19628 (output_skeleton_debug_sections): Likewise.
19629 (output_rnglists, output_line_info): Likewise.
19630 (output_pubnames, output_aranges): Update version comments.
19631 (output_one_line_info_table): Output view numbers in asm comments.
19632 (dw_loc_list): Determine current endview, pass it to new_loc_list.
19633 Call maybe_gen_llsym.
19634 (loc_list_from_tree_1): Adjust.
19635 (add_AT_location_description): Create view list attribute if
19636 needed, check it's absent otherwise.
19637 (convert_cfa_to_fb_loc_list): Adjust.
19638 (maybe_emit_file): Call output_asm_line_debug_info for test.
19639 (dwarf2out_var_location): Reset views as needed. Precompute
19640 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
19641 attribute. Set view.
19642 (new_line_info_table): Reset next view.
19643 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
19644 (dwarf2out_source_line): Likewise. Output view resets and labels to
19645 the assembler, or select appropriate line info opcodes.
19646 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
19647 (optimize_string_length): Catch it. Adjust.
19648 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
19649 dw_val_class_view_list, and remove it if no longer needed.
19650 (hash_loc_list): Hash view numbers.
19651 (loc_list_hasher::equal): Compare them.
19652 (optimize_location_lists): Check whether a view list symbol is
19653 needed, and whether the locview attribute is present, and
19654 whether they match. Remove the locview attribute if no longer
19656 (index_location_lists): Call skip_loc_list_entry for test.
19657 (dwarf2out_finish): Call output_asm_line_debug_info for test.
19658 Use output_dwarf_version.
19659 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
19660 (struct dw_val_node): Add val_view_list.
19661 * final.c (SEEN_NEXT_VIEW): New.
19662 (set_next_view_needed): New.
19663 (clear_next_view_needed): New.
19664 (maybe_output_next_view): New.
19665 (final_start_function): Rename to...
19666 (final_start_function_1): ... this. Take pointer to FIRST,
19667 add SEEN parameter. Emit param bindings in the initial view.
19668 (final_start_function): Reintroduce SEEN-less interface.
19669 (final): Rename to...
19670 (final_1): ... this. Take SEEN parameter. Output final pending
19671 next view at the end.
19672 (final): Reintroduce seen-less interface.
19673 (final_scan_insn): Output pending next view before switching
19674 sections or ending a block. Mark the next view as needed when
19675 outputting variable locations. Notify debug backend of section
19676 changes, and of location view changes.
19677 (rest_of_handle_final): Adjust.
19678 * toplev.c (process_options): Autodetect value for debug variable
19679 location views option. Warn on incompat5 without -gdwarf-5.
19680 * doc/invoke.texi (gvariable-location-views): New.
19681 (gvariable-location-views=incompat5): New.
19682 (gno-variable-location-views): New.
19684 2018-02-08 David Malcolm <dmalcolm@redhat.com>
19686 PR tree-optimization/84136
19687 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
19688 that the result of find_edge is non-NULL.
19690 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
19693 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
19694 storing integer register in SImode. Fix cost of 256 and 512
19695 byte aligned SSE register store.
19697 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
19699 * config/i386/i386.c (ix86_multiplication_cost): Fix
19700 multiplication cost for TARGET_AVX512DQ.
19702 2018-02-08 Marek Polacek <polacek@redhat.com>
19704 PR tree-optimization/84238
19705 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
19708 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19710 PR tree-optimization/84265
19711 * tree-vect-stmts.c (vectorizable_store): Don't treat
19712 VMAT_CONTIGUOUS accesses as grouped.
19713 (vectorizable_load): Likewise.
19715 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19717 PR tree-optimization/81635
19718 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
19719 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
19720 (test_round_for_mask): New functions.
19721 (wide_int_cc_tests): Call test_round_for_mask.
19722 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
19723 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
19724 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
19725 range returned by get_range_info.
19727 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
19730 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
19731 * symtab.c: Include builtins.h
19732 (symtab_node::output_to_lto_symbol_table_p): Move here
19733 from lto-streamer-out.c:output_symbol_p.
19734 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
19735 (output_symbol_p): Move all logic to symtab.c
19736 (produce_symtab): Update.
19738 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19740 * config/s390/s390-opts.h (enum indirect_branch): Define.
19741 * config/s390/s390-protos.h (s390_return_addr_from_memory)
19742 (s390_indirect_branch_via_thunk)
19743 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
19744 (enum s390_indirect_branch_type): Define.
19745 * config/s390/s390.c (struct s390_frame_layout, struct
19746 machine_function): Remove.
19747 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
19748 (indirect_branch_table_label_no, indirect_branch_table_name):
19750 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
19751 (enum s390_indirect_branch_option): Define.
19752 (s390_return_addr_from_memory): New function.
19753 (s390_handle_string_attribute): New function.
19754 (s390_attribute_table): Add new attribute handler.
19755 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
19756 (s390_indirect_branch_via_thunk): New function.
19757 (s390_indirect_branch_via_inline_thunk): New function.
19758 (s390_function_ok_for_sibcall): When jumping via thunk disallow
19759 sibling call optimization for non z10 compiles.
19760 (s390_emit_call): Force indirect branch target to be a single
19761 register. Add r1 clobber for non-z10 compiles.
19762 (s390_emit_epilogue): Emit return jump via return_use expander.
19763 (s390_reorg): Handle JUMP_INSNs as execute targets.
19764 (s390_option_override_internal): Perform validity checks for the
19765 new command line options.
19766 (s390_indirect_branch_attrvalue): New function.
19767 (s390_indirect_branch_settings): New function.
19768 (s390_set_current_function): Invoke s390_indirect_branch_settings.
19769 (s390_output_indirect_thunk_function): New function.
19770 (s390_code_end): Implement target hook.
19771 (s390_case_values_threshold): Implement target hook.
19772 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
19774 * config/s390/s390.h (struct s390_frame_layout)
19775 (struct machine_function): Move here from s390.c.
19776 (TARGET_INDIRECT_BRANCH_NOBP_RET)
19777 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
19778 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
19779 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
19780 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
19781 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
19782 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
19783 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
19784 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
19785 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
19786 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
19787 (mnemonic attribute): Add values which aren't recognized
19789 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
19790 pattern for branch conversion. Fix mnemonic attribute.
19791 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
19792 indirect branch via thunk if requested.
19793 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
19794 ("*indirect_jump"): Disable for branch conversion using out of
19796 ("indirect_jump_via_thunk<mode>_z10")
19797 ("indirect_jump_via_thunk<mode>")
19798 ("indirect_jump_via_inlinethunk<mode>_z10")
19799 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
19800 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
19801 ("casesi_jump_via_inlinethunk<mode>_z10")
19802 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
19803 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
19804 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
19805 ("*indirect2_jump"): Disable for branch conversion.
19806 ("casesi_jump"): Turn into expander and expand patterns for branch
19808 ("return_use"): New expander.
19809 ("*return"): Emit return via thunk and rename it to ...
19810 ("*return<mode>"): ... this one.
19811 * config/s390/s390.opt: Add new options and and enum for the
19814 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19816 * lra-constraints.c (match_reload): Unconditionally use
19817 gen_lowpart_SUBREG, rather than selecting between that
19818 and equivalent gen_rtx_SUBREG code.
19820 2018-02-08 Richard Biener <rguenther@suse.de>
19822 PR tree-optimization/84233
19823 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
19824 changed flag instead of boguously re-using phi_inserted.
19826 2018-02-08 Martin Jambor <mjambor@suse.cz>
19828 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
19829 static local variables.
19831 2018-02-08 Richard Biener <rguenther@suse.de>
19833 PR tree-optimization/84278
19834 * tree-vect-stmts.c (vectorizable_store): When looking for
19835 smaller vector types to perform grouped strided loads/stores
19836 make sure the mode is supported by the target.
19837 (vectorizable_load): Likewise.
19839 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
19841 * config/aarch64/aarch64.c (aarch64_components_for_bb):
19842 Increase LDP/STP opportunities by adding adjacent callee-saves.
19844 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
19846 PR rtl-optimization/84068
19847 PR rtl-optimization/83459
19848 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
19850 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
19852 PR tree-optimization/84224
19853 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
19854 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
19855 non-zero arguments.
19857 2018-02-07 Iain Sandoe <iain@codesourcery.com>
19860 * config/rs6000/altivec.md (*restore_world): Remove LR use.
19861 * config/rs6000/predicates.md (restore_world_operation): Adjust op
19862 count, remove one USE.
19864 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
19866 * doc/install.texi (Configuration): Document the
19867 --with-long-double-format={ibm,ieee} PowerPC configuration
19871 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
19872 Convert from define_expand to be define_insn_and_split. Rework
19873 float/double/_Float128 conversions to QI/HI/SImode to work with
19874 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
19875 conversions to QI/HImode types did a store and then a load to
19876 truncate the value. For conversions to VSX registers, don't split
19877 the insn, instead emit the code directly. Use the code iterator
19878 any_fix to combine signed and unsigned conversions.
19879 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
19880 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
19881 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19882 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
19883 (fix_<mode>di2_hw): Likewise.
19884 (fixuns_<mode>di2_hw): Likewise.
19885 (fix_<mode>si2_hw): Likewise.
19886 (fixuns_<mode>si2_hw): Likewise.
19887 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
19888 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19889 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
19890 fix<uns>_trunc<SFDF:mode>si2_p8.
19891 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
19892 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
19893 (fix<uns>_<mode>_mem): Likewise.
19894 (fctiw<u>z_<mode>_mem): Likewise.
19895 (fix<uns>_<mode>_mem): Likewise.
19896 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
19897 the register allocator from doing a direct move to the GPRs to do
19898 a store, and instead use the ISA 3.0 store byte/half-word from
19899 vector register instruction. For IEEE 128-bit floating point,
19900 also optimize stores of 32-bit ints.
19901 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
19903 2018-02-07 Alan Hayward <alan.hayward@arm.com>
19905 * genextract.c (push_pathstr_operand): New function to support
19907 (walk_rtx): Call push_pathstr_operand.
19908 (print_path): Support [a-zA-Z].
19910 2018-02-07 Richard Biener <rguenther@suse.de>
19912 PR tree-optimization/84037
19913 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
19914 (cse_and_gimplify_to_preheader): Declare.
19915 (vect_get_place_in_interleaving_chain): Likewise.
19916 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19918 (_loop_vec_info::~_loop_vec_info): Delete it.
19919 (cse_and_gimplify_to_preheader): New function.
19920 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
19921 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
19922 (vectorizable_load): Likewise. For grouped stores always base
19923 the IV on the first element.
19924 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
19925 condition before gimplifying.
19927 2018-02-07 Jakub Jelinek <jakub@redhat.com>
19929 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
19930 *DIV_EXPR and *MOD_EXPR.
19932 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
19935 * config/i386/i386.c (ix86_option_override_internal): Mask out
19936 the CF_SET bit when checking -fcf-protection.
19938 2018-02-07 Tom de Vries <tom@codesourcery.com>
19941 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
19944 2018-02-07 Richard Biener <rguenther@suse.de>
19946 PR tree-optimization/84204
19947 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
19950 PR tree-optimization/84205
19951 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
19952 special-case isl_ast_op_zdiv_r.
19954 PR tree-optimization/84223
19955 * graphite-scop-detection.c (gather_bbs::before_dom_children):
19956 Only add conditions from within the region.
19957 (gather_bbs::after_dom_children): Adjust.
19959 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
19962 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
19963 * config/avr/avr.md: Only post-reload split REG-REG moves if
19964 either register is GENERAL_REG_P.
19966 2018-02-07 Jakub Jelinek <jakub@redhat.com>
19968 PR tree-optimization/84235
19969 * tree-ssa-scopedtables.c
19970 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
19971 if the subtraction is performed in floating point type where NaNs are
19972 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
19973 build 1. Formatting fix.
19975 2018-02-06 Jakub Jelinek <jakub@redhat.com>
19978 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
19979 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
19980 and skip it regardless of bb boundaries. Use CALL_P macro,
19981 don't test INSN_P (insn) together with CALL_P or JUMP_P check
19982 unnecessarily, formatting fix.
19984 2018-02-06 Michael Collison <michael.collison@arm.com>
19986 * config/arm/thumb2.md:
19987 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
19988 (*thumb_mov_notscc): Ditto.
19990 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
19993 * config/rs6000/rs6000.md (su code attribute): Use "u" for
19994 unsigned_fix, not "s".
19996 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19998 * configure.ac (gcc_fn_eh_frame_ro): New function.
19999 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
20000 correct .eh_frame permissions.
20001 * configure: Regenerate.
20003 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
20005 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
20006 irrelevant options.
20008 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20010 * config/rs6000/rs6000.c (rs6000_option_override_internal):
20011 Display warning message for -mno-speculate-indirect-jumps.
20013 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
20015 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
20017 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
20019 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
20021 PR tree-optimization/84225
20022 * tree-eh.c (find_trapping_overflow): Only call
20023 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
20025 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
20028 * config/i386/i386.c: Reimplement the check of possible options
20029 -mibt/-mshstk conbination. Change error messages.
20030 * doc/invoke.texi: Fix a typo: remove extra '='.
20032 2018-02-06 Marek Polacek <polacek@redhat.com>
20034 PR tree-optimization/84228
20035 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
20037 2018-02-06 Tamar Christina <tamar.christina@arm.com>
20040 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
20041 emitted arch directives.
20042 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
20043 __ARM_FEATURE_COPROC before changing architectures.
20045 2018-02-06 Richard Biener <rguenther@suse.de>
20047 * config/i386/i386.c (print_reg): Fix typo.
20048 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
20050 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
20052 * configure: Regenerate.
20054 2018-02-05 Martin Sebor <msebor@redhat.com>
20056 PR tree-optimization/83369
20057 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
20060 2018-02-05 Martin Liska <mliska@suse.cz>
20062 * doc/invoke.texi: Cherry-pick upstream r323995.
20064 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
20066 * ira.c (ira_init_register_move_cost): Adjust comment.
20068 2018-02-05 Martin Liska <mliska@suse.cz>
20070 PR gcov-profile/84137
20071 * doc/gcov.texi: Fix typo in documentation.
20073 2018-02-05 Martin Liska <mliska@suse.cz>
20075 PR gcov-profile/83879
20076 * doc/gcov.texi: Document necessity of --dynamic-list-data when
20077 using dlopen functionality.
20079 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
20081 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
20082 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
20083 _mm_maskz_range_ss, _mm_mask_range_round_ss,
20084 _mm_maskz_range_round_ss): New intrinsics.
20085 (__builtin_ia32_rangesd128_round)
20086 (__builtin_ia32_rangess128_round): Remove.
20087 (__builtin_ia32_rangesd128_mask_round,
20088 __builtin_ia32_rangess128_mask_round): New builtins.
20089 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
20090 __builtin_ia32_rangess128_round): Remove.
20091 (__builtin_ia32_rangesd128_mask_round,
20092 __builtin_ia32_rangess128_mask_round): New builtins.
20093 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
20094 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
20095 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
20096 "<round_saeonly_constraint>")): Changed to ...
20097 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
20098 "<round_saeonly_scalar_constraint>")): ... this.
20099 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
20100 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
20101 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
20102 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
20103 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
20105 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
20107 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
20109 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
20110 Remove all values except native, 8540 and 8548.
20112 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
20114 * config/i386/i386.c (ix86_output_function_return): Pass
20115 INVALID_REGNUM, instead of -1, as invalid register number to
20116 indirect_thunk_name and output_indirect_thunk.
20118 2018-02-02 Julia Koval <julia.koval@intel.com>
20120 * config.gcc: Add -march=icelake.
20121 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
20122 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
20123 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
20124 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
20125 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
20126 (processor_target_table): Add icelake.
20127 (ix86_option_override_internal): Handle new PTAs.
20128 (get_builtin_code_for_version): Handle icelake.
20129 (M_INTEL_COREI7_ICELAKE): New.
20130 (fold_builtin_cpu): Handle icelake.
20131 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
20132 * doc/invoke.texi: Add -march=icelake.
20134 2018-02-02 Julia Koval <julia.koval@intel.com>
20136 * config/i386/i386.c (ix86_option_override_internal): Change flags type
20137 to wide_int_bitmask.
20138 * wide-int-bitmask.h: New.
20140 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
20143 * config/i386/i386.md: Replace Pmode with word_mode in
20144 builtin_setjmp_setup and builtin_longjmp to support x32.
20146 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
20150 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
20152 (rs6000_supported_cpu_names): New static variable.
20153 (linux_cpu_translation_table): Likewise.
20154 (elf_platform) <cpu>: Define new static variable and use it.
20155 Translate kernel AT_PLATFORM name to canonical name if needed.
20156 Error if platform name is unknown.
20158 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
20161 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
20163 2018-02-01 Jeff Law <law@redhat.com>
20166 * config/i386/i386.c (release_scratch_register_on_entry): Add new
20167 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
20168 the scratch if RELEASE_VIA_POP is false.
20169 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
20170 If we have to save a temporary register, decrement SIZE appropriately.
20171 Pass new arguments to release_scratch_register_on_entry.
20172 (ix86_adjust_stack_and_probe): Likewise.
20173 (ix86_emit_probe_stack_range): Pass new arguments to
20174 release_scratch_register_on_entry.
20176 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
20178 PR rtl-optimization/84157
20179 * combine.c (change_zero_ext): Use REG_P predicate in
20180 front of HARD_REGISTER_P predicate.
20182 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
20184 * config/avr/avr.c (avr_option_override): Move disabling of
20185 -fdelete-null-pointer-checks to...
20186 * common/config/avr/avr-common.c (avr_option_optimization_table):
20189 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20191 PR tree-optimization/81635
20192 * tree-data-ref.c (split_constant_offset_1): For types that
20193 wrap on overflow, try to use range info to prove that wrapping
20196 2018-02-01 Renlin Li <renlin.li@arm.com>
20199 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
20200 TAILCALL_ADDR_REGS.
20201 (aarch64_register_move_cost): Likewise.
20202 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
20203 TAILCALL_ADDR_REGS.
20204 (REG_CLASS_NAMES): Likewise.
20205 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
20206 TAILCALL_ADDR_REGS. Remove IP registers.
20207 * config/aarch64/aarch64.md (Ucs): Update register constraint.
20209 2018-02-01 Richard Biener <rguenther@suse.de>
20211 * domwalk.h (dom_walker::dom_walker): Add additional constructor
20212 for specifying RPO order and allow NULL for that.
20213 * domwalk.c (dom_walker::dom_walker): Likewise.
20214 (dom_walker::walk): Handle NULL RPO order.
20215 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
20217 (rewrite_update_dom_walker): Likewise.
20218 (mark_def_dom_walker): Likewise.
20220 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20222 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
20223 (aarch64_maybe_expand_sve_subreg_move): Declare.
20224 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
20225 * config/aarch64/predicates.md (aarch64_any_register_operand): New
20227 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
20228 that are semantically a reverse operation.
20229 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
20230 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
20231 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
20233 (aarch64_can_change_mode_class): For big-endian, forbid changes
20234 between two SVE modes if they have different element sizes.
20236 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20238 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
20239 the TImode handling for big-endian targets.
20241 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20243 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
20244 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
20246 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
20247 Remove BSWAP handing for big-endian targets and use the form of
20248 LD1RQ appropariate for the mode.
20250 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20252 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
20253 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
20254 duplicated element.
20256 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20259 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
20260 check for operands that need to go through aarch64_sve_reload_be.
20262 2018-02-01 Jakub Jelinek <jakub@redhat.com>
20264 PR tree-optimization/81661
20265 PR tree-optimization/84117
20266 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
20267 * tree-eh.c: Include gimplify.h.
20268 (find_trapping_overflow, replace_trapping_overflow,
20269 rewrite_to_non_trapping_overflow): New functions.
20270 * tree-vect-loop.c: Include tree-eh.h.
20271 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
20272 * tree-data-ref.c: Include tree-eh.h.
20273 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
20275 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
20277 PR rtl-optimization/84123
20278 * combine.c (change_zero_ext): Check if hard register satisfies
20279 can_change_dest_mode before calling gen_lowpart_SUBREG.
20281 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
20284 * ira.c (ira_init_register_move_cost): Remove assert.
20286 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
20288 PR rtl-optimization/84071
20289 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
20290 * doc/tm.texi: Regenerate.
20292 2018-01-31 Richard Biener <rguenther@suse.de>
20294 PR tree-optimization/84132
20295 * tree-data-ref.c (analyze_miv_subscript): Properly
20296 check whether evolution_function_is_affine_multivariate_p
20297 before calling gcd_of_steps_may_divide_p.
20299 2018-01-31 Julia Koval <julia.koval@intel.com>
20302 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
20303 * config/i386/i386.md (rdpid_rex64) New.
20304 (rdpid): Make 32bit only.
20306 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
20309 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
20310 an IDENTIFIER_NODE for FUNCTION_TYPE's.
20312 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
20315 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20317 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
20319 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
20321 PR rtl-optimization/84071
20322 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
20323 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
20325 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
20327 * config/arc/arc.c (arc_handle_aux_attribute): New function.
20328 (arc_attribute_table): Add 'aux' attribute.
20329 (arc_in_small_data_p): Consider aux like variables.
20330 (arc_is_aux_reg_p): New function.
20331 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
20332 (arc_get_aux_arg): New function.
20333 (prepare_move_operands): Handle aux-register access.
20334 (arc_handle_aux_attribute): New function.
20335 * doc/extend.texi (ARC Variable attributes): Add subsection.
20337 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
20339 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
20340 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
20341 (arc_attribute_table): Add 'uncached' attribute.
20342 (arc_print_operand): Print '.di' flag for uncached memory
20344 (arc_in_small_data_p): Do not consider for small data the uncached
20346 (arc_is_uncached_mem_p): New function.
20347 * config/arc/predicates.md (compact_store_memory_operand): Check
20348 for uncached memory accesses.
20349 (nonvol_nonimm_operand): Likewise.
20350 * doc/extend.texi (ARC Type Attribute): New subsection.
20352 2018-01-31 Jakub Jelinek <jakub@redhat.com>
20355 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
20356 falign-loops=): Add Optimization flag.
20358 2018-01-30 Jeff Law <law@redhat.com>
20361 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
20362 INT_REGISTERS_SAVED. Check it prior to calling
20363 get_scratch_register_on_entry.
20364 (ix86_adjust_stack_and_probe): Similarly.
20365 (ix86_emit_probe_stack_range): Similarly.
20366 (ix86_expand_prologue): Corresponding changes.
20368 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20371 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
20372 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
20374 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
20377 * lra-constraints.c (curr_insn_transform): Process AND in the
20380 2018-01-30 Jakub Jelinek <jakub@redhat.com>
20382 PR rtl-optimization/83986
20383 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
20384 dependence against last_pending_memory_flush in addition to
20385 pending_jump_insns.
20387 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
20389 PR tree-optimization/81611
20390 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
20393 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20396 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
20399 2018-01-30 Richard Biener <rguenther@suse.de>
20400 Jakub Jelinek <jakub@redhat.com>
20402 PR tree-optimization/84111
20403 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
20404 inner loops added during recursion, as they don't have up-to-date
20407 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
20410 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
20411 (can_inline_edge_by_limits_p): ... here.
20412 (can_early_inline_edge_p, check_callers,
20413 update_caller_keys, update_callee_keys, recursive_inlining,
20414 add_new_edges_to_heap, speculation_useful_p,
20415 inline_small_functions,
20416 inline_small_functions, flatten_function,
20417 inline_to_all_callers_1): Update.
20419 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
20421 * profile-count.c (profile_count::combine_with_ipa_count): Handle
20424 2018-01-30 Richard Biener <rguenther@suse.de>
20426 PR tree-optimization/83008
20427 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
20428 invariant and constant vector uses in stmts when they need
20429 more than one stmt.
20431 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20434 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
20435 * configure: Regenerate.
20437 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20439 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
20441 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
20442 Use gen_rtx_REG rather than gen_lowpart.
20444 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20446 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
20447 rather than 0 when creating partial subregs.
20449 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20451 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
20454 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
20457 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
20458 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
20459 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
20460 flags. This restores the settings used before the 2017-07-24.
20461 Turning off pre increment/decrement/modify allows IVOPTS to
20462 optimize DF/SF loops where the index is an int.
20464 2018-01-29 Richard Biener <rguenther@suse.de>
20465 Kelvin Nilsen <kelvin@gcc.gnu.org>
20468 * tree-vect-stmts.c (vectorizable_call): Don't call
20469 targetm.vectorize_builtin_md_vectorized_function if callee is
20472 2018-01-22 Carl Love <cel@us.ibm.com>
20474 * doc/extend.tex: Fix typo in second arg in
20475 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
20477 2018-01-29 Richard Biener <rguenther@suse.de>
20479 PR tree-optimization/84086
20480 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
20481 (flush_ssaname_freelist): When SSA names were released reset
20482 the SCEV hash table.
20484 2018-01-29 Richard Biener <rguenther@suse.de>
20486 PR tree-optimization/84057
20487 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
20488 removed paths when removing edges.
20490 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
20492 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
20493 -mfunction-return=@var{choice}.
20495 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
20497 PR diagnostic/84034
20498 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
20499 Handle CR like TAB.
20500 (layout::print_source_line): Likewise.
20501 (test_get_line_width_without_trailing_whitespace): Add test cases.
20503 2018-01-27 Jakub Jelinek <jakub@redhat.com>
20505 PR middle-end/84040
20506 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
20509 2018-01-26 Jim Wilson <jimw@sifive.com>
20511 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
20513 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
20516 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20518 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
20519 and CMP + SUB-immediate -> SUBS.
20521 2018-01-26 Martin Sebor <msebor@redhat.com>
20523 PR tree-optimization/83896
20524 * tree-ssa-strlen.c (get_string_len): Rename...
20525 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
20526 Avoid assuming length is constant.
20527 (handle_char_store): Use HOST_WIDE_INT for string length.
20529 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
20532 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
20533 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
20535 2018-01-26 Richard Biener <rguenther@suse.de>
20537 PR rtl-optimization/84003
20538 * dse.c (record_store): Only record redundant stores when
20539 the earlier store aliases at least all accesses the later one does.
20541 2018-01-26 Jakub Jelinek <jakub@redhat.com>
20543 PR rtl-optimization/83985
20544 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
20545 REG_CFA_RESTORE insns.
20546 (delete_unmarked_insns): Don't ignore separate shrink wrapping
20547 REG_CFA_RESTORE insns here.
20550 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
20551 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
20553 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20555 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
20556 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
20557 (arc_init): Likewise.
20558 (arc_override_options): Likewise.
20559 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
20561 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
20563 * config/arc/arc.h (TARGET_DBNZ): Define.
20564 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
20565 properly set the tune attribute.
20566 (dbnz): Use TARGET_DBNZ guard.
20567 * config/arc/arc.opt (mtune): Add core3 option.
20569 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20571 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
20572 recognize new pic like addresses.
20573 (arc_delegitimize_address): Clean up.
20575 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20577 * config/arc/arc-arches.def: Option mrf16 valid for all
20579 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
20580 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
20581 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
20582 * config/arc/arc-tables.opt: Regenerate.
20583 * config/arc/arc.c (arc_conditional_register_usage): Handle
20584 reduced register file case.
20585 (arc_file_start): Set must have build attributes.
20586 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
20587 mrf16 option value.
20588 * config/arc/arc.opt (mrf16): Add new option.
20589 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
20590 * config/arc/genmultilib.awk: Handle new mrf16 option.
20591 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
20592 * config/arc/t-multilib: Regenerate.
20593 * doc/invoke.texi (ARC Options): Document mrf16 option.
20595 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20597 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
20598 * config/arc/arc.c (arc_handle_secure_attribute): New function.
20599 (arc_attribute_table): Add 'secure_call' attribute.
20600 (arc_print_operand): Print secure call operand.
20601 (arc_function_ok_for_sibcall): Don't optimize tail calls when
20603 (arc_is_secure_call_p): New function. * config/arc/arc.md
20604 (call_i): Add support for sjli instruction.
20605 (call_value_i): Likewise.
20606 * config/arc/constraints.md (Csc): New constraint.
20608 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20609 John Eric Martin <John.Martin@emmicro-us.com>
20611 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
20612 * config/arc/arc.c (_arc_jli_section): New struct.
20613 (arc_jli_section): New type.
20614 (rc_jli_sections): New static variable.
20615 (arc_handle_jli_attribute): New function.
20616 (arc_attribute_table): Add jli_always and jli_fixed attribute.
20617 (arc_file_end): New function.
20618 (TARGET_ASM_FILE_END): Define.
20619 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
20620 (arc_add_jli_section): New function.
20621 (jli_call_scan): Likewise.
20622 (arc_reorg): Call jli_call_scan.
20623 (arc_output_addsi): Remove 'S' from printing asm operand.
20624 (arc_is_jli_call_p): New function.
20625 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
20627 (movhi_insn): Likewise.
20628 (movsi_insn): Likewise.
20629 (movsi_set_cc_insn): Likewise.
20630 (loadqi_update): Likewise.
20631 (load_zeroextendqisi_update): Likewise.
20632 (load_signextendqisi_update): Likewise.
20633 (loadhi_update): Likewise.
20634 (load_zeroextendhisi_update): Likewise.
20635 (load_signextendhisi_update): Likewise.
20636 (loadsi_update): Likewise.
20637 (loadsf_update): Likewise.
20638 (movsicc_insn): Likewise.
20639 (bset_insn): Likewise.
20640 (bxor_insn): Likewise.
20641 (bclr_insn): Likewise.
20642 (bmsk_insn): Likewise.
20643 (bicsi3_insn): Likewise.
20644 (cmpsi_cc_c_insn): Likewise.
20645 (movsi_ne): Likewise.
20646 (movsi_cond_exec): Likewise.
20647 (clrsbsi2): Likewise.
20648 (norm_f): Likewise.
20658 (call_i): Remove 'S' asm letter, add jli instruction.
20659 (call_value_i): Likewise.
20660 * config/arc/arc.op (mjli-always): New option.
20661 * config/arc/constraints.md (Cji): New constraint.
20662 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
20664 (subsf3_fpx): Likewise.
20665 (mulsf3_fpx): Likewise.
20666 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
20668 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
20669 function attrbutes.
20670 * doc/invoke.texi (ARC): Document mjli-always option.
20672 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
20674 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
20675 avoid addition with 0 and use incw and decw where possible.
20677 2018-01-26 Richard Biener <rguenther@suse.de>
20679 PR tree-optimization/81082
20680 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
20681 association if it requires casting to unsigned.
20682 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
20683 from fold_plusminus_mult_expr to catch important cases late when
20684 range info is available.
20686 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20688 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
20689 * configure.ac (hidden_linkonce): New test.
20690 * configure: Regenerate.
20691 * config.in: Regenerate.
20693 2018-01-26 Julia Koval <julia.koval@intel.com>
20695 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
20696 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
20697 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
20698 _mm_mask_bitshuffle_epi64_mask): Fix type.
20699 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
20700 USI_FTYPE_V4DI_V4DI_USI): Remove.
20701 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
20702 __builtin_ia32_vpshufbitqmb256_mask,
20703 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
20704 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
20705 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
20707 2018-01-26 Alan Modra <amodra@gmail.com>
20710 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
20711 UNSPEC_VBPERMQ. Sort other unspecs.
20713 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
20715 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
20717 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
20719 PR middle-end/83055
20720 * predict.c (drop_profile): Do not push/pop cfun; update also
20722 (handle_missing_profiles): Fix logic looking for zero profiles.
20724 2018-01-25 Jakub Jelinek <jakub@redhat.com>
20726 PR middle-end/83977
20727 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
20728 on functions with #pragma omp declare simd or functions with simd
20730 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
20731 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
20732 Remove trailing \n from warning_at calls.
20734 2018-01-25 Tom de Vries <tom@codesourcery.com>
20737 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20738 for neutered workers.
20740 2018-01-24 Joseph Myers <joseph@codesourcery.com>
20743 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
20744 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
20746 2018-01-24 Jeff Law <law@redhat.com>
20749 * i386.c (get_probe_interval): Move to earlier point.
20750 (ix86_compute_frame_layout): If -fstack-clash-protection and
20751 the frame is larger than the probe interval, then use pushes
20752 to save registers rather than reg->mem moves.
20753 (ix86_expand_prologue): Remove conditional for int_registers_saved
20756 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
20759 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
20760 min/max for never referenced object.
20762 2018-01-24 Jakub Jelinek <jakub@redhat.com>
20764 PR middle-end/83977
20765 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
20767 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
20768 attributes from DECL_ATTRIBUTES (decl) without affecting
20769 DECL_ATTRIBUTES (current_function_decl).
20770 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
20771 functions with non-NULL DECL_ABSTRACT_ORIGIN.
20773 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
20775 PR tree-optimization/83979
20776 * fold-const.c (fold_comparison): Use constant_boolean_node
20777 instead of boolean_{true,false}_node.
20779 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
20781 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
20784 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20786 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20787 Simplify the clause that sets the length attribute.
20788 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20789 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
20790 clause that sets the length attribute.
20791 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20793 2018-01-24 Tom de Vries <tom@codesourcery.com>
20796 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
20797 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
20798 Add strict parameter.
20799 (prevent_branch_around_nothing): Insert dummy insn between branch to
20800 label and label with no ptx insn inbetween.
20801 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
20803 2018-01-24 Tom de Vries <tom@codesourcery.com>
20806 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20807 for neutered threads in warp.
20808 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
20810 2018-01-24 Richard Biener <rguenther@suse.de>
20812 PR tree-optimization/83176
20813 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
20816 2018-01-24 Richard Biener <rguenther@suse.de>
20818 PR tree-optimization/82819
20819 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
20820 code generating pluses that are no-ops in the target precision.
20822 2018-01-24 Richard Biener <rguenther@suse.de>
20824 PR middle-end/84000
20825 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
20827 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20829 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
20830 to merge probabilities.
20831 * predict.c (probably_never_executed): Also mark as cold functions
20832 with global 0 profile and guessed local profile.
20833 * profile-count.c (profile_probability::combine_with_count): New
20835 * profile-count.h (profile_probability::operator*,
20836 profile_probability::operator*=, profile_probability::operator/,
20837 profile_probability::operator/=): Reduce precision to adjusted
20838 and set value to guessed on contradictory divisions.
20839 (profile_probability::combine_with_freq): Remove.
20840 (profile_probability::combine_wiht_count): Declare.
20841 (profile_count::force_nonzero):: Set to adjusted.
20842 (profile_count::probability_in):: Set quality to adjusted.
20843 * tree-ssa-tail-merge.c (replace_block_by): Use
20844 combine_with_count.
20846 2018-01-23 Andrew Waterman <andrew@sifive.com>
20847 Jim Wilson <jimw@sifive.com>
20849 * config/riscv/riscv.c (riscv_stack_boundary): New.
20850 (riscv_option_override): Set riscv_stack_boundary. Handle
20851 riscv_preferred_stack_boundary_arg.
20852 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
20853 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
20854 (STACK_BOUNDARY): Set to riscv_stack_boundary.
20855 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
20856 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
20857 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
20859 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
20862 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
20863 of struct ix86_frame.
20864 (ix86_expand_epilogue): Likewise. Add a local variable for
20865 the reg_save_offset field in struct ix86_frame.
20867 2018-01-23 Bin Cheng <bin.cheng@arm.com>
20869 PR tree-optimization/82604
20870 * tree-loop-distribution.c (enum partition_kind): New enum item
20871 PKIND_PARTIAL_MEMSET.
20872 (partition_builtin_p): Support above new enum item.
20873 (generate_code_for_partition): Ditto.
20874 (compute_access_range): Differentiate cases that equality can be
20875 proven at all loops, the innermost loops or no loops.
20876 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
20877 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
20878 (finalize_partitions, distribute_loop): Don't fuse partition of
20879 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
20880 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
20881 parloop is enabled.
20883 2018-01-23 Martin Liska <mliska@suse.cz>
20885 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
20886 order to ignore the predictor.
20887 (PRED_POLYMORPHIC_CALL): Likewise.
20888 (PRED_RECURSIVE_CALL): Likewise.
20890 2018-01-23 Martin Liska <mliska@suse.cz>
20892 * tree-profile.c (tree_profiling): Print function header to
20893 aware reader which function we are working on.
20894 * value-prof.c (gimple_find_values_to_profile): Do not print
20895 not interesting value histograms.
20897 2018-01-23 Martin Liska <mliska@suse.cz>
20899 * profile-count.h (enum profile_quality): Add
20900 profile_uninitialized as the first value. Do not number values
20901 as they are zero based.
20902 (profile_count::verify): Update sanity check.
20903 (profile_probability::verify): Likewise.
20905 2018-01-23 Nathan Sidwell <nathan@acm.org>
20907 * doc/invoke.texi (ffor-scope): Deprecate.
20909 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20911 PR tree-optimization/83510
20912 * domwalk.c (set_all_edges_as_executable): New function.
20913 (dom_walker::dom_walker): Convert bool param
20914 "skip_unreachable_blocks" to enum reachability. Move setup of
20915 edge flags to set_all_edges_as_executable and only do it when
20916 reachability is REACHABLE_BLOCKS.
20917 * domwalk.h (enum dom_walker::reachability): New enum.
20918 (dom_walker::dom_walker): Convert bool param
20919 "skip_unreachable_blocks" to enum reachability.
20920 (set_all_edges_as_executable): New decl.
20921 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
20922 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
20924 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
20925 but converting true to REACHABLE_BLOCKS.
20926 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
20928 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
20929 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
20930 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
20932 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
20933 if check_all_array_refs will be called.
20935 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20937 * tree.c (selftest::test_location_wrappers): Add more test
20940 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20942 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
20943 (selftest::test_bit_in_range): Likewise.
20945 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
20948 * doc/sourcebuild.texi (vect_float): Say that the selector
20949 only describes the situation when -funsafe-math-optimizations is on.
20950 (vect_float_strict): Document.
20952 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
20954 PR tree-optimization/83965
20955 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
20956 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
20957 instead of checking only for a reduction.
20958 (vect_recog_widen_sum_pattern): Likewise.
20960 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20962 * predict.c (probably_never_executed): Only use precise profile info.
20963 (compute_function_frequency): Skip after inlining hack since we now
20964 have quality checking.
20966 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20968 * profile-count.h (profile_probability::very_unlikely,
20969 profile_probability::unlikely, profile_probability::even): Set
20970 precision to guessed.
20972 2018-01-23 Richard Biener <rguenther@suse.de>
20974 PR tree-optimization/83963
20975 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
20976 Properly terminate dominator walk when crossing the exit edge not
20977 when visiting its source block.
20979 2018-01-23 Jakub Jelinek <jakub@redhat.com>
20982 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
20983 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
20985 2018-01-22 Jakub Jelinek <jakub@redhat.com>
20987 PR tree-optimization/83957
20988 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
20989 semicolon after for body surrounded by braces.
20991 PR tree-optimization/83081
20992 * profile-count.h (profile_probability::split): New method.
20993 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
20994 Use profile_probability::split.
20995 (do_compare_rtx_and_jump): Fix adjustment of probabilities
20996 when splitting a single conditional jump into 2.
20998 2018-01-22 David Malcolm <dmalcolm@redhat.com>
21000 PR tree-optimization/69452
21001 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
21004 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21006 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
21007 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
21008 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
21010 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21012 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
21014 * config/rl78/rl78.md (movdi): New define_expand.
21015 * config/rl78/rl78.c (rl78_split_movdi): New function.
21017 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
21020 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
21022 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
21023 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
21024 128-bit to produce an UNSPEC move to get the double word with the
21025 signbit and then a shift directly to do signbit.
21026 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
21027 implementation with a new version that just does either a direct
21028 move or a regular move. Move memory interface to separate insns.
21029 Move insns so they are next to the expander.
21030 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
21031 with signbit move. Split big and little endian case.
21032 (signbit<mode>2_dm_mem_le): Likewise.
21033 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
21034 (signbit<mode>2_dm2): Likewise.
21036 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21038 * config/rl78/rl78.md (anddi3): New define_expand.
21040 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21042 * config/rl78/rl78.md (umindi3): New define_expand.
21044 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21046 * config/rl78/rl78.md (smindi3): New define_expand.
21048 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21050 * config/rl78/rl78.md (smaxdi3): New define_expand.
21052 2018-01-22 Carl Love <cel@us.ibm.com>
21054 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
21055 LVX_V1TI): Add macro expansion.
21056 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
21057 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
21058 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
21059 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
21060 Change check to determine if the instruction is a byte reversing
21061 entry. Fix typo in comment.
21062 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
21063 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
21064 Add def_builtin calls for new builtins.
21065 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
21066 Add define_insn expansion.
21068 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21070 * config/rl78/rl78.md (umaxdi3): New define_expand.
21072 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21074 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
21075 for non-QImode registers.
21077 2018-01-22 Richard Biener <rguenther@suse.de>
21079 PR tree-optimization/83963
21080 * graphite-scop-detection.c (scop_detection::get_sese): Delay
21081 including the loop exit block.
21082 (scop_detection::merge_sese): Likewise.
21083 (scop_detection::add_scop): Do it here instead.
21085 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21087 * doc/sourcebuild.texi (arm_softfloat): Document.
21089 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
21092 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
21093 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
21094 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
21096 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21097 David Edelsohn <dje.gcc@gmail.com>
21100 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21101 Change "crset eq" to "crset 2".
21102 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21103 (*call_indirect_aix<mode>_nospec): Likewise.
21104 (*call_value_indirect_aix<mode>_nospec): Likewise.
21105 (*call_indirect_elfv2<mode>_nospec): Likewise.
21106 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
21107 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
21108 change assembly output from . to $.
21109 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21110 (indirect_jump<mode>_nospec): Change assembly output from . to $.
21111 (*tablejump<mode>_internal1_nospec): Likewise.
21113 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
21116 * config/sh/sh_optimize_sett_clrt.cc:
21117 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
21119 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
21121 PR tree-optimization/83940
21122 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
21123 offset_dt to vect_constant_def rather than vect_unknown_def_type.
21124 (vect_check_load_store_mask): Add a mask_dt_out parameter and
21125 use it to pass back the definition type.
21126 (vect_check_store_rhs): Likewise rhs_dt_out.
21127 (vect_build_gather_load_calls): Add a mask_dt argument and use
21128 it instead of a call to vect_is_simple_use.
21129 (vectorizable_store): Update calls to vect_check_load_store_mask
21130 and vect_check_store_rhs. Use the dt returned by the latter instead
21131 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
21132 instead of calls to vect_is_simple_use. Pass the scalar rather
21133 than the vector operand to vect_is_simple_use when handling
21134 second and subsequent copies of an rhs value.
21135 (vectorizable_load): Update calls to vect_check_load_store_mask
21136 and vect_build_gather_load_calls. Use the cached mask_dt and
21137 gs_info.offset_dt instead of calls to vect_is_simple_use.
21139 2018-01-20 Jakub Jelinek <jakub@redhat.com>
21141 PR middle-end/83945
21142 * tree-emutls.c: Include gimplify.h.
21143 (lower_emutls_2): New function.
21144 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
21145 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
21146 it before further processing.
21149 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
21150 UINTVAL (trueop1) instead of INTVAL (op1).
21152 2018-01-19 Jakub Jelinek <jakub@redhat.com>
21156 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
21157 INCOMING_FRAME_SP_OFFSET if not defined.
21158 (scan_trace): Add ENTRY argument. If true and
21159 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
21160 emit a note to adjust the CFA offset.
21161 (create_cfi_notes): Adjust scan_trace callers.
21162 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
21163 INCOMING_FRAME_SP_OFFSET in the CIE.
21164 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
21165 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
21167 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
21168 * doc/tm.texi: Regenerated.
21170 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
21172 PR rtl-optimization/83147
21173 * lra-constraints.c (remove_inheritance_pseudos): Use
21174 lra_substitute_pseudo_within_insn.
21176 2018-01-19 Tom de Vries <tom@codesourcery.com>
21177 Cesar Philippidis <cesar@codesourcery.com>
21180 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
21182 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
21185 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
21186 spaces for function labels.
21188 2018-01-19 Martin Liska <mliska@suse.cz>
21190 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
21191 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
21192 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
21193 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
21194 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
21195 (PRED_CONST_RETURN): Change from 69 to 65.
21196 (PRED_NULL_RETURN): Change from 91 to 71.
21197 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
21198 (PRED_LOOP_GUARD): Change from 66 to 73.
21200 2018-01-19 Martin Liska <mliska@suse.cz>
21202 * predict.c (predict_insn_def): Add new assert.
21203 (struct branch_predictor): Change type to signed integer.
21204 (test_prediction_value_range): Amend test to cover
21205 PROB_UNINITIALIZED.
21206 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
21207 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
21208 (PRED_LOOP_ITERATIONS_MAX): Likewise.
21209 (PRED_LOOP_IV_COMPARE): Likewise.
21210 * predict.h (PROB_UNINITIALIZED): Define new constant.
21212 2018-01-19 Martin Liska <mliska@suse.cz>
21214 * predict.c (dump_prediction): Add new format for
21215 analyze_brprob.py script which is enabled with -details
21217 * profile-count.h (precise_p): New function.
21219 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
21221 PR tree-optimization/83922
21222 * tree-vect-loop.c (vect_verify_full_masking): Return false if
21223 there are no statements that need masking.
21224 (vect_active_double_reduction_p): New function.
21225 (vect_analyze_loop_operations): Use it when handling phis that
21226 are not in the loop header.
21228 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
21230 PR tree-optimization/83914
21231 * tree-vect-loop.c (vectorizable_induction): Don't convert
21232 init_expr or apply the peeling adjustment for inductions
21233 that are nested within the vectorized loop.
21235 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21237 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
21240 2018-01-18 Jakub Jelinek <jakub@redhat.com>
21244 * function.h (gimplify_parameters): Add gimple_seq * argument.
21245 * function.c: Include gimple.h and options.h.
21246 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
21247 for the added local temporaries if needed.
21248 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
21249 if there are any parameter cleanups, wrap whole body into a
21250 try/finally with the cleanups.
21252 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
21255 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
21256 Use GET_MODE_CLASS for scalar floating point.
21258 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
21262 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
21263 Fix call of call_cgraph_insertion_hooks.
21265 2018-01-18 Martin Sebor <msebor@redhat.com>
21267 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
21269 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
21272 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
21275 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
21278 * common.opt: (-ffile-prefix-map): New option.
21279 * opts.c (common_handle_option): Defer it.
21280 * opts-global.c (handle_common_deferred_options): Handle it.
21281 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
21282 * file-prefix-map.h: New file.
21283 (remap_debug_filename, add_debug_prefix_map): ...here.
21284 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
21285 * final.c (debug_prefix_map, add_debug_prefix_map
21286 remap_debug_filename): Move to...
21287 * file-prefix-map.c: New file.
21288 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
21289 generalize, get rid of alloca(), use strrchr() instead of strchr().
21290 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
21291 Implement in terms of add_prefix_map().
21292 (remap_macro_filename, remap_debug_filename): Implement in term of
21294 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
21295 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
21296 * dbxout.c: Include file-prefix-map.h.
21297 * varasm.c: Likewise.
21298 * vmsdbgout.c: Likewise.
21299 * xcoffout.c: Likewise.
21300 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
21301 * doc/cppopts.texi (-fmacro-prefix-map): Document.
21302 * doc/invoke.texi (-ffile-prefix-map): Document.
21303 (-fdebug-prefix-map): Update description.
21305 2018-01-18 Martin Liska <mliska@suse.cz>
21307 * config/i386/i386.c (indirect_thunk_name): Document that also
21309 (output_indirect_thunk): Document why both instructions
21310 (pause and lfence) are generated.
21312 2018-01-18 Richard Biener <rguenther@suse.de>
21314 PR tree-optimization/83887
21315 * graphite-scop-detection.c
21316 (scop_detection::get_nearest_dom_with_single_entry): Remove.
21317 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
21318 (scop_detection::merge_sese): Re-implement with a flood-fill
21319 algorithm that properly finds a SESE region if it exists.
21321 2018-01-18 Jakub Jelinek <jakub@redhat.com>
21324 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
21325 pointer_diff optimizations use view_convert instead of convert.
21327 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21329 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21330 Generate different code for -mno-speculate-indirect-jumps.
21331 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21332 (*call_indirect_aix<mode>): Disable for
21333 -mno-speculate-indirect-jumps.
21334 (*call_indirect_aix<mode>_nospec): New define_insn.
21335 (*call_value_indirect_aix<mode>): Disable for
21336 -mno-speculate-indirect-jumps.
21337 (*call_value_indirect_aix<mode>_nospec): New define_insn.
21338 (*sibcall_nonlocal_sysv<mode>): Generate different code for
21339 -mno-speculate-indirect-jumps.
21340 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21342 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
21344 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
21345 long double type, set the flags for noting the default long double
21346 type, even if we don't pass or return a long double type.
21348 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
21351 * ipa-inline.c (flatten_function): Do not overwrite final inlining
21354 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
21356 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
21357 support for merge[hl].
21358 (fold_mergehl_helper): New helper function.
21359 (tree-vector-builder.h): New #include for tree_vector_builder usage.
21360 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
21361 (altivec_vmrglw_direct): Add xxmrglw insn.
21363 2018-01-17 Andrew Waterman <andrew@sifive.com>
21365 * config/riscv/riscv.c (riscv_conditional_register_usage): If
21366 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
21368 2018-01-17 David Malcolm <dmalcolm@redhat.com>
21371 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
21372 call the lto_location_cache before reading the
21373 DECL_SOURCE_LOCATION of the types.
21375 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
21376 Richard Sandiford <richard.sandiford@linaro.org>
21378 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
21379 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
21380 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
21381 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
21382 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
21384 * config/aarch64/constraints.md (aarch64_movti_operand):
21386 * config/aarch64/predicates.md (Uti): Add new constraint.
21388 2018-01-17 Carl Love <cel@us.ibm.com>
21390 * config/rs6000/vsx.md (define_expand xl_len_r,
21391 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
21392 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
21394 (define_expand, define_insn): Move the shift left from the
21395 define_insn to the define_expand for lxvl and stxvl instructions.
21396 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
21397 and XL_LEN_R definitions to PURE.
21399 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
21401 * config/i386/i386.c (indirect_thunk_name): Declare regno
21402 as unsigned int. Compare regno with INVALID_REGNUM.
21403 (output_indirect_thunk): Ditto.
21404 (output_indirect_thunk_function): Ditto.
21405 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
21406 in the call to output_indirect_thunk_function.
21408 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
21410 PR middle-end/83884
21411 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
21412 rather than the size of inner_type to determine the stack slot size
21413 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
21415 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
21418 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
21421 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
21423 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
21424 endian Linux systems to optionally enable multilibs for selecting
21425 the long double type if the user configured an explicit type.
21426 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
21427 have no long double multilibs if not defined.
21428 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
21429 warn if the user used -mabi={ieee,ibm}longdouble and we built
21430 multilibs for long double.
21431 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
21432 appropriate multilib option.
21433 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
21435 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
21436 for building long double multilibs.
21437 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
21439 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
21441 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
21444 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
21446 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
21449 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
21452 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
21455 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
21457 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
21458 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
21460 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
21462 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
21463 different rtl trees depending on TARGET_64BIT.
21464 (rs6000_gen_lvx): Likewise.
21466 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
21468 * config/visium/visium.md (nop): Tweak comment.
21469 (hazard_nop): Likewise.
21471 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21473 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
21474 -mspeculate-indirect-jumps.
21475 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
21476 for -mno-speculate-indirect-jumps.
21477 (*call_indirect_elfv2<mode>_nospec): New define_insn.
21478 (*call_value_indirect_elfv2<mode>): Disable for
21479 -mno-speculate-indirect-jumps.
21480 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
21481 (indirect_jump): Emit different RTL for
21482 -mno-speculate-indirect-jumps.
21483 (*indirect_jump<mode>): Disable for
21484 -mno-speculate-indirect-jumps.
21485 (*indirect_jump<mode>_nospec): New define_insn.
21486 (tablejump): Emit different RTL for
21487 -mno-speculate-indirect-jumps.
21488 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
21489 (tablejumpsi_nospec): New define_expand.
21490 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
21491 (tablejumpdi_nospec): New define_expand.
21492 (*tablejump<mode>_internal1): Disable for
21493 -mno-speculate-indirect-jumps.
21494 (*tablejump<mode>_internal1_nospec): New define_insn.
21495 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
21498 2018-01-16 Artyom Skrobov tyomitch@gmail.com
21500 * caller-save.c (insert_save): Drop unnecessary parameter. All
21503 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21504 Richard Biener <rguenth@suse.de>
21507 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
21508 return early, inline manually is_gimple_sizepos. Make sure if we
21509 call gimplify_expr we don't end up with a gimple constant.
21510 * tree.c (variably_modified_type_p): Don't return true for
21511 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
21512 * gimplify.h (is_gimple_sizepos): Remove.
21514 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21516 PR tree-optimization/83857
21517 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
21518 vectorizable_live_operation for pure SLP statements.
21519 (vectorizable_live_operation): Handle PHIs.
21521 2018-01-16 Richard Biener <rguenther@suse.de>
21523 PR tree-optimization/83867
21524 * tree-vect-stmts.c (vect_transform_stmt): Precompute
21525 nested_in_vect_loop_p since the scalar stmt may get invalidated.
21527 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21530 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
21531 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
21532 If off is not INTEGER_CST, issue a may not be aligned warning
21533 rather than isn't aligned. Use isn%'t rather than isn't.
21534 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
21536 <case MULT_EXPR>: Improve the case when bottom and one of the
21537 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
21538 operand, in that case check if the other operand is multiple of
21539 bottom divided by the INTEGER_CST operand.
21541 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21544 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
21545 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
21546 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
21547 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
21548 * config/pa/pa.c (pa_function_arg_advance): Likewise.
21549 (pa_function_arg, pa_arg_partial_bytes): Likewise.
21550 (pa_function_arg_size): New function.
21552 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21554 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
21555 in a separate statement.
21557 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21559 PR tree-optimization/83847
21560 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
21561 group gathers and scatters.
21563 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21565 PR rtl-optimization/86620
21566 * params.def (max-sched-ready-insns): Bump minimum value to 1.
21568 PR rtl-optimization/83213
21569 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
21570 to last if both are JUMP_INSNs.
21572 PR tree-optimization/83843
21573 * gimple-ssa-store-merging.c
21574 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
21575 store_immediate_info for bswap/nop orig_stores.
21577 2018-01-15 Andrew Waterman <andrew@sifive.com>
21579 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
21581 <UDIV>: Increase cost if !TARGET_DIV.
21583 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
21585 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
21586 (define_attr "cr_logical_3op"): New.
21587 (cceq_ior_compare): Adjust.
21588 (cceq_ior_compare_complement): Adjust.
21589 (*cceq_rev_compare): Adjust.
21590 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
21591 (is_cracked_insn): Adjust.
21592 (insn_must_be_first_in_group): Adjust.
21593 * config/rs6000/40x.md: Adjust.
21594 * config/rs6000/440.md: Adjust.
21595 * config/rs6000/476.md: Adjust.
21596 * config/rs6000/601.md: Adjust.
21597 * config/rs6000/603.md: Adjust.
21598 * config/rs6000/6xx.md: Adjust.
21599 * config/rs6000/7450.md: Adjust.
21600 * config/rs6000/7xx.md: Adjust.
21601 * config/rs6000/8540.md: Adjust.
21602 * config/rs6000/cell.md: Adjust.
21603 * config/rs6000/e300c2c3.md: Adjust.
21604 * config/rs6000/e500mc.md: Adjust.
21605 * config/rs6000/e500mc64.md: Adjust.
21606 * config/rs6000/e5500.md: Adjust.
21607 * config/rs6000/e6500.md: Adjust.
21608 * config/rs6000/mpc.md: Adjust.
21609 * config/rs6000/power4.md: Adjust.
21610 * config/rs6000/power5.md: Adjust.
21611 * config/rs6000/power6.md: Adjust.
21612 * config/rs6000/power7.md: Adjust.
21613 * config/rs6000/power8.md: Adjust.
21614 * config/rs6000/power9.md: Adjust.
21615 * config/rs6000/rs64.md: Adjust.
21616 * config/rs6000/titan.md: Adjust.
21618 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21620 * config/i386/predicates.md (indirect_branch_operand): Rewrite
21621 ix86_indirect_branch_register logic.
21623 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21625 * config/i386/constraints.md (Bs): Update
21626 ix86_indirect_branch_register check. Don't check
21627 ix86_indirect_branch_register with GOT_memory_operand.
21629 * config/i386/predicates.md (GOT_memory_operand): Don't check
21630 ix86_indirect_branch_register here.
21631 (GOT32_symbol_operand): Likewise.
21633 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21635 * config/i386/predicates.md (constant_call_address_operand):
21636 Rewrite ix86_indirect_branch_register logic.
21637 (sibcall_insn_operand): Likewise.
21639 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21641 * config/i386/constraints.md (Bs): Replace
21642 ix86_indirect_branch_thunk_register with
21643 ix86_indirect_branch_register.
21645 * config/i386/i386.md (indirect_jump): Likewise.
21646 (tablejump): Likewise.
21647 (*sibcall_memory): Likewise.
21648 (*sibcall_value_memory): Likewise.
21649 Peepholes of indirect call and jump via memory: Likewise.
21650 * config/i386/i386.opt: Likewise.
21651 * config/i386/predicates.md (indirect_branch_operand): Likewise.
21652 (GOT_memory_operand): Likewise.
21653 (call_insn_operand): Likewise.
21654 (sibcall_insn_operand): Likewise.
21655 (GOT32_symbol_operand): Likewise.
21657 2018-01-15 Jakub Jelinek <jakub@redhat.com>
21659 PR middle-end/83837
21660 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
21661 type rather than type addr's type points to.
21662 (expand_omp_atomic_mutex): Likewise.
21663 (expand_omp_atomic): Likewise.
21665 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21668 * config/i386/i386.c (output_indirect_thunk_function): Use
21669 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
21670 for __x86_return_thunk.
21672 2018-01-15 Richard Biener <rguenther@suse.de>
21674 PR middle-end/83850
21675 * expmed.c (extract_bit_field_1): Fix typo.
21677 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21680 * config/arm/iterators.md (VF): New mode iterator.
21681 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
21682 Remove integer-related logic from pattern.
21683 (neon_vabd<mode>_3): Likewise.
21685 2018-01-15 Jakub Jelinek <jakub@redhat.com>
21687 PR middle-end/82694
21688 * common.opt (fstrict-overflow): No longer an alias.
21689 (fwrapv-pointer): New option.
21690 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
21691 also for pointer types based on flag_wrapv_pointer.
21692 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
21693 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
21694 opts->x_flag_wrapv got set.
21695 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
21696 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
21697 POINTER_TYPE_OVERFLOW_UNDEFINED.
21698 * match.pd: Likewise in address comparison pattern.
21699 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
21701 2018-01-15 Richard Biener <rguenther@suse.de>
21704 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
21705 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
21706 Reset type names to their identifier if their TYPE_DECL doesn't
21707 have linkage (and thus is used for ODR and devirt).
21708 (save_debug_info_for_decl): Remove.
21709 (save_debug_info_for_type): Likewise.
21710 (add_tree_to_fld_list): Adjust.
21711 * tree-pretty-print.c (dump_generic_node): Make dumping of
21712 type names more robust.
21714 2018-01-15 Richard Biener <rguenther@suse.de>
21716 * BASE-VER: Bump to 8.0.1.
21718 2018-01-14 Martin Sebor <msebor@redhat.com>
21721 * builtins.c (check_access): Avoid warning when the no-warning bit
21724 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
21726 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
21727 * ira-color (allocno_hard_regs_compare): Likewise.
21729 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
21732 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
21733 Use .pushsection/.popsection.
21735 2018-01-14 Martin Sebor <msebor@redhat.com>
21738 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
21740 2018-01-14 Jakub Jelinek <jakub@redhat.com>
21742 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
21743 entry from extra_headers.
21744 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
21745 extra_headers, make the list bitwise identical to the i?86-*-* one.
21747 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21749 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
21750 -mcmodel=large with -mindirect-branch=thunk,
21751 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
21752 -mfunction-return=thunk-extern.
21753 * doc/invoke.texi: Document -mcmodel=large is incompatible with
21754 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
21755 -mfunction-return=thunk and -mfunction-return=thunk-extern.
21757 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21759 * config/i386/i386.c (print_reg): Print the name of the full
21760 integer register without '%'.
21761 (ix86_print_operand): Handle 'V'.
21762 * doc/extend.texi: Document 'V' modifier.
21764 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21766 * config/i386/constraints.md (Bs): Disallow memory operand for
21767 -mindirect-branch-register.
21769 * config/i386/predicates.md (indirect_branch_operand): Likewise.
21770 (GOT_memory_operand): Likewise.
21771 (call_insn_operand): Likewise.
21772 (sibcall_insn_operand): Likewise.
21773 (GOT32_symbol_operand): Likewise.
21774 * config/i386/i386.md (indirect_jump): Call convert_memory_address
21775 for -mindirect-branch-register.
21776 (tablejump): Likewise.
21777 (*sibcall_memory): Likewise.
21778 (*sibcall_value_memory): Likewise.
21779 Disallow peepholes of indirect call and jump via memory for
21780 -mindirect-branch-register.
21781 (*call_pop): Replace m with Bw.
21782 (*call_value_pop): Likewise.
21783 (*sibcall_pop_memory): Replace m with Bs.
21784 * config/i386/i386.opt (mindirect-branch-register): New option.
21785 * doc/invoke.texi: Document -mindirect-branch-register option.
21787 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21789 * config/i386/i386-protos.h (ix86_output_function_return): New.
21790 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
21791 set function_return_type.
21792 (indirect_thunk_name): Add ret_p to indicate thunk for function
21794 (output_indirect_thunk_function): Pass false to
21795 indirect_thunk_name.
21796 (ix86_output_indirect_branch_via_reg): Likewise.
21797 (ix86_output_indirect_branch_via_push): Likewise.
21798 (output_indirect_thunk_function): Create alias for function
21799 return thunk if regno < 0.
21800 (ix86_output_function_return): New function.
21801 (ix86_handle_fndecl_attribute): Handle function_return.
21802 (ix86_attribute_table): Add function_return.
21803 * config/i386/i386.h (machine_function): Add
21804 function_return_type.
21805 * config/i386/i386.md (simple_return_internal): Use
21806 ix86_output_function_return.
21807 (simple_return_internal_long): Likewise.
21808 * config/i386/i386.opt (mfunction-return=): New option.
21809 (indirect_branch): Mention -mfunction-return=.
21810 * doc/extend.texi: Document function_return function attribute.
21811 * doc/invoke.texi: Document -mfunction-return= option.
21813 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21815 * config/i386/i386-opts.h (indirect_branch): New.
21816 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
21817 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
21818 with local indirect jump when converting indirect call and jump.
21819 (ix86_set_indirect_branch_type): New.
21820 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
21821 (indirectlabelno): New.
21822 (indirect_thunk_needed): Likewise.
21823 (indirect_thunk_bnd_needed): Likewise.
21824 (indirect_thunks_used): Likewise.
21825 (indirect_thunks_bnd_used): Likewise.
21826 (INDIRECT_LABEL): Likewise.
21827 (indirect_thunk_name): Likewise.
21828 (output_indirect_thunk): Likewise.
21829 (output_indirect_thunk_function): Likewise.
21830 (ix86_output_indirect_branch_via_reg): Likewise.
21831 (ix86_output_indirect_branch_via_push): Likewise.
21832 (ix86_output_indirect_branch): Likewise.
21833 (ix86_output_indirect_jmp): Likewise.
21834 (ix86_code_end): Call output_indirect_thunk_function if needed.
21835 (ix86_output_call_insn): Call ix86_output_indirect_branch if
21837 (ix86_handle_fndecl_attribute): Handle indirect_branch.
21838 (ix86_attribute_table): Add indirect_branch.
21839 * config/i386/i386.h (machine_function): Add indirect_branch_type
21840 and has_local_indirect_jump.
21841 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
21843 (tablejump): Likewise.
21844 (*indirect_jump): Use ix86_output_indirect_jmp.
21845 (*tablejump_1): Likewise.
21846 (simple_return_indirect_internal): Likewise.
21847 * config/i386/i386.opt (mindirect-branch=): New option.
21848 (indirect_branch): New.
21851 (thunk-inline): Likewise.
21852 (thunk-extern): Likewise.
21853 * doc/extend.texi: Document indirect_branch function attribute.
21854 * doc/invoke.texi: Document -mindirect-branch= option.
21856 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
21859 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
21861 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
21863 * ipa-inline.c (want_inline_small_function_p): Return false if
21864 inlining has already failed with CIF_FINAL_ERROR.
21865 (update_caller_keys): Call want_inline_small_function_p before
21867 (update_callee_keys): Likewise.
21869 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
21871 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
21873 (rs6000_quadword_masked_address_p): Likewise.
21874 (quad_aligned_load_p): Likewise.
21875 (quad_aligned_store_p): Likewise.
21876 (const_load_sequence_p): Add comment to describe the outer-most loop.
21877 (mimic_memory_attributes_and_flags): New function.
21878 (rs6000_gen_stvx): Likewise.
21879 (replace_swapped_aligned_store): Likewise.
21880 (rs6000_gen_lvx): Likewise.
21881 (replace_swapped_aligned_load): Likewise.
21882 (replace_swapped_load_constant): Capitalize argument name in
21883 comment describing this function.
21884 (rs6000_analyze_swaps): Add a third pass to search for vector loads
21885 and stores that access quad-word aligned addresses and replace
21886 with stvx or lvx instructions when appropriate.
21887 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
21888 New function prototype.
21889 (rs6000_quadword_masked_address_p): Likewise.
21890 (rs6000_gen_lvx): Likewise.
21891 (rs6000_gen_stvx): Likewise.
21892 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
21893 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
21894 when memory address is aligned.
21895 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
21896 this split to select lvx instruction when memory address is aligned.
21897 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
21898 instruction when memory address is aligned.
21899 (*vsx_le_perm_load_v16qi): Likewise.
21900 (four unnamed splitters): Modify to select the stvx instruction
21901 when memory is aligned.
21903 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
21905 * predict.c (determine_unlikely_bbs): Handle correctly BBs
21906 which appears in the queue multiple times.
21908 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21909 Alan Hayward <alan.hayward@arm.com>
21910 David Sherwood <david.sherwood@arm.com>
21912 * tree-vectorizer.h (vec_lower_bound): New structure.
21913 (_loop_vec_info): Add check_nonzero and lower_bounds.
21914 (LOOP_VINFO_CHECK_NONZERO): New macro.
21915 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
21916 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
21917 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
21918 fields. Make seg_len the distance travelled, not including the
21920 (dr_direction_indicator): Declare.
21921 (dr_zero_step_indicator): Likewise.
21922 (dr_known_forward_stride_p): Likewise.
21923 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
21925 (runtime_alias_check_p): Allow runtime alias checks with
21927 (operator ==): Compare access_size and align.
21928 (prune_runtime_alias_test_list): Rework for new distinction between
21929 the access_size and seg_len.
21930 (create_intersect_range_checks_index): Likewise. Cope with polynomial
21932 (get_segment_min_max): New function.
21933 (create_intersect_range_checks): Use it.
21934 (dr_step_indicator): New function.
21935 (dr_direction_indicator): Likewise.
21936 (dr_zero_step_indicator): Likewise.
21937 (dr_known_forward_stride_p): Likewise.
21938 * tree-loop-distribution.c (data_ref_segment_size): Return
21939 DR_STEP * (niters - 1).
21940 (compute_alias_check_pairs): Update call to the dr_with_seg_len
21942 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
21943 (vect_preserves_scalar_order_p): New function, split out from...
21944 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
21945 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
21946 (vect_vfa_access_size): New function.
21947 (vect_vfa_align): Likewise.
21948 (vect_compile_time_alias): Take access_size_a and access_b arguments.
21949 (dump_lower_bound): New function.
21950 (vect_check_lower_bound): Likewise.
21951 (vect_small_gap_p): Likewise.
21952 (vectorizable_with_step_bound_p): Likewise.
21953 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
21954 depencies if the vectorization factor is 1. Convert the checks
21955 for nonzero steps into checks on the bounds of DR_STEP. Try using
21956 a bunds check for variable steps if the minimum required step is
21957 relatively small. Update calls to the dr_with_seg_len
21958 constructor and to vect_compile_time_alias.
21959 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
21961 (vect_loop_versioning): Call it.
21962 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
21964 (vect_estimate_min_profitable_iters): Account for any bounds checks.
21966 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21967 Alan Hayward <alan.hayward@arm.com>
21968 David Sherwood <david.sherwood@arm.com>
21970 * doc/sourcebuild.texi (vect_scatter_store): Document.
21971 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
21973 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
21975 * genopinit.c (main): Add supports_vec_scatter_store and
21976 supports_vec_scatter_store_cached to target_optabs.
21977 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
21978 IFN_MASK_SCATTER_STORE.
21979 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
21981 * internal-fn.h (internal_store_fn_p): Declare.
21982 (internal_fn_stored_value_index): Likewise.
21983 * internal-fn.c (scatter_store_direct): New macro.
21984 (expand_scatter_store_optab_fn): New function.
21985 (direct_scatter_store_optab_supported_p): New macro.
21986 (internal_store_fn_p): New function.
21987 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
21988 IFN_MASK_SCATTER_STORE.
21989 (internal_fn_mask_index): Likewise.
21990 (internal_fn_stored_value_index): New function.
21991 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
21992 for scatter stores.
21993 * optabs-query.h (supports_vec_scatter_store_p): Declare.
21994 * optabs-query.c (supports_vec_scatter_store_p): New function.
21995 * tree-vectorizer.h (vect_get_store_rhs): Declare.
21996 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
21997 true for scatter stores.
21998 (vect_gather_scatter_fn_p): Handle scatter stores too.
21999 (vect_check_gather_scatter): Consider using scatter stores if
22000 supports_vec_scatter_store_p.
22001 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
22002 scatter stores too.
22003 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
22004 internal_fn_stored_value_index.
22005 (check_load_store_masking): Handle scatter stores too.
22006 (vect_get_store_rhs): Make public.
22007 (vectorizable_call): Use internal_store_fn_p.
22008 (vectorizable_store): Handle scatter store internal functions.
22009 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
22010 when deciding whether the end of the group has been reached.
22011 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
22012 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
22013 (mask_scatter_store<mode>): New insns.
22015 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22016 Alan Hayward <alan.hayward@arm.com>
22017 David Sherwood <david.sherwood@arm.com>
22019 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
22020 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
22021 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
22023 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
22024 Use vect_truncate_gather_scatter_offset if we can't treat the
22025 operation as a normal gather load or scatter store.
22026 (get_group_load_store_type): Take the gather_scatter_info
22027 as argument. Try using a gather load or scatter store for
22028 single-element groups.
22029 (get_load_store_type): Update calls to get_group_load_store_type
22030 and vect_use_strided_gather_scatters_p.
22032 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22033 Alan Hayward <alan.hayward@arm.com>
22034 David Sherwood <david.sherwood@arm.com>
22036 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
22037 optional tree argument.
22038 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
22040 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
22041 but continue to use the current value as a fallback.
22042 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
22043 to compare the updates.
22044 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
22045 (get_load_store_type): Use it when handling a strided access.
22046 (vect_get_strided_load_store_ops): New function.
22047 (vect_get_data_ptr_increment): Likewise.
22048 (vectorizable_load): Handle strided gather loads. Always pass
22049 a step to vect_create_data_ref_ptr and bump_vector_ptr.
22051 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22052 Alan Hayward <alan.hayward@arm.com>
22053 David Sherwood <david.sherwood@arm.com>
22055 * doc/md.texi (gather_load@var{m}): Document.
22056 (mask_gather_load@var{m}): Likewise.
22057 * genopinit.c (main): Add supports_vec_gather_load and
22058 supports_vec_gather_load_cached to target_optabs.
22059 * optabs-tree.c (init_tree_optimization_optabs): Use
22060 ggc_cleared_alloc to allocate target_optabs.
22061 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
22062 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
22064 * internal-fn.h (internal_load_fn_p): Declare.
22065 (internal_gather_scatter_fn_p): Likewise.
22066 (internal_fn_mask_index): Likewise.
22067 (internal_gather_scatter_fn_supported_p): Likewise.
22068 * internal-fn.c (gather_load_direct): New macro.
22069 (expand_gather_load_optab_fn): New function.
22070 (direct_gather_load_optab_supported_p): New macro.
22071 (direct_internal_fn_optab): New function.
22072 (internal_load_fn_p): Likewise.
22073 (internal_gather_scatter_fn_p): Likewise.
22074 (internal_fn_mask_index): Likewise.
22075 (internal_gather_scatter_fn_supported_p): Likewise.
22076 * optabs-query.c (supports_at_least_one_mode_p): New function.
22077 (supports_vec_gather_load_p): Likewise.
22078 * optabs-query.h (supports_vec_gather_load_p): Declare.
22079 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
22080 and memory_type field.
22081 (NUM_PATTERNS): Bump to 15.
22082 * tree-vect-data-refs.c: Include internal-fn.h.
22083 (vect_gather_scatter_fn_p): New function.
22084 (vect_describe_gather_scatter_call): Likewise.
22085 (vect_check_gather_scatter): Try using internal functions for
22086 gather loads. Recognize existing calls to a gather load function.
22087 (vect_analyze_data_refs): Consider using gather loads if
22088 supports_vec_gather_load_p.
22089 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
22090 (vect_get_gather_scatter_offset_type): Likewise.
22091 (vect_convert_mask_for_vectype): Likewise.
22092 (vect_add_conversion_to_patterm): Likewise.
22093 (vect_try_gather_scatter_pattern): Likewise.
22094 (vect_recog_gather_scatter_pattern): New pattern recognizer.
22095 (vect_vect_recog_func_ptrs): Add it.
22096 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
22097 internal_fn_mask_index and internal_gather_scatter_fn_p.
22098 (check_load_store_masking): Take the gather_scatter_info as an
22099 argument and handle gather loads.
22100 (vect_get_gather_scatter_ops): New function.
22101 (vectorizable_call): Check internal_load_fn_p.
22102 (vectorizable_load): Likewise. Handle gather load internal
22104 (vectorizable_store): Update call to check_load_store_masking.
22105 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
22106 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
22107 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
22108 (aarch64_gather_scale_operand_d): New predicates.
22109 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
22110 (mask_gather_load<mode>): New insns.
22112 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22113 Alan Hayward <alan.hayward@arm.com>
22114 David Sherwood <david.sherwood@arm.com>
22116 * optabs.def (fold_left_plus_optab): New optab.
22117 * doc/md.texi (fold_left_plus_@var{m}): Document.
22118 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
22119 * internal-fn.c (fold_left_direct): Define.
22120 (expand_fold_left_optab_fn): Likewise.
22121 (direct_fold_left_optab_supported_p): Likewise.
22122 * fold-const-call.c (fold_const_fold_left): New function.
22123 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
22124 * tree-parloops.c (valid_reduction_p): New function.
22125 (gather_scalar_reductions): Use it.
22126 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
22127 (vect_finish_replace_stmt): Declare.
22128 * tree-vect-loop.c (fold_left_reduction_fn): New function.
22129 (needs_fold_left_reduction_p): New function, split out from...
22130 (vect_is_simple_reduction): ...here. Accept reductions that
22131 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
22132 (vect_force_simple_reduction): Also store the reduction type in
22133 the assignment's STMT_VINFO_REDUC_TYPE.
22134 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
22135 (merge_with_identity): New function.
22136 (vect_expand_fold_left): Likewise.
22137 (vectorize_fold_left_reduction): Likewise.
22138 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
22139 scalar phi in place for it. Check for target support and reject
22140 cases that would reassociate the operation. Defer the transform
22141 phase to vectorize_fold_left_reduction.
22142 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
22143 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
22144 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
22146 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22148 * tree-if-conv.c (predicate_mem_writes): Remove redundant
22149 call to ifc_temp_var.
22151 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22152 Alan Hayward <alan.hayward@arm.com>
22153 David Sherwood <david.sherwood@arm.com>
22155 * target.def (legitimize_address_displacement): Take the original
22156 offset as a poly_int.
22157 * targhooks.h (default_legitimize_address_displacement): Update
22159 * targhooks.c (default_legitimize_address_displacement): Likewise.
22160 * doc/tm.texi: Regenerate.
22161 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
22162 as an argument, moving assert of ad->disp == ad->disp_term to...
22163 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
22164 Try calling targetm.legitimize_address_displacement before expanding
22165 the address rather than afterwards, and adjust for the new interface.
22166 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
22167 Match the new hook interface. Handle SVE addresses.
22168 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
22169 new hook interface.
22171 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22173 * Makefile.in (OBJS): Add early-remat.o.
22174 * target.def (select_early_remat_modes): New hook.
22175 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
22176 * doc/tm.texi: Regenerate.
22177 * targhooks.h (default_select_early_remat_modes): Declare.
22178 * targhooks.c (default_select_early_remat_modes): New function.
22179 * timevar.def (TV_EARLY_REMAT): New timevar.
22180 * passes.def (pass_early_remat): New pass.
22181 * tree-pass.h (make_pass_early_remat): Declare.
22182 * early-remat.c: New file.
22183 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
22185 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
22187 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22188 Alan Hayward <alan.hayward@arm.com>
22189 David Sherwood <david.sherwood@arm.com>
22191 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
22192 vfm1 with a bound_epilog parameter.
22193 (vect_do_peeling): Update calls accordingly, and move the prologue
22194 call earlier in the function. Treat the base bound_epilog as 0 for
22195 fully-masked loops and retain vf - 1 for other loops. Add 1 to
22196 this base when peeling for gaps.
22197 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
22198 with fully-masked loops.
22199 (vect_estimate_min_profitable_iters): Handle the single peeled
22200 iteration in that case.
22202 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22203 Alan Hayward <alan.hayward@arm.com>
22204 David Sherwood <david.sherwood@arm.com>
22206 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
22207 single-element interleaving even if the size is not a power of 2.
22208 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
22209 accesses for single-element interleaving if the group size is
22212 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22213 Alan Hayward <alan.hayward@arm.com>
22214 David Sherwood <david.sherwood@arm.com>
22216 * doc/md.texi (fold_extract_last_@var{m}): Document.
22217 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
22218 * optabs.def (fold_extract_last_optab): New optab.
22219 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
22220 * internal-fn.c (fold_extract_direct): New macro.
22221 (expand_fold_extract_optab_fn): Likewise.
22222 (direct_fold_extract_optab_supported_p): Likewise.
22223 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
22224 * tree-vect-loop.c (vect_model_reduction_cost): Handle
22225 EXTRACT_LAST_REDUCTION.
22226 (get_initial_def_for_reduction): Do not create an initial vector
22227 for EXTRACT_LAST_REDUCTION reductions.
22228 (vectorizable_reduction): Leave the scalar phi in place for
22229 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
22230 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
22231 epilogue code for EXTRACT_LAST_REDUCTION and defer the
22232 transform phase to vectorizable_condition.
22233 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
22235 (vect_finish_stmt_generation): ...here.
22236 (vect_finish_replace_stmt): New function.
22237 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
22238 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
22240 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
22242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22243 Alan Hayward <alan.hayward@arm.com>
22244 David Sherwood <david.sherwood@arm.com>
22246 * doc/md.texi (extract_last_@var{m}): Document.
22247 * optabs.def (extract_last_optab): New optab.
22248 * internal-fn.def (EXTRACT_LAST): New internal function.
22249 * internal-fn.c (cond_unary_direct): New macro.
22250 (expand_cond_unary_optab_fn): Likewise.
22251 (direct_cond_unary_optab_supported_p): Likewise.
22252 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
22253 loops using EXTRACT_LAST.
22254 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
22255 (extract_last_<mode>): ...this optab.
22256 (vec_extract<mode><Vel>): Update accordingly.
22258 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22259 Alan Hayward <alan.hayward@arm.com>
22260 David Sherwood <david.sherwood@arm.com>
22262 * target.def (empty_mask_is_expensive): New hook.
22263 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
22264 * doc/tm.texi: Regenerate.
22265 * targhooks.h (default_empty_mask_is_expensive): Declare.
22266 * targhooks.c (default_empty_mask_is_expensive): New function.
22267 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
22268 if the target says that empty masks are expensive.
22269 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
22271 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
22273 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22274 Alan Hayward <alan.hayward@arm.com>
22275 David Sherwood <david.sherwood@arm.com>
22277 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
22278 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
22279 (vect_use_loop_mask_for_alignment_p): New function.
22280 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
22281 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
22282 niters_skip argument. Make sure that the first niters_skip elements
22283 of the first iteration are inactive.
22284 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
22285 Update call to vect_set_loop_masks_directly.
22286 (get_misalign_in_elems): New function, split out from...
22287 (vect_gen_prolog_loop_niters): ...here.
22288 (vect_update_init_of_dr): Take a code argument that specifies whether
22289 the adjustment should be added or subtracted.
22290 (vect_update_init_of_drs): Likewise.
22291 (vect_prepare_for_masked_peels): New function.
22292 (vect_do_peeling): Skip prologue peeling if we're using a mask
22293 instead. Update call to vect_update_inits_of_drs.
22294 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22296 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
22297 alignment. Do not include the number of peeled iterations in
22298 the minimum threshold in that case.
22299 (vectorizable_induction): Adjust the start value down by
22300 LOOP_VINFO_MASK_SKIP_NITERS iterations.
22301 (vect_transform_loop): Call vect_prepare_for_masked_peels.
22302 Take the number of skipped iterations into account when calculating
22304 * tree-vect-stmts.c (vect_gen_while_not): New function.
22306 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22307 Alan Hayward <alan.hayward@arm.com>
22308 David Sherwood <david.sherwood@arm.com>
22310 * doc/sourcebuild.texi (vect_fully_masked): Document.
22311 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
22312 default value to 0.
22313 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
22315 (vect_analyze_loop_2): ...here. Don't check the vectorization
22316 factor against the number of loop iterations if the loop is
22319 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22320 Alan Hayward <alan.hayward@arm.com>
22321 David Sherwood <david.sherwood@arm.com>
22323 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
22324 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
22325 (dump_groups): Update accordingly.
22326 (iv_use::mem_type): New member variable.
22327 (address_p): New function.
22328 (record_use): Add a mem_type argument and initialize the new
22330 (record_group_use): Add a mem_type argument. Use address_p.
22331 Remove obsolete null checks of base_object. Update call to record_use.
22332 (find_interesting_uses_op): Update call to record_group_use.
22333 (find_interesting_uses_cond): Likewise.
22334 (find_interesting_uses_address): Likewise.
22335 (get_mem_type_for_internal_fn): New function.
22336 (find_address_like_use): Likewise.
22337 (find_interesting_uses_stmt): Try find_address_like_use before
22338 calling find_interesting_uses_op.
22339 (addr_offset_valid_p): Use the iv mem_type field as the type
22340 of the addressed memory.
22341 (add_autoinc_candidates): Likewise.
22342 (get_address_cost): Likewise.
22343 (split_small_address_groups_p): Use address_p.
22344 (split_address_groups): Likewise.
22345 (add_iv_candidate_for_use): Likewise.
22346 (autoinc_possible_for_pair): Likewise.
22347 (rewrite_groups): Likewise.
22348 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
22349 (determine_group_iv_cost): Update after split of USE_ADDRESS.
22350 (get_alias_ptr_type_for_ptr_address): New function.
22351 (rewrite_use_address): Rewrite address uses in calls that were
22352 identified by find_address_like_use.
22354 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22355 Alan Hayward <alan.hayward@arm.com>
22356 David Sherwood <david.sherwood@arm.com>
22358 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
22360 * gimple-expr.h (is_gimple_addressable: Likewise.
22361 * gimple-expr.c (is_gimple_address): Likewise.
22362 * internal-fn.c (expand_call_mem_ref): New function.
22363 (expand_mask_load_optab_fn): Use it.
22364 (expand_mask_store_optab_fn): Likewise.
22366 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22367 Alan Hayward <alan.hayward@arm.com>
22368 David Sherwood <david.sherwood@arm.com>
22370 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
22371 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
22372 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
22373 (cond_umax@var{mode}): Document.
22374 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
22375 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
22376 (cond_umin_optab, cond_umax_optab): New optabs.
22377 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
22378 (COND_IOR, COND_XOR): New internal functions.
22379 * internal-fn.h (get_conditional_internal_fn): Declare.
22380 * internal-fn.c (cond_binary_direct): New macro.
22381 (expand_cond_binary_optab_fn): Likewise.
22382 (direct_cond_binary_optab_supported_p): Likewise.
22383 (get_conditional_internal_fn): New function.
22384 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
22385 Cope with reduction statements that are vectorized as calls rather
22387 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
22388 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
22389 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
22390 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
22391 (UNSPEC_COND_EOR): New unspecs.
22392 (optab): Add mappings for them.
22393 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
22394 (sve_int_op, sve_fp_op): New int attributes.
22396 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22397 Alan Hayward <alan.hayward@arm.com>
22398 David Sherwood <david.sherwood@arm.com>
22400 * optabs.def (while_ult_optab): New optab.
22401 * doc/md.texi (while_ult@var{m}@var{n}): Document.
22402 * internal-fn.def (WHILE_ULT): New internal function.
22403 * internal-fn.h (direct_internal_fn_supported_p): New override
22404 that takes two types as argument.
22405 * internal-fn.c (while_direct): New macro.
22406 (expand_while_optab_fn): New function.
22407 (convert_optab_supported_p): Likewise.
22408 (direct_while_optab_supported_p): New macro.
22409 * wide-int.h (wi::udiv_ceil): New function.
22410 * tree-vectorizer.h (rgroup_masks): New structure.
22411 (vec_loop_masks): New typedef.
22412 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
22413 and fully_masked_p.
22414 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
22415 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
22416 (vect_max_vf): New function.
22417 (slpeel_make_loop_iterate_ntimes): Delete.
22418 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
22419 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
22420 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
22421 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
22422 internal-fn.h, stor-layout.h and optabs-query.h.
22423 (vect_set_loop_mask): New function.
22424 (add_preheader_seq): Likewise.
22425 (add_header_seq): Likewise.
22426 (interleave_supported_p): Likewise.
22427 (vect_maybe_permute_loop_masks): Likewise.
22428 (vect_set_loop_masks_directly): Likewise.
22429 (vect_set_loop_condition_masked): Likewise.
22430 (vect_set_loop_condition_unmasked): New function, split out from
22431 slpeel_make_loop_iterate_ntimes.
22432 (slpeel_make_loop_iterate_ntimes): Rename to..
22433 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
22434 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
22435 (vect_do_peeling): Update call accordingly.
22436 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
22438 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22439 mask_compare_type, can_fully_mask_p and fully_masked_p.
22440 (release_vec_loop_masks): New function.
22441 (_loop_vec_info): Use it to free the loop masks.
22442 (can_produce_all_loop_masks_p): New function.
22443 (vect_get_max_nscalars_per_iter): Likewise.
22444 (vect_verify_full_masking): Likewise.
22445 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
22446 retries, and free the mask rgroups before retrying. Check loop-wide
22447 reasons for disallowing fully-masked loops. Make the final decision
22448 about whether use a fully-masked loop or not.
22449 (vect_estimate_min_profitable_iters): Do not assume that peeling
22450 for the number of iterations will be needed for fully-masked loops.
22451 (vectorizable_reduction): Disable fully-masked loops.
22452 (vectorizable_live_operation): Likewise.
22453 (vect_halve_mask_nunits): New function.
22454 (vect_double_mask_nunits): Likewise.
22455 (vect_record_loop_mask): Likewise.
22456 (vect_get_loop_mask): Likewise.
22457 (vect_transform_loop): Handle the case in which the final loop
22458 iteration might handle a partial vector. Call vect_set_loop_condition
22459 instead of slpeel_make_loop_iterate_ntimes.
22460 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
22461 (check_load_store_masking): New function.
22462 (prepare_load_store_mask): Likewise.
22463 (vectorizable_store): Handle fully-masked loops.
22464 (vectorizable_load): Likewise.
22465 (supportable_widening_operation): Use vect_halve_mask_nunits for
22467 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
22468 (vect_gen_while): New function.
22469 * config/aarch64/aarch64.md (umax<mode>3): New expander.
22470 (aarch64_uqdec<mode>): New insn.
22472 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22473 Alan Hayward <alan.hayward@arm.com>
22474 David Sherwood <david.sherwood@arm.com>
22476 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
22477 (reduc_xor_scal_optab): New optabs.
22478 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
22479 (reduc_xor_scal_@var{m}): Document.
22480 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
22481 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
22482 internal functions.
22483 * fold-const-call.c (fold_const_call): Handle them.
22484 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
22485 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
22486 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
22487 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
22488 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
22489 (UNSPEC_XORV): New unspecs.
22490 (optab): Add entries for them.
22491 (BITWISEV): New int iterator.
22492 (bit_reduc_op): New int attributes.
22494 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22495 Alan Hayward <alan.hayward@arm.com>
22496 David Sherwood <david.sherwood@arm.com>
22498 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
22499 * internal-fn.def (VEC_SHL_INSERT): New internal function.
22500 * optabs.def (vec_shl_insert_optab): New optab.
22501 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
22502 (duplicate_and_interleave): Likewise.
22503 * tree-vect-loop.c: Include internal-fn.h.
22504 (neutral_op_for_slp_reduction): New function, split out from
22505 get_initial_defs_for_reduction.
22506 (get_initial_def_for_reduction): Handle option 2 for variable-length
22507 vectors by loading the neutral value into a vector and then shifting
22508 the initial value into element 0.
22509 (get_initial_defs_for_reduction): Replace the code argument with
22510 the neutral value calculated by neutral_op_for_slp_reduction.
22511 Use gimple_build_vector for constant-length vectors.
22512 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
22513 but the first group_size elements have a neutral value.
22514 Use duplicate_and_interleave otherwise.
22515 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
22516 Update call to get_initial_defs_for_reduction. Handle SLP
22517 reductions for variable-length vectors by creating one vector
22518 result for each scalar result, with the elements associated
22519 with other scalar results stubbed out with the neutral value.
22520 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
22521 Require IFN_VEC_SHL_INSERT for double reductions on
22522 variable-length vectors, or SLP reductions that have
22523 a neutral value. Require can_duplicate_and_interleave_p
22524 support for variable-length unchained SLP reductions if there
22525 is no neutral value, such as for MIN/MAX reductions. Also require
22526 the number of vector elements to be a multiple of the number of
22527 SLP statements when doing variable-length unchained SLP reductions.
22528 Update call to vect_create_epilog_for_reduction.
22529 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
22530 and remove initial values.
22531 (duplicate_and_interleave): Make public.
22532 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
22533 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
22535 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22536 Alan Hayward <alan.hayward@arm.com>
22537 David Sherwood <david.sherwood@arm.com>
22539 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
22540 (can_duplicate_and_interleave_p): New function.
22541 (vect_get_and_check_slp_defs): Take the vector of statements
22542 rather than just the current one. Remove excess parentheses.
22543 Restriction rejectinon of vect_constant_def and vect_external_def
22544 for variable-length vectors to boolean types, or types for which
22545 can_duplicate_and_interleave_p is false.
22546 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
22547 (duplicate_and_interleave): New function.
22548 (vect_get_constant_vectors): Use gimple_build_vector for
22549 constant-length vectors and suitable variable-length constant
22550 vectors. Use duplicate_and_interleave for other variable-length
22551 vectors. Don't defer the update when inserting new statements.
22553 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22554 Alan Hayward <alan.hayward@arm.com>
22555 David Sherwood <david.sherwood@arm.com>
22557 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
22558 min_profitable_iters doesn't go negative.
22560 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22561 Alan Hayward <alan.hayward@arm.com>
22562 David Sherwood <david.sherwood@arm.com>
22564 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
22565 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
22566 * optabs.def (vec_mask_load_lanes_optab): New optab.
22567 (vec_mask_store_lanes_optab): Likewise.
22568 * internal-fn.def (MASK_LOAD_LANES): New internal function.
22569 (MASK_STORE_LANES): Likewise.
22570 * internal-fn.c (mask_load_lanes_direct): New macro.
22571 (mask_store_lanes_direct): Likewise.
22572 (expand_mask_load_optab_fn): Handle masked operations.
22573 (expand_mask_load_lanes_optab_fn): New macro.
22574 (expand_mask_store_optab_fn): Handle masked operations.
22575 (expand_mask_store_lanes_optab_fn): New macro.
22576 (direct_mask_load_lanes_optab_supported_p): Likewise.
22577 (direct_mask_store_lanes_optab_supported_p): Likewise.
22578 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
22580 (vect_load_lanes_supported): Likewise.
22581 * tree-vect-data-refs.c (strip_conversion): New function.
22582 (can_group_stmts_p): Likewise.
22583 (vect_analyze_data_ref_accesses): Use it instead of checking
22584 for a pair of assignments.
22585 (vect_store_lanes_supported): Take a masked_p parameter.
22586 (vect_load_lanes_supported): Likewise.
22587 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
22588 vect_store_lanes_supported and vect_load_lanes_supported.
22589 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
22590 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
22591 parameter. Don't allow gaps for masked accesses.
22592 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
22593 and vect_load_lanes_supported.
22594 (get_load_store_type): Take a masked_p parameter and update
22595 call to get_group_load_store_type.
22596 (vectorizable_store): Update call to get_load_store_type.
22597 Handle IFN_MASK_STORE_LANES.
22598 (vectorizable_load): Update call to get_load_store_type.
22599 Handle IFN_MASK_LOAD_LANES.
22601 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22602 Alan Hayward <alan.hayward@arm.com>
22603 David Sherwood <david.sherwood@arm.com>
22605 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
22607 * config/aarch64/aarch64-protos.h
22608 (aarch64_sve_struct_memory_operand_p): Declare.
22609 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
22610 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
22611 (VPRED, vpred): Handle SVE structure modes.
22612 * config/aarch64/constraints.md (Utx): New constraint.
22613 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
22614 (aarch64_sve_struct_nonimmediate_operand): New predicates.
22615 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
22616 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
22617 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
22618 structure modes. Split into pieces after RA.
22619 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
22620 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
22622 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
22623 SVE structure modes.
22624 (aarch64_classify_address): Likewise.
22625 (sizetochar): Move earlier in file.
22626 (aarch64_print_operand): Handle SVE register lists.
22627 (aarch64_array_mode): New function.
22628 (aarch64_sve_struct_memory_operand_p): Likewise.
22629 (TARGET_ARRAY_MODE): Redefine.
22631 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22632 Alan Hayward <alan.hayward@arm.com>
22633 David Sherwood <david.sherwood@arm.com>
22635 * target.def (array_mode): New target hook.
22636 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
22637 * doc/tm.texi: Regenerate.
22638 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
22639 * hooks.c (hook_optmode_mode_uhwi_none): New function.
22640 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
22641 targetm.array_mode.
22642 * stor-layout.c (mode_for_array): Likewise. Support polynomial
22645 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22646 Alan Hayward <alan.hayward@arm.com>
22647 David Sherwood <david.sherwood@arm.com>
22649 * fold-const.c (fold_binary_loc): Check the argument types
22650 rather than the result type when testing for a vector operation.
22652 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22654 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
22655 * doc/tm.texi: Regenerate.
22657 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22658 Alan Hayward <alan.hayward@arm.com>
22659 David Sherwood <david.sherwood@arm.com>
22661 * doc/invoke.texi (-msve-vector-bits=): Document new option.
22662 (sve): Document new AArch64 extension.
22663 * doc/md.texi (w): Extend the description of the AArch64
22664 constraint to include SVE vectors.
22665 (Upl, Upa): Document new AArch64 predicate constraints.
22666 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
22668 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
22669 (msve-vector-bits=): New option.
22670 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
22671 SVE when these are disabled.
22672 (sve): New extension.
22673 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
22674 modes. Adjust their number of units based on aarch64_sve_vg.
22675 (MAX_BITSIZE_MODE_ANY_MODE): Define.
22676 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
22677 aarch64_addr_query_type.
22678 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
22679 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
22680 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
22681 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
22682 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
22683 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
22684 (aarch64_simd_imm_zero_p): Delete.
22685 (aarch64_check_zero_based_sve_index_immediate): Declare.
22686 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22687 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22688 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22689 (aarch64_sve_float_mul_immediate_p): Likewise.
22690 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22691 rather than an rtx.
22692 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
22693 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
22694 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
22695 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
22696 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
22697 (aarch64_regmode_natural_size): Likewise.
22698 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
22699 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
22701 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
22702 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
22703 for VG and the SVE predicate registers.
22704 (V_ALIASES): Add a "z"-prefixed alias.
22705 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
22706 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
22707 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
22708 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
22709 (REG_CLASS_NAMES): Add entries for them.
22710 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
22711 and the predicate registers.
22712 (aarch64_sve_vg): Declare.
22713 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
22714 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
22715 (REGMODE_NATURAL_SIZE): Define.
22716 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
22718 * config/aarch64/aarch64.c: Include cfgrtl.h.
22719 (simd_immediate_info): Add a constructor for series vectors,
22720 and an associated step field.
22721 (aarch64_sve_vg): New variable.
22722 (aarch64_dbx_register_number): Handle VG and the predicate registers.
22723 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
22724 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
22725 (VEC_ANY_DATA, VEC_STRUCT): New constants.
22726 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
22727 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
22728 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
22729 (aarch64_get_mask_mode): New functions.
22730 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
22731 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
22732 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
22733 predicate modes and predicate registers. Explicitly restrict
22734 GPRs to modes of 16 bytes or smaller. Only allow FP registers
22735 to store a vector mode if it is recognized by
22736 aarch64_classify_vector_mode.
22737 (aarch64_regmode_natural_size): New function.
22738 (aarch64_hard_regno_caller_save_mode): Return the original mode
22740 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
22741 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
22742 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
22743 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
22745 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
22746 does not overlap dest if the function is frame-related. Handle
22748 (aarch64_split_add_offset): New function.
22749 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
22750 them aarch64_add_offset.
22751 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
22752 and update call to aarch64_sub_sp.
22753 (aarch64_add_cfa_expression): New function.
22754 (aarch64_expand_prologue): Pass extra temporary registers to the
22755 functions above. Handle the case in which we need to emit new
22756 DW_CFA_expressions for registers that were originally saved
22757 relative to the stack pointer, but now have to be expressed
22758 relative to the frame pointer.
22759 (aarch64_output_mi_thunk): Pass extra temporary registers to the
22761 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
22762 IP0 and IP1 values for SVE frames.
22763 (aarch64_expand_vec_series): New function.
22764 (aarch64_expand_sve_widened_duplicate): Likewise.
22765 (aarch64_expand_sve_const_vector): Likewise.
22766 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
22767 Handle SVE constants. Use emit_move_insn to move a force_const_mem
22768 into the register, rather than emitting a SET directly.
22769 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
22770 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
22771 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
22772 (offset_9bit_signed_scaled_p): New functions.
22773 (aarch64_replicate_bitmask_imm): New function.
22774 (aarch64_bitmask_imm): Use it.
22775 (aarch64_cannot_force_const_mem): Reject expressions involving
22776 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
22777 (aarch64_classify_index): Handle SVE indices, by requiring
22778 a plain register index with a scale that matches the element size.
22779 (aarch64_classify_address): Handle SVE addresses. Assert that
22780 the mode of the address is VOIDmode or an integer mode.
22781 Update call to aarch64_classify_symbol.
22782 (aarch64_classify_symbolic_expression): Update call to
22783 aarch64_classify_symbol.
22784 (aarch64_const_vec_all_in_range_p): New function.
22785 (aarch64_print_vector_float_operand): Likewise.
22786 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
22787 "vN" for FP registers with SVE modes. Handle (const ...) vectors
22788 and the FP immediates 1.0 and 0.5.
22789 (aarch64_print_address_internal): Handle SVE addresses.
22790 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
22791 (aarch64_regno_regclass): Handle predicate registers.
22792 (aarch64_secondary_reload): Handle big-endian reloads of SVE
22794 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
22795 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
22796 (aarch64_convert_sve_vector_bits): New function.
22797 (aarch64_override_options): Use it to handle -msve-vector-bits=.
22798 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22799 rather than an rtx.
22800 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
22801 Handle SVE vector and predicate modes. Accept VL-based constants
22802 that need only one temporary register, and VL offsets that require
22803 no temporary registers.
22804 (aarch64_conditional_register_usage): Mark the predicate registers
22805 as fixed if SVE isn't available.
22806 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
22807 Return true for SVE vector and predicate modes.
22808 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
22809 rather than an unsigned int. Handle SVE modes.
22810 (aarch64_preferred_simd_mode): Update call accordingly. Handle
22812 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
22814 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22815 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22816 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22817 (aarch64_sve_float_mul_immediate_p): New functions.
22818 (aarch64_sve_valid_immediate): New function.
22819 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
22820 Explicitly reject structure modes. Check for INDEX constants.
22821 Handle PTRUE and PFALSE constants.
22822 (aarch64_check_zero_based_sve_index_immediate): New function.
22823 (aarch64_simd_imm_zero_p): Delete.
22824 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
22825 vector modes. Accept constants in the range of CNT[BHWD].
22826 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
22827 ask for an Advanced SIMD mode.
22828 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
22829 (aarch64_simd_vector_alignment): Handle SVE predicates.
22830 (aarch64_vectorize_preferred_vector_alignment): New function.
22831 (aarch64_simd_vector_alignment_reachable): Use it instead of
22833 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
22834 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
22836 (MAX_VECT_LEN): Delete.
22837 (expand_vec_perm_d): Add a vec_flags field.
22838 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
22839 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
22840 (aarch64_evpc_ext): Don't apply a big-endian lane correction
22842 (aarch64_evpc_rev): Rename to...
22843 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
22844 (aarch64_evpc_rev_global): New function.
22845 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
22846 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
22848 (aarch64_evpc_sve_tbl): New function.
22849 (aarch64_expand_vec_perm_const_1): Update after rename of
22850 aarch64_evpc_rev. Handle SVE permutes too, trying
22851 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
22852 than aarch64_evpc_tbl.
22853 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
22854 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
22855 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
22856 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
22857 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
22858 (aarch64_expand_sve_vcond): New functions.
22859 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
22860 of aarch64_vector_mode_p.
22861 (aarch64_dwarf_poly_indeterminate_value): New function.
22862 (aarch64_compute_pressure_classes): Likewise.
22863 (aarch64_can_change_mode_class): Likewise.
22864 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
22865 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
22866 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
22867 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
22868 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
22869 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
22870 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
22871 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
22873 (Dn, Dl, Dr): Accept const as well as const_vector.
22874 (Dz): Likewise. Compare against CONST0_RTX.
22875 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
22876 of "vector" where appropriate.
22877 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
22878 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
22879 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
22880 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
22881 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
22882 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
22883 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
22884 (v_int_equiv): Extend to SVE modes.
22885 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
22887 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
22888 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
22889 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
22890 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
22891 (SVE_COND_FP_CMP): New int iterators.
22892 (perm_hilo): Handle the new unpack unspecs.
22893 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
22895 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
22896 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
22897 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
22898 (aarch64_equality_operator, aarch64_constant_vector_operand)
22899 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
22900 (aarch64_sve_nonimmediate_operand): Likewise.
22901 (aarch64_sve_general_operand): Likewise.
22902 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
22903 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
22904 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
22905 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
22906 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
22907 (aarch64_sve_float_arith_immediate): Likewise.
22908 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
22909 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
22910 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
22911 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
22912 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
22913 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
22914 (aarch64_sve_float_arith_operand): Likewise.
22915 (aarch64_sve_float_arith_with_sub_operand): Likewise.
22916 (aarch64_sve_float_mul_operand): Likewise.
22917 (aarch64_sve_vec_perm_operand): Likewise.
22918 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
22919 (aarch64_mov_operand): Accept const_poly_int and const_vector.
22920 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
22921 as well as const_vector.
22922 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
22923 in file. Use CONST0_RTX and CONSTM1_RTX.
22924 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
22925 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
22926 Use aarch64_simd_imm_zero.
22927 * config/aarch64/aarch64-sve.md: New file.
22928 * config/aarch64/aarch64.md: Include it.
22929 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
22930 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
22931 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
22932 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
22933 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
22934 (sve): New attribute.
22935 (enabled): Disable instructions with the sve attribute unless
22937 (movqi, movhi): Pass CONST_POLY_INT operaneds through
22938 aarch64_expand_mov_immediate.
22939 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
22940 CNT[BHSD] immediates.
22941 (movti): Split CONST_POLY_INT moves into two halves.
22942 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
22943 Split additions that need a temporary here if the destination
22944 is the stack pointer.
22945 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
22946 (*add<mode>3_poly_1): New instruction.
22947 (set_clobber_cc): New expander.
22949 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22951 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
22952 parameter and use it instead of GET_MODE_SIZE (innermode). Use
22953 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
22954 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
22955 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
22956 Change innermode from fixed_mode_size to machine_mode.
22957 (simplify_subreg): Update call accordingly. Handle a constant-sized
22958 subreg of a variable-length CONST_VECTOR.
22960 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22961 Alan Hayward <alan.hayward@arm.com>
22962 David Sherwood <david.sherwood@arm.com>
22964 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
22965 (add_offset_to_base): New function, split out from...
22966 (create_mem_ref): ...here. When handling a scale other than 1,
22967 check first whether the address is valid without the offset.
22968 Add it into the base if so, leaving the index and scale as-is.
22970 2018-01-12 Jakub Jelinek <jakub@redhat.com>
22973 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
22974 fold_for_warn before checking if arg2 is INTEGER_CST.
22976 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
22978 * config/rs6000/predicates.md (load_multiple_operation): Delete.
22979 (store_multiple_operation): Delete.
22980 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
22981 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
22982 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
22983 guarded by TARGET_STRING.
22984 (rs6000_output_load_multiple): Delete.
22985 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
22986 OPTION_MASK_STRING / TARGET_STRING handling.
22987 (print_operand) <'N', 'O'>: Add comment that these are unused now.
22988 (const rs6000_opt_masks) <"string">: Change mask to 0.
22989 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
22990 (MASK_STRING): Delete.
22991 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
22993 (load_multiple): Delete.
23000 (store_multiple): Delete.
23007 (movmemsi_8reg): Delete.
23008 (corresponding unnamed define_insn): Delete.
23009 (movmemsi_6reg): Delete.
23010 (corresponding unnamed define_insn): Delete.
23011 (movmemsi_4reg): Delete.
23012 (corresponding unnamed define_insn): Delete.
23013 (movmemsi_2reg): Delete.
23014 (corresponding unnamed define_insn): Delete.
23015 (movmemsi_1reg): Delete.
23016 (corresponding unnamed define_insn): Delete.
23017 * config/rs6000/rs6000.opt (mno-string): New.
23018 (mstring): Replace by deprecation warning stub.
23019 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
23021 2018-01-12 Jakub Jelinek <jakub@redhat.com>
23023 * regrename.c (regrename_do_replace): If replacing the same
23024 reg multiple times, try to reuse last created gen_raw_REG.
23027 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
23028 main to workaround a bug in GDB.
23030 2018-01-12 Tom de Vries <tom@codesourcery.com>
23033 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
23035 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
23037 PR rtl-optimization/80481
23038 * ira-color.c (get_cap_member): New function.
23039 (allocnos_conflict_by_live_ranges_p): Use it.
23040 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
23041 (setup_slot_coalesced_allocno_live_ranges): Ditto.
23043 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
23046 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
23047 (*saddl_se_1): Ditto.
23048 (*ssubsi_1): Ditto.
23049 (*ssubl_se_1): Ditto.
23051 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
23053 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
23054 rather than wi::to_widest for DR_INITs.
23055 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
23056 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
23057 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
23059 (vect_analyze_group_access_1): Note that here.
23061 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
23063 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
23064 polynomial type sizes.
23066 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
23068 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
23069 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
23070 (gimple_add_tmp_var): Likewise.
23072 2018-01-12 Martin Liska <mliska@suse.cz>
23074 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
23075 (gimple_alloc_sizes): Likewise.
23076 (dump_gimple_statistics): Use PRIu64 in printf format.
23077 * gimple.h: Change uint64_t to int.
23079 2018-01-12 Martin Liska <mliska@suse.cz>
23081 * tree-core.h: Use uint64_t instead of int.
23082 * tree.c (tree_node_counts): Likewise.
23083 (tree_node_sizes): Likewise.
23084 (dump_tree_statistics): Use PRIu64 in printf format.
23086 2018-01-12 Martin Liska <mliska@suse.cz>
23088 * Makefile.in: As qsort_chk is implemented in vec.c, add
23089 vec.o to linkage of gencfn-macros.
23090 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
23091 passing the info to record_node_allocation_statistics.
23092 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
23094 * ggc-common.c (struct ggc_usage): Add operator== and use
23095 it in operator< and compare function.
23096 * mem-stats.h (struct mem_usage): Likewise.
23097 * vec.c (struct vec_usage): Remove operator< and compare
23098 function. Can be simply inherited.
23100 2018-01-12 Martin Jambor <mjambor@suse.cz>
23103 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
23104 * tree-ssa-math-opts.c: Include domwalk.h.
23105 (convert_mult_to_fma_1): New function.
23106 (fma_transformation_info): New type.
23107 (fma_deferring_state): Likewise.
23108 (cancel_fma_deferring): New function.
23109 (result_of_phi): Likewise.
23110 (last_fma_candidate_feeds_initial_phi): Likewise.
23111 (convert_mult_to_fma): Added deferring logic, split actual
23112 transformation to convert_mult_to_fma_1.
23113 (math_opts_dom_walker): New type.
23114 (math_opts_dom_walker::after_dom_children): New method, body moved
23115 here from pass_optimize_widening_mul::execute, added deferring logic
23117 (pass_optimize_widening_mul::execute): Moved most of code to
23118 math_opts_dom_walker::after_dom_children.
23119 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
23120 * config/i386/i386.c (ix86_option_override_internal): Added
23121 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
23123 2018-01-12 Richard Biener <rguenther@suse.de>
23126 * dwarf2out.c (gen_variable_die): Do not reset old_die for
23127 inline instance vars.
23129 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
23132 * config/rx/rx.c (rx_is_restricted_memory_address):
23133 Handle SUBREG case.
23135 2018-01-12 Richard Biener <rguenther@suse.de>
23137 PR tree-optimization/80846
23138 * target.def (split_reduction): New target hook.
23139 * targhooks.c (default_split_reduction): New function.
23140 * targhooks.h (default_split_reduction): Declare.
23141 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
23142 target requests first reduce vectors by combining low and high
23144 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
23145 (get_vectype_for_scalar_type_and_size): Export.
23146 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
23147 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
23148 * doc/tm.texi: Regenerate.
23149 * config/i386/i386.c (ix86_split_reduction): Implement
23150 TARGET_VECTORIZE_SPLIT_REDUCTION.
23152 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
23155 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
23156 in PIC mode except for TARGET_VXWORKS_RTP.
23157 * config/sparc/sparc.c: Include cfgrtl.h.
23158 (TARGET_INIT_PIC_REG): Define.
23159 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
23160 (sparc_pic_register_p): New predicate.
23161 (sparc_legitimate_address_p): Use it.
23162 (sparc_legitimize_pic_address): Likewise.
23163 (sparc_delegitimize_address): Likewise.
23164 (sparc_mode_dependent_address_p): Likewise.
23165 (gen_load_pcrel_sym): Remove 4th parameter.
23166 (load_got_register): Adjust call to above. Remove obsolete stuff.
23167 (sparc_expand_prologue): Do not call load_got_register here.
23168 (sparc_flat_expand_prologue): Likewise.
23169 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
23170 (sparc_use_pseudo_pic_reg): New function.
23171 (sparc_init_pic_reg): Likewise.
23172 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
23173 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
23175 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
23177 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
23178 Add item for branch_cost.
23180 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
23182 PR rtl-optimization/83565
23183 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
23184 not extend the result to a larger mode for rotate operations.
23185 (num_sign_bit_copies1): Likewise.
23187 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23190 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
23192 Use values-Xc.o for -pedantic.
23193 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
23195 2018-01-12 Martin Liska <mliska@suse.cz>
23198 * ipa-devirt.c (final_warning_record::grow_type_warnings):
23200 (possible_polymorphic_call_targets): Use it.
23201 (ipa_devirt): Likewise.
23203 2018-01-12 Martin Liska <mliska@suse.cz>
23205 * profile-count.h (enum profile_quality): Use 0 as invalid
23206 enum value of profile_quality.
23208 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
23210 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
23211 -mext-string options.
23213 2018-01-12 Richard Biener <rguenther@suse.de>
23215 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
23216 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
23217 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
23219 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
23221 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
23223 * configure.ac (--with-long-double-format): Add support for the
23224 configuration option to change the default long double format on
23226 * config.gcc (powerpc*-linux*-*): Likewise.
23227 * configure: Regenerate.
23228 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
23229 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
23230 used without modification.
23232 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23234 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
23235 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
23236 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
23237 MISC_BUILTIN_SPEC_BARRIER.
23238 (rs6000_init_builtins): Likewise.
23239 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
23241 (speculation_barrier): New define_insn.
23242 * doc/extend.texi: Document __builtin_speculation_barrier.
23244 2018-01-11 Jakub Jelinek <jakub@redhat.com>
23247 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
23248 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
23249 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
23251 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
23252 integral modes instead of "ss" and "sd".
23253 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
23254 vectors with 32-bit and 64-bit elements.
23255 (vecdupssescalarmodesuffix): New mode attribute.
23256 (vec_dup<mode>): Use it.
23258 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
23261 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
23262 frame if argument is passed on stack.
23264 2018-01-11 Jakub Jelinek <jakub@redhat.com>
23267 * ree.c (combine_reaching_defs): Optimize also
23268 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
23269 reg2=any_extend(exp); reg1=reg2;, formatting fix.
23271 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
23273 PR middle-end/83189
23274 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
23276 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
23278 PR middle-end/83718
23279 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
23280 after they are computed.
23282 2018-01-11 Bin Cheng <bin.cheng@arm.com>
23284 PR tree-optimization/83695
23285 * gimple-loop-linterchange.cc
23286 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
23287 reset cached scev information after interchange.
23288 (pass_linterchange::execute): Remove call to scev_reset_htab.
23290 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23292 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
23293 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
23294 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
23295 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
23296 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
23297 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
23298 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
23299 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
23300 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
23301 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
23302 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
23303 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
23304 (V_lane_reg): Likewise.
23305 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
23307 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
23308 (vfmal_lane_low<mode>_intrinsic,
23309 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
23310 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
23311 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
23312 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
23313 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
23314 vfmsl_lane_high<mode>_intrinsic): New define_insns.
23316 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23318 * config/arm/arm-cpus.in (fp16fml): New feature.
23319 (ALL_SIMD): Add fp16fml.
23320 (armv8.2-a): Add fp16fml as an option.
23321 (armv8.3-a): Likewise.
23322 (armv8.4-a): Add fp16fml as part of fp16.
23323 * config/arm/arm.h (TARGET_FP16FML): Define.
23324 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
23326 * config/arm/arm-modes.def (V2HF): Define.
23327 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
23328 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
23329 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
23330 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
23331 vfmsl_low, vfmsl_high): New set of builtins.
23332 * config/arm/iterators.md (PLUSMINUS): New code iterator.
23333 (vfml_op): New code attribute.
23334 (VFMLHALVES): New int iterator.
23335 (VFML, VFMLSEL): New mode attributes.
23336 (V_reg): Define mapping for V2HF.
23337 (V_hi, V_lo): New mode attributes.
23338 (VF_constraint): Likewise.
23339 (vfml_half, vfml_half_selector): New int attributes.
23340 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
23342 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
23343 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
23345 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
23346 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
23347 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
23348 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
23350 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
23351 Document new effective target and option set.
23353 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23355 * config/arm/arm-cpus.in (armv8_4): New feature.
23356 (ARMv8_4a): New fgroup.
23357 (armv8.4-a): New arch.
23358 * config/arm/arm-tables.opt: Regenerate.
23359 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
23360 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
23361 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
23362 Add matching rules for -march=armv8.4-a and extensions.
23363 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
23365 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
23368 * config/rx/rx.md (BW): New mode attribute.
23369 (sync_lock_test_and_setsi): Add mode suffix to insn output.
23371 2018-01-11 Richard Biener <rguenther@suse.de>
23373 PR tree-optimization/83435
23374 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
23375 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
23376 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
23378 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23379 Alan Hayward <alan.hayward@arm.com>
23380 David Sherwood <david.sherwood@arm.com>
23382 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
23384 (aarch64_classify_address): Initialize it. Track polynomial offsets.
23385 (aarch64_print_address_internal): Use it to check for a zero offset.
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-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
23392 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
23393 Return a poly_int64 rather than a HOST_WIDE_INT.
23394 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
23395 rather than a HOST_WIDE_INT.
23396 * config/aarch64/aarch64.h (aarch64_frame): Protect with
23397 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
23398 hard_fp_offset, frame_size, initial_adjust, callee_offset and
23399 final_offset from HOST_WIDE_INT to poly_int64.
23400 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
23401 to_constant when getting the number of units in an Advanced SIMD
23403 (aarch64_builtin_vectorized_function): Check for a constant number
23405 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
23407 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
23408 attribute instead of GET_MODE_NUNITS.
23409 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
23410 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
23411 GET_MODE_SIZE for fixed-size registers.
23412 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
23413 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
23414 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
23415 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
23416 (aarch64_print_operand, aarch64_print_address_internal)
23417 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
23418 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
23419 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
23420 Handle polynomial GET_MODE_SIZE.
23421 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
23422 wider than SImode without modification.
23423 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
23424 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
23425 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
23426 passing and returning SVE modes.
23427 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
23428 rather than GEN_INT.
23429 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
23430 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
23431 (aarch64_allocate_and_probe_stack_space): Likewise.
23432 (aarch64_layout_frame): Cope with polynomial offsets.
23433 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
23434 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
23435 polynomial offsets.
23436 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
23437 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
23438 poly_int64 rather than a HOST_WIDE_INT.
23439 (aarch64_get_separate_components, aarch64_process_components)
23440 (aarch64_expand_prologue, aarch64_expand_epilogue)
23441 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
23442 (aarch64_anchor_offset): New function, split out from...
23443 (aarch64_legitimize_address): ...here.
23444 (aarch64_builtin_vectorization_cost): Handle polynomial
23445 TYPE_VECTOR_SUBPARTS.
23446 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
23448 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
23449 number of elements from the PARALLEL rather than the mode.
23450 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
23451 rather than GET_MODE_BITSIZE.
23452 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
23453 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
23454 (aarch64_expand_vec_perm_const_1): Handle polynomial
23455 d->perm.length () and d->perm elements.
23456 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
23457 Apply to_constant to d->perm elements.
23458 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
23459 polynomial CONST_VECTOR_NUNITS.
23460 (aarch64_move_pointer): Take amount as a poly_int64 rather
23462 (aarch64_progress_pointer): Avoid temporary variable.
23463 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
23464 the mode attribute instead of GET_MODE.
23466 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23467 Alan Hayward <alan.hayward@arm.com>
23468 David Sherwood <david.sherwood@arm.com>
23470 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
23471 x exists before using it.
23472 (aarch64_add_constant_internal): Rename to...
23473 (aarch64_add_offset_1): ...this. Replace regnum with separate
23474 src and dest rtxes. Handle the case in which they're different,
23475 including when the offset is zero. Replace scratchreg with an rtx.
23476 Use 2 additions if there is no spare register into which we can
23477 move a 16-bit constant.
23478 (aarch64_add_constant): Delete.
23479 (aarch64_add_offset): Replace reg with separate src and dest
23480 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
23481 Use aarch64_add_offset_1.
23482 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
23483 an rtx rather than an int. Take the delta as a poly_int64
23484 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
23485 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
23486 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
23487 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
23488 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
23489 and aarch64_add_sp.
23490 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
23491 aarch64_add_constant.
23493 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23495 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
23496 Use scalar_float_mode.
23498 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23500 * config/aarch64/aarch64-simd.md
23501 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
23502 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
23503 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
23504 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
23505 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
23506 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
23507 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
23508 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
23509 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
23510 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
23512 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23515 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
23516 targ_options->x_arm_arch_string is non NULL.
23518 2018-01-11 Tamar Christina <tamar.christina@arm.com>
23520 * config/aarch64/aarch64.h
23521 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
23523 2018-01-11 Sudakshina Das <sudi.das@arm.com>
23526 * expmed.c (emit_store_flag_force): Swap if const op0
23527 and change VOIDmode to mode of op0.
23529 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23531 PR rtl-optimization/83761
23532 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
23533 than bytes to mode_for_size.
23535 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23537 PR middle-end/83189
23538 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
23539 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
23542 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23544 PR middle-end/83575
23545 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
23546 when in layout mode.
23547 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
23548 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
23551 2018-01-10 Michael Collison <michael.collison@arm.com>
23553 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
23554 * config/aarch64/aarch64-option-extension.def: Add
23555 AARCH64_OPT_EXTENSION of 'fp16fml'.
23556 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23557 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
23558 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
23559 * config/aarch64/constraints.md (Ui7): New constraint.
23560 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
23561 (VFMLA_SEL_W): Ditto.
23564 (VFMLA16_LOW): New int iterator.
23565 (VFMLA16_HIGH): Ditto.
23566 (UNSPEC_FMLAL): New unspec.
23567 (UNSPEC_FMLSL): Ditto.
23568 (UNSPEC_FMLAL2): Ditto.
23569 (UNSPEC_FMLSL2): Ditto.
23570 (f16mac): New code attribute.
23571 * config/aarch64/aarch64-simd-builtins.def
23572 (aarch64_fmlal_lowv2sf): Ditto.
23573 (aarch64_fmlsl_lowv2sf): Ditto.
23574 (aarch64_fmlalq_lowv4sf): Ditto.
23575 (aarch64_fmlslq_lowv4sf): Ditto.
23576 (aarch64_fmlal_highv2sf): Ditto.
23577 (aarch64_fmlsl_highv2sf): Ditto.
23578 (aarch64_fmlalq_highv4sf): Ditto.
23579 (aarch64_fmlslq_highv4sf): Ditto.
23580 (aarch64_fmlal_lane_lowv2sf): Ditto.
23581 (aarch64_fmlsl_lane_lowv2sf): Ditto.
23582 (aarch64_fmlal_laneq_lowv2sf): Ditto.
23583 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
23584 (aarch64_fmlalq_lane_lowv4sf): Ditto.
23585 (aarch64_fmlsl_lane_lowv4sf): Ditto.
23586 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
23587 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
23588 (aarch64_fmlal_lane_highv2sf): Ditto.
23589 (aarch64_fmlsl_lane_highv2sf): Ditto.
23590 (aarch64_fmlal_laneq_highv2sf): Ditto.
23591 (aarch64_fmlsl_laneq_highv2sf): Ditto.
23592 (aarch64_fmlalq_lane_highv4sf): Ditto.
23593 (aarch64_fmlsl_lane_highv4sf): Ditto.
23594 (aarch64_fmlalq_laneq_highv4sf): Ditto.
23595 (aarch64_fmlsl_laneq_highv4sf): Ditto.
23596 * config/aarch64/aarch64-simd.md:
23597 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
23598 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23599 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
23600 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23601 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
23602 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
23603 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
23604 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
23605 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
23606 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
23607 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
23608 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
23609 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
23610 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
23611 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
23612 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
23613 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
23614 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
23615 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
23616 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
23617 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
23618 (vfmlsl_low_u32): Ditto.
23619 (vfmlalq_low_u32): Ditto.
23620 (vfmlslq_low_u32): Ditto.
23621 (vfmlal_high_u32): Ditto.
23622 (vfmlsl_high_u32): Ditto.
23623 (vfmlalq_high_u32): Ditto.
23624 (vfmlslq_high_u32): Ditto.
23625 (vfmlal_lane_low_u32): Ditto.
23626 (vfmlsl_lane_low_u32): Ditto.
23627 (vfmlal_laneq_low_u32): Ditto.
23628 (vfmlsl_laneq_low_u32): Ditto.
23629 (vfmlalq_lane_low_u32): Ditto.
23630 (vfmlslq_lane_low_u32): Ditto.
23631 (vfmlalq_laneq_low_u32): Ditto.
23632 (vfmlslq_laneq_low_u32): Ditto.
23633 (vfmlal_lane_high_u32): Ditto.
23634 (vfmlsl_lane_high_u32): Ditto.
23635 (vfmlal_laneq_high_u32): Ditto.
23636 (vfmlsl_laneq_high_u32): Ditto.
23637 (vfmlalq_lane_high_u32): Ditto.
23638 (vfmlslq_lane_high_u32): Ditto.
23639 (vfmlalq_laneq_high_u32): Ditto.
23640 (vfmlslq_laneq_high_u32): Ditto.
23641 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
23642 (AARCH64_FL_FOR_ARCH8_4): New.
23643 (AARCH64_ISA_F16FML): New ISA flag.
23644 (TARGET_F16FML): New feature flag for fp16fml.
23645 (doc/invoke.texi): Document new fp16fml option.
23647 2018-01-10 Michael Collison <michael.collison@arm.com>
23649 * config/aarch64/aarch64-builtins.c:
23650 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
23651 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23652 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
23653 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
23654 (AARCH64_ISA_SHA3): New ISA flag.
23655 (TARGET_SHA3): New feature flag for sha3.
23656 * config/aarch64/iterators.md (sha512_op): New int attribute.
23657 (CRYPTO_SHA512): New int iterator.
23658 (UNSPEC_SHA512H): New unspec.
23659 (UNSPEC_SHA512H2): Ditto.
23660 (UNSPEC_SHA512SU0): Ditto.
23661 (UNSPEC_SHA512SU1): Ditto.
23662 * config/aarch64/aarch64-simd-builtins.def
23663 (aarch64_crypto_sha512hqv2di): New builtin.
23664 (aarch64_crypto_sha512h2qv2di): Ditto.
23665 (aarch64_crypto_sha512su0qv2di): Ditto.
23666 (aarch64_crypto_sha512su1qv2di): Ditto.
23667 (aarch64_eor3qv8hi): Ditto.
23668 (aarch64_rax1qv2di): Ditto.
23669 (aarch64_xarqv2di): Ditto.
23670 (aarch64_bcaxqv8hi): Ditto.
23671 * config/aarch64/aarch64-simd.md:
23672 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
23673 (aarch64_crypto_sha512su0qv2di): Ditto.
23674 (aarch64_crypto_sha512su1qv2di): Ditto.
23675 (aarch64_eor3qv8hi): Ditto.
23676 (aarch64_rax1qv2di): Ditto.
23677 (aarch64_xarqv2di): Ditto.
23678 (aarch64_bcaxqv8hi): Ditto.
23679 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
23680 (vsha512h2q_u64): Ditto.
23681 (vsha512su0q_u64): Ditto.
23682 (vsha512su1q_u64): Ditto.
23683 (veor3q_u16): Ditto.
23684 (vrax1q_u64): Ditto.
23685 (vxarq_u64): Ditto.
23686 (vbcaxq_u16): Ditto.
23687 * config/arm/types.md (crypto_sha512): New type attribute.
23688 (crypto_sha3): Ditto.
23689 (doc/invoke.texi): Document new sha3 option.
23691 2018-01-10 Michael Collison <michael.collison@arm.com>
23693 * config/aarch64/aarch64-builtins.c:
23694 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
23695 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23696 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
23697 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
23698 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
23699 (AARCH64_ISA_SM4): New ISA flag.
23700 (TARGET_SM4): New feature flag for sm4.
23701 * config/aarch64/aarch64-simd-builtins.def
23702 (aarch64_sm3ss1qv4si): Ditto.
23703 (aarch64_sm3tt1aq4si): Ditto.
23704 (aarch64_sm3tt1bq4si): Ditto.
23705 (aarch64_sm3tt2aq4si): Ditto.
23706 (aarch64_sm3tt2bq4si): Ditto.
23707 (aarch64_sm3partw1qv4si): Ditto.
23708 (aarch64_sm3partw2qv4si): Ditto.
23709 (aarch64_sm4eqv4si): Ditto.
23710 (aarch64_sm4ekeyqv4si): Ditto.
23711 * config/aarch64/aarch64-simd.md:
23712 (aarch64_sm3ss1qv4si): Ditto.
23713 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
23714 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
23715 (aarch64_sm4eqv4si): Ditto.
23716 (aarch64_sm4ekeyqv4si): Ditto.
23717 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
23718 (sm3part_op): Ditto.
23719 (CRYPTO_SM3TT): Ditto.
23720 (CRYPTO_SM3PART): Ditto.
23721 (UNSPEC_SM3SS1): New unspec.
23722 (UNSPEC_SM3TT1A): Ditto.
23723 (UNSPEC_SM3TT1B): Ditto.
23724 (UNSPEC_SM3TT2A): Ditto.
23725 (UNSPEC_SM3TT2B): Ditto.
23726 (UNSPEC_SM3PARTW1): Ditto.
23727 (UNSPEC_SM3PARTW2): Ditto.
23728 (UNSPEC_SM4E): Ditto.
23729 (UNSPEC_SM4EKEY): Ditto.
23730 * config/aarch64/constraints.md (Ui2): New constraint.
23731 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
23732 * config/arm/types.md (crypto_sm3): New type attribute.
23733 (crypto_sm4): Ditto.
23734 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
23735 (vsm3tt1aq_u32): Ditto.
23736 (vsm3tt1bq_u32): Ditto.
23737 (vsm3tt2aq_u32): Ditto.
23738 (vsm3tt2bq_u32): Ditto.
23739 (vsm3partw1q_u32): Ditto.
23740 (vsm3partw2q_u32): Ditto.
23741 (vsm4eq_u32): Ditto.
23742 (vsm4ekeyq_u32): Ditto.
23743 (doc/invoke.texi): Document new sm4 option.
23745 2018-01-10 Michael Collison <michael.collison@arm.com>
23747 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
23748 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
23749 (AARCH64_FL_FOR_ARCH8_4): New.
23750 (AARCH64_FL_V8_4): New flag.
23751 (doc/invoke.texi): Document new armv8.4-a option.
23753 2018-01-10 Michael Collison <michael.collison@arm.com>
23755 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23756 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
23757 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
23758 * config/aarch64/aarch64-option-extension.def: Add
23759 AARCH64_OPT_EXTENSION of 'sha2'.
23760 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
23761 (crypto): Disable sha2 and aes if crypto disabled.
23762 (crypto): Enable aes and sha2 if enabled.
23763 (simd): Disable sha2 and aes if simd disabled.
23764 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
23766 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
23767 (TARGET_SHA2): New feature flag for sha2.
23768 (TARGET_AES): New feature flag for aes.
23769 * config/aarch64/aarch64-simd.md:
23770 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
23771 conditional on TARGET_AES.
23772 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
23773 (aarch64_crypto_sha1hsi): Make pattern conditional
23775 (aarch64_crypto_sha1hv4si): Ditto.
23776 (aarch64_be_crypto_sha1hv4si): Ditto.
23777 (aarch64_crypto_sha1su1v4si): Ditto.
23778 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
23779 (aarch64_crypto_sha1su0v4si): Ditto.
23780 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
23781 (aarch64_crypto_sha256su0v4si): Ditto.
23782 (aarch64_crypto_sha256su1v4si): Ditto.
23783 (doc/invoke.texi): Document new aes and sha2 options.
23785 2018-01-10 Martin Sebor <msebor@redhat.com>
23787 PR tree-optimization/83781
23788 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
23791 2018-01-11 Martin Sebor <msebor@gmail.com>
23792 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23794 PR tree-optimization/83501
23795 PR tree-optimization/81703
23797 * tree-ssa-strlen.c (get_string_cst): Rename...
23798 (get_string_len): ...to this. Handle global constants.
23799 (handle_char_store): Adjust.
23801 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
23802 Jim Wilson <jimw@sifive.com>
23804 * config/riscv/riscv-protos.h (riscv_output_return): New.
23805 * config/riscv/riscv.c (struct machine_function): New naked_p field.
23806 (riscv_attribute_table, riscv_output_return),
23807 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
23808 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
23809 (riscv_compute_frame_info): Only compute frame->mask if not a naked
23811 (riscv_expand_prologue): Add early return for naked function.
23812 (riscv_expand_epilogue): Likewise.
23813 (riscv_function_ok_for_sibcall): Return false for naked function.
23814 (riscv_set_current_function): New.
23815 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
23816 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
23817 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
23818 * doc/extend.texi (RISC-V Function Attributes): New.
23820 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
23822 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
23823 check for 128-bit long double before checking TCmode.
23824 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
23825 128-bit long doubles before checking TFmode or TCmode.
23826 (FLOAT128_IBM_P): Likewise.
23828 2018-01-10 Martin Sebor <msebor@redhat.com>
23830 PR tree-optimization/83671
23831 * builtins.c (c_strlen): Unconditionally return zero for the empty
23833 Use -Warray-bounds for warnings.
23834 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
23835 for non-constant array indices with COMPONENT_REF, arrays of
23836 arrays, and pointers to arrays.
23837 (gimple_fold_builtin_strlen): Determine and set length range for
23838 non-constant character arrays.
23840 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
23842 PR middle-end/81897
23843 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
23846 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
23848 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
23850 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
23853 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
23854 VECTOR_MEM_ALTIVEC_OR_VSX_P.
23855 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
23856 indexed_or_indirect_operand predicate.
23857 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
23858 (*vsx_le_perm_load_v8hi): Likewise.
23859 (*vsx_le_perm_load_v16qi): Likewise.
23860 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
23861 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
23862 (*vsx_le_perm_store_v8hi): Likewise.
23863 (*vsx_le_perm_store_v16qi): Likewise.
23864 (eight unnamed splitters): Likewise.
23866 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
23868 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
23869 * config/rs6000/emmintrin.h: Likewise.
23870 * config/rs6000/mmintrin.h: Likewise.
23871 * config/rs6000/xmmintrin.h: Likewise.
23873 2018-01-10 David Malcolm <dmalcolm@redhat.com>
23876 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
23878 * tree.c (tree_nop_conversion): Return true for location wrapper
23880 (maybe_wrap_with_location): New function.
23881 (selftest::check_strip_nops): New function.
23882 (selftest::test_location_wrappers): New function.
23883 (selftest::tree_c_tests): Call it.
23884 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
23885 (maybe_wrap_with_location): New decl.
23886 (EXPR_LOCATION_WRAPPER_P): New macro.
23887 (location_wrapper_p): New inline function.
23888 (tree_strip_any_location_wrapper): New inline function.
23890 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
23893 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
23894 stack_realign_offset for the largest alignment of stack slot
23896 (ix86_find_max_used_stack_alignment): New function.
23897 (ix86_finalize_stack_frame_flags): Use it. Set
23898 max_used_stack_alignment if we don't realign stack.
23899 * config/i386/i386.h (machine_function): Add
23900 max_used_stack_alignment.
23902 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
23904 * config/arm/arm.opt (-mbranch-cost): New option.
23905 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
23908 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
23911 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
23912 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
23914 2018-01-10 Richard Biener <rguenther@suse.de>
23917 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
23918 early out so it also covers the case where we have a non-NULL
23921 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
23923 PR tree-optimization/83753
23924 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
23925 for non-strided grouped accesses if the number of elements is 1.
23927 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23930 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
23931 * i386.h (TARGET_USE_GATHER): Define.
23932 * x86-tune.def (X86_TUNE_USE_GATHER): New.
23934 2018-01-10 Martin Liska <mliska@suse.cz>
23937 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
23938 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
23940 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
23941 CLEANUP_NO_PARTITIONING is not set.
23943 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
23945 * doc/rtl.texi: Remove documentation of (const ...) wrappers
23946 for vectors, as a partial revert of r254296.
23947 * rtl.h (const_vec_p): Delete.
23948 (const_vec_duplicate_p): Don't test for vector CONSTs.
23949 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
23950 * expmed.c (make_tree): Likewise.
23953 * common.md (E, F): Use CONSTANT_P instead of checking for
23955 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
23956 checking for CONST_VECTOR.
23958 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
23960 PR middle-end/83575
23961 * predict.c (force_edge_cold): Handle in more sane way edges
23962 with no prediction.
23964 2018-01-09 Carl Love <cel@us.ibm.com>
23966 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
23968 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
23969 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
23970 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
23971 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
23972 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
23973 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
23974 * config/rs6000/rs6000-protos.h: Add extern defition for
23975 rs6000_generate_float2_double_code.
23976 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
23978 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
23979 (float2_v2df): Add define_expand.
23981 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
23984 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
23985 op_mode in the force_to_mode call.
23987 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
23989 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
23990 instead of checking each element individually.
23991 (aarch64_evpc_uzp): Likewise.
23992 (aarch64_evpc_zip): Likewise.
23993 (aarch64_evpc_ext): Likewise.
23994 (aarch64_evpc_rev): Likewise.
23995 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
23996 instead of checking each element individually. Return true without
23998 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
23999 whether all selected elements come from the same input, instead of
24000 checking each element individually. Remove calls to gen_rtx_REG,
24001 start_sequence and end_sequence and instead assert that no rtl is
24004 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
24006 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
24007 order of HIGH and CONST checks.
24009 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
24011 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
24012 if the destination isn't an SSA_NAME.
24014 2018-01-09 Richard Biener <rguenther@suse.de>
24016 PR tree-optimization/83668
24017 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
24019 (canonicalize_loop_form): ... here, renamed from ...
24020 (canonicalize_loop_closed_ssa_form): ... this and amended to
24021 swap successor edges for loop exit blocks to make us use
24022 the RPO order we need for initial schedule generation.
24024 2018-01-09 Joseph Myers <joseph@codesourcery.com>
24026 PR tree-optimization/64811
24027 * match.pd: When optimizing comparisons with Inf, avoid
24028 introducing or losing exceptions from comparisons with NaN.
24030 2018-01-09 Martin Liska <mliska@suse.cz>
24033 * asan.c (shadow_mem_size): Add gcc_assert.
24035 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
24037 Don't save registers in main().
24040 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
24041 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
24042 * config/avr/avr.c (avr_set_current_function): Don't error if
24043 naked, OS_task or OS_main are specified at the same time.
24044 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
24046 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
24048 * common/config/avr/avr-common.c (avr_option_optimization_table):
24049 Switch on -mmain-is-OS_task for optimizing compilations.
24051 2018-01-09 Richard Biener <rguenther@suse.de>
24053 PR tree-optimization/83572
24054 * graphite.c: Include cfganal.h.
24055 (graphite_transform_loops): Connect infinite loops to exit
24056 and remove fake edges at the end.
24058 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
24060 * ipa-inline.c (edge_badness): Revert accidental checkin.
24062 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
24065 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
24066 symbols; not inline clones.
24068 2018-01-09 Jakub Jelinek <jakub@redhat.com>
24071 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
24072 hard registers. Formatting fixes.
24074 PR preprocessor/83722
24075 * gcc.c (try_generate_repro): Pass
24076 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
24077 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
24080 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
24081 Kito Cheng <kito.cheng@gmail.com>
24083 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
24084 (riscv_leaf_function_p): Delete.
24085 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
24087 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24089 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
24091 (do_ifelse): New function.
24092 (do_isel): New function.
24093 (do_sub3): New function.
24094 (do_add3): New function.
24095 (do_load_mask_compare): New function.
24096 (do_overlap_load_compare): New function.
24097 (expand_compare_loop): New function.
24098 (expand_block_compare): Call expand_compare_loop() when appropriate.
24099 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
24100 option description.
24101 (-mblock-compare-inline-loop-limit): New option.
24103 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24106 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
24107 Reverse order of second and third operands in first alternative.
24108 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
24109 of first and second elements in UNSPEC_VPERMR vector.
24110 (altivec_expand_vec_perm_le): Likewise.
24112 2018-01-08 Jeff Law <law@redhat.com>
24114 PR rtl-optimizatin/81308
24115 * tree-switch-conversion.c (cfg_altered): New file scoped static.
24116 (process_switch): If group_case_labels makes a change, then set
24118 (pass_convert_switch::execute): If a switch is converted, then
24119 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
24121 PR rtl-optimization/81308
24122 * recog.c (split_all_insns): Conditionally cleanup the CFG after
24125 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
24127 PR target/83663 - Revert r255946
24128 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
24129 generation for cases where splatting a value is not useful.
24130 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
24131 across a vec_duplicate and a paradoxical subreg forming a vector
24132 mode to a vec_concat.
24134 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24136 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
24137 -march=armv8.3-a variants.
24138 * config/arm/t-multilib: Likewise.
24139 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
24141 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24143 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
24145 (cceq_ior_compare_complement): Give it a name so I can use it, and
24146 change boolean_or_operator predicate to boolean_operator so it can
24147 be used to generate a crand.
24148 (eqne): New code iterator.
24149 (bd/bd_neg): New code_attrs.
24150 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
24151 a single define_insn.
24152 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
24153 decrement (bdnzt/bdnzf/bdzt/bdzf).
24154 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
24155 with the new names of the branch decrement patterns, and added the
24156 names of the branch decrement conditional patterns.
24158 2018-01-08 Richard Biener <rguenther@suse.de>
24160 PR tree-optimization/83563
24161 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
24164 2018-01-08 Richard Biener <rguenther@suse.de>
24166 PR middle-end/83713
24167 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
24169 2018-01-08 Richard Biener <rguenther@suse.de>
24171 PR tree-optimization/83685
24172 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
24173 references to abnormals.
24175 2018-01-08 Richard Biener <rguenther@suse.de>
24178 * dwarf2out.c (output_indirect_strings): Handle empty
24179 skeleton_debug_str_hash.
24180 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
24182 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
24184 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
24185 (emit_store_direct): Likewise.
24186 (arc_trampoline_adjust_address): Likewise.
24187 (arc_asm_trampoline_template): New function.
24188 (arc_initialize_trampoline): Use asm_trampoline_template.
24189 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
24190 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
24191 * config/arc/arc.md (flush_icache): Delete pattern.
24193 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
24195 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
24196 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
24199 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
24202 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
24203 by not USED_FOR_TARGET.
24204 (make_pass_resolve_sw_modes): Likewise.
24206 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
24208 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
24211 2018-01-08 Richard Biener <rguenther@suse.de>
24213 PR middle-end/83580
24214 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
24216 2018-01-08 Richard Biener <rguenther@suse.de>
24218 PR middle-end/83517
24219 * match.pd ((t * 2) / 2) -> t): Add missing :c.
24221 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
24223 PR middle-end/81897
24224 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
24225 basic blocks with a small number of successors.
24226 (convert_control_dep_chain_into_preds): Improve handling of
24228 (dump_predicates): Split apart into...
24229 (dump_pred_chain): ...here...
24230 (dump_pred_info): ...and here.
24231 (can_one_predicate_be_invalidated_p): Add debugging printfs.
24232 (can_chain_union_be_invalidated_p): Improve check for invalidation
24234 (uninit_uses_cannot_happen): Avoid unnecessary if
24235 convert_control_dep_chain_into_preds yielded nothing.
24237 2018-01-06 Martin Sebor <msebor@redhat.com>
24239 PR tree-optimization/83640
24240 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
24241 subtracting negative offset from size.
24242 (builtin_access::overlap): Adjust offset bounds of the access to fall
24243 within the size of the object if possible.
24245 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
24247 PR rtl-optimization/83699
24248 * expmed.c (extract_bit_field_1): Restrict the vector usage of
24249 extract_bit_field_as_subreg to cases in which the extracted
24250 value is also a vector.
24252 * lra-constraints.c (process_alt_operands): Test for the equivalence
24253 substitutions when detecting a possible reload cycle.
24255 2018-01-06 Jakub Jelinek <jakub@redhat.com>
24258 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
24259 by default if flag_selective_schedling{,2}. Formatting fixes.
24261 PR rtl-optimization/83682
24262 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
24263 if it has non-VECTOR_MODE element mode.
24264 (vec_duplicate_p): Likewise.
24266 PR middle-end/83694
24267 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
24268 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
24270 2018-01-05 Jakub Jelinek <jakub@redhat.com>
24273 * config/i386/i386-builtin.def
24274 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
24275 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
24276 Require also OPTION_MASK_ISA_AVX512F in addition to
24277 OPTION_MASK_ISA_GFNI.
24278 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
24279 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
24280 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
24281 to OPTION_MASK_ISA_GFNI.
24282 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
24283 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
24284 OPTION_MASK_ISA_AVX512BW.
24285 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
24286 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
24287 addition to OPTION_MASK_ISA_GFNI.
24288 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
24289 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
24290 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
24291 to OPTION_MASK_ISA_GFNI.
24292 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
24293 a requirement for all ISAs rather than any of them with a few
24295 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
24297 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
24298 bitmasks to be enabled with 3 exceptions, instead of requiring any
24299 enabled ISA with lots of exceptions.
24300 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
24301 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
24302 Change avx512bw in isa attribute to avx512f.
24303 * config/i386/sgxintrin.h: Add license boilerplate.
24304 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
24305 to __AVX512F__ and __AVX512VL to __AVX512VL__.
24306 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
24307 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
24309 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
24310 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
24311 temporarily sse2 rather than sse if not enabled already.
24314 * config/i386/sse.md (VI248_VLBW): Rename to ...
24315 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
24316 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
24317 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
24318 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
24319 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
24320 mode iterator instead of VI248_VLBW.
24322 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
24324 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
24325 (record_modified): Skip clobbers; add debug output.
24326 (param_change_prob): Use sreal frequencies.
24328 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
24330 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
24331 punt for user-aligned variables.
24333 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
24335 * tree-chrec.c (chrec_contains_symbols): Return true for
24338 2018-01-05 Sudakshina Das <sudi.das@arm.com>
24341 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
24342 of (x|y) == x for BICS pattern.
24344 2018-01-05 Jakub Jelinek <jakub@redhat.com>
24346 PR tree-optimization/83605
24347 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
24348 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
24351 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
24353 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
24354 * config/epiphany/rtems.h: New file.
24356 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24357 Uros Bizjak <ubizjak@gmail.com>
24360 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
24361 QIreg_operand instead of register_operand predicate.
24362 * config/i386/i386.c (ix86_rop_should_change_byte_p,
24363 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
24364 comments instead of -fmitigate[-_]rop.
24366 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24369 * cgraphunit.c (symbol_table::compile): Switch to text_section
24370 before calling assembly_start debug hook.
24371 * run-rtl-passes.c (run_rtl_passes): Likewise.
24374 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24376 * tree-vrp.c (extract_range_from_binary_expr_1): Check
24377 range_int_cst_p rather than !symbolic_range_p before calling
24378 extract_range_from_multiplicative_op_1.
24380 2018-01-04 Jeff Law <law@redhat.com>
24382 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
24383 redundant test in assertion.
24385 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24387 * doc/rtl.texi: Document machine_mode wrapper classes.
24389 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24391 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
24392 using tree_to_uhwi.
24394 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24396 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
24397 the VEC_PERM_EXPR fold to fail.
24399 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24402 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
24403 to switched_sections.
24405 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24408 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
24409 test for d.testing.
24411 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
24414 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
24415 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
24417 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24420 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
24421 is BLKmode and bitpos not zero or mode change is needed.
24423 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24426 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
24429 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
24432 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
24433 instead of MULT rtx. Update all corresponding splitters.
24434 (*saddl_se): Ditto.
24435 (*ssub<modesuffix>): Ditto.
24436 (*ssubl_se): Ditto.
24437 (*cmp_sadd_di): Update split patterns.
24438 (*cmp_sadd_si): Ditto.
24439 (*cmp_sadd_sidi): Ditto.
24440 (*cmp_ssub_di): Ditto.
24441 (*cmp_ssub_si): Ditto.
24442 (*cmp_ssub_sidi): Ditto.
24443 * config/alpha/predicates.md (const23_operand): New predicate.
24444 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
24445 Look for ASHIFT, not MULT inner operand.
24446 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
24448 2018-01-04 Martin Liska <mliska@suse.cz>
24450 PR gcov-profile/83669
24451 * gcov.c (output_intermediate_file): Add version to intermediate
24453 * doc/gcov.texi: Document new field 'version' in intermediate
24454 file format. Fix location of '-k' option of gcov command.
24456 2018-01-04 Martin Liska <mliska@suse.cz>
24459 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
24461 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24463 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
24465 2018-01-03 Martin Sebor <msebor@redhat.com>
24467 PR tree-optimization/83655
24468 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
24469 checking calls with invalid arguments.
24471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24473 * tree-vect-stmts.c (vect_get_store_rhs): New function.
24474 (vectorizable_mask_load_store): Delete.
24475 (vectorizable_call): Return false for masked loads and stores.
24476 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
24477 instead of gimple_assign_rhs1.
24478 (vectorizable_load): Handle IFN_MASK_LOAD.
24479 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
24481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24483 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
24485 (vectorizable_mask_load_store): ...here.
24486 (vectorizable_load): ...and here.
24488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24490 * tree-vect-stmts.c (vect_build_all_ones_mask)
24491 (vect_build_zero_merge_argument): New functions, split out from...
24492 (vectorizable_load): ...here.
24494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24496 * tree-vect-stmts.c (vect_check_store_rhs): New function,
24498 (vectorizable_mask_load_store): ...here.
24499 (vectorizable_store): ...and here.
24501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24503 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
24505 (vectorizable_mask_load_store): ...here.
24507 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24509 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
24510 (vect_model_store_cost): Take a vec_load_store_type instead of a
24512 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
24513 (vect_model_store_cost): Take a vec_load_store_type instead of a
24515 (vectorizable_mask_load_store): Update accordingly.
24516 (vectorizable_store): Likewise.
24517 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
24519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24521 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
24522 IFN_MASK_LOAD calls here rather than...
24523 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
24525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24526 Alan Hayward <alan.hayward@arm.com>
24527 David Sherwood <david.sherwood@arm.com>
24529 * expmed.c (extract_bit_field_1): For vector extracts,
24530 fall back to extract_bit_field_as_subreg if vec_extract
24533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24534 Alan Hayward <alan.hayward@arm.com>
24535 David Sherwood <david.sherwood@arm.com>
24537 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
24538 they are variable or constant sized.
24539 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
24540 slots for constant-sized data.
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 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
24547 handling COND_EXPRs with boolean comparisons, try to find a better
24548 basis for the mask type than the boolean itself.
24550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24552 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
24553 is calculated and how it can be overridden.
24554 * genmodes.c (max_bitsize_mode_any_mode): New variable.
24555 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
24557 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
24560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24561 Alan Hayward <alan.hayward@arm.com>
24562 David Sherwood <david.sherwood@arm.com>
24564 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
24565 Remove the mode argument.
24566 (aarch64_simd_valid_immediate): Remove the mode and inverse
24568 * config/aarch64/iterators.md (bitsize): New iterator.
24569 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
24570 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
24571 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
24572 aarch64_simd_valid_immediate.
24573 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
24574 (aarch64_reg_or_bic_imm): Likewise.
24575 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
24576 with an insn_type enum and msl with a modifier_type enum.
24577 Replace element_width with a scalar_mode. Change the shift
24578 to unsigned int. Add constructors for scalar_float_mode and
24579 scalar_int_mode elements.
24580 (aarch64_vect_float_const_representable_p): Delete.
24581 (aarch64_can_const_movi_rtx_p)
24582 (aarch64_simd_scalar_immediate_valid_for_move)
24583 (aarch64_simd_make_constant): Update call to
24584 aarch64_simd_valid_immediate.
24585 (aarch64_advsimd_valid_immediate_hs): New function.
24586 (aarch64_advsimd_valid_immediate): Likewise.
24587 (aarch64_simd_valid_immediate): Remove mode and inverse
24588 arguments. Rewrite to use the above. Use const_vec_duplicate_p
24589 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
24590 and aarch64_float_const_representable_p on the result.
24591 (aarch64_output_simd_mov_immediate): Remove mode argument.
24592 Update call to aarch64_simd_valid_immediate and use of
24593 simd_immediate_info.
24594 (aarch64_output_scalar_simd_mov_immediate): Update call
24597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24598 Alan Hayward <alan.hayward@arm.com>
24599 David Sherwood <david.sherwood@arm.com>
24601 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
24602 (mode_nunits): Likewise CONST_MODE_NUNITS.
24603 * machmode.def (ADJUST_NUNITS): Document.
24604 * genmodes.c (mode_data::need_nunits_adj): New field.
24605 (blank_mode): Update accordingly.
24606 (adj_nunits): New variable.
24607 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
24609 (emit_mode_size_inline): Set need_bytesize_adj for all modes
24610 listed in adj_nunits.
24611 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
24612 listed in adj_nunits. Don't emit case statements for such modes.
24613 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
24614 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
24615 nothing if adj_nunits is nonnull.
24616 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
24617 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
24618 (emit_mode_fbit): Update use of print_maybe_const_decl.
24619 (emit_move_size): Likewise. Treat the array as non-const
24621 (emit_mode_adjustments): Handle adj_nunits.
24623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24625 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
24626 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
24627 (VECTOR_MODES): Use it.
24628 (make_vector_modes): Take the prefix as an argument.
24630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24631 Alan Hayward <alan.hayward@arm.com>
24632 David Sherwood <david.sherwood@arm.com>
24634 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
24635 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
24636 for MODE_VECTOR_BOOL.
24637 * machmode.def (VECTOR_BOOL_MODE): Document.
24638 * genmodes.c (VECTOR_BOOL_MODE): New macro.
24639 (make_vector_bool_mode): New function.
24640 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
24642 * lto-streamer-in.c (lto_input_mode_table): Likewise.
24643 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
24645 * stor-layout.c (int_mode_for_mode): Likewise.
24646 * tree.c (build_vector_type_for_mode): Likewise.
24647 * varasm.c (output_constant_pool_2): Likewise.
24648 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
24649 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
24650 for MODE_VECTOR_BOOL.
24651 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
24652 of mode class checks.
24653 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
24654 instead of a list of mode class checks.
24655 (expand_vector_scalar_condition): Likewise.
24656 (type_for_widest_vector_mode): Handle BImode as an inner mode.
24658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24659 Alan Hayward <alan.hayward@arm.com>
24660 David Sherwood <david.sherwood@arm.com>
24662 * machmode.h (mode_size): Change from unsigned short to
24664 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
24665 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24666 or if measurement_type is not polynomial.
24667 (fixed_size_mode::includes_p): Check for constant-sized modes.
24668 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
24669 return a poly_uint16 rather than an unsigned short.
24670 (emit_mode_size): Change the type of mode_size from unsigned short
24671 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
24672 (emit_mode_adjustments): Cope with polynomial vector sizes.
24673 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24675 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24677 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
24678 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
24679 * caller-save.c (setup_save_areas): Likewise.
24680 (replace_reg_with_saved_mem): Likewise.
24681 * calls.c (emit_library_call_value_1): Likewise.
24682 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
24683 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
24684 (gen_lowpart_for_combine): Likewise.
24685 * convert.c (convert_to_integer_1): Likewise.
24686 * cse.c (equiv_constant, cse_insn): Likewise.
24687 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
24688 (cselib_subst_to_values): Likewise.
24689 * dce.c (word_dce_process_block): Likewise.
24690 * df-problems.c (df_word_lr_mark_ref): Likewise.
24691 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
24692 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
24693 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
24694 (rtl_for_decl_location): Likewise.
24695 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
24696 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
24697 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
24698 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
24699 (expand_expr_real_1): Likewise.
24700 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
24701 (pad_below): Likewise.
24702 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24703 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
24704 * ira.c (get_subreg_tracking_sizes): Likewise.
24705 * ira-build.c (ira_create_allocno_objects): Likewise.
24706 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
24707 (ira_sort_regnos_for_alter_reg): Likewise.
24708 * ira-costs.c (record_operand_costs): Likewise.
24709 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
24710 (resolve_simple_move): Likewise.
24711 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
24712 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
24713 (lra_constraints): Likewise.
24714 (CONST_POOL_OK_P): Reject variable-sized modes.
24715 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
24716 (add_pseudo_to_slot, lra_spill): Likewise.
24717 * omp-low.c (omp_clause_aligned_alignment): Likewise.
24718 * optabs-query.c (get_best_extraction_insn): Likewise.
24719 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24720 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
24721 (expand_mult_highpart, valid_multiword_target_p): Likewise.
24722 * recog.c (offsettable_address_addr_space_p): Likewise.
24723 * regcprop.c (maybe_mode_change): Likewise.
24724 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
24725 * regrename.c (build_def_use): Likewise.
24726 * regstat.c (dump_reg_info): Likewise.
24727 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
24728 (find_reloads, find_reloads_subreg_address): Likewise.
24729 * reload1.c (eliminate_regs_1): Likewise.
24730 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
24731 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
24732 (simplify_binary_operation_1, simplify_subreg): Likewise.
24733 * targhooks.c (default_function_arg_padding): Likewise.
24734 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
24735 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
24736 (verify_gimple_assign_ternary): Likewise.
24737 * tree-inline.c (estimate_move_cost): Likewise.
24738 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24739 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
24740 (get_address_cost_ainc): Likewise.
24741 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
24742 (vect_supportable_dr_alignment): Likewise.
24743 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24744 (vectorizable_reduction): Likewise.
24745 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
24746 (vectorizable_operation, vectorizable_load): Likewise.
24747 * tree.c (build_same_sized_truth_vector_type): Likewise.
24748 * valtrack.c (cleanup_auto_inc_dec): Likewise.
24749 * var-tracking.c (emit_note_insn_var_location): Likewise.
24750 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
24751 (ADDR_VEC_ALIGN): Likewise.
24753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24754 Alan Hayward <alan.hayward@arm.com>
24755 David Sherwood <david.sherwood@arm.com>
24757 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
24759 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24760 or if measurement_type is polynomial.
24761 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
24762 * combine.c (make_extraction): Likewise.
24763 * dse.c (find_shift_sequence): Likewise.
24764 * dwarf2out.c (mem_loc_descriptor): Likewise.
24765 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
24766 (extract_bit_field, extract_low_bits): Likewise.
24767 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
24768 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
24769 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
24770 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
24771 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24772 * reload.c (find_reloads): Likewise.
24773 * reload1.c (alter_reg): Likewise.
24774 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
24775 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
24776 * tree-if-conv.c (predicate_mem_writes): Likewise.
24777 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
24778 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
24779 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
24780 * valtrack.c (dead_debug_insert_temp): Likewise.
24781 * varasm.c (mergeable_constant_section): Likewise.
24782 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
24784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24785 Alan Hayward <alan.hayward@arm.com>
24786 David Sherwood <david.sherwood@arm.com>
24788 * expr.c (expand_assignment): Cope with polynomial mode sizes
24789 when assigning to a CONCAT.
24791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24792 Alan Hayward <alan.hayward@arm.com>
24793 David Sherwood <david.sherwood@arm.com>
24795 * machmode.h (mode_precision): Change from unsigned short to
24797 (mode_to_precision): Return a poly_uint16 rather than an unsigned
24799 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
24800 or if measurement_type is not polynomial.
24801 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
24802 in which the mode is already known to be a scalar_int_mode.
24803 * genmodes.c (emit_mode_precision): Change the type of mode_precision
24804 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
24806 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24807 for GET_MODE_PRECISION.
24808 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24809 for GET_MODE_PRECISION.
24810 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
24812 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
24813 (expand_field_assignment, make_extraction): Likewise.
24814 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
24815 (get_last_value): Likewise.
24816 * convert.c (convert_to_integer_1): Likewise.
24817 * cse.c (cse_insn): Likewise.
24818 * expr.c (expand_expr_real_1): Likewise.
24819 * lra-constraints.c (simplify_operand_subreg): Likewise.
24820 * optabs-query.c (can_atomic_load_p): Likewise.
24821 * optabs.c (expand_atomic_load): Likewise.
24822 (expand_atomic_store): Likewise.
24823 * ree.c (combine_reaching_defs): Likewise.
24824 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
24825 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
24826 * tree.h (type_has_mode_precision_p): Likewise.
24827 * ubsan.c (instrument_si_overflow): Likewise.
24829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24830 Alan Hayward <alan.hayward@arm.com>
24831 David Sherwood <david.sherwood@arm.com>
24833 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
24834 polynomial numbers of units.
24835 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
24836 (valid_vector_subparts_p): New function.
24837 (build_vector_type): Remove temporary shim and take the number
24838 of units as a poly_uint64 rather than an int.
24839 (build_opaque_vector_type): Take the number of units as a
24840 poly_uint64 rather than an int.
24841 * tree.c (build_vector_from_ctor): Handle polynomial
24842 TYPE_VECTOR_SUBPARTS.
24843 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
24844 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
24845 (build_vector_from_val): If the number of units is variable,
24846 use build_vec_duplicate_cst for constant operands and
24847 VEC_DUPLICATE_EXPR otherwise.
24848 (make_vector_type): Remove temporary is_constant ().
24849 (build_vector_type, build_opaque_vector_type): Take the number of
24850 units as a poly_uint64 rather than an int.
24851 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
24853 * cfgexpand.c (expand_debug_expr): Likewise.
24854 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
24855 (store_constructor, expand_expr_real_1): Likewise.
24856 (const_scalar_mask_from_tree): Likewise.
24857 * fold-const-call.c (fold_const_reduction): Likewise.
24858 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
24859 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
24860 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
24861 (fold_relational_const): Likewise.
24862 (native_interpret_vector): Likewise. Change the size from an
24863 int to an unsigned int.
24864 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
24865 TYPE_VECTOR_SUBPARTS.
24866 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
24867 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
24868 duplicating a non-constant operand into a variable-length vector.
24869 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
24870 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
24871 * ipa-icf.c (sem_variable::equals): Likewise.
24872 * match.pd: Likewise.
24873 * omp-simd-clone.c (simd_clone_subparts): Likewise.
24874 * print-tree.c (print_node): Likewise.
24875 * stor-layout.c (layout_type): Likewise.
24876 * targhooks.c (default_builtin_vectorization_cost): Likewise.
24877 * tree-cfg.c (verify_gimple_comparison): Likewise.
24878 (verify_gimple_assign_binary): Likewise.
24879 (verify_gimple_assign_ternary): Likewise.
24880 (verify_gimple_assign_single): Likewise.
24881 * tree-pretty-print.c (dump_generic_node): Likewise.
24882 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24883 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
24884 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
24885 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
24886 (vect_shift_permute_load_chain): Likewise.
24887 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
24888 (expand_vector_condition, optimize_vector_constructor): Likewise.
24889 (lower_vec_perm, get_compute_type): Likewise.
24890 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24891 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
24892 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
24893 (vect_recog_mask_conversion_pattern): Likewise.
24894 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
24895 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
24896 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24897 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
24898 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
24899 (vectorizable_shift, vectorizable_operation, vectorizable_store)
24900 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
24901 (supportable_widening_operation): Likewise.
24902 (supportable_narrowing_operation): Likewise.
24903 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
24905 * varasm.c (output_constant): Likewise.
24907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24908 Alan Hayward <alan.hayward@arm.com>
24909 David Sherwood <david.sherwood@arm.com>
24911 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
24912 so that both the length == 3 and length != 3 cases set up their
24913 own permute vectors. Add comments explaining why we know the
24914 number of elements is constant.
24915 (vect_permute_load_chain): Likewise.
24917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24918 Alan Hayward <alan.hayward@arm.com>
24919 David Sherwood <david.sherwood@arm.com>
24921 * machmode.h (mode_nunits): Change from unsigned char to
24923 (ONLY_FIXED_SIZE_MODES): New macro.
24924 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
24925 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
24926 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
24928 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
24929 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
24930 or if measurement_type is not polynomial.
24931 * genmodes.c (ZERO_COEFFS): New macro.
24932 (emit_mode_nunits_inline): Make mode_nunits_inline return a
24934 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
24935 Use ZERO_COEFFS when emitting initializers.
24936 * data-streamer.h (bp_pack_poly_value): New function.
24937 (bp_unpack_poly_value): Likewise.
24938 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24939 for GET_MODE_NUNITS.
24940 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24941 for GET_MODE_NUNITS.
24942 * tree.c (make_vector_type): Remove temporary shim and make
24943 the real function take the number of units as a poly_uint64
24944 rather than an int.
24945 (build_vector_type_for_mode): Handle polynomial nunits.
24946 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
24947 * emit-rtl.c (const_vec_series_p_1): Likewise.
24948 (gen_rtx_CONST_VECTOR): Likewise.
24949 * fold-const.c (test_vec_duplicate_folding): Likewise.
24950 * genrecog.c (validate_pattern): Likewise.
24951 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
24952 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24953 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
24954 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
24955 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
24956 * rtlanal.c (subreg_get_info): Likewise.
24957 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24958 (vect_grouped_load_supported): Likewise.
24959 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
24960 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
24961 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
24962 (simplify_const_unary_operation, simplify_binary_operation_1)
24963 (simplify_const_binary_operation, simplify_ternary_operation)
24964 (test_vector_ops_duplicate, test_vector_ops): Likewise.
24965 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
24966 instead of CONST_VECTOR_NUNITS.
24967 * varasm.c (output_constant_pool_2): Likewise.
24968 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
24969 explicit-encoded elements in the XVEC for variable-length vectors.
24971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24973 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
24975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24976 Alan Hayward <alan.hayward@arm.com>
24977 David Sherwood <david.sherwood@arm.com>
24979 * coretypes.h (fixed_size_mode): Declare.
24980 (fixed_size_mode_pod): New typedef.
24981 * builtins.h (target_builtins::x_apply_args_mode)
24982 (target_builtins::x_apply_result_mode): Change type to
24983 fixed_size_mode_pod.
24984 * builtins.c (apply_args_size, apply_result_size, result_vector)
24985 (expand_builtin_apply_args_1, expand_builtin_apply)
24986 (expand_builtin_return): Update accordingly.
24988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24990 * cse.c (hash_rtx_cb): Hash only the encoded elements.
24991 * cselib.c (cselib_hash_rtx): Likewise.
24992 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
24993 CONST_VECTOR encoding.
24995 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24996 Jeff Law <law@redhat.com>
24999 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
25000 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
25001 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
25002 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
25005 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
25006 explicitly probe *sp in a noreturn function if there were any callee
25007 register saves or frame pointer is needed.
25009 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25012 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
25013 BLKmode for ternary, binary or unary expressions.
25016 * var-tracking.c (delete_vta_debug_insn): New inline function.
25017 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
25018 insns from get_insns () to NULL instead of each bb separately.
25019 Use delete_vta_debug_insn. No longer static.
25020 (vt_debug_insns_local, variable_tracking_main_1): Adjust
25021 delete_vta_debug_insns callers.
25022 * rtl.h (delete_vta_debug_insns): Declare.
25023 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
25024 instead of variable_tracking_main.
25026 2018-01-03 Martin Sebor <msebor@redhat.com>
25028 PR tree-optimization/83603
25029 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
25030 arguments past the endof the argument list in functions declared
25031 without a prototype.
25032 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
25033 Avoid checking when arguments are null.
25035 2018-01-03 Martin Sebor <msebor@redhat.com>
25038 * doc/extend.texi (attribute const): Fix a typo.
25039 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
25040 issuing -Wsuggest-attribute for void functions.
25042 2018-01-03 Martin Sebor <msebor@redhat.com>
25044 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
25045 offset_int::from instead of wide_int::to_shwi.
25046 (maybe_diag_overlap): Remove assertion.
25047 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
25048 * gimple-ssa-sprintf.c (format_directive): Same.
25049 (parse_directive): Same.
25050 (sprintf_dom_walker::compute_format_length): Same.
25051 (try_substitute_return_value): Same.
25053 2018-01-03 Jeff Law <law@redhat.com>
25055 PR middle-end/83654
25056 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
25057 non-constant residual for zero at runtime and avoid probing in
25058 that case. Reorganize code for trailing problem to mirror handling
25061 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
25063 PR tree-optimization/83501
25064 * tree-ssa-strlen.c (get_string_cst): New.
25065 (handle_char_store): Call get_string_cst.
25067 2018-01-03 Martin Liska <mliska@suse.cz>
25069 PR tree-optimization/83593
25070 * tree-ssa-strlen.c: Include tree-cfg.h.
25071 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
25072 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
25073 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
25075 (strlen_dom_walker::before_dom_children): Call
25076 gimple_purge_dead_eh_edges. Dump tranformation with details
25078 (strlen_dom_walker::before_dom_children): Update call by adding
25079 new argument cleanup_eh.
25080 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
25082 2018-01-03 Martin Liska <mliska@suse.cz>
25085 * cif-code.def (VARIADIC_THUNK): New enum value.
25086 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
25089 2018-01-03 Jan Beulich <jbeulich@suse.com>
25091 * sse.md (mov<mode>_internal): Tighten condition for when to use
25092 vmovdqu<ssescalarsize> for TI and OI modes.
25094 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25096 Update copyright years.
25098 2018-01-03 Martin Liska <mliska@suse.cz>
25101 * ipa-visibility.c (function_and_variable_visibility): Skip
25102 functions with noipa attribure.
25104 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25106 * gcc.c (process_command): Update copyright notice dates.
25107 * gcov-dump.c (print_version): Ditto.
25108 * gcov.c (print_version): Ditto.
25109 * gcov-tool.c (print_version): Ditto.
25110 * gengtype.c (create_file): Ditto.
25111 * doc/cpp.texi: Bump @copying's copyright year.
25112 * doc/cppinternals.texi: Ditto.
25113 * doc/gcc.texi: Ditto.
25114 * doc/gccint.texi: Ditto.
25115 * doc/gcov.texi: Ditto.
25116 * doc/install.texi: Ditto.
25117 * doc/invoke.texi: Ditto.
25119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25121 * vector-builder.h (vector_builder::m_full_nelts): Change from
25122 unsigned int to poly_uint64.
25123 (vector_builder::full_nelts): Update prototype accordingly.
25124 (vector_builder::new_vector): Likewise.
25125 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
25126 (vector_builder::operator ==): Likewise.
25127 (vector_builder::finalize): Likewise.
25128 * int-vector-builder.h (int_vector_builder::int_vector_builder):
25129 Take the number of elements as a poly_uint64 rather than an
25131 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
25132 from unsigned int to poly_uint64.
25133 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
25134 (vec_perm_indices::new_vector): Likewise.
25135 (vec_perm_indices::length): Likewise.
25136 (vec_perm_indices::nelts_per_input): Likewise.
25137 (vec_perm_indices::input_nelts): Likewise.
25138 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
25139 number of elements per input as a poly_uint64 rather than an
25140 unsigned int. Use the original encoding for variable-length
25141 vectors, rather than clamping each individual element.
25142 For the second and subsequent elements in each pattern,
25143 clamp the step and base before clamping their sum.
25144 (vec_perm_indices::series_p): Handle polynomial element counts.
25145 (vec_perm_indices::all_in_range_p): Likewise.
25146 (vec_perm_indices_to_tree): Likewise.
25147 (vec_perm_indices_to_rtx): Likewise.
25148 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
25149 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
25150 (tree_vector_builder::new_binary_operation): Handle polynomial
25151 element counts. Return false if we need to know the number
25152 of elements at compile time.
25153 * fold-const.c (fold_vec_perm): Punt if the number of elements
25154 isn't known at compile time.
25156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25158 * vec-perm-indices.h (vec_perm_builder): Change element type
25159 from HOST_WIDE_INT to poly_int64.
25160 (vec_perm_indices::element_type): Update accordingly.
25161 (vec_perm_indices::clamp): Handle polynomial element_types.
25162 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25163 (vec_perm_indices::all_in_range_p): Likewise.
25164 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
25166 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
25167 polynomial vec_perm_indices element types.
25168 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
25169 * fold-const.c (fold_vec_perm): Likewise.
25170 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
25171 * tree-vect-generic.c (lower_vec_perm): Likewise.
25172 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25173 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
25174 element type to HOST_WIDE_INT.
25176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25177 Alan Hayward <alan.hayward@arm.com>
25178 David Sherwood <david.sherwood@arm.com>
25180 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
25181 rather than an int. Use plus_constant.
25182 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
25183 Take the offset "c" as a poly_int64 rather than a 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 * calls.c (emit_call_1, expand_call): Change struct_value_size from
25190 a HOST_WIDE_INT to a poly_int64.
25192 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25193 Alan Hayward <alan.hayward@arm.com>
25194 David Sherwood <david.sherwood@arm.com>
25196 * calls.c (load_register_parameters): Cope with polynomial
25197 mode sizes. Require a constant size for BLKmode parameters
25198 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
25199 forces a parameter to be padded at the lsb end in order to
25200 fill a complete number of words, require the parameter size
25201 to be ordered wrt UNITS_PER_WORD.
25203 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25204 Alan Hayward <alan.hayward@arm.com>
25205 David Sherwood <david.sherwood@arm.com>
25207 * reload1.c (spill_stack_slot_width): Change element type
25208 from unsigned int to poly_uint64_pod.
25209 (alter_reg): Treat mode sizes as polynomial.
25211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25212 Alan Hayward <alan.hayward@arm.com>
25213 David Sherwood <david.sherwood@arm.com>
25215 * reload.c (complex_word_subreg_p): New function.
25216 (reload_inner_reg_of_subreg, push_reload): Use it.
25218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25219 Alan Hayward <alan.hayward@arm.com>
25220 David Sherwood <david.sherwood@arm.com>
25222 * lra-constraints.c (process_alt_operands): Reject matched
25223 operands whose sizes aren't ordered.
25224 (match_reload): Refer to this check here.
25226 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25227 Alan Hayward <alan.hayward@arm.com>
25228 David Sherwood <david.sherwood@arm.com>
25230 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
25231 that the mode size is in the set {1, 2, 4, 8, 16}.
25233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25234 Alan Hayward <alan.hayward@arm.com>
25235 David Sherwood <david.sherwood@arm.com>
25237 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
25238 Use plus_constant instead of gen_rtx_PLUS.
25240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25241 Alan Hayward <alan.hayward@arm.com>
25242 David Sherwood <david.sherwood@arm.com>
25244 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
25245 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
25246 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
25247 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
25248 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
25249 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
25250 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
25251 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
25252 * config/i386/i386.c (ix86_push_rounding): ...this new function.
25253 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
25255 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
25256 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
25257 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
25258 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
25259 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
25260 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
25261 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
25262 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
25263 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
25264 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
25266 * expr.c (emit_move_resolve_push): Treat the input and result
25267 of PUSH_ROUNDING as a poly_int64.
25268 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
25269 (emit_push_insn): Likewise.
25270 * lra-eliminations.c (mark_not_eliminable): Likewise.
25271 * recog.c (push_operand): Likewise.
25272 * reload1.c (elimination_effects): Likewise.
25273 * rtlanal.c (nonzero_bits1): Likewise.
25274 * calls.c (store_one_arg): Likewise. Require the padding to be
25275 known at compile time.
25277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25278 Alan Hayward <alan.hayward@arm.com>
25279 David Sherwood <david.sherwood@arm.com>
25281 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
25282 Use plus_constant instead of gen_rtx_PLUS.
25284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25285 Alan Hayward <alan.hayward@arm.com>
25286 David Sherwood <david.sherwood@arm.com>
25288 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
25289 rather than an int.
25291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25292 Alan Hayward <alan.hayward@arm.com>
25293 David Sherwood <david.sherwood@arm.com>
25295 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
25296 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
25297 via stack temporaries. Treat the mode size as polynomial too.
25299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25300 Alan Hayward <alan.hayward@arm.com>
25301 David Sherwood <david.sherwood@arm.com>
25303 * expr.c (expand_expr_real_2): When handling conversions involving
25304 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
25305 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
25306 as a poly_uint64 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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
25314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25315 Alan Hayward <alan.hayward@arm.com>
25316 David Sherwood <david.sherwood@arm.com>
25318 * combine.c (can_change_dest_mode): Handle polynomial
25319 REGMODE_NATURAL_SIZE.
25320 * expmed.c (store_bit_field_1): Likewise.
25321 * expr.c (store_constructor): Likewise.
25322 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
25323 and polynomial REGMODE_NATURAL_SIZE.
25324 (gen_lowpart_common): Likewise.
25325 * reginfo.c (record_subregs_of_mode): Likewise.
25326 * rtlanal.c (read_modify_subreg_p): Likewise.
25328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25329 Alan Hayward <alan.hayward@arm.com>
25330 David Sherwood <david.sherwood@arm.com>
25332 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
25333 numbers of elements.
25335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25336 Alan Hayward <alan.hayward@arm.com>
25337 David Sherwood <david.sherwood@arm.com>
25339 * match.pd: Cope with polynomial numbers of vector elements.
25341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25342 Alan Hayward <alan.hayward@arm.com>
25343 David Sherwood <david.sherwood@arm.com>
25345 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
25346 in a POINTER_PLUS_EXPR.
25348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25349 Alan Hayward <alan.hayward@arm.com>
25350 David Sherwood <david.sherwood@arm.com>
25352 * omp-simd-clone.c (simd_clone_subparts): New function.
25353 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
25354 (ipa_simd_modify_function_body): Likewise.
25356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25357 Alan Hayward <alan.hayward@arm.com>
25358 David Sherwood <david.sherwood@arm.com>
25360 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
25361 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
25362 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
25363 (expand_vector_condition, vector_element): Likewise.
25364 (subparts_gt): New function.
25365 (get_compute_type): Use subparts_gt.
25366 (count_type_subparts): Delete.
25367 (expand_vector_operations_1): Use subparts_gt instead of
25368 count_type_subparts.
25370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25371 Alan Hayward <alan.hayward@arm.com>
25372 David Sherwood <david.sherwood@arm.com>
25374 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
25375 (vect_compile_time_alias): ...this new function. Do the calculation
25376 on poly_ints rather than trees.
25377 (vect_prune_runtime_alias_test_list): Update call accordingly.
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-slp.c (vect_build_slp_tree_1): Handle polynomial
25385 (vect_schedule_slp_instance): Likewise.
25387 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25388 Alan Hayward <alan.hayward@arm.com>
25389 David Sherwood <david.sherwood@arm.com>
25391 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
25392 constant and extern definitions for variable-length vectors.
25393 (vect_get_constant_vectors): Note that the number of units
25394 is known to be constant.
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-stmts.c (vectorizable_conversion): Treat the number
25401 of units as polynomial. Choose between WIDE and NARROW based
25404 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25405 Alan Hayward <alan.hayward@arm.com>
25406 David Sherwood <david.sherwood@arm.com>
25408 * tree-vect-stmts.c (simd_clone_subparts): New function.
25409 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
25411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25412 Alan Hayward <alan.hayward@arm.com>
25413 David Sherwood <david.sherwood@arm.com>
25415 * tree-vect-stmts.c (vectorizable_call): Treat the number of
25416 vectors as polynomial. Use build_index_vector for
25417 IFN_GOMP_SIMD_LANE.
25419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25420 Alan Hayward <alan.hayward@arm.com>
25421 David Sherwood <david.sherwood@arm.com>
25423 * tree-vect-stmts.c (get_load_store_type): Treat the number of
25424 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
25425 for variable-length vectors.
25426 (vectorizable_mask_load_store): Treat the number of units as
25427 polynomial, asserting that it is constant if the condition has
25428 already been enforced.
25429 (vectorizable_store, vectorizable_load): Likewise.
25431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25432 Alan Hayward <alan.hayward@arm.com>
25433 David Sherwood <david.sherwood@arm.com>
25435 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
25436 of units as polynomial. Punt if we can't tell at compile time
25437 which vector contains the final result.
25439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25440 Alan Hayward <alan.hayward@arm.com>
25441 David Sherwood <david.sherwood@arm.com>
25443 * tree-vect-loop.c (vectorizable_induction): Treat the number
25444 of units as polynomial. Punt on SLP inductions. Use an integer
25445 VEC_SERIES_EXPR for variable-length integer reductions. Use a
25446 cast of such a series for variable-length floating-point
25449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25450 Alan Hayward <alan.hayward@arm.com>
25451 David Sherwood <david.sherwood@arm.com>
25453 * tree.h (build_index_vector): Declare.
25454 * tree.c (build_index_vector): New function.
25455 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
25456 of units as polynomial, forcibly converting it to a constant if
25457 vectorizable_reduction has already enforced the condition.
25458 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
25459 to create a {1,2,3,...} vector.
25460 (vectorizable_reduction): Treat the number of units as polynomial.
25461 Choose vectype_in based on the largest scalar element size rather
25462 than the smallest number of units. Enforce the restrictions
25465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25466 Alan Hayward <alan.hayward@arm.com>
25467 David Sherwood <david.sherwood@arm.com>
25469 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
25470 number of units as polynomial.
25472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25473 Alan Hayward <alan.hayward@arm.com>
25474 David Sherwood <david.sherwood@arm.com>
25476 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
25477 * target.def (autovectorize_vector_sizes): Return the vector sizes
25478 by pointer, using vector_sizes rather than a bitmask.
25479 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
25480 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
25481 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
25483 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
25484 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
25485 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
25486 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
25487 * omp-general.c (omp_max_vf): Likewise.
25488 * omp-low.c (omp_clause_aligned_alignment): Likewise.
25489 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
25490 * tree-vect-loop.c (vect_analyze_loop): Likewise.
25491 * tree-vect-slp.c (vect_slp_bb): Likewise.
25492 * doc/tm.texi: Regenerate.
25493 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
25495 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
25496 the vector size as a poly_uint64 rather than an unsigned int.
25497 (current_vector_size): Change from an unsigned int to a poly_uint64.
25498 (get_vectype_for_scalar_type): Update accordingly.
25499 * tree.h (build_truth_vector_type): Take the size and number of
25500 units as a poly_uint64 rather than an unsigned int.
25501 (build_vector_type): Add a temporary overload that takes
25502 the number of units as a poly_uint64 rather than an unsigned int.
25503 * tree.c (make_vector_type): Likewise.
25504 (build_truth_vector_type): Take the number of units as a poly_uint64
25505 rather than an unsigned int.
25507 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25508 Alan Hayward <alan.hayward@arm.com>
25509 David Sherwood <david.sherwood@arm.com>
25511 * target.def (get_mask_mode): Take the number of units and length
25512 as poly_uint64s rather than unsigned ints.
25513 * targhooks.h (default_get_mask_mode): Update accordingly.
25514 * targhooks.c (default_get_mask_mode): Likewise.
25515 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
25516 * doc/tm.texi: Regenerate.
25518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25519 Alan Hayward <alan.hayward@arm.com>
25520 David Sherwood <david.sherwood@arm.com>
25522 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
25523 * omp-general.c (omp_max_vf): Likewise.
25524 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
25525 (expand_omp_simd): Handle polynomial safelen.
25526 * omp-low.c (omplow_simd_context): Add a default constructor.
25527 (omplow_simd_context::max_vf): Change from int to poly_uint64.
25528 (lower_rec_simd_input_clauses): Update accordingly.
25529 (lower_rec_input_clauses): Likewise.
25531 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25532 Alan Hayward <alan.hayward@arm.com>
25533 David Sherwood <david.sherwood@arm.com>
25535 * tree-vectorizer.h (vect_nunits_for_cost): New function.
25536 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
25537 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
25538 (vect_analyze_slp_cost): Likewise.
25539 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
25540 (vect_model_load_cost): Likewise.
25542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25543 Alan Hayward <alan.hayward@arm.com>
25544 David Sherwood <david.sherwood@arm.com>
25546 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
25547 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
25548 from an unsigned int * to a poly_uint64_pod *.
25549 (calculate_unrolling_factor): New function.
25550 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
25552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25553 Alan Hayward <alan.hayward@arm.com>
25554 David Sherwood <david.sherwood@arm.com>
25556 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
25557 from an unsigned int to a poly_uint64.
25558 (_loop_vec_info::slp_unrolling_factor): Likewise.
25559 (_loop_vec_info::vectorization_factor): Change from an int
25561 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
25562 (vect_get_num_vectors): New function.
25563 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
25564 (vect_get_num_copies): Use vect_get_num_vectors.
25565 (vect_analyze_data_ref_dependences): Change max_vf from an int *
25566 to an unsigned int *.
25567 (vect_analyze_data_refs): Change min_vf from an int * to a
25569 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25570 than an unsigned HOST_WIDE_INT.
25571 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
25572 (vect_analyze_data_ref_dependence): Change max_vf from an int *
25573 to an unsigned int *.
25574 (vect_analyze_data_ref_dependences): Likewise.
25575 (vect_compute_data_ref_alignment): Handle polynomial vf.
25576 (vect_enhance_data_refs_alignment): Likewise.
25577 (vect_prune_runtime_alias_test_list): Likewise.
25578 (vect_shift_permute_load_chain): Likewise.
25579 (vect_supportable_dr_alignment): Likewise.
25580 (dependence_distance_ge_vf): Take the vectorization factor as a
25581 poly_uint64 rather than an unsigned HOST_WIDE_INT.
25582 (vect_analyze_data_refs): Change min_vf from an int * to a
25584 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
25585 vfm1 as a poly_uint64 rather than an int. Make the same change
25586 for the returned bound_scalar.
25587 (vect_gen_vector_loop_niters): Handle polynomial vf.
25588 (vect_do_peeling): Likewise. Update call to
25589 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
25590 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
25592 * tree-vect-loop.c (vect_determine_vectorization_factor)
25593 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
25594 (vect_get_known_peeling_cost): Likewise.
25595 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
25596 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
25597 (vect_transform_loop): Likewise. Use the lowest possible VF when
25598 updating the upper bounds of the loop.
25599 (vect_min_worthwhile_factor): Make static. Return an unsigned int
25600 rather than an int.
25601 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
25602 polynomial unroll factors.
25603 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
25604 (vect_make_slp_decision): Likewise.
25605 (vect_supported_load_permutation_p): Likewise, and polynomial
25607 (vect_analyze_slp_cost): Handle polynomial vf.
25608 (vect_slp_analyze_node_operations): Likewise.
25609 (vect_slp_analyze_bb_1): Likewise.
25610 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25611 than an unsigned HOST_WIDE_INT.
25612 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
25613 (vectorizable_load): Handle polynomial vf.
25614 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
25616 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
25618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25619 Alan Hayward <alan.hayward@arm.com>
25620 David Sherwood <david.sherwood@arm.com>
25622 * match.pd: Handle bit operations involving three constants
25623 and try to fold one pair.
25625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25627 * tree-vect-loop-manip.c: Include gimple-fold.h.
25628 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
25629 niters_maybe_zero parameters. Handle other cases besides a step of 1.
25630 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
25631 Add a path that uses a step of VF instead of 1, but disable it
25633 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
25634 and niters_no_overflow parameters. Update calls to
25635 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
25636 Create a new SSA name if the latter choses to use a ste other
25637 than zero, and return it via niters_vector_mult_vf_var.
25638 * tree-vect-loop.c (vect_transform_loop): Update calls to
25639 vect_do_peeling, vect_gen_vector_loop_niters and
25640 slpeel_make_loop_iterate_ntimes.
25641 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
25642 (vect_gen_vector_loop_niters): Update declarations after above changes.
25644 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
25646 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
25647 128-bit round to integer instructions.
25648 (ceil<mode>2): Likewise.
25649 (btrunc<mode>2): Likewise.
25650 (round<mode>2): Likewise.
25652 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25654 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
25655 unaligned VSX load/store on P8/P9.
25656 (expand_block_clear): Allow the use of unaligned VSX
25657 load/store on P8/P9.
25659 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25661 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
25663 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
25664 swap associated with both a load and a store.
25666 2018-01-02 Andrew Waterman <andrew@sifive.com>
25668 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
25669 * config/riscv/riscv.md (clear_cache): Use it.
25671 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
25673 * web.c: Remove out-of-date comment.
25675 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25677 * expr.c (fixup_args_size_notes): Check that any existing
25678 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
25679 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
25680 (emit_single_push_insn): ...here.
25682 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25684 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
25685 (const_vector_encoded_nelts): New function.
25686 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
25687 (const_vector_int_elt, const_vector_elt): Declare.
25688 * emit-rtl.c (const_vector_int_elt_1): New function.
25689 (const_vector_elt): Likewise.
25690 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
25691 of CONST_VECTOR_ELT.
25693 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25695 * expr.c: Include rtx-vector-builder.h.
25696 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
25697 directly on the tree encoding.
25698 (const_vector_from_tree): Likewise.
25699 * optabs.c: Include rtx-vector-builder.h.
25700 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
25701 sequence of "u" values.
25702 * vec-perm-indices.c: Include rtx-vector-builder.h.
25703 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
25704 directly on the vec_perm_indices encoding.
25706 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25708 * doc/rtl.texi (const_vector): Describe new encoding scheme.
25709 * Makefile.in (OBJS): Add rtx-vector-builder.o.
25710 * rtx-vector-builder.h: New file.
25711 * rtx-vector-builder.c: Likewise.
25712 * rtl.h (rtx_def::u2): Add a const_vector field.
25713 (CONST_VECTOR_NPATTERNS): New macro.
25714 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
25715 (CONST_VECTOR_DUPLICATE_P): Likewise.
25716 (CONST_VECTOR_STEPPED_P): Likewise.
25717 (CONST_VECTOR_ENCODED_ELT): Likewise.
25718 (const_vec_duplicate_p): Check for a duplicated vector encoding.
25719 (unwrap_const_vec_duplicate): Likewise.
25720 (const_vec_series_p): Check for a non-duplicated vector encoding.
25721 Say that the function only returns true for integer vectors.
25722 * emit-rtl.c: Include rtx-vector-builder.h.
25723 (gen_const_vec_duplicate_1): Delete.
25724 (gen_const_vector): Call gen_const_vec_duplicate instead of
25725 gen_const_vec_duplicate_1.
25726 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
25727 (gen_const_vec_duplicate): Use rtx_vector_builder.
25728 (gen_const_vec_series): Likewise.
25729 (gen_rtx_CONST_VECTOR): Likewise.
25730 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
25731 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25732 Build a new vector rather than modifying a CONST_VECTOR in-place.
25733 (handle_special_swappables): Update call accordingly.
25734 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
25735 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25736 Build a new vector rather than modifying a CONST_VECTOR in-place.
25737 (handle_special_swappables): Update call accordingly.
25739 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25741 * simplify-rtx.c (simplify_const_binary_operation): Use
25742 CONST_VECTOR_ELT instead of XVECEXP.
25744 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25746 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
25747 the selector elements to be different from the data elements
25748 if the selector is a VECTOR_CST.
25749 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
25750 ssizetype for the selector.
25752 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25754 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
25755 before testing each element individually.
25756 * tree-vect-generic.c (lower_vec_perm): Likewise.
25758 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25760 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
25761 * selftest-run-tests.c (selftest::run_tests): Call it.
25762 * vector-builder.h (vector_builder::operator ==): New function.
25763 (vector_builder::operator !=): Likewise.
25764 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
25765 (vec_perm_indices::all_from_input_p): New function.
25766 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25767 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
25768 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
25769 instead of reading the VECTOR_CST directly. Detect whether both
25770 vector inputs are the same before constructing the vec_perm_indices,
25771 and update the number of inputs argument accordingly. Use the
25772 utility functions added above. Only construct sel2 if we need to.
25774 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25776 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
25777 the broadcast of the low byte.
25778 (expand_mult_highpart): Use an explicit encoding for the permutes.
25779 * optabs-query.c (can_mult_highpart_p): Likewise.
25780 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25781 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25782 (vectorizable_bswap): Likewise.
25783 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
25784 explicit encoding for the power-of-2 permutes.
25785 (vect_permute_store_chain): Likewise.
25786 (vect_grouped_load_supported): Likewise.
25787 (vect_permute_load_chain): Likewise.
25789 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25791 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
25792 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
25793 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
25794 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25795 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
25796 (vect_gen_perm_mask_any): Likewise.
25798 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25800 * int-vector-builder.h: New file.
25801 * vec-perm-indices.h: Include int-vector-builder.h.
25802 (vec_perm_indices): Redefine as an int_vector_builder.
25803 (auto_vec_perm_indices): Delete.
25804 (vec_perm_builder): Redefine as a stand-alone class.
25805 (vec_perm_indices::vec_perm_indices): New function.
25806 (vec_perm_indices::clamp): Likewise.
25807 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
25808 (vec_perm_indices::new_vector): New function.
25809 (vec_perm_indices::new_expanded_vector): Update for new
25810 vec_perm_indices class.
25811 (vec_perm_indices::rotate_inputs): New function.
25812 (vec_perm_indices::all_in_range_p): Operate directly on the
25813 encoded form, without computing elided elements.
25814 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
25815 encoding. Update for new vec_perm_indices class.
25816 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
25817 the given vec_perm_builder.
25818 (expand_vec_perm_var): Update vec_perm_builder constructor.
25819 (expand_mult_highpart): Use vec_perm_builder instead of
25820 auto_vec_perm_indices.
25821 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
25822 vec_perm_indices instead of auto_vec_perm_indices. Use a single
25823 or double series encoding as appropriate.
25824 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
25825 vec_perm_indices instead of auto_vec_perm_indices.
25826 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25827 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25828 (vect_permute_store_chain): Likewise.
25829 (vect_grouped_load_supported): Likewise.
25830 (vect_permute_load_chain): Likewise.
25831 (vect_shift_permute_load_chain): Likewise.
25832 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
25833 (vect_transform_slp_perm_load): Likewise.
25834 (vect_schedule_slp_instance): Likewise.
25835 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25836 (vectorizable_mask_load_store): Likewise.
25837 (vectorizable_bswap): Likewise.
25838 (vectorizable_store): Likewise.
25839 (vectorizable_load): Likewise.
25840 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
25841 vec_perm_indices instead of auto_vec_perm_indices. Use
25842 tree_to_vec_perm_builder to read the vector from a tree.
25843 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
25844 vec_perm_builder instead of a vec_perm_indices.
25845 (have_whole_vector_shift): Use vec_perm_builder and
25846 vec_perm_indices instead of auto_vec_perm_indices. Leave the
25847 truncation to calc_vec_perm_mask_for_shift.
25848 (vect_create_epilog_for_reduction): Likewise.
25849 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
25850 from auto_vec_perm_indices to vec_perm_indices.
25851 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25852 instead of changing individual elements.
25853 (aarch64_vectorize_vec_perm_const): Use new_vector to install
25854 the vector in d.perm.
25855 * config/arm/arm.c (expand_vec_perm_d::perm): Change
25856 from auto_vec_perm_indices to vec_perm_indices.
25857 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25858 instead of changing individual elements.
25859 (arm_vectorize_vec_perm_const): Use new_vector to install
25860 the vector in d.perm.
25861 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
25862 Update vec_perm_builder constructor.
25863 (rs6000_expand_interleave): Likewise.
25864 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
25865 (rs6000_expand_interleave): Likewise.
25867 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25869 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
25870 to qimode could truncate the indices.
25871 * optabs.c (expand_vec_perm_var): Likewise.
25873 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25875 * Makefile.in (OBJS): Add vec-perm-indices.o.
25876 * vec-perm-indices.h: New file.
25877 * vec-perm-indices.c: Likewise.
25878 * target.h (vec_perm_indices): Replace with a forward class
25880 (auto_vec_perm_indices): Move to vec-perm-indices.h.
25881 * optabs.h: Include vec-perm-indices.h.
25882 (expand_vec_perm): Delete.
25883 (selector_fits_mode_p, expand_vec_perm_var): Declare.
25884 (expand_vec_perm_const): Declare.
25885 * target.def (vec_perm_const_ok): Replace with...
25886 (vec_perm_const): ...this new hook.
25887 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
25888 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
25889 * doc/tm.texi: Regenerate.
25890 * optabs.def (vec_perm_const): Delete.
25891 * doc/md.texi (vec_perm_const): Likewise.
25892 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
25893 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
25894 expand_vec_perm for constant permutation vectors. Assert that
25895 the mode of variable permutation vectors is the integer equivalent
25896 of the mode that is being permuted.
25897 * optabs-query.h (selector_fits_mode_p): Declare.
25898 * optabs-query.c: Include vec-perm-indices.h.
25899 (selector_fits_mode_p): New function.
25900 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
25901 is defined, instead of checking whether the vec_perm_const_optab
25902 exists. Use targetm.vectorize.vec_perm_const instead of
25903 targetm.vectorize.vec_perm_const_ok. Check whether the indices
25904 fit in the vector mode before using a variable permute.
25905 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
25906 vec_perm_indices instead of an rtx.
25907 (expand_vec_perm): Replace with...
25908 (expand_vec_perm_const): ...this new function. Take the selector
25909 as a vec_perm_indices rather than an rtx. Also take the mode of
25910 the selector. Update call to shift_amt_for_vec_perm_mask.
25911 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
25912 Use vec_perm_indices::new_expanded_vector to expand the original
25913 selector into bytes. Check whether the indices fit in the vector
25914 mode before using a variable permute.
25915 (expand_vec_perm_var): Make global.
25916 (expand_mult_highpart): Use expand_vec_perm_const.
25917 * fold-const.c: Includes vec-perm-indices.h.
25918 * tree-ssa-forwprop.c: Likewise.
25919 * tree-vect-data-refs.c: Likewise.
25920 * tree-vect-generic.c: Likewise.
25921 * tree-vect-loop.c: Likewise.
25922 * tree-vect-slp.c: Likewise.
25923 * tree-vect-stmts.c: Likewise.
25924 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
25926 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
25927 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
25928 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
25929 (aarch64_vectorize_vec_perm_const): ...this new function.
25930 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25931 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25932 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
25933 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
25934 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25935 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25936 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
25938 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
25939 check for NEON modes.
25940 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
25941 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
25942 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
25943 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
25945 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
25946 the old VEC_PERM_CONST conditions.
25947 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
25948 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
25949 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
25950 (ia64_vectorize_vec_perm_const_ok): Merge into...
25951 (ia64_vectorize_vec_perm_const): ...this new function.
25952 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
25953 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
25954 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
25955 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
25956 * config/mips/mips.c (mips_expand_vec_perm_const)
25957 (mips_vectorize_vec_perm_const_ok): Merge into...
25958 (mips_vectorize_vec_perm_const): ...this new function.
25959 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
25960 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
25961 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
25962 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
25963 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
25964 (rs6000_expand_vec_perm_const): Delete.
25965 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
25967 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25968 (altivec_expand_vec_perm_const_le): Take each operand individually.
25969 Operate on constant selectors rather than rtxes.
25970 (altivec_expand_vec_perm_const): Likewise. Update call to
25971 altivec_expand_vec_perm_const_le.
25972 (rs6000_expand_vec_perm_const): Delete.
25973 (rs6000_vectorize_vec_perm_const_ok): Delete.
25974 (rs6000_vectorize_vec_perm_const): New function.
25975 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
25976 an element count and rtx array.
25977 (rs6000_expand_extract_even): Update call accordingly.
25978 (rs6000_expand_interleave): Likewise.
25979 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
25980 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
25981 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
25982 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
25983 (rs6000_expand_vec_perm_const): Delete.
25984 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25985 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25986 (altivec_expand_vec_perm_const_le): Take each operand individually.
25987 Operate on constant selectors rather than rtxes.
25988 (altivec_expand_vec_perm_const): Likewise. Update call to
25989 altivec_expand_vec_perm_const_le.
25990 (rs6000_expand_vec_perm_const): Delete.
25991 (rs6000_vectorize_vec_perm_const_ok): Delete.
25992 (rs6000_vectorize_vec_perm_const): New function. Remove stray
25993 reference to the SPE evmerge intructions.
25994 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
25995 an element count and rtx array.
25996 (rs6000_expand_extract_even): Update call accordingly.
25997 (rs6000_expand_interleave): Likewise.
25998 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
25999 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
26001 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26003 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26005 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
26006 vector mode and that that mode matches the mode of the data
26008 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
26009 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
26010 directly using expand_vec_perm_1 when forcing selectors into
26012 (expand_vec_perm_var): New function, split out from expand_vec_perm.
26014 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26016 * optabs-query.h (can_vec_perm_p): Delete.
26017 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
26018 * optabs-query.c (can_vec_perm_p): Split into...
26019 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
26020 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
26021 particular selector is valid.
26022 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26023 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26024 (vect_grouped_load_supported): Likewise.
26025 (vect_shift_permute_load_chain): Likewise.
26026 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
26027 (vect_transform_slp_perm_load): Likewise.
26028 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26029 (vectorizable_bswap): Likewise.
26030 (vect_gen_perm_mask_checked): Likewise.
26031 * fold-const.c (fold_ternary_loc): Likewise. Don't take
26032 implementations of variable permutation vectors into account
26033 when deciding which selector to use.
26034 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
26035 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
26036 with a false third argument.
26037 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
26038 to test whether the constant selector is valid and can_vec_perm_var_p
26039 to test whether a variable selector is valid.
26041 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26043 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
26044 * optabs-query.c (can_vec_perm_p): Likewise.
26045 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
26046 instead of vec_perm_indices.
26047 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
26048 (vect_gen_perm_mask_checked): Likewise,
26049 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
26050 (vect_gen_perm_mask_checked): Likewise,
26052 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26054 * optabs-query.h (qimode_for_vec_perm): Declare.
26055 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
26056 (qimode_for_vec_perm): ...this new function.
26057 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
26059 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26061 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
26062 does not have a conditional at the top.
26064 2018-01-02 Richard Biener <rguenther@suse.de>
26066 * ipa-inline.c (big_speedup_p): Fix expression.
26068 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
26071 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
26074 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
26078 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
26079 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
26080 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
26081 cond_taken_branch_cost 3->4.
26083 2018-01-01 Jakub Jelinek <jakub@redhat.com>
26085 PR tree-optimization/83581
26086 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
26087 TODO_cleanup_cfg if any changes have been made.
26089 PR middle-end/83608
26090 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
26091 convert_modes if target mode has the right side, but different mode
26094 PR middle-end/83609
26095 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
26096 last argument when extracting from CONCAT. If either from_real or
26097 from_imag is NULL, use expansion through memory. If result is not
26098 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
26099 the parts directly to inner mode, if even that fails, use expansion
26102 PR middle-end/83623
26103 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
26104 check for bswap in mode rather than HImode and use that in expand_unop
26107 Copyright (C) 2018 Free Software Foundation, Inc.
26109 Copying and distribution of this file, with or without modification,
26110 are permitted in any medium without royalty provided the copyright
26111 notice and this notice are preserved.