1 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3 * poly-int.h (can_div_trunc_p): Add new overload in which all values
5 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
6 (memrefs_conflict_p): Likewise.
7 (init_alias_analysis): Likewise.
8 * cfgexpand.c (expand_debug_expr): Likewise.
9 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
10 * cse.c (fold_rtx): Likewise.
11 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
12 * expr.c (emit_block_move_hints): Likewise.
13 (clear_storage_hints, push_block, emit_push_insn): Likewise.
14 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
15 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
16 (emit_group_store): Likewise.
17 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
18 to read the PRE/POST_MODIFY increment.
19 * calls.c (store_one_arg): Use strip_offset.
20 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
22 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
24 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
25 (simplify_binary_operation_1): Extend CONST_INT handling to
27 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
29 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
31 (adjust_mems, add_stores): Update accodingly.
32 (vt_canonicalize_addr): Track polynomial offsets.
33 (emit_note_insn_var_location): Likewise.
34 (vt_add_function_parameter): Likewise.
35 (vt_initialize): Likewise.
37 2018-06-12 Jeff Law <law@redhat.com>
39 * config.gcc (alpha*-*-freebsd*): Remove.
40 * config/alpha/freebsd.h: Remove.
42 2018-06-12 David Malcolm <dmalcolm@redhat.com>
45 * spellcheck-tree.c (levenshtein_distance): Rename to...
46 (get_edit_distance): ...this, and update for underlying renaming.
47 * spellcheck-tree.h (levenshtein_distance): Rename to...
48 (get_edit_distance): ...this.
49 * spellcheck.c (levenshtein_distance): Rename to...
50 (get_edit_distance): ...this. Convert from Levenshtein distance
51 to Damerau-Levenshtein distance by supporting transpositions of
52 adjacent characters. Rename "v1" to "v_next" and "v0" to
54 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
55 (selftest::test_edit_distance_unit_test_oneway): ...this, and
56 update for underlying renaming.
57 (selftest::levenshtein_distance_unit_test): Rename to...
58 (selftest::test_get_edit_distance_unit): ...this, and update for
60 (selftest::test_find_closest_string): Add example from PR 69968
61 where transposition helps
62 (selftest::test_metric_conditions): Update for renaming.
63 (selftest::test_metric_conditions): Likewise.
64 (selftest::spellcheck_c_tests): Likewise.
65 * spellcheck.h (levenshtein_distance): Rename both overloads to...
66 (get_edit_distance): ...this.
67 (best_match::consider): Update for renaming.
69 2018-06-12 Martin Sebor <msebor@redhat.com>
71 PR tree-optimization/85259
72 * builtins.c (compute_objsize): Handle constant offsets.
73 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
74 true iff a warning has been issued.
75 * gimple.h (gimple_nonartificial_location): New function.
76 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
77 gimple_nonartificial_location and handle -Wno-system-headers.
78 (handle_builtin_stxncpy): Same.
80 2018-06-12 Martin Sebor <msebor@redhat.com>
83 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
85 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
87 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
88 BUILTIN_VEC_XST entries for pointer to double and long long.
90 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
93 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
95 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
98 2018-06-12 Martin Liska <mliska@suse.cz>
100 * doc/options.texi: Document IntegerRange.
102 2018-06-12 Martin Liska <mliska@suse.cz>
104 * config/i386/i386.opt: Make MPX-related options as Deprecated.
105 * opt-functions.awk: Handle Deprecated flag.
106 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
108 (read_cmdline_option): Report warning for a deprecated option.
109 * opts.h (struct cl_option): Add new field cl_deprecated.
110 (CL_ERR_DEPRECATED): New.
112 2018-06-12 Martin Liska <mliska@suse.cz>
114 * doc/options.texi: Document Deprecated option flag.
116 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
118 * config/arc/arc-arch.h (arc_extras): New enum.
119 (arc_cpu_t):Add field extra.
120 (arc_cpu_types): Consider the extras.
121 * config/arc/arc-cpus.def: Add extras info.
122 * config/arc/arc-opts.h (processor_type): Consider extra field.
123 * config/arc/arc.c (arc_override_options): Handle extra field.
125 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
127 * config/arc/arc-arch.h: Update ARC_OPTX macro.
128 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
130 * config/arc/arc.c (arc_init): Update pic warning.
131 (irq_range): Update irq range parsing warnings.
132 (arc_override_options): Update various warning messages.
133 (arc_handle_aux_attribute): Likewise.
135 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
137 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
139 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
141 * doc/sourcebuild.texi: Document usage of line number 0 in verify
142 compiler messages directives.
144 2018-06-12 Matthew Fortune <mfortune@gmail.com>
146 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
147 * config/mips/mips-tables.opt: Regenerate.
148 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
150 * doc/invoke.texi: Document -march=i6500.
152 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
154 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
155 (i6400_gpmul): Add cpu_unit.
156 (i6400_gpdiv): Likewise.
157 (i6400_msa_add_d): Update reservations.
158 (i6400_msa_int_add) Likewise.
159 (i6400_msa_short_logic3) Likewise.
160 (i6400_msa_short_logic2) Likewise.
161 (i6400_msa_short_logic) Likewise.
162 (i6400_msa_move) Likewise.
163 (i6400_msa_cmp) Likewise.
164 (i6400_msa_short_float2) Likewise.
165 (i6400_msa_div_d) Likewise.
166 (i6400_msa_long_logic1) Likewise.
167 (i6400_msa_long_logic2) Likewise.
168 (i6400_msa_mult) Likewise.
169 (i6400_msa_long_float2) Likewise.
170 (i6400_msa_long_float4) Likewise.
171 (i6400_msa_long_float5) Likewise.
172 (i6400_msa_long_float8) Likewise.
173 (i6400_fpu_fadd): Include frint type.
174 (i6400_fpu_store): New define_insn_reservation.
175 (i6400_fpu_load): Likewise.
176 (i6400_fpu_move): Likewise.
177 (i6400_fpu_fcmp): Likewise.
178 (i6400_fpu_fmadd): Likewise.
179 (i6400_int_mult): Include imul3nc type and update reservation.
180 (i6400_int_div): Include idiv3 type and update reservation.
181 (i6400_int_load): Update to check type not move_type.
182 (i6400_int_store): Likewise.
183 (i6400_int_prefetch): Set zero latency.
185 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
187 * gcc.c: Document new %@{...} sequence.
188 (LINK_COMMAND_SPEC): Use it for the -L switches.
189 (cpp_unique_options): Use it for the -I switches.
190 (at_file_argbuf): New global variable.
191 (in_at_file): Likewise.
192 (alloc_args): Create at_file_argbuf.
193 (clear_args): Truncate at_file_argbuf.
194 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
195 (open_at_file): New function.
196 (close_at_file): Likewise.
197 (create_at_file): Delete.
198 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
201 (validate_switches_from_spec): Deal with %@{...} sequence.
202 (validate_switches): Likewise.
203 (driver::finalize): Call clear_args.
205 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
207 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
209 2018-06-11 Martin Sebor <msebor@redhat.com>
211 * doc/invoke.texi (-Wall): List -Wc++17-compat.
212 (Wno-class-memaccess): Add @opindex.
213 (Wno-templates, Wno-multiple-inheritance): Same.
214 (Wno-virtual-inheritance, Wno-namespaces): Same.
215 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
216 (Wno-format-overflow, Wno-format-truncation): Same.
217 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
218 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
219 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
220 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
221 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
222 (Wno-misspelled-isr): Same.
224 2018-06-11 Martin Sebor <msebor@redhat.com>
226 * PR tree-optimization/86083
227 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
229 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
231 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
233 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
236 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
237 on the correct operand.
238 (*movdi_internal64): Ditto.
240 2018-06-11 Martin Liska <mliska@suse.cz>
242 PR tree-optimization/86089
243 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
245 2018-06-11 Julia Koval <julia.koval@intel.com>
247 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
248 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
249 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
251 2018-06-11 Olivier Hainque <hainque@adacore.com>
253 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
254 for Ada with strict dwarf2.
256 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
259 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
262 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
264 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
266 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
268 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
271 2018-06-08 Martin Liska <mliska@suse.cz>
273 * tree-cfg.h (debug_function): Fix argument type to match
276 2018-06-08 Martin Liska <mliska@suse.cz>
278 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
279 Remove usage of MPX-related (and removed) fields.
280 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
282 2018-06-08 David Malcolm <dmalcolm@redhat.com>
284 * cfg.c (debug): Use TDF_NONE rather than 0.
285 * cfghooks.c (debug): Likewise.
286 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
287 (struct dump_option_value_info): Convert to...
288 (struct kv_pair): ...this template type.
289 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
291 (optinfo_verbosity_options): Likewise.
292 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
294 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
295 than int for "optgroup_flags" param.
296 (dump_generic_expr_loc): Use dump_flags_t rather than int for
298 (dump_dec): Likewise.
299 (dump_finish): Use TDF_NONE rather than 0.
300 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
301 rather than int for "optgroup_flags" param. Use TDF_NONE rather
302 than 0. Update for change to option_ptr.
303 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
304 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
305 0. Update for changes to optinfo_verbosity_options and
307 (opt_info_switch_p): Convert optgroup_flags from int to
309 (dump_basic_block): Use dump_flags_t rather than int
310 for "dump_kind" param.
311 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
312 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
313 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
314 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
315 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
316 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
317 TDF_NONE): Convert from macros to...
318 (enum dump_flag): ...this new enum.
319 (dump_flags_t): Update to use enum.
320 (operator|, operator&, operator~, operator|=, operator&=):
321 Implement for dump_flags_t.
322 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
323 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
324 Convert from macros to...
325 (enum optgroup_flag): ...this new enum.
326 (optgroup_flags_t): New typedef.
327 (operator|, operator|=): Implement for optgroup_flags_t.
328 (struct dump_file_info): Convert field "alt_flags" to
329 dump_flags_t. Convert field "optgroup_flags" to
331 (dump_basic_block): Use dump_flags_t rather than int for param.
332 (dump_generic_expr_loc): Likewise.
333 (dump_dec): Likewise.
334 (dump_register): Convert param "optgroup_flags" to
336 (opt_info_enable_passes): Likewise.
337 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
339 * gimple-pretty-print.c (debug): Likewise.
340 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
341 (merged_store_group::apply_stores): Likewise.
342 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
343 * gimple.c (verify_gimple_pp): Likewise.
344 * graphite-poly.c (print_pbb_body): Likewise.
345 * passes.c (pass_manager::register_one_dump_file): Convert
346 local "optgroup_flags" to optgroup_flags_t.
347 * print-tree.c (print_node): Use TDF_NONE rather than 0.
349 (debug_body): Likewise.
350 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
352 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
354 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
355 (convert_mult_to_fma): Likewise.
356 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
357 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
358 * tree-vect-data-refs.c (dump_lower_bound): Convert param
359 "dump_kind" to dump_flags_t.
361 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
363 * config/rs6000/rs6000.c (min, max): Delete.
365 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
367 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
368 -mabi=spe and -mabi=no-spe.
370 2018-06-08 Martin Liska <mliska@suse.cz>
372 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
373 where we expect an existing summary.
375 2018-06-08 Martin Liska <mliska@suse.cz>
377 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
378 * ipa-inline.h (estimate_edge_growth): Likewise.
380 2018-06-08 Martin Liska <mliska@suse.cz>
382 * cgraph.c (function_version_hasher::hash): Use
383 cgraph_node::get_uid ().
384 (function_version_hasher::equal):
385 * cgraph.h (cgraph_node::get_uid): New method.
386 * ipa-inline.c (update_caller_keys): Use
387 cgraph_node::get_uid ().
388 (update_callee_keys): Likewise.
389 * ipa-utils.c (searchc): Likewise.
390 (ipa_reduced_postorder): Likewise.
391 * lto-cgraph.c (input_node): Likewise.
392 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
393 * symbol-summary.h (symtab_insertion): Likewise.
394 (symtab_removal): Likewise.
395 (symtab_duplication): Likewise.
396 * tree-pretty-print.c (dump_function_header): Likewise.
397 * tree-sra.c (convert_callers_for_node): Likewise.
399 2018-06-08 Martin Liska <mliska@suse.cz>
401 * cgraph.c (symbol_table::create_edge): Always assign a new
403 (symbol_table::free_edge): Do not recycle numbers.
404 * cgraph.h (cgraph_edge::get): New method.
405 * symbol-summary.h (symtab_removal): Use it.
406 (symtab_duplication): Likewise.
407 (call_summary::hashable_uid): Remove.
409 2018-06-08 Martin Liska <mliska@suse.cz>
411 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
412 (initialize_growth_caches): Remove.
413 (free_growth_caches): Likewise.
414 (do_estimate_edge_time): Use edge_growth_cache.
415 (do_estimate_edge_size): Likewise.
416 (do_estimate_edge_hints): Likewise.
417 * ipa-inline.c (reset_edge_caches): Likewise.
418 (recursive_inlining): Likewise.
419 (inline_small_functions): Likewise.
420 * ipa-inline.h (initialize_growth_caches): Remove.
421 (estimate_edge_size): Likewise.
422 (estimate_edge_time): Likewise.
423 (estimate_edge_hints): Likewise.
424 (reset_edge_growth_cache): Likewise.
425 * symbol-summary.h (call_summary::remove): New method.
427 2018-06-08 Martin Liska <mliska@suse.cz>
429 * ipa-cp.c (class edge_clone_summary): New summary.
430 (grow_edge_clone_vectors): Remove.
431 (ipcp_edge_duplication_hook): Remove.
432 (class edge_clone_summary_t): New call_summary class.
433 (ipcp_edge_removal_hook): Remove.
434 (edge_clone_summary_t::duplicate): New function.
435 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
436 (create_specialized_node): Likewise.
437 (ipcp_driver): Initialize edge_clone_summaries and do not
440 2018-06-08 Martin Liska <mliska@suse.cz>
442 * symbol-summary.h (get): New function.
443 (call_summary::m_initialize_when_cloning): New class member.
445 2018-06-08 Martin Liska <mliska@suse.cz>
447 * cgraph.c (cgraph_node::remove): Do not recycle uid.
448 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
449 (symbol_table::allocate_cgraph_symbol): Do not set uid.
450 * passes.c (uid_hash_t): Record removed_nodes by their uids.
451 (remove_cgraph_node_from_order): Use the removed_nodes set.
452 (do_per_function_toporder): Likwise.
453 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
454 instead of summary_uid.
455 (symtab_removal): Likewise.
456 (symtab_duplication): Likewise.
458 2018-06-08 Martin Liska <mliska@suse.cz>
460 * ipa-cp.c (ipcp_store_bits_results): Use
461 ipcp_transformation_sum.
462 (ipcp_store_vr_results): Likewise.
463 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
465 (ipcp_transformation_initialize): ... this.
466 (ipa_set_node_agg_value_chain):
467 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
468 (write_ipcp_transformation_info): Likewise.
469 (read_ipcp_transformation_info): Likewise.
470 (ipcp_update_bits): Likewise.
471 (ipcp_update_vr): Likewise.
472 (ipcp_transform_function): Likewise.
473 * ipa-prop.h: Rename ipcp_transformation_summary to
475 (class ipcp_transformation_t): New function summary.
476 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
477 (ipa_get_agg_replacements_for_node): Likewise.
479 2018-06-08 Martin Liska <mliska@suse.cz>
481 * ipa-pure-const.c (struct funct_state_d): Do it class instead
483 (class funct_state_summary_t): New function_summary class.
484 (has_function_state): Remove.
485 (get_function_state): Likewise.
486 (set_function_state): Likewise.
487 (add_new_function): Likewise.
488 (funct_state_summary_t::insert): New function.
489 (duplicate_node_data): Remove.
490 (remove_node_data): Remove.
491 (funct_state_summary_t::duplicate): New function.
492 (register_hooks): Create new funct_state_summaries.
493 (pure_const_generate_summary): Use it.
494 (pure_const_write_summary): Likewise.
495 (pure_const_read_summary): Likewise.
496 (propagate_pure_const): Likewise.
497 (propagate_nothrow): Likewise.
498 (dump_malloc_lattice): Likewise.
499 (propagate_malloc): Likewise.
500 (execute): Do not register hooks, just remove summary
502 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
505 2018-06-08 Martin Liska <mliska@suse.cz>
507 * ipa-reference.c (remove_node_data): Remove.
508 (duplicate_node_data): Likewise.
509 (class ipa_ref_var_info_summary_t): New class.
510 (class ipa_ref_opt_summary_t): Likewise.
511 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
512 (get_reference_optimization_summary): Use
513 ipa_ref_opt_sum_summaries.
514 (set_reference_vars_info): Remove.
515 (set_reference_optimization_summary): Likewise.
516 (ipa_init): Create summaries.
517 (init_function_info): Use function summary.
518 (ipa_ref_opt_summary_t::duplicate): New function.
519 (ipa_ref_opt_summary_t::remove): New function.
520 (get_read_write_all_from_node): Fix GNU coding style.
521 (propagate): Use function summary.
522 (write_node_summary_p): Fix GNU coding style.
523 (stream_out_bitmap): Likewise.
524 (ipa_reference_read_optimization_summary): Use function summary.
525 (ipa_reference_c_finalize): Do not release hooks.
527 2018-06-08 Martin Liska <mliska@suse.cz>
529 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
530 (analyze_function_body): Extract multiple calls of get_create.
531 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
532 * ipa-inline.c (recursive_inlining): Use ::get method.
533 * ipa-inline.h (estimate_edge_growth): Likewise.
535 2018-06-08 Martin Liska <mliska@suse.cz>
537 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
539 (hsa_function_summary::hsa_function_summary): Use the new enum
541 (hsa_gpu_implementation_p): Use hsa_summaries::get.
542 * hsa-gen.c (hsa_get_host_function): Likewise.
543 (get_brig_function_name): Likewise.
544 * ipa-hsa.c (process_hsa_functions): Likewise.
545 (ipa_hsa_write_summary): Likewise.
546 * symbol-summary.h (symtab_duplication): Use ::get function/
549 2018-06-08 Martin Liska <mliska@suse.cz>
551 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
553 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
554 (hsa_register_kernel): Likewise.
555 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
556 * hsa-gen.c (hsa_get_host_function): Likewise.
557 (get_brig_function_name): Likewise.
558 (generate_hsa): Likewise.
559 (pass_gen_hsail::execute): Likewise.
560 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
561 (devirtualization_time_bonus): Likewise.
562 (ipcp_propagate_stage): Likewise.
563 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
564 (edge_set_predicate): Likewise.
565 (evaluate_conditions_for_known_args): Likewise.
566 (evaluate_properties_for_edge): Likewise.
567 (ipa_fn_summary::reset): Likewise.
568 (ipa_fn_summary_t::duplicate): Likewise.
569 (dump_ipa_call_summary): Likewise.
570 (ipa_dump_fn_summary): Likewise.
571 (analyze_function_body): Likewise.
572 (compute_fn_summary): Likewise.
573 (estimate_edge_devirt_benefit): Likewise.
574 (estimate_edge_size_and_time): Likewise.
575 (estimate_calls_size_and_time): Likewise.
576 (estimate_node_size_and_time): Likewise.
577 (inline_update_callee_summaries): Likewise.
578 (remap_edge_change_prob): Likewise.
579 (remap_edge_summaries): Likewise.
580 (ipa_merge_fn_summary_after_inlining): Likewise.
581 (ipa_update_overall_fn_summary): Likewise.
582 (read_ipa_call_summary): Likewise.
583 (inline_read_section): Likewise.
584 (write_ipa_call_summary): Likewise.
585 (ipa_fn_summary_write): Likewise.
586 (ipa_free_fn_summary): Likewise.
587 * ipa-hsa.c (process_hsa_functions): Likewise.
588 (ipa_hsa_write_summary): Likewise.
589 (ipa_hsa_read_section): Likewise.
590 * ipa-icf.c (sem_function::merge): Likewise.
591 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
592 (do_estimate_edge_time): Likewise.
593 (estimate_size_after_inlining): Likewise.
594 (estimate_growth): Likewise.
595 (growth_likely_positive): Likewise.
596 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
597 (inline_call): Likewise.
598 * ipa-inline.c (caller_growth_limits): Likewise.
599 (can_inline_edge_p): Likewise.
600 (can_inline_edge_by_limits_p): Likewise.
601 (compute_uninlined_call_time): Likewise.
602 (compute_inlined_call_time): Likewise.
603 (want_inline_small_function_p): Likewise.
604 (edge_badness): Likewise.
605 (update_caller_keys): Likewise.
606 (update_callee_keys): Likewise.
607 (recursive_inlining): Likewise.
608 (inline_small_functions): Likewise.
609 (inline_to_all_callers_1): Likewise.
610 (dump_overall_stats): Likewise.
611 (early_inline_small_functions): Likewise.
612 (early_inliner): Likewise.
613 * ipa-inline.h (estimate_edge_growth): Likewise.
614 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
615 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
616 * ipa-prop.h (IPA_NODE_REF): Likewise.
617 (IPA_EDGE_REF): Likewise.
618 * ipa-pure-const.c (malloc_candidate_p): Likewise.
619 (propagate_malloc): Likewise.
620 * ipa-split.c (execute_split_functions): Likewise.
621 * symbol-summary.h: Rename get to get_create.
623 (get_create): Likewise.
624 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
626 2018-06-08 Martin Liska <mliska@suse.cz>
628 * symbol-summary.h (release): Move definition out of class
630 (symtab_removal): Likewise.
631 (symtab_duplication): Likewise.
633 2018-06-08 Martin Liska <mliska@suse.cz>
635 * symbol-summary.h (function_summary): Move constructor
636 implementation out of class declaration.
638 (symtab_insertion): Likewise.
639 (symtab_removal): Likewise.
640 (symtab_duplication): Likewise.
643 2018-06-08 Martin Liska <mliska@suse.cz>
645 * Makefile.in: Remove support for MPX (macros, related functions,
646 fields in cgraph_node, ...).
647 * builtin-types.def (BT_BND): Likewise.
648 (BT_FN_BND_CONST_PTR): Likewise.
649 (BT_FN_CONST_PTR_BND): Likewise.
650 (BT_FN_VOID_PTR_BND): Likewise.
651 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
652 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
653 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
654 (expand_builtin_mempcpy_with_bounds): Likewise.
655 (expand_builtin_memset_with_bounds): Likewise.
656 (expand_builtin_memset_args): Likewise.
657 (std_expand_builtin_va_start): Likewise.
658 (expand_builtin): Likewise.
659 (expand_builtin_with_bounds): Likewise.
660 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
661 (DEF_LIB_BUILTIN_CHKP): Likewise.
662 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
663 (DEF_CHKP_BUILTIN): Likewise.
664 (BUILT_IN_MEMCPY): Likewise.
665 (BUILT_IN_MEMMOVE): Likewise.
666 (BUILT_IN_MEMPCPY): Likewise.
667 (BUILT_IN_MEMSET): Likewise.
668 (BUILT_IN_STPCPY): Likewise.
669 (BUILT_IN_STRCAT): Likewise.
670 (BUILT_IN_STRCHR): Likewise.
671 (BUILT_IN_STRCPY): Likewise.
672 (BUILT_IN_STRLEN): Likewise.
673 (BUILT_IN_MEMCPY_CHK): Likewise.
674 (BUILT_IN_MEMMOVE_CHK): Likewise.
675 (BUILT_IN_MEMPCPY_CHK): Likewise.
676 (BUILT_IN_MEMSET_CHK): Likewise.
677 (BUILT_IN_STPCPY_CHK): Likewise.
678 (BUILT_IN_STRCAT_CHK): Likewise.
679 (BUILT_IN_STRCPY_CHK): Likewise.
680 * calls.c (store_bounds): Likewise.
681 (emit_call_1): Likewise.
682 (special_function_p): Likewise.
683 (maybe_warn_nonstring_arg): Likewise.
684 (initialize_argument_information): Likewise.
685 (finalize_must_preallocate): Likewise.
686 (compute_argument_addresses): Likewise.
687 (expand_call): Likewise.
688 * cfgexpand.c (expand_call_stmt): Likewise.
689 (expand_return): Likewise.
690 (expand_gimple_stmt_1): Likewise.
691 (pass_expand::execute): Likewise.
692 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
693 (cgraph_node::remove): Likewise.
694 (cgraph_node::dump): Likewise.
695 (cgraph_node::verify_node): Likewise.
696 * cgraph.h (chkp_function_instrumented_p): Likewise.
697 (symtab_node::get_alias_target): Likewise.
698 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
699 (cgraph_local_p): Likewise.
700 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
701 (cgraph_edge::rebuild_references): Likewise.
702 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
703 (walk_polymorphic_call_targets): Likewise.
704 (cgraph_node::expand_thunk): Likewise.
705 (symbol_table::output_weakrefs): Likewise.
706 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
707 (ix86_handle_option): Likewise.
708 * config/i386/constraints.md: Likewise.
709 * config/i386/i386-builtin-types.def (BND): Likewise.
713 * config/i386/i386-builtin.def (BDESC_END): Likewise.
714 (BDESC_FIRST): Likewise.
716 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
717 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
718 * config/i386/i386.c (enum reg_class): Likewise.
719 (ix86_target_string): Likewise.
720 (ix86_option_override_internal): Likewise.
721 (ix86_conditional_register_usage): Likewise.
722 (ix86_valid_target_attribute_inner_p): Likewise.
723 (ix86_set_indirect_branch_type): Likewise.
724 (ix86_set_current_function): Likewise.
725 (ix86_function_arg_regno_p): Likewise.
726 (init_cumulative_args): Likewise.
727 (ix86_function_arg_advance): Likewise.
728 (ix86_function_arg): Likewise.
729 (ix86_pass_by_reference): Likewise.
730 (ix86_function_value_regno_p): Likewise.
731 (ix86_function_value_1): Likewise.
732 (ix86_function_value_bounds): Likewise.
733 (ix86_return_in_memory): Likewise.
734 (ix86_setup_incoming_vararg_bounds): Likewise.
735 (ix86_va_start): Likewise.
736 (indirect_thunk_need_prefix): Likewise.
737 (print_reg): Likewise.
738 (ix86_print_operand): Likewise.
739 (ix86_expand_call): Likewise.
740 (ix86_output_function_return): Likewise.
741 (reg_encoded_number): Likewise.
742 (BDESC_VERIFYS): Likewise.
743 (ix86_init_mpx_builtins): Likewise.
744 (ix86_init_builtins): Likewise.
745 (ix86_emit_cmove): Likewise.
746 (ix86_emit_move_max): Likewise.
747 (ix86_expand_builtin): Likewise.
748 (ix86_builtin_mpx_function): Likewise.
749 (ix86_get_arg_address_for_bt): Likewise.
750 (ix86_load_bounds): Likewise.
751 (ix86_store_bounds): Likewise.
752 (ix86_load_returned_bounds): Likewise.
753 (ix86_store_returned_bounds): Likewise.
754 (ix86_class_likely_spilled_p): Likewise.
755 (ix86_hard_regno_mode_ok): Likewise.
756 (x86_order_regs_for_local_alloc): Likewise.
757 (ix86_mitigate_rop): Likewise.
758 (ix86_bnd_prefixed_insn_p): Likewise.
759 (ix86_mpx_bound_mode): Likewise.
760 (ix86_make_bounds_constant): Likewise.
761 (ix86_initialize_bounds): Likewise.
762 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
763 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
764 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
765 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
766 (TARGET_CHKP_BOUND_MODE): Likewise.
767 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
768 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
769 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
770 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
771 * config/i386/i386.h (TARGET_MPX): Likewise.
772 (TARGET_MPX_P): Likewise.
773 (VALID_BND_REG_MODE): Likewise.
774 (FIRST_BND_REG): Likewise.
775 (LAST_BND_REG): Likewise.
776 (enum reg_class): Likewise.
777 (BND_REG_P): Likewise.
778 (BND_REGNO_P): Likewise.
780 (ADJUST_INSN_LENGTH): Likewise.
781 * config/i386/i386.md: Likewise.
782 * config/i386/i386.opt: Likewise.
783 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
785 (LINK_MPX): Likewise.
786 (MPX_SPEC): Likewise.
787 (LIBMPX_SPEC): Likewise.
788 (LIBMPXWRAPPERS_SPEC): Likewise.
789 (CHKP_SPEC): Likewise.
790 * config/i386/predicates.md: Likewise.
791 * dbxout.c (dbxout_type): Likewise.
792 * doc/extend.texi: Likewise.
793 * doc/invoke.texi: Likewise.
794 * doc/md.texi: Likewise.
795 * doc/tm.texi: Likewise.
796 * doc/tm.texi.in: Likewise.
797 * dwarf2out.c (is_base_type): Likewise.
798 (gen_formal_types_die): Likewise.
799 (gen_subprogram_die): Likewise.
800 (gen_type_die_with_usage): Likewise.
801 (gen_decl_die): Likewise.
802 (dwarf2out_late_global_decl): Likewise.
803 * expr.c (expand_assignment): Likewise.
804 (emit_storent_insn): Likewise.
805 (store_expr_with_bounds): Likewise.
806 (store_expr): Likewise.
807 (expand_expr_real_1): Likewise.
808 * expr.h (store_expr_with_bounds): Likewise.
809 * function.c (use_register_for_decl): Likewise.
810 (struct bounds_parm_data): Likewise.
811 (assign_parms_augmented_arg_list): Likewise.
812 (assign_parm_find_entry_rtl): Likewise.
813 (assign_parm_is_stack_parm): Likewise.
814 (assign_parm_load_bounds): Likewise.
815 (assign_bounds): Likewise.
816 (assign_parms): Likewise.
817 (expand_function_start): Likewise.
818 * gcc.c (CHKP_SPEC): Likewise.
819 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
820 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
821 (wrestrict_dom_walker::check_call): Likewise.
822 * gimple.c (gimple_build_call_from_tree): Likewise.
823 * gimple.h (enum gf_mask): Likewise.
824 (gimple_call_with_bounds_p): Likewise.
825 (gimple_call_set_with_bounds): Likewise.
826 * gimplify.c (gimplify_init_constructor): Likewise.
827 * ipa-cp.c (initialize_node_lattices): Likewise.
828 (propagate_constants_across_call): Likewise.
829 (find_more_scalar_values_for_callers_subset): Likewise.
830 * ipa-hsa.c (process_hsa_functions): Likewise.
831 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
832 * ipa-icf.c (sem_function::merge): Likewise.
833 * ipa-inline.c (early_inliner): Likewise.
834 * ipa-pure-const.c (warn_function_noreturn): Likewise.
835 (warn_function_cold): Likewise.
836 (propagate_pure_const): Likewise.
837 * ipa-ref.h (enum GTY): Likewise.
838 * ipa-split.c (find_retbnd): Likewise.
839 (consider_split): Likewise.
840 (split_function): Likewise.
841 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
842 * ipa.c (walk_polymorphic_call_targets): Likewise.
843 (symbol_table::remove_unreachable_nodes): Likewise.
844 (process_references): Likewise.
845 (cgraph_build_static_cdtor_1): Likewise.
846 * lto-cgraph.c (lto_output_node): Likewise.
847 (output_refs): Likewise.
848 (compute_ltrans_boundary): Likewise.
849 (input_overwrite_node): Likewise.
850 (input_node): Likewise.
851 (input_cgraph_1): Likewise.
852 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
853 * passes.c (pass_manager::execute_early_local_passes): Likewise.
854 (class pass_chkp_instrumentation_passes): Likewise.
855 (make_pass_chkp_instrumentation_passes): Likewise.
856 * passes.def: Likewise.
857 * rtl.h (struct GTY): Likewise.
858 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
859 * stor-layout.c (layout_type): Likewise.
860 * symtab.c: Likewise.
861 * target.def: Likewise.
862 * targhooks.c (default_chkp_bound_type): Likewise.
863 (default_chkp_bound_mode): Likewise.
864 (default_builtin_chkp_function): Likewise.
865 (default_chkp_function_value_bounds): Likewise.
866 (default_chkp_make_bounds_constant): Likewise.
867 (default_chkp_initialize_bounds): Likewise.
868 * targhooks.h (default_chkp_bound_type): Likewise.
869 (default_chkp_bound_mode): Likewise.
870 (default_builtin_chkp_function): Likewise.
871 (default_chkp_function_value_bounds): Likewise.
872 (default_chkp_make_bounds_constant): Likewise.
873 (default_chkp_initialize_bounds): Likewise.
874 * toplev.c (compile_file): Likewise.
875 (process_options): Likewise.
876 * tree-core.h (DEF_BUILTIN): Likewise.
877 (DEF_BUILTIN_CHKP): Likewise.
878 * tree-inline.c (declare_return_variable): Likewise.
879 (remap_gimple_stmt): Likewise.
881 (initialize_inlined_parameters): Likewise.
882 (expand_call_inline): Likewise.
883 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
884 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
885 (make_pass_ipa_chkp_produce_thunks): Likewise.
886 (make_pass_chkp): Likewise.
887 (make_pass_chkp_opt): Likewise.
888 (make_pass_chkp_instrumentation_passes): Likewise.
889 * tree-pretty-print.c (dump_generic_node): Likewise.
890 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
891 * tree-ssa-dce.c (propagate_necessity): Likewise.
892 (eliminate_unnecessary_stmts): Likewise.
893 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
894 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
895 * tree-ssa-sccvn.h: Likewise.
896 * tree-ssa-strlen.c (get_string_length): Likewise.
897 (valid_builtin_call): Likewise.
898 (adjust_last_stmt): Likewise.
899 (handle_builtin_strchr): Likewise.
900 (handle_builtin_strcpy): Likewise.
901 (handle_builtin_stxncpy): Likewise.
902 (handle_builtin_memcpy): Likewise.
903 (handle_builtin_strcat): Likewise.
904 (strlen_check_and_optimize_stmt): Likewise.
905 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
906 * tree-streamer-in.c: Likewise.
907 * tree-streamer.c (record_common_node): Likewise.
908 * tree.c (tree_code_size): Likewise.
909 (wide_int_to_tree_1): Likewise.
910 (type_contains_placeholder_1): Likewise.
911 (build_common_tree_nodes): Likewise.
912 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
913 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
914 (POINTER_BOUNDS_P): Likewise.
915 (BOUNDED_TYPE_P): Likewise.
916 (BOUNDED_P): Likewise.
917 (CALL_WITH_BOUNDS_P): Likewise.
918 (pointer_bounds_type_node): Likewise.
919 * value-prof.c (gimple_ic): Likewise.
920 * var-tracking.c (vt_add_function_parameters): Likewise.
921 * varasm.c (make_decl_rtl): Likewise.
922 (assemble_start_function): Likewise.
923 (output_constant): Likewise.
924 (maybe_assemble_visibility): Likewise.
925 * varpool.c (ctor_for_folding): Likewise.
926 * chkp-builtins.def: Remove.
927 * ipa-chkp.c: Remove.
928 * ipa-chkp.h: Remove.
929 * rtl-chkp.c: Remove.
930 * rtl-chkp.h: Remove.
931 * tree-chkp-opt.c: Remove.
932 * tree-chkp.c: Remove.
933 * tree-chkp.h: Remove.
935 2018-06-07 Carl Love <cel@us.ibm.com>
937 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
938 vextract_fp_from_shortl): Add BE support.
940 2018-06-07 Paul Koning <ni1d@arrl.net>
942 * compare-elim.c (try_merge_compare): Don't merge compare if
943 address contains a side effect.
944 (try_eliminate_compare): Likewise.
946 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
948 * config.gcc: Support "tremont".
949 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
950 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
952 * config/i386/i386.c (m_TREMONT): Define.
953 (processor_target_table): Add "tremont".
954 (PTA_TREMONT): Define.
955 (ix86_lea_outperforms): Add TARGET_TREMONT.
956 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
957 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
958 and M_INTEL_GOLDMONT_PLUS.
959 (fold_builtin_cpu): Add "tremont".
960 (ix86_add_stmt_cost): Add TARGET_TREMONT.
961 (ix86_option_override_internal): Add "tremont".
962 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
963 (processor_type): Add PROCESSOR_TREMONT.
964 * config/i386/x86-tune.def: Add m_TREMONT.
965 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
967 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
969 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
970 symbol defined for msp430i* devices to be lower case.
972 gcc/testsuite/gcc.target/msp430/
973 * msp430i-device-symbol.c: New test.
974 * msp430f-device-symbol.c: New test.
975 * msp430.h: New test header file.
977 2018-06-07 Richard Biener <rguenther@suse.de>
979 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
980 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
981 Properly wrap signed arithmetic if overflow wraps.
983 2018-06-07 Jakub Jelinek <jakub@redhat.com>
985 PR tree-optimization/69615
986 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
987 of a cast from a same precision integral SSA_NAME in a bb dominated
988 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
989 cast to utype if rhs2 has already a compatible type.
991 2018-06-07 Richard Biener <rguenther@suse.de>
993 PR tree-optimization/85935
994 * graphite-scop-detection.c (find_params_in_bb): Analyze
995 condition operands with respect to the correct loop. Assert
996 the analysis doesn't fail.
998 2018-06-04 Carl Love <cel@us.ibm.com>
1000 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
1001 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
1002 as it is slightly cheaper.
1003 (first_match_or_eos_index_<mode>):
1004 Calculate index using natural element order.
1005 (first_match_index_<mode>):
1006 Calculate index using natural element order.
1007 (first_match_or_eos_index_<mode>):
1008 Calculate index using natural order.
1009 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
1010 for BE and LE modes.
1011 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
1012 P9V_BUILTIN_VCLZLSBB_V16QI.
1013 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
1016 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1018 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
1019 indentation and line wrap for many prototypes. Add missing
1020 @smallexample directives around block of prototypes for vec_xl and
1023 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
1025 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
1026 track if we pass or return IEEE 128-bit floating point.
1027 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
1028 C++ mangling that is compatible with GCC 8.1.
1029 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
1030 (init_cumulative_args): Note if we pass or return IEEE 128-bit
1031 floating point types.
1032 (rs6000_function_arg_advance_1): Likewise.
1033 (rs6000_mangle_type): Optionally generate mangled names that match
1034 what GCC 8.1 generated for IEEE 128-bit floating point types.
1035 (rs6000_globalize_decl_name): If we have an external function that
1036 passes or returns IEEE 128-bit types, generate a weak reference
1037 from the mangled name used in GCC 8.1 to the current mangled
1039 (rs6000_init_builtins): Make __ibm128 use the long double type if
1040 long double is IBM extended double. Make __float128 use the long
1041 double type if long double is IEEE 128-bit.
1044 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
1045 macro for __ibm128 built-in functions.
1046 (PACK_IF): Add __ibm128 pack/unpack functions.
1047 (UNPACK_IF): Likewise.
1048 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
1049 enable long double built-in functions if long double is IEEE
1050 128-bit floating point.
1051 (rs6000_invalid_builtin): Update long double built-in function
1053 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
1054 functions, adjust the built-in function to use the long double
1055 built-in function if __ibm128 and long double are the same type.
1056 * doc/extend.texi (PowerPC builtins): Update documention for
1057 __builtin_{,un}pack_longdouble. Add documentation for
1058 __builtin_{,un}pack_ibm128.
1060 2018-06-06 Jim Wilson <jimw@sifive.com>
1062 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
1063 (struct machine_function): New field interrupt_mode.
1064 (riscv_handle_type_attribute): New function. Add forward declaration.
1065 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
1066 (riscv_expand_epilogue): Check interrupt_mode field.
1067 (riscv_set_current_function): Check interrupt attribute args and
1068 set interrupt_mode field.
1069 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
1070 (riscv_sret, riscv_uret): New.
1071 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
1072 new arguments to interrupt attribute.
1074 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
1077 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
1078 Don't handle -mcpu=power8 if -mpower9-vector is also used.
1080 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1082 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
1083 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1084 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1085 several redundant entries.
1087 2018-06-06 David Malcolm <dmalcolm@redhat.com>
1089 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1090 type from "rtx" to "rtx_insn *".
1091 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1092 for local "call_insn", removing cast.
1093 (ix86_expand_call): Likewise, introducing a "call_insn" local.
1095 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
1097 PR tree-optimization/86066
1098 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1099 for BIT_INSERT_EXPR stores.
1101 2018-06-06 Richard Biener <rguenther@suse.de>
1103 PR tree-optimization/86062
1104 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1105 component refs ontop
1106 of to be offsetted base.
1108 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1110 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1111 to be static and remove check on interrupt attribute name.
1113 gcc/testsuite/gcc.target/msp430/
1114 * function-attributes-4.c: New test.
1115 * static-interrupts.c: New test.
1117 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1119 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1120 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1122 2018-06-05 Steve Ellcey <sellcey@cavium.com>
1125 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1127 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1128 Remove second argument, change how error is called.
1129 (aarch64_layout_arg): Remove second argument from
1130 aarch64_err_no_fpadvsimd call.
1131 (aarch64_init_cumulative_args): Ditto.
1132 (aarch64_gimplify_va_arg_expr): Ditto.
1133 * config/aarch64/aarch64.md (mov<mode>): Ditto.
1135 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
1137 * config/i386/i386.md (simple_return_indirect_internal): New expander.
1138 (*simple_return_indirect_internal<mode>): Rename from
1139 simple_return_indirect_internal. Use W mode iterator.
1140 (rstorssp): New expander.
1141 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
1142 (clrssbsy): New expander.
1143 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
1145 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
1147 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1149 (cmse_check_pointed_object): Likewise.
1151 2018-06-05 Martin Liska <mliska@suse.cz>
1153 PR gcov-profile/47618
1154 * doc/invoke.texi: Document how -fprofile-dir format
1157 2018-06-05 Richard Biener <rguenther@suse.de>
1159 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1160 removal pretend DOM info isn't available so we do not update
1161 it and only remove edges, not dominated blocks. Actually free
1162 DOM info in case we removed something. Remove unreachable blocks.
1163 (mfb_keep_latches): Work with either DOM info or marked backedges.
1164 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1165 first. Mark backedges if DOM info isn't available.
1166 (Re-)compute DOM info after cleanup_control_flow_pre.
1168 2018-06-05 Richard Biener <rguenther@suse.de>
1170 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1171 (locus_discrim_hasher::hash): Adjust.
1172 (locus_discrim_hasher::equal): Likewise.
1173 (next_discriminator_for_locus): Work on line directly.
1174 (same_line_p): Pass in expanded locus1 as well.
1175 (assign_discriminators): Avoid redundant location expansions.
1177 2018-06-05 Richard Biener <rguenther@suse.de>
1179 PR tree-optimization/86046
1180 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1181 if required after clearing TREE_ADDRESSABLE.
1183 2018-06-05 Richard Biener <rguenther@suse.de>
1185 PR tree-optimization/86047
1186 * tree-ssa-loop.c (for_each_index): Glob handling of all
1187 decls and constants and really handle all of them.
1189 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1192 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1193 qualifier_void_pointer and qualifier_const_void_pointer.
1194 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1195 (arm_init_builtins): Handle the above.
1196 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1197 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1198 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1201 2018-06-05 Martin Liska <mliska@suse.cz>
1203 * auto-profile.c (read_autofdo_file): Do not use
1204 gcov_ctr_summary struct.
1205 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1206 * coverage.c (struct counts_entry): Likewise.
1207 (read_counts_file): Read just single summary entry.
1208 (get_coverage_counts): Use gcov_summary struct.
1209 * coverage.h (get_coverage_counts): Likewise.
1210 * gcov-dump.c (dump_working_sets): Likewise.
1211 (tag_summary): Dump just single summary.
1212 * gcov-io.c (gcov_write_summary): Write just histogram
1214 (gcov_read_summary): Read just single summary.
1215 (compute_working_sets): Use gcov_summary struct.
1216 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1217 of GCOV_COUNTERS_SUMMABLE.
1218 (GCOV_COUNTERS_SUMMABLE): Remove.
1219 (GCOV_FIRST_VALUE_COUNTER): Replace with
1220 GCOV_COUNTER_V_INTERVAL.
1221 (struct gcov_ctr_summary): Remove.
1222 (struct gcov_summary): Directly use fields of former
1224 (compute_working_sets): Use gcov_summary struct.
1225 * gcov.c (read_count_file): Do not use ctrs fields.
1226 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1228 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1230 * profile.c: Likewise.
1231 * profile.h: Likewise.
1233 2018-06-05 Martin Liska <mliska@suse.cz>
1235 PR gcov-profile/84846
1236 * gcov.c (output_lines): Print working directory only
1237 in intermediate format.
1239 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
1241 * config/s390/s390-builtin-types.def: Add void function type.
1242 * config/s390/s390-builtins.def: Use the function type for the
1245 2018-06-04 Jim Wilson <jimw@sifive.com>
1247 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
1249 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
1250 and maybe_eh_return. Change regno to unsigned int. Use new args to
1251 handle EH_RETURN_DATA_REGNO registers properly.
1252 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
1253 (riscv_expand_epilogue): Update comment. Change argument name and
1254 type. Update code to use new name and type. Pass new args to
1255 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
1257 * config/riscv/riscv.md (NORMAL_RETURN): New.
1258 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
1259 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
1260 (eh_return): Call gen_eh_return_internal and emit barrier.
1261 (eh_return_internal): Call riscv_expand_epilogue.
1263 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
1265 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
1266 bit_insertion field and declare can_be_merged_into method.
1267 (merged_store_group::can_be_merged_into): New method.
1268 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
1269 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
1270 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
1272 2018-06-04 Richard Biener <rguenther@suse.de>
1274 PR tree-optimization/85955
1275 * builtins.c (fold_builtin_sincos): Convert pointers to
1276 destination to appropriate type before dereferencing.
1278 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
1280 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
1282 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
1284 * expr.c (expand_expr_real_1): Force the operand into memory if
1285 its TYPE_MODE is BLKmode and if there is no integer mode for
1286 the number of bits being extracted.
1288 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1292 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1293 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
1295 2018-06-04 Richard Biener <rguenther@suse.de>
1297 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
1298 (cleanup_tree_cfg_noloop): ... single caller. Do
1299 start_recording_case_labels later.
1301 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
1303 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
1304 to _IMMINTRIN_H_INCLUDED.
1305 * config/i386/pconfigintrin.h: Ditto.
1306 * config/i386/waitpkgintrin.h: Ditto.
1307 * config/i386/immintrin.h: Add includes for sgxintrin.h,
1308 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
1309 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
1310 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
1311 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
1312 waitpkgintrin.h and cldemoteintrin.h.
1314 2018-06-04 Richard Biener <rguenther@suse.de>
1316 PR tree-optimization/86038
1317 * tracer.c (find_best_successor): Check probability for
1318 being initialized, bail out if not.
1320 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
1323 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
1324 of bits to ignore when comparing architectures.
1326 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1328 PR tree-optimization/69615
1329 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
1330 maximum or minimum of the type, try to merge it also as if
1331 range1 is + [-, x - 1] or + [x + 1, -].
1334 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
1336 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
1338 PR tree-optimization/86034
1339 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
1340 the unsigned bitfield type in a bit insertion sequence if it does not
1341 have a larger precision than the bitfield size.
1342 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
1344 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
1346 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
1348 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1350 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
1351 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
1352 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
1353 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
1355 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1357 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
1358 Disable -fdelete-null-pointer-checks for ELF toolchain.
1360 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1361 Kito Cheng <kito.cheng@gmail.com>
1363 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
1364 (nds32le-*-*, nds32be-*-*): Integrate checking process.
1365 (nds32*-*-*): Add glibc and uclibc conditions.
1366 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
1367 (TARGET_EXCEPT_UNWIND_INFO): Define.
1368 * config/nds32/elf.h: New file.
1369 * config/nds32/linux.h: New file.
1370 * config/nds32/nds32-elf.opt: New file.
1371 * config/nds32/nds32-linux.opt: New file.
1372 * config/nds32/nds32-fp-as-gp.c
1373 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
1374 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
1376 (nds32_asm_file_end): Ditto.
1377 (nds32_print_operand): Ditto.
1378 (nds32_insert_attributes): Ditto.
1379 (nds32_init_libfuncs): New function.
1380 (TARGET_HAVE_TLS): Define.
1381 (TARGET_INIT_LIBFUNCS): Define.
1382 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
1384 (TARGET_ELF): Apply different mcmodel setting.
1385 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
1386 been migrated into elf.h and linux.h files.
1387 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
1388 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
1389 (mcmodel): The content has been migrated into nds32-elf.opt and
1390 nds32-linux.opt files.
1391 * config/nds32/t-elf: New file.
1392 * config/nds32/t-linux: New file.
1394 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1395 Shiva Chen <shiva0217@gmail.com>
1397 * config/nds32/constants.md (unspec_volatile_element): Add
1398 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
1399 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
1401 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
1402 (make_pass_nds32_fp_as_gp): Declare.
1403 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
1405 (nds32_asm_function_end_prologue): Remove unused asm output.
1406 (nds32_asm_function_begin_epilogue): Remove unused asm output.
1407 (nds32_asm_file_start): Output necessary fp_as_gp information.
1408 (nds32_option_override): Adjust register usage.
1409 (nds32_expand_prologue): Consider fp_as_gp situation.
1410 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
1411 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
1414 (simple_return): Ditto.
1415 (omit_fp_begin): Output special directive for fp_as_gp.
1416 (omit_fp_end): Output special directive for fp_as_gp.
1417 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
1418 mforbid-fp-as-gp): New options.
1420 2018-06-01 Mark Wielaard <mark@klomp.org>
1422 * dwarf2out.c (dwarf2out_finish): Remove generation of
1423 DW_AT_loclists_base.
1425 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
1427 * gimple-ssa-store-merging.c: Include gimple-fold.h.
1428 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
1429 (struct merged_store_group): Add bit_insertion field.
1430 (dump_char_array): Use standard hexadecimal format.
1431 (merged_store_group::merged_store_group): Set bit_insertion to false.
1432 (merged_store_group::apply_stores): Use optimal buffer size. Deal
1433 with BIT_INSERT_EXPR stores. Move up code updating the mask and
1434 also print the mask in the dump file.
1435 (pass_store_merging::gate): Minor tweak.
1436 (imm_store_chain_info::coalesce_immediate): Fix wrong association
1437 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
1438 stores with INTEGER_CST stores.
1439 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
1440 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
1441 and use it throughout. Generate bit insertion sequences if need be.
1442 (pass_store_merging::process_store): Remove redundant condition.
1443 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
1445 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
1447 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
1448 the 128-bit floating point types. Fix function comment.
1450 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1452 * config/aarch64/aarch64-simd.md
1453 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
1455 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
1458 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1460 PR tree-optimization/85989
1461 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
1463 (backprop::intersect_uses): Check it when deciding whether this
1464 is a backedge reference.
1465 (backprop::process_block): Add each phi to m_visited_phis
1466 after visiting it, then clear it at the end.
1468 2018-06-01 Richard Biener <rguenther@suse.de>
1470 * tree-vectorizer.h (vect_dr_stmt): New function.
1471 (vect_get_load_cost): Adjust.
1472 (vect_get_store_cost): Likewise.
1473 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
1474 Use vect_dr_stmt instead of DR_SMTT.
1475 (vect_record_base_alignments): Likewise.
1476 (vect_calculate_target_alignment): Likewise.
1477 (vect_compute_data_ref_alignment): Likewise and make static.
1478 (vect_update_misalignment_for_peel): Likewise.
1479 (vect_verify_datarefs_alignment): Likewise.
1480 (vector_alignment_reachable_p): Likewise.
1481 (vect_get_data_access_cost): Likewise. Pass down
1482 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
1483 (vect_get_peeling_costs_all_drs): Likewise.
1484 (vect_peeling_hash_get_lowest_cost): Likewise.
1485 (vect_enhance_data_refs_alignment): Likewise.
1486 (vect_find_same_alignment_drs): Likewise.
1487 (vect_analyze_data_refs_alignment): Likewise.
1488 (vect_analyze_group_access_1): Likewise.
1489 (vect_analyze_group_access): Likewise.
1490 (vect_analyze_data_ref_access): Likewise.
1491 (vect_analyze_data_ref_accesses): Likewise.
1492 (vect_vfa_segment_size): Likewise.
1493 (vect_small_gap_p): Likewise.
1494 (vectorizable_with_step_bound_p): Likewise.
1495 (vect_prune_runtime_alias_test_list): Likewise.
1496 (vect_analyze_data_refs): Likewise.
1497 (vect_supportable_dr_alignment): Likewise.
1498 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1499 (vect_gen_prolog_loop_niters): Likewise.
1500 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1501 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
1503 (vect_recog_mask_conversion_pattern): Likewise.
1504 (vect_try_gather_scatter_pattern): Likewise.
1505 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
1506 to vect_get_store_cost.
1507 (vect_get_store_cost): Get stmt_info instead of DR.
1508 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
1509 (vect_get_load_cost): Get stmt_info instead of DR.
1511 2018-06-01 Richard Biener <rguenther@suse.de>
1514 * gimple-fold.c (var_decl_component_p): Also allow offsetted
1515 vars wrapped in MEM_REFs.
1517 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1519 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1520 Fix subreg tests so that we only return a choice between
1521 GENERAL_REGS and FP_REGS if the original classes included both.
1523 2018-06-01 Richard Biener <rguenther@suse.de>
1526 * tree-ssa-structalias.c (get_function_part_constraint):
1527 Handle NULL fi->decl.
1528 (find_func_aliases_for_call): Properly handle indirect
1529 fi from direct call.
1530 (find_func_clobbers): Likewise.
1531 (ipa_pta_execute): Likewise.
1532 (create_variable_info_for): For functions that are ifunc_resolver
1533 resolve to a varinfo that contains the result of the resolver call.
1534 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
1537 2018-05-31 Michael Collison <michael.collison@arm.com>
1539 * config/aarch64/aarch64.md:
1540 (*fix_to_zero_extenddfdi2): New pattern.
1541 * gcc.target/aarch64/fix_extend1.c: New testcase.
1543 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
1547 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
1548 and BUILT_IN_STRNCMP_EQ.
1549 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
1550 BUILT_IN_STRNCMP_EQ.
1551 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
1552 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1553 (gimple_fold_builtin): Likewise.
1554 * tree-ssa-strlen.c (compute_string_length): New function.
1555 (determine_min_obsize): New function.
1556 (handle_builtin_string_cmp): New function to handle calls to
1557 string compare functions.
1558 (strlen_optimize_stmt): Add handling to builtin string compare
1560 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1561 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1562 * tree.c (build_common_builtin_nodes): Add new defines of
1563 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
1565 2018-05-31 Jakub Jelinek <jakub@redhat.com>
1568 * bb-reorder.c (pass_partition_blocks::gate): Return false for
1569 functions with naked attribute.
1571 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
1573 * config/i386/sse.md (avx_vec_concat<mode>):
1574 Substitute concat_tg_mode mode attribute with xtg_mode.
1575 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
1576 (concat_tg_mode): Remove mode attribute.
1578 2018-05-31 Martin Sebor <msebor@redhat.com>
1581 * calls.c (alloc_max_size): Correct a logic error/typo.
1582 Treat excessive arguments as infinite. Warn for invalid arguments.
1583 * doc/invoke.texi (-Walloc-size-larger-than): Update.
1585 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
1588 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
1589 and movx for Haswell.
1591 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
1592 Cesar Philippidis <cesar@codesourcery.com>
1595 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
1596 when emitting error on private/firstprivate reductions.
1597 * omp-low.c (lower_omp_target): Avoid reference-type processing
1598 on pointers for firstprivate clause.
1600 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1602 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
1605 * config/aarch64/aarch64-simd.md
1606 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
1607 (aarch64_ld1_x3_<mode>): Likewise
1608 (aarch64_st1x2<VALLDIF:mode>): Likewise
1609 (aarch64_st1_x2_<mode>): Likewise
1610 (aarch64_st1x3<VALLDIF:mode>): Likewise
1611 (aarch64_st1_x3_<mode>): Likewise
1612 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
1613 (vld1_s8_x3): Likewise.
1614 (vld1_u16_x3): Likewise.
1615 (vld1_s16_x3): Likewise.
1616 (vld1_u32_x3): Likewise.
1617 (vld1_s32_x3): Likewise.
1618 (vld1_u64_x3): Likewise.
1619 (vld1_s64_x3): Likewise.
1620 (vld1_f16_x3): Likewise.
1621 (vld1_f32_x3): Likewise.
1622 (vld1_f64_x3): Likewise.
1623 (vld1_p8_x3): Likewise.
1624 (vld1_p16_x3): Likewise.
1625 (vld1_p64_x3): Likewise.
1626 (vld1q_u8_x3): Likewise.
1627 (vld1q_s8_x3): Likewise.
1628 (vld1q_u16_x3): Likewise.
1629 (vld1q_s16_x3): Likewise.
1630 (vld1q_u32_x3): Likewise.
1631 (vld1q_s32_x3): Likewise.
1632 (vld1q_u64_x3): Likewise.
1633 (vld1q_s64_x3): Likewise.
1634 (vld1q_f16_x3): Likewise.
1635 (vld1q_f32_x3): Likewise.
1636 (vld1q_f64_x3): Likewise.
1637 (vld1q_p8_x3): Likewise.
1638 (vld1q_p16_x3): Likewise.
1639 (vld1q_p64_x3): Likewise.
1640 (vst1_s64_x2): Likewise.
1641 (vst1_u64_x2): Likewise.
1642 (vst1_f64_x2): Likewise.
1643 (vst1_s8_x2): Likewise.
1644 (vst1_p8_x2): Likewise.
1645 (vst1_s16_x2): Likewise.
1646 (vst1_p16_x2): Likewise.
1647 (vst1_s32_x2): Likewise.
1648 (vst1_u8_x2): Likewise.
1649 (vst1_u16_x2): Likewise.
1650 (vst1_u32_x2): Likewise.
1651 (vst1_f16_x2): Likewise.
1652 (vst1_f32_x2): Likewise.
1653 (vst1_p64_x2): Likewise.
1654 (vst1q_s8_x2): Likewise.
1655 (vst1q_p8_x2): Likewise.
1656 (vst1q_s16_x2): Likewise.
1657 (vst1q_p16_x2): Likewise.
1658 (vst1q_s32_x2): Likewise.
1659 (vst1q_s64_x2): Likewise.
1660 (vst1q_u8_x2): Likewise.
1661 (vst1q_u16_x2): Likewise.
1662 (vst1q_u32_x2): Likewise.
1663 (vst1q_u64_x2): Likewise.
1664 (vst1q_f16_x2): Likewise.
1665 (vst1q_f32_x2): Likewise.
1666 (vst1q_f64_x2): Likewise.
1667 (vst1q_p64_x2): Likewise.
1668 (vst1_s64_x3): Likewise.
1669 (vst1_u64_x3): Likewise.
1670 (vst1_f64_x3): Likewise.
1671 (vst1_s8_x3): Likewise.
1672 (vst1_p8_x3): Likewise.
1673 (vst1_s16_x3): Likewise.
1674 (vst1_p16_x3): Likewise.
1675 (vst1_s32_x3): Likewise.
1676 (vst1_u8_x3): Likewise.
1677 (vst1_u16_x3): Likewise.
1678 (vst1_u32_x3): Likewise.
1679 (vst1_f16_x3): Likewise.
1680 (vst1_f32_x3): Likewise.
1681 (vst1_p64_x3): Likewise.
1682 (vst1q_s8_x3): Likewise.
1683 (vst1q_p8_x3): Likewise.
1684 (vst1q_s16_x3): Likewise.
1685 (vst1q_p16_x3): Likewise.
1686 (vst1q_s32_x3): Likewise.
1687 (vst1q_s64_x3): Likewise.
1688 (vst1q_u8_x3): Likewise.
1689 (vst1q_u16_x3): Likewise.
1690 (vst1q_u32_x3): Likewise.
1691 (vst1q_u64_x3): Likewise.
1692 (vst1q_f16_x3): Likewise.
1693 (vst1q_f32_x3): Likewise.
1694 (vst1q_f64_x3): Likewise.
1695 (vst1q_p64_x3): Likewise.
1697 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1699 * config/msp430/msp430.c (msp430_output_labelref): Prepend
1700 user_label_prefix to name.
1702 * tree-core.h: Update comment about the format of NAME string
1703 passed to handler in attribute_spec.
1705 * config/msp430/msp430.md: Remove erroneous subreg expression from
1706 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
1707 zero_extend{q,h}isi2.
1709 2018-05-30 Borislav Petkov <bp@suse.de>
1711 * doc/extend.texi: Document some architecture specific
1712 constraints and sort entries.
1714 2018-05-30 Martin Sebor <msebor@redhat.com>
1717 * builtins.c (expand_builtin_stpcpy_1): New function.
1718 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
1719 only if the former succeeds.
1721 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1723 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
1726 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1728 * doc/invoke.texi (-flinker-output): Document
1730 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1732 * passes.c (ipa_write_summaries): Only modify statements if body
1734 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
1735 incrementally linking.
1736 (ipa_passes): Likewise.
1737 * lto-cgraph.c (lto_output_node): When incrementally linking do not
1738 pass down resolution info.
1739 * common.opt (flag_incremental_link): Update info.
1740 * gcc.c (plugin specs): Turn flinker-output=* to
1741 -plugin-opt=-linker-output-known
1742 * toplev.c (compile_file): Also cut compilation when doing incremental
1744 * flag-types. (enum lto_partition_model): Add
1745 LTO_LINKER_OUTPUT_NOLTOREL.
1746 (invoke.texi): Add -flinker-output docs.
1747 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
1748 link same way as WPA; do not stream in dead initializers.
1750 * dwarf2out.c (dwarf2out_die_ref_for_decl,
1751 darf2out_register_external_decl): Support incremental link.
1753 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1755 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
1757 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1759 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
1760 it down to simple_object_copy_lto_debug_sections.
1761 (run_gcc): Determine incremental LTO link time and configure
1762 lto1 into non-wpa mode, disable renaming of debug sections.
1764 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
1766 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1767 descriptions of various incorrectly documented functions.
1769 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
1772 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1773 address check not strict.
1775 2018-05-30 Richard Biener <rguenther@suse.de>
1777 PR tree-optimization/85964
1778 * tracer.c (better_p): Drop initialized count check, we only
1779 call the function with initialized counts now.
1780 (find_best_successor): Do find a best edge if one
1781 has uninitialized count.
1782 (find_best_predecessor): Likewise. Do BB frequency check only
1783 if count is initialized.
1785 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
1787 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
1788 (aarch64_ldrstr_offset_compare): New.
1789 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
1790 load/store orderings.
1791 (aarch64_gen_adjusted_ldpstp): Likewise.
1793 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
1795 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1796 Check for subset of GENERAL_REGS and FP_REGS.
1797 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
1800 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
1802 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
1803 and wi::to_poly_offset. Add the current offset and then check
1804 whether the sum fits, rather than using an unchecked addition of
1805 a checked term. Check for a shwi rather than a uhwi.
1806 * expr.c (get_bit_range): Use tree_to_poly_uint64.
1807 (store_constructor): Use poly_int_tree_p.
1808 (expand_expr_real_1): Likewise.
1809 * function.c (assign_temp): Likewise.
1810 * fold-const.c (const_binop): Use poly_int_tree_p and
1812 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
1814 * ipa-icf-gimple.c (func_checker::compare_operand): Use
1815 to_poly_offset for MEM offsets.
1816 * ipa-icf.c (sem_variable::equals): Likewise.
1817 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
1818 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
1819 wi::to_poly_offset for BIT_FIELD_REF offsets.
1820 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
1822 * var-tracking.c (emit_note_insn_var_location): Use
1823 tree_to_poly_uint64.
1825 2018-05-29 Jim Wilson <jimw@sifive.com>
1827 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
1829 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
1832 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
1833 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
1835 (sse4_1_round<mode>2): Use nonimmediate_operand
1836 for operand 1 predicate.
1838 2018-05-29 Martin Sebor <msebor@redhat.com>
1839 Richard Biener <rguenther@suse.de>
1842 * calls.c (get_size_range): Call determine_value_range instead
1843 of get_value_range..
1844 * tree-vrp.h (determine_value_range): Declared new function.
1845 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
1847 2018-05-29 Richard Biener <rguenther@suse.de>
1849 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
1850 sure to use non-pattern stmts for get_earlier_stmt arguments.
1851 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
1852 called on pattern stmts.
1853 (get_later_stmt): Likewise.
1855 2018-05-29 Martin Liska <mliska@suse.cz>
1857 PR gcov-profile/85759
1858 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
1861 2018-05-29 Jakub Jelinek <jakub@redhat.com>
1863 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
1865 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
1869 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
1870 VEC_PACK_FLOAT_EXPR): New tree codes.
1871 * tree-pretty-print.c (op_code_prio): Handle
1872 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
1873 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
1874 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
1875 * tree-inline.c (estimate_operator_cost): Likewise.
1876 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
1877 * fold-const.c (const_binop): Likewise.
1878 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
1879 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
1880 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1881 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
1882 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
1883 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
1884 * expr.c (expand_expr_real_2): Likewise.
1885 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
1886 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
1887 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
1889 * optabs.c (expand_widen_pattern_expr): For
1890 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
1891 sign from result type rather than operand's type.
1892 (expand_binop_directly): For vec_packu_float_optab and
1893 vec_packs_float_optab allow result type to be different from operand's
1895 * optabs-tree.c (optab_for_tree_code): Handle
1896 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
1897 VEC_PACK_FLOAT_EXPR. Formatting fixes.
1898 * tree-vect-generic.c (expand_vector_operations_1): Handle
1899 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
1900 VEC_PACK_FLOAT_EXPR.
1901 * tree-vect-stmts.c (supportable_widening_operation): Handle
1903 (supportable_narrowing_operation): Handle FLOAT_EXPR.
1904 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
1905 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
1906 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
1907 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
1909 (vec_pack<floatprefix>_float_<mode>): New expander.
1910 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
1912 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
1913 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
1914 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
1915 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
1916 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
1918 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
1919 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
1920 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
1921 VEC_PACK_FLOAT_EXPR): Document.
1923 2018-05-29 Richard Biener <rguenther@suse.de>
1925 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
1927 (stmt_vec_info_vec): Make pointer.
1928 (init_stmt_vec_info_vec): Remove.
1929 (free_stmt_vec_info_vec): Likewise.
1930 (set_stmt_vec_info_vec): New function.
1931 (free_stmt_vec_infos): Likewise.
1932 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
1933 (set_vinfo_for_stmt): Likewise.
1934 (get_earlier_stmt): Likewise.
1935 (get_later_stmt): Likewise.
1936 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
1937 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
1938 (vec_info::~vec_info): Free stmt_vec_infos.
1939 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
1940 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
1941 (pass_slp_vectorize::execute): Likewise.
1942 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
1943 (free_stmt_vec_info_vec): Likewise.
1944 (set_stmt_vec_info_vec): New function.
1945 (free_stmt_vec_infos): Likewise.
1946 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
1947 the global stmt_vec_info_vec.
1948 * tree-parloops.c (gather_scalar_reductions): Use
1949 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
1952 2018-05-29 Richard Biener <rguenther@suse.de>
1954 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
1956 2018-05-29 Martin Liska <mliska@suse.cz>
1957 David Malcolm <dmalcolm@redhat.com>
1959 * vec.c (test_reverse): New.
1960 (vec_c_tests): Add new test.
1961 * vec.h (vl_ptr>::reverse): New function.
1963 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
1965 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
1967 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
1970 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
1972 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
1974 2018-05-28 Richard Biener <rguenther@suse.de>
1976 PR tree-optimization/85933
1977 * tree-vect-data-refs.c (vect_record_base_alignments): Only
1978 look at stmts marked as vectorizable.
1980 2018-05-28 Richard Biener <rguenther@suse.de>
1982 PR tree-optimization/85934
1983 * tree-vect-generic.c (expand_vector_operations_1): Hoist
1984 vector boolean check before scalar optimization.
1986 2018-05-28 Jakub Jelinek <jakub@redhat.com>
1988 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
1991 2018-05-28 Mark Wielaard <mark@klomp.org>
1993 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
1994 if it is an expression containing a minus sign.
1996 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
1998 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2000 2018-05-27 Paul Koning <ni1d@arrl.net>
2002 * config/pdp11/pdp11.md (truncsihi2): Remove.
2004 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
2005 Chung-Ju Wu <jasonwucj@gmail.com>
2007 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
2009 (unaligned_store_dw): Ditto.
2010 * config/nds32/nds32-memory-manipulation.c
2011 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
2012 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
2013 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
2014 (emit_setmem_word_loop): Rename to ...
2015 (emit_setmem_doubleword_loop): ... this.
2016 (nds32_gen_dup_4_byte_to_word_value): New function.
2017 (nds32_gen_dup_8_byte_to_double_word_value): New function.
2018 (nds32_expand_setmem_loop): Refine implementation.
2019 (nds32_expand_setmem_loop_v3m): Ditto.
2020 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
2023 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2025 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
2027 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2029 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
2030 (nds32_init_machine_status): Initialize machine->attr_naked_p and
2031 machine->attr_no_prologue_p.
2032 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
2033 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
2034 (nds32_expand_epilogue): Consider attr_naked_p.
2035 (nds32_expand_epilogue_v3pop): Likewise.
2036 (nds32_can_use_return_insn): Likewise.
2037 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
2038 attr_no_prologue_p fields.
2039 * config/nds32/nds32.opt (mret-in-naked-func): New option.
2041 2018-05-27 Jakub Jelinek <jakub@redhat.com>
2044 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
2046 * config/i386/sse.md
2047 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
2049 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
2051 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
2053 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
2055 (*<floatsuffix>floatv2div2sf2): Rename to ...
2056 (*float<floatunssuffix>v2div2sf2): ... this.
2057 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
2058 (float<floatunssuffix>v2div2sf2_mask): ... this.
2059 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
2060 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
2061 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
2063 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
2065 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2067 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2069 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2071 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2073 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
2074 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
2075 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
2076 gen_ufix_truncv8dfv8si2.
2077 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
2078 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
2079 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
2080 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
2081 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
2082 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
2083 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2084 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2086 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2090 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2092 2018-05-25 Jim Wilson <jimw@sifive.com>
2094 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2095 * config/riscv/riscv.c (struct machine_function): Add
2096 interrupt_handler_p and attribute_checked_p fields.
2097 (riscv_attribute_table): Add interrupt.
2098 (riscv_interrupt_type_p): New.
2099 (riscv_save_reg_p): Save extra regs for interrupt handler.
2100 (riscv_use_save_libcall): Return false for interrupt handler.
2101 (riscv_first_stack_step): Add forward declaration.
2102 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
2103 for interrupt handler with large frame. Use it for saved reg list.
2104 (riscv_expand_prologue): Move flag_stack_usage_info support to
2105 eliminate duplication.
2106 (riscv_expand_epilogue): Generate mret for interrupt handler.
2107 (riscv_epilogue_uses): New.
2108 (riscv_can_use_return_insn): Return false for interrupt handler.
2109 (riscv_function_ok_for_sibcall): Likewise.
2110 (riscv_set_current_function): Add interrupt handler support.
2111 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2112 * config/riscv/riscv.md (UNSPECV_MRET): New.
2114 (riscv_frflags, riscv_fsflags): Use tab after opcode.
2116 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2118 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
2120 PR tree-optimization/85712
2121 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2122 this candidate has already been replaced in-situ by a copy.
2124 2018-05-25 Jason Merrill <jason@redhat.com>
2126 PR c++/80485 - inline function non-zero address.
2127 * symtab.c (nonzero_address): Check DECL_COMDAT.
2129 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
2132 * config/alpha/alpha.md (ashlsi3): New insn pattern.
2133 (*ashlsi_se): Rename from *ashldi_se. Define as sign
2134 extension of SImode operation. Use const123_operand predicate.
2135 (*saddsi_1): Remove.
2136 (*saddl_se_1): Ditto.
2138 (*ssubl_se_1): Ditto.
2139 * config/alpha/predicates.md (const123_operand): New predicate.
2140 * config/alpha/constraints.md (P): Use IN_RANGE.
2142 2018-05-25 Richard Biener <rguenther@suse.de>
2144 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2146 (ref_maybe_used_by_stmt_p): Likewise.
2147 (stmt_may_clobber_ref_p): Likewise.
2148 (stmt_may_clobber_ref_p_1): Likewise.
2149 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2151 (ref_maybe_used_by_stmt_p): Likewise.
2152 (stmt_may_clobber_ref_p): Likewise.
2153 (stmt_may_clobber_ref_p_1): Likewise.
2154 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2155 the alias oracle to disambiguate DRs with stmts DR analysis
2157 (vect_analyze_data_refs): Do not give up on not analyzable
2158 DRs for BB vectorization. Remove code truncating the dataref
2161 2018-05-25 Jakub Jelinek <jakub@redhat.com>
2164 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2165 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
2166 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2168 2018-05-25 Richard Biener <rguenther@suse.de>
2170 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2171 function, combining stmt data ref gathering and fatal analysis
2173 (vect_analyze_data_refs): Remove now redudnant code and simplify.
2174 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2175 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2176 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2177 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2179 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2181 PR tree-optimization/85720
2182 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2183 SCC if all partitions are builtins.
2184 (version_loop_by_alias_check): New parameter. Generate cancelable
2185 runtime alias check if all partitions are builtins.
2186 (distribute_loop): Update call to above function.
2188 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2190 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2191 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2192 (parm_default_def_partition_arg): Ditto.
2193 (set_parm_default_def_partition): Ditto.
2194 (get_parm_default_def_partitions): Ditto and make it static.
2195 (get_undefined_value_partitions): Ditto and make it static.
2196 (remove_ssa_form): Refactor call to init_var_map here.
2197 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2198 computation for loop region.
2199 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2200 (register_default_def): Delete.
2201 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2202 (parm_default_def_partition_arg): Ditto.
2203 (set_parm_default_def_partition): Ditto.
2204 (get_parm_default_def_partitions): Ditto and make it static.
2205 (get_undefined_value_partitions): Ditto and make it static.
2206 (coalesce_with_default, coalesce_with_default): Update comment.
2207 (create_coalesce_list_for_region): New func factored out from
2208 create_outofssa_var_map.
2209 (populate_coalesce_list_for_outofssa): New func factored out from
2210 create_outofssa_var_map and coalesce_ssa_name.
2211 (create_outofssa_var_map): Delete.
2212 (coalesce_ssa_name): Refactor to support live range computation.
2213 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2214 (get_parm_default_def_partitions): Delete.
2215 (get_undefined_value_partitions): Ditto.
2216 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2217 computation for loop region.
2218 (new_tree_live_info, loe_visit_block): Ditto.
2219 (live_worklist, set_var_live_on_entry): Ditto.
2220 (calculate_live_on_exit, verify_live_on_entry): Ditto.
2221 * tree-ssa-live.h (struct _var_map): New fields.
2222 (init_var_map): Change decl.
2223 (region_contains_p): New.
2225 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2227 * tree-ssa-live.h (live_merge_and_clear): Delete.
2229 2018-05-25 Richard Biener <rguenther@suse.de>
2232 * tree-dump.c (dequeue_and_dump): Remove access to removed
2233 operand 2 of a SWITCH_EXPR.
2235 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2237 * doc/sourcebuild.texi (vect_double_cond_arith): Include
2238 multiplication and division.
2239 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2240 (cond_udiv@var{m}, cond_umod@var{m}): Document.
2241 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2242 (cond_udiv_optab, cond_umod_optab): New optabs.
2243 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2244 (IFN_COND_RDIV): New internal functions.
2245 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
2246 TRUNC_MOD_EXPR and RDIV_EXPR.
2247 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
2248 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
2250 (SVE_INT_BINARY): Include mult.
2251 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
2252 (optab, sve_int_op): Handle mult.
2253 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
2255 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
2256 for SVE_INT_BINARY_SD.
2258 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2260 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
2261 (optab, sve_int_op): Handle div and udiv.
2262 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
2263 for SVE_INT_BINARY_SD.
2264 (*<optab><mode>3): New insn for the same.
2266 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2268 * tree-vect-patterns.c: Include predict.h.
2269 (vect_recog_divmod_pattern): Restrict check for division support
2270 to when optimizing for size.
2272 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2274 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
2275 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
2276 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
2277 (gimple_match_op::set_op): Likewise.
2278 (gimple_resimplify4): Declare.
2279 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
2280 (expr::gen_transform): Likewise.
2281 (decision_tree::gen): Generate a simplification routine for 4 operands.
2282 * gimple-match-head.c (gimple_simplify): Add an overload for
2283 4 operands. In the top-level function, handle up to 4 call
2284 arguments and call gimple_resimplify4.
2285 (gimple_resimplify4): New function.
2286 (build_call_internal): Pass a fourth operand.
2287 (maybe_push_to_seq): Likewise.
2288 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
2289 Fold VEC_COND_EXPRs of an operation and a default value into
2290 an IFN_COND_* function if possible.
2291 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
2293 (SVE_COND_FP_BINARY): Include them.
2294 (optab, sve_fp_op): Handle them.
2295 (SVE_INT_BINARY_REV): New code iterator.
2296 (SVE_COND_FP_BINARY_REV): New int iterator.
2297 (commutative): New int attribute.
2298 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
2300 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
2302 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
2303 (*cond_<optab><mode>): New patterns for reversed operands.
2305 2018-05-25 Richard Biener <rguenther@suse.de>
2307 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
2308 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
2309 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
2310 (STMT_VINFO_GROUPED_ACCESS): Adjust.
2311 * tree-vect-data-refs.c (everywhere): Adjust users.
2312 * tree-vect-loop.c (everywhere): Likewise.
2313 * tree-vect-slp.c (everywhere): Likewise.
2314 * tree-vect-stmts.c (everywhere): Likewise.
2315 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2317 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2319 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
2321 (gcc_cv_as_section_exclude): ... this.
2322 Try Solaris as #exclude syntax.
2323 * configure: Regenerate.
2324 * config.in: Regenerate.
2325 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
2327 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
2328 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
2330 * varasm.c (default_elf_asm_named_section): Don't check if
2331 HAVE_GAS_SECTION_EXCLUDE is defined.
2333 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2335 * doc/md.texi: Update the documentation of the cond_* optabs
2336 to mention the new final operand. Fix GET_MODE_NUNITS call.
2337 Describe the scalar case too.
2338 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
2339 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
2341 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
2342 (get_conditional_internal_fn): Update comment.
2343 * tree-vect-loop.c (vectorizable_reduction): Pass the original
2344 accumulator value as a final argument to conditional functions.
2345 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
2346 a define_expand and add an "else" operand. Assert for now that
2347 the else operand is equal to operand 2. Use SVE_INT_BINARY and
2348 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
2349 (*cond_<optab><mode>): New patterns.
2350 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
2351 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
2352 (UNSPEC_COND_EOR): Delete.
2353 (optab): Remove associated mappings.
2354 (SVE_INT_BINARY): New code iterator.
2355 (sve_int_op): Remove int attribute and add "minus" to the code
2357 (SVE_COND_INT_OP): Delete.
2358 (SVE_COND_FP_OP): Rename to...
2359 (SVE_COND_FP_BINARY): ...this.
2361 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2363 * optabs.c (can_reuse_operands_p): New function.
2364 (maybe_legitimize_operands): Try to reuse the results for
2367 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2369 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
2370 Add {q} suffix to insn mnemonic.
2372 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2374 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
2375 (msp430_warn_func_return): New.
2377 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
2379 * fold-const.c (tree_nonzero_bits): New function.
2380 * fold-const.h (tree_nonzero_bits): Likewise.
2381 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
2382 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
2384 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2388 * varasm.c (assemble_alias): Check ifunc_resolver only on
2391 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2394 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
2395 when memory input operand is handled.
2397 2018-05-24 Luis Machado <luis.machado@linaro.org>
2399 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
2401 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
2403 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2405 * match.pd: Delay FMA folds until after vectorization.
2407 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
2410 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2411 address check not strict.
2413 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2415 * gimple-match.h (gimple_match_op): New class.
2416 (mprts_hook): Replace parameters with a gimple_match_op *.
2417 (maybe_build_generic_op): Likewise.
2418 (gimple_simplified_result_is_gimple_val): Replace parameters with
2419 a const gimple_match_op *.
2420 (gimple_simplify): Replace code_helper * and tree * parameters with
2421 a gimple_match_op * parameter.
2422 (gimple_resimplify1): Replace code_helper *, tree and tree *
2423 parameters with a gimple_match_op * parameter.
2424 (gimple_resimplify2): Likewise.
2425 (gimple_resimplify3): Likewise.
2426 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
2427 parameters with a gimple_match_op * parameter.
2428 * gimple-match-head.c (gimple_simplify): Change prototypes of
2429 auto-generated functions to take a gimple_match_op * instead of
2430 separate code_helper * and tree * parameters. Make the same
2431 change in the top-level overload and update calls to the
2432 gimple_resimplify routines. Update calls to the auto-generated
2433 functions and to maybe_push_res_to_seq in the publicly-facing
2434 operation-specific gimple_simplify overloads.
2435 (gimple_match_op::MAX_NUM_OPS): Define.
2436 (gimple_resimplify1): Replace rcode and ops with a single res_op
2437 parameter. Update call to gimple_simplify.
2438 (gimple_resimplify2): Likewise.
2439 (gimple_resimplify3): Likewise.
2440 (mprts_hook): Replace parameters with a gimple_match_op *.
2441 (maybe_build_generic_op): Likewise.
2442 (build_call_internal): Replace type, nargs and ops with
2443 a gimple_match_op *.
2444 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
2445 with a single gimple_match_op *. Update calls to mprts_hook,
2446 build_call_internal and gimple_simplified_result_is_gimple_val.
2447 Factor out code that is common to the tree_code and combined_fn cases.
2448 * genmatch.c (expr::gen_transform): Replace tem_code and
2449 tem_ops with a gimple_match_op called tem_op. Update calls
2450 to the gimple_resimplify functions and maybe_push_res_to_seq.
2451 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
2452 res_ops. Update call to the gimple_resimplify functions.
2453 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
2454 (decision_tree::gen): Make the functions take a gimple_match_op *
2455 called res_op instead of separate res_code and res_ops parameters.
2456 Update call accordingly.
2457 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
2458 and ops with a single res_op parameter. Update calls to
2459 maybe_build_generic_op and maybe_push_res_to_seq.
2460 (fold_stmt_1): Update calls to gimple_simplify and
2461 replace_stmt_with_simplification.
2462 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
2463 and gimple_simplified_result_is_gimple_val.
2464 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
2466 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
2467 with a gimple_match_op *.
2468 (vn_nary_build_or_lookup): Likewise. Update call to
2469 vn_nary_build_or_lookup_1.
2470 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
2471 gimple_match_op *. Update calls to the gimple_resimplify routines
2472 and to gimple_simplified_result_is_gimple_val.
2473 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
2474 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
2475 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
2476 (visit_nary_op): Likewise.
2477 (visit_reference_op_load): Likewise.
2479 2018-05-23 Luis Machado <luis.machado@linaro.org>
2481 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
2482 modifier for printing the step amount.
2484 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
2487 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
2490 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
2492 * doc/sourcebuild.texi (Endianness): New subsubsection.
2494 2018-05-23 Luis Machado <luis.machado@linaro.org>
2496 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2497 <prefetch_dynamic_strides>: New const bool field.
2498 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2499 prefetch_dynamic_strides.
2500 (exynosm1_prefetch_tune): Likewise.
2501 (thunderxt88_prefetch_tune): Likewise.
2502 (thunderx_prefetch_tune): Likewise.
2503 (thunderx2t99_prefetch_tune): Likewise.
2504 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
2506 (aarch64_override_options_internal): Update to set
2507 PARAM_PREFETCH_DYNAMIC_STRIDES.
2508 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2509 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2510 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2511 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2512 prefetch-dynamic-strides setting.
2514 2018-05-23 Luis Machado <luis.machado@linaro.org>
2516 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2517 <minimum_stride>: New const int field.
2518 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2519 minimum_stride field defaulting to -1.
2520 (exynosm1_prefetch_tune): Likewise.
2521 (thunderxt88_prefetch_tune): Likewise.
2522 (thunderx_prefetch_tune): Likewise.
2523 (thunderx2t99_prefetch_tune): Likewise.
2524 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
2525 <default_opt_level>: Set to 3.
2526 (aarch64_override_options_internal): Update to set
2527 PARAM_PREFETCH_MINIMUM_STRIDE.
2528 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2529 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2530 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2531 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2532 stride is constant and is below the minimum stride threshold.
2534 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2536 * config/arm/arm-cpus.in (mode26): Delete.
2537 (armv4): Delete mode26 reference.
2538 * config/arm/arm.c (arm_configure_build_target): Delete use of
2541 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2543 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
2545 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
2546 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
2548 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
2549 (floatunsdidf2): Ditto.
2551 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2553 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
2554 (fixuns_trunc<mode>si2_avx512f): Ditto.
2555 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
2556 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
2557 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
2559 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2561 PR rtl-optimization/79985
2562 * df-scan.c (df_insn_refs_collect): Remove special case for
2563 global registers and asm statements.
2565 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2567 * extend.texi (Global Register Variables): Rewrite the bullet list.
2568 Note that the register is available for allocation. Note that access
2569 via inline asm must use constraints. Add note about async-signal
2570 handlers. Remove paragraph about automagic register selection.
2572 2018-05-23 Richard Biener <rguenther@suse.de>
2574 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
2575 of fixed offset from memset VN.
2577 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
2579 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
2581 (alloc_cand_and_find_basis): Initialize first_interp field.
2582 (slsr_process_mul): Modify first_interp field.
2583 (slsr_process_add): Likewise.
2584 (slsr_process_cast): Modify first_interp field for each new
2586 (slsr_process_copy): Likewise.
2587 (dump_candidate): Dump first_interp field.
2588 (replace_mult_candidate): Process all interpretations, not just
2590 (replace_rhs_if_not_dup): Likewise.
2591 (replace_one_candidate): Likewise.
2593 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
2595 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
2597 (aarch64_needs_frame_chain): New function.
2598 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
2600 2018-05-23 Sudakshina Das <sudi.das@arm.com>
2603 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
2604 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
2605 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
2606 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
2607 as true for strict-align.
2608 (aarch64_can_inline_p): Perform checks even when callee has no
2609 attributes to check for strict alignment.
2610 * doc/extend.texi (AArch64 Function Attributes): Document
2612 * doc/invoke.texi: (AArch64 Options): Likewise.
2614 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
2616 PR tree-optimization/85853
2617 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
2618 the handling of the root of the node to...
2619 (vect_slp_analyze_node_operations_1): ...this new function,
2620 and run the whole thing with the child nodes' def types
2621 set according to their SLP node's def type.
2623 2018-05-23 Richard Biener <rguenther@suse.de>
2626 * tree-data-ref.c (create_runtime_alias_checks): Defer
2627 and ignore overflow warnings.
2629 2018-05-23 Yury Gribov <tetra2005@gmail.com>
2631 PR tree-optimization/85822
2632 * tree-vrp.c (is_masked_range_test): Fix handling of negative
2635 2018-05-23 Richard Biener <rguenther@suse.de>
2637 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2638 memset constants via native_interpret_expr.
2640 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
2643 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
2645 (cgraph_node::create_alias): Likewise.
2646 (cgraph_node::get_availability): Check ifunc_resolver instead
2647 of looking up ifunc attribute.
2648 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2649 * varasm.c (do_assemble_alias): Likewise.
2650 (assemble_alias): Likewise.
2651 (default_binds_local_p_3): Likewise.
2652 * cgraph.h (cgraph_node): Add ifunc_resolver.
2653 (cgraph_node::only_called_directly_or_aliased_p): Return false
2655 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
2657 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
2658 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
2659 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
2660 instead of looking up ifunc attribute.
2662 2018-05-22 Luis Machado <luis.machado@linaro.org>
2664 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
2666 2018-05-22 Martin Sebor <msebor@redhat.com>
2669 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
2670 only when expasion succeeds.
2671 (expand_builtin_strcmp): Same.
2672 (expand_builtin_strncmp): Same.
2674 2018-05-22 Martin Sebor <msebor@redhat.com>
2676 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
2678 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2679 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2681 * config/aarch64/aarch64-ldpstp.md: Replace uses of
2682 aarch64_mem_pair_operand with memory_operand and delete operand swapping
2684 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2685 Add check for legitimate_address.
2686 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
2687 (aarch64_swap_ldrstr_operands): New.
2688 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
2691 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2692 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2694 * config/aarch64/aarch64.md: New patterns to generate stp
2696 (store_pair_sw, store_pair_dw): New patterns to generate stp for
2697 single words and double words.
2698 (load_pair_sw, load_pair_dw): Likewise.
2699 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
2701 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
2703 * config/aarch64/aarch64-ldpstp.md: Modify peephole
2704 for different mode ldpstp and add peephole for merged zero stores.
2706 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2708 (aarch64_gen_store_pair): Rename calls to match new patterns.
2709 (aarch64_gen_load_pair): Rename calls to match new patterns.
2710 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
2711 (load_pair<DREG:mode><DREG2:mode>): ... This.
2712 (store_pair<mode>): Rename to...
2713 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
2714 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
2716 (V_INT_EQUIV): Handle SImode.
2717 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
2720 2018-05-22 Martin Sebor <msebor@redhat.com>
2723 * calls.c (maybe_warn_nonstring_arg): Use string length to set
2724 or ajust the presumed bound on an operation to avoid unnecessary
2727 2018-05-22 Martin Sebor <msebor@redhat.com>
2729 PR tree-optimization/85826
2730 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
2731 assuming that a DECL necesarily has a constant size.
2733 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
2736 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
2738 2018-05-22 Richard Biener <rguenther@suse.de>
2740 PR tree-optimization/85834
2741 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
2742 non-constant and non-zero memset arguments.
2744 2018-05-22 Martin Liska <mliska@suse.cz>
2747 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
2749 2018-05-22 Richard Biener <rguenther@suse.de>
2751 PR tree-optimization/85863
2752 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
2753 comparisons when vectype is specified.
2754 (vectorizable_condition): Do not specify vectype for
2755 vect_is_simple_cond when SLP vectorizing.
2757 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
2760 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
2761 define __ibm128 as long double.
2762 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
2763 as a distinct type when IEEE 128-bit support is enabled.
2764 (init_float128_ieee): Fix up conversions between IFmode and IEEE
2765 128-bit types to use the correct functions.
2766 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
2767 convert between 128-bit floating point types that have different
2768 modes but the same representation, instead of using gen_lowpart to
2770 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
2772 (IFKF_reg): New attributes to give the register constraints for
2774 (extend<mode>tf2_internal): New insns to mark an explicit
2775 conversion between 128-bit floating point types that have a
2776 different mode but share the same representation.
2778 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
2780 PR tree-optimization/85814
2781 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
2782 a null return from get_strinfo when unsharing the next
2783 strinfo in the chain.
2785 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2788 * varasm.c (weak_finish): Clean up weak_decls.
2790 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2792 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
2793 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
2794 UNSPEC_UADALP values.
2795 * config/aarch64/iterators.md (ABAL): New int iterator.
2798 (sur): Add mappings for the above.
2799 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
2801 (aarch64_<sur>abal<mode>_4): Likewise.
2802 (aarch64_<sur>adalp<mode>_3): Likewise.
2803 (<sur>sadv16qi): New define_expand.
2805 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
2807 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
2808 (*movdf_internal): Ditto.
2809 (*rcpsf2_sse): Ditto.
2810 (*rsqrtsf2_sse): Ditto.
2811 (*sqrt<mode>2_sse): Ditto.
2813 2018-05-21 Tamar Christina <tamar.christina@arm.com>
2815 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
2817 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
2818 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
2819 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
2820 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
2822 * config/aarch64/arm_neon.h: Likewise.
2823 * config/aarch64/iterators.md (VQ_I): New.
2825 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
2827 * config.gcc: Add arc/t-multilib-linux to tmake_file for
2829 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
2832 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
2834 * config/nds32/constraints.md (S): New constraint.
2835 * config/nds32/nds32.md (call_internal): Use constraint S.
2836 (call_value_internal): Likewise.
2837 (sibcall_internal): Likewise.
2838 (sibcall_value_internal): Likewise.
2840 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
2841 Chung-Ju Wu <jasonwucj@gmail.com>
2843 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
2846 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
2847 Chung-Ju Wu <jasonwucj@gmail.com>
2849 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
2850 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
2851 (nds32_rtx_costs_impl): Simplify.
2852 (nds32_address_cost_impl): Simplify.
2853 (nds32_init_rtx_costs): New function.
2854 (nds32_rtx_costs_speed_prefer): Likewise.
2855 (nds32_rtx_costs_size_prefer): Likewise.
2856 (nds32_address_cost_speed_prefer): Likewise.
2857 (nds32_address_cost_speed_fwprop): Likewise.
2858 (nds32_address_cost_size_prefer): Likewise.
2859 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
2860 * config/nds32/nds32.c (nds32_option_override): Use
2861 nds32_init_rtx_costs function.
2863 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
2865 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
2866 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
2867 (TARGET_PIPELINE_N8): Likewise.
2868 (TARGET_PIPELINE_N10): Likewise.
2869 (TARGET_PIPELINE_N13): Likewise.
2870 (TARGET_PIPELINE_GRAYWOLF): Likewise.
2872 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
2874 * config/nds32/nds32-fpu.md: Update copyright year.
2876 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2878 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
2880 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2882 * config/nds32/nds32.c
2883 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
2884 * config/nds32/nds32.opt (minline-asm-r15): New option.
2886 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2888 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
2890 * config/nds32/nds32.md (abssi2): New pattern.
2892 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
2894 * config/i386/i386.md (rex64namesuffix): New mode attribute.
2895 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
2896 Merge insn pattern from sse_cvtsi2ss<round_name> and
2897 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
2898 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
2899 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
2900 using SWI48 mode iterator.
2901 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
2902 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
2903 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
2904 pattern from sse_cvttss2si<round_saeonly_name>
2905 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
2906 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
2907 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
2908 using SWI48 mode iterator.
2909 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
2910 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
2911 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
2912 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
2913 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
2914 using SWI48 mode iterator.
2915 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
2916 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
2917 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
2918 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
2919 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
2920 SWI48 mode iterator.
2921 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
2922 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
2923 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
2924 pattern from sse_cvttsd2si<round_saeonly_name>
2925 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
2927 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2929 * config/nds32/nds32-md-auxiliary.c
2930 (nds32_valid_smw_lwm_base_p): Refine.
2931 (nds32_output_smw_single_word): Refine.
2932 (nds32_output_smw_double_word): New.
2933 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
2935 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2937 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
2938 (nds32_output_stack_pop): Refine.
2939 (nds32_expand_unaligned_load): Refine.
2940 (nds32_expand_unaligned_store): Refine.
2942 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2943 Chung-Ju Wu <jasonwucj@gmail.com>
2945 * config/nds32/constants.md: Add TP_REGNUM constant.
2946 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
2947 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
2949 * config/nds32/nds32-doubleword.md: Consider flag_pic.
2950 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
2951 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
2952 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
2953 and PIC code generation.
2954 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
2956 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
2958 * config/nds32/nds32.md: Support TLS and PIC.
2959 * config/nds32/nds32.c: Support TLS and PIC.
2960 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
2961 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
2964 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2966 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
2967 mode with E_ prefix.
2969 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2970 Chung-Ju Wu <jasonwucj@gmail.com>
2972 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
2973 * config/nds32/nds32-md-auxiliary.c
2974 (symbolic_reference_mentioned_p): New.
2975 (nds32_legitimize_ict_address): New.
2976 (nds32_expand_ict_move): New.
2977 (nds32_indirect_call_referenced_p): New.
2978 (nds32_symbol_binds_local_p): Delete.
2979 (nds32_long_call_p): Modify.
2980 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
2981 * config/nds32/nds32-protos.h
2982 (symbolic_reference_mentioned_p): Declare.
2983 (nds32_legitimize_ict_address): Declare.
2984 (nds32_expand_ict_move): Declare.
2985 (nds32_indirect_call_referenced_p): Declare.
2986 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
2987 (nds32_relax_group): Use nds32_ict_const_p as condition.
2988 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
2989 (nds32_asm_file_start): Output ict_model directive in asm code.
2990 (nds32_legitimate_address_p): Consider indirect call.
2991 (nds32_print_operand): Consider indirect call.
2992 (nds32_print_operand_address): Consider indirect call.
2993 (nds32_insert_attributes): Handle "indirect_call" attribute.
2994 (TARGET_LEGITIMATE_ADDRESS_P): Define.
2995 (TARGET_LEGITIMATE_CONSTANT_P): Define.
2996 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
2997 (TARGET_DELEGITIMIZE_ADDRESS): Define.
2998 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
2999 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
3000 (TARGET_ICT_MODEL_SMALL): Define.
3001 (TARGET_ICT_MODEL_LARGE): Define.
3002 * config/nds32/nds32.md (movsi): Consider ict model.
3003 (call, call_value): Consider ict model.
3004 (sibcall, sibcall_value): Consider ict model.
3005 * config/nds32/nds32.opt (mict-model): New option.
3006 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
3009 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
3010 Monk Chiang <sh.chiang04@gmail.com>
3011 Jim Wilson <jimw@sifive.com>
3013 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
3014 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
3015 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
3016 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
3017 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
3018 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
3019 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
3020 compute save_libcall_adjustment properly.
3021 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
3022 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
3023 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
3024 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
3025 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
3026 (ABI_SPEC): Handle mabi=ilp32e.
3027 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
3028 (RVE): Add RVE mask.
3029 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
3030 <-march>: Add rv32e as an example.
3032 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
3035 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
3036 (intra_create_variable_infos): Handle C++ constructors.
3038 2018-05-18 Martin Liska <mliska@suse.cz>
3040 * passes.def: Remove a redundant pass.
3042 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
3045 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
3047 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3049 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
3051 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
3053 (armv2, armv2a, armv3, armv3m): Delete architectures.
3054 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
3055 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
3056 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
3058 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
3059 (*mulsidi3adddi): Likewise.
3060 (mulsidi3): Likewise.
3061 (*mulsidi3_nov6): Likewise.
3062 (umulsidi3): Likewise.
3063 (umulsidi3_nov6): Likewise.
3064 (umaddsidi4): Likewise.
3065 (*umulsidi3adddi): Likewise.
3066 (smulsi3_highpart): Likewise.
3067 (*smulsi3_highpart_nov6): Likewise.
3068 (umulsi3_highpart): Likewise.
3069 (*umulsi3_highpart_nov6): Likewise.
3070 * config/arm/arm.h (arm_arch3m): Delete.
3071 * config/arm/arm.c (arm_arch3m): Delete.
3072 (arm_option_override_internal): Update armv3-related comment.
3073 (arm_configure_build_target): Delete use of isa_bit_mode32.
3074 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
3075 (arm_rtx_costs_internal): Delete check of arm_arch3m.
3076 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
3078 (mulusa3): Likewise.
3079 * config/arm/arm-protos.h (arm_arch3m): Delete.
3080 * config/arm/arm-tables.opt: Regenerate.
3081 * config/arm/arm-tune.md: Likewise.
3082 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
3083 deleted architectures.
3085 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3087 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3088 (armv5t, armv5te): New features.
3089 (ARMv5, ARMv5e): Delete fgroups.
3090 (ARMv5t, ARMv5te): Adjust for above changes.
3092 (armv5, armv5e): Delete arches.
3093 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3095 (*call_reg_arm): Likewise.
3096 (*call_value_reg_armv5): Likewise.
3097 (*call_value_reg_arm): Likewise.
3098 (*call_symbol): Likewise.
3099 (*call_value_symbol): Likewise.
3100 (*sibcall_insn): Likewise.
3101 (*sibcall_value_insn): Likewise.
3103 (prefetch): Likewise.
3104 (define_split and define_peephole2 dependent on arm_arch5):
3106 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3108 (TARGET_ARM_QBIT): Likewise.
3109 (TARGET_DSP_MULTIPLY): Likewise.
3110 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3111 (arm_arch5, arm_arch5e): Delete.
3112 (arm_arch5t, arm_arch5te): Declare.
3113 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3114 (arm_arch5t): Declare.
3115 (arm_option_reconfigure_globals): Update for the above.
3116 (arm_options_perform_arch_sanity_checks): Update comment, replace
3117 use of arm_arch5 with arm_arch5t.
3118 (use_return_insn): Likewise.
3119 (arm_emit_call_insn): Likewise.
3120 (output_return_instruction): Likewise.
3121 (arm_final_prescan_insn): Likewise.
3122 (arm_coproc_builtin_available): Likewise.
3123 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3124 arm_arch5e with arm_arch5t and arm_arch5te.
3125 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3126 (arm_arch5t, arm_arch5te): Declare.
3127 * config/arm/arm-tables.opt: Regenerate.
3128 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3129 * config/arm/t-multilib: Likewise.
3130 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3131 instead of arm_arch5.
3132 (*call_reg_thumb1): Likewise.
3133 (*call_value_reg_thumb1_v5): Likewise.
3134 (*call_value_reg_thumb1): Likewise.
3135 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3137 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3139 2018-05-18 Martin Liska <mliska@suse.cz>
3141 PR gcov-profile/84846
3142 * doc/gcov.texi: Document -t option of gcov tool.
3144 2018-05-18 Martin Liska <mliska@suse.cz>
3146 PR gcov-profile/84846
3147 * gcov.c (print_usage): Add new -t option.
3148 (process_args): Handle the option.
3149 (generate_results): Use stdout as output when requested by
3152 2018-05-18 Martin Liska <mliska@suse.cz>
3154 PR gcov-profile/84846
3155 * coverage.c (coverage_init): Write PWD to .gcno file.
3156 * doc/gcov.texi: Document how working directory is printed.
3157 * gcov-dump.c (dump_gcov_file): Print PWD.
3158 * gcov.c (output_intermediate_file): Likewise.
3159 (read_graph_file): Read PWD string.
3160 (output_lines): Print PWD.
3162 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3165 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3166 for retval and return false if all args to phi are zero.
3168 2018-05-18 Richard Biener <rguenther@suse.de>
3170 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3172 (evrp_dom_walker::before_dom_children): Call it.
3174 2018-05-18 Richard Biener <rguenther@suse.de>
3176 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3177 results when processing array refs with variable index.
3179 2018-05-18 Toon Moene <toon@moene.org>
3181 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3182 directly after that of -floop-interchange. Indicate that both
3183 options are enabled by default when specifying -O3.
3185 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3187 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3188 iterator. Delete separate integer-mode vec_set<mode> expander.
3189 (aarch64_simd_vec_setv2di): Delete.
3190 (vec_setv2di): Delete.
3191 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3192 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
3193 the "w, r" alternative.
3195 2018-05-18 Martin Liska <mliska@suse.cz>
3197 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3198 * tree-pass.h (make_pass_lower_switch_O0): New function.
3199 * tree-switch-conversion.c (node_has_low_bound): Remove.
3200 (node_has_high_bound): Likewise.
3201 (node_is_bounded): Likewise.
3202 (class pass_lower_switch): Make it a template type and create
3204 (pass_lower_switch::execute): Add template argument.
3205 (make_pass_lower_switch): New function.
3206 (make_pass_lower_switch_O0): New function.
3207 (do_jump_if_equal): Remove.
3208 (emit_case_nodes): Simplify to just handle all 3 cases and leave
3209 all the hard work to tree optimization passes.
3211 2018-05-18 Martin Liska <mliska@suse.cz>
3213 * dbgcnt.c (limit_low): Renamed from limit.
3214 (limit_high): New variable.
3215 (dbg_cnt_is_enabled): Check for upper limit.
3216 (dbg_cnt): Adjust dumping.
3217 (dbg_cnt_set_limit_by_index): Add new argument for high
3219 (dbg_cnt_set_limit_by_name): Likewise.
3220 (dbg_cnt_process_single_pair): Parse new format.
3221 (dbg_cnt_process_opt): Use strtok.
3222 (dbg_cnt_list_all_counters): Remove 'value' and add
3224 * doc/invoke.texi: Document changes.
3226 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
3228 * doc/sourcebuild.texi (scalar_all_fma): Document.
3229 * tree.def (FMA_EXPR): Delete.
3230 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3231 * internal-fn.c (ternary_direct): New macro.
3232 (expand_ternary_optab_fn): Likewise.
3233 (direct_ternary_optab_supported_p): Likewise.
3234 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3235 * builtins.c (fold_builtin_fma): Delete.
3236 (fold_builtin_3): Don't call it.
3237 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3238 * expr.c (expand_expr_real_2): Likewise.
3239 * fold-const.c (operand_equal_p): Likewise.
3240 (fold_ternary_loc): Likewise.
3241 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3242 * gimple.c (DEFTREECODE): Likewise.
3243 * gimplify.c (gimplify_expr): Likewise.
3244 * optabs-tree.c (optab_for_tree_code): Likewise.
3245 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
3246 * tree-eh.c (operation_could_trap_p): Likewise.
3247 (stmt_could_throw_1_p): Likewise.
3248 * tree-inline.c (estimate_operator_cost): Likewise.
3249 * tree-pretty-print.c (dump_generic_node): Likewise.
3250 (op_code_prio): Likewise.
3251 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3252 * tree-ssa-operands.c (get_expr_operands): Likewise.
3253 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
3254 * fold-const-call.h (fold_fma): Delete.
3255 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
3256 CFN_FNMA and CFN_FNMS.
3258 * genmatch.c (combined_fn): New enum.
3259 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
3260 (commutative_op): New function.
3261 (commutate): Use it. Handle more than 2 operands.
3262 (dt_operand::gen_gimple_expr): Use commutative_op.
3263 (parser::parse_expr): Allow :c to be used with non-binary
3264 operators if the commutative operand is known.
3265 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
3266 CFN_FMS, CFN_FNMA and CFN_FNMS.
3267 (backprop::process_assign_use): Remove FMA_EXPR handling.
3268 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
3269 (gen_hsa_fma): New function.
3270 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
3271 IFN_FNMA and IFN_FNMS.
3272 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
3273 * gimple-fold.h (follow_all_ssa_edges): Declare.
3274 * gimple-fold.c (follow_all_ssa_edges): New function.
3275 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
3276 gimple_build interface and use follow_all_ssa_edges to fold the result.
3277 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
3278 instead of checking for optabs directly.
3279 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
3280 rather than FMA_EXPRs.
3281 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
3282 call to IFN_FMA instead of an FMA_EXPR.
3284 2018-05-17 Jim Wilson <jimw@sifive.com>
3286 * expr.c (do_tablejump): When converting index to Pmode, if we have a
3287 sign extended promoted subreg, and the range does not have the sign bit
3288 set, then do a sign extend.
3290 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
3291 test, check for sign extended subreg and/or constant operands, and
3292 do a sign extend in that case.
3294 2018-05-17 Steve Ellcey <sellcey@cavium.com>
3296 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
3297 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
3299 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
3300 Change logics_shift_reg to logics_shift_imm.
3301 (thunderx2t99_fp_loadpair_basic): Delete.
3302 (thunderx2t99_fp_storepair_basic): Delete.
3303 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
3304 (thunderx2t99_asimd_polynomial): Delete.
3305 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
3306 and neon_fp_mul_d_scalar_q.
3307 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
3308 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
3309 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
3310 (thunderx2t99_asimd_lut): Add missing tbl types.
3311 (thunderx2t99_asimd_ext): Delete.
3312 (thunderx2t99_asimd_load1_1_mult): Delete.
3313 (thunderx2t99_asimd_load1_2_mult): Delete.
3314 (thunderx2t99_asimd_load1_ldp): New.
3315 (thunderx2t99_asimd_load1): New.
3316 (thunderx2t99_asimd_load2): Add missing *load2* types.
3317 (thunderx2t99_asimd_load3): New.
3318 (thunderx2t99_asimd_load4): New.
3319 (thunderx2t99_asimd_store1_1_mult): Delete.
3320 (thunderx2t99_asimd_store1_2_mult): Delete.
3321 (thunderx2t99_asimd_store2_mult): Delete.
3322 (thunderx2t99_asimd_store2_onelane): Delete.
3323 (thunderx2t99_asimd_store_stp): New.
3324 (thunderx2t99_asimd_store1): New.
3325 (thunderx2t99_asimd_store2): New.
3326 (thunderx2t99_asimd_store3): New.
3327 (thunderx2t99_asimd_store4): New.
3329 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
3331 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
3332 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
3334 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
3335 Segher Boessenkool <segher@kernel.crashing.org>
3338 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
3341 2018-05-17 Richard Biener <rguenther@suse.de>
3343 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
3344 for pruning loop and prune defs feeding only already visited PHIs.
3346 2018-05-17 Richard Biener <rguenther@suse.de>
3348 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
3350 2018-05-17 Bin Cheng <bin.cheng@arm.com>
3351 Richard Biener <rguenther@suse.de>
3353 PR tree-optimization/85793
3354 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
3355 for VMAT_ELEMENTWISE.
3357 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3359 * internal-fn.h (lookup_internal_fn): Declare
3360 * internal-fn.c (lookup_internal_fn): New function.
3361 * gimple.c (gimple_build_call_from_tree): Handle calls to
3363 * gimple-pretty-print.c (dump_gimple_call): Print "." before
3364 internal function names.
3365 * tree-pretty-print.c (dump_generic_node): Likewise.
3366 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
3368 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3370 * gimple-fold.h (gimple_build): Make the function forms take
3371 combined_fn rather than built_in_function.
3372 (gimple_simplify): Likewise.
3373 * gimple-match-head.c (gimple_simplify): Likewise.
3374 * gimple-fold.c (gimple_build): Likewise.
3375 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
3376 rather than gimple_build_call_internal.
3377 (get_initial_defs_for_reduction): Likewise.
3378 (vect_create_epilog_for_reduction): Likewise.
3379 (vectorizable_live_operation): Likewise.
3381 2018-05-17 Martin Liska <mliska@suse.cz>
3383 * gimple-ssa-sprintf.c (format_directive): Do not use
3384 space in between 'G_' and '('.
3386 2018-05-17 Jakub Jelinek <jakub@redhat.com>
3389 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
3390 even if the mask is not all ones.
3393 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
3395 (ix86_gimple_fold_builtin): Likewise.
3398 * config/i386/i386.c: Include tree-vector-builder.h.
3399 (ix86_vector_shift_count): New function.
3400 (ix86_fold_builtin): Fold shift builtins by scalar count.
3401 (ix86_gimple_fold_builtin): Likewise.
3403 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
3404 _mm512_setzero): New intrinsics.
3406 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
3407 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3409 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
3410 code generation for cases where splatting a value is not useful.
3411 * simplify-rtx.c (simplify_ternary_operation): Simplify
3412 vec_merge across a vec_duplicate and a paradoxical subreg forming
3413 a vector mode to a vec_concat.
3415 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
3417 * config.gcc: Support "goldmont-plus".
3418 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3420 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3421 PROCESSOR_GOLDMONT_PLUS.
3422 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
3423 (processor_target_table): Add "goldmont-plus".
3424 (PTA_GOLDMONT_PLUS): Define.
3425 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
3426 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
3427 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
3428 (fold_builtin_cpu): Add "goldmont-plus".
3429 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
3430 (ix86_option_override_internal): Add "goldmont-plus".
3431 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
3432 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
3433 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
3434 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
3436 2018-05-17 Richard Biener <rguenther@suse.de>
3438 PR tree-optimization/85757
3439 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
3440 remove defs that only feed that PHI from further processing.
3442 2018-05-16 Jim Wilson <jimw@sifive.com>
3444 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
3446 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
3448 2018-05-16 Mark Wielaard <mark@klomp.org>
3450 * dwarf2out.c (count_index_strings): New function.
3451 (output_indirect_strings): Call count_index_strings and generate
3452 header for dwarf_version >= 5.
3454 2018-05-16 Mark Wielaard <mark@klomp.org>
3456 * dwarf2out.c (dwarf_FORM): New function.
3457 (set_indirect_string): Use dwarf_FORM.
3458 (reset_indirect_string): Likewise.
3459 (size_of_die): Likewise.
3460 (value_format): Likewise.
3461 (output_die): Likewise.
3462 (add_skeleton_AT_string): Likewise.
3463 (output_macinfo_op): Likewise.
3464 (index_string): Likewise.
3465 (output_index_string_offset): Likewise.
3466 (output_index_string): Likewise.
3467 (count_index_strings): Likewise.
3469 2018-05-16 Carl Love <cel@us.ibm.com>
3471 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3472 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
3474 2018-05-16 Martin Jambor <mjambor@suse.cz>
3476 * ipa-prop.c (ipa_free_all_edge_args): Remove.
3477 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
3479 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
3481 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
3482 (fnma<mode>4): Likewise.
3483 (fms<mode>4): Likewise.
3484 (fnms<mode>4): Likewise.
3485 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
3486 (aarch64_fnma<mode>4): Likewise.
3487 (aarch64_fms<mode>4): Likewise.
3488 (aarch64_fnms<mode>4): Likewise.
3489 (aarch64_fnmadd<mode>4): Likewise.
3491 2018-05-16 Jason Merrill <jason@redhat.com>
3493 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
3495 2018-05-16 Richard Biener <rguenther@suse.de>
3497 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
3498 (dump_stmt_cost): Declare.
3499 (add_stmt_cost): Dump cost we add.
3500 (add_stmt_costs): New function.
3501 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
3503 (vect_analyze_stmt): Adjust prototype.
3504 (vectorizable_condition): Likewise.
3505 (vectorizable_live_operation): Likewise.
3506 (vectorizable_reduction): Likewise.
3507 (vectorizable_induction): Likewise.
3508 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
3509 cost vector to pass to vectorizable_ and record afterwards.
3510 (vect_model_reduction_cost): Take cost vector argument and adjust.
3511 (vect_model_induction_cost): Likewise.
3512 (vectorizable_reduction): Likewise.
3513 (vectorizable_induction): Likewise.
3514 (vectorizable_live_operation): Likewise.
3515 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
3516 SLP_TREE_NUMBER_OF_VEC_STMTS.
3517 (vect_analyze_slp_cost_1): Remove.
3518 (vect_analyze_slp_cost): Likewise.
3519 (vect_slp_analyze_node_operations): Take visited args and
3520 a target cost vector. Avoid processing already visited stmt sets.
3521 (vect_slp_analyze_operations): Use a local cost vector to gather
3522 costs and register those of non-discarded instances.
3523 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
3524 (vect_schedule_slp_instance): Remove copying of
3525 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
3527 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
3528 adding cost. Record cost entry location.
3529 (vect_prologue_cost_for_slp_op): Function to compute cost of
3530 a constant or invariant generated for SLP vect in the prologue,
3531 split out from vect_analyze_slp_cost_1.
3532 (vect_model_simple_cost): Make static. Adjust for SLP costing.
3533 (vect_model_promotion_demotion_cost): Likewise.
3534 (vect_model_store_cost): Likewise, make static.
3535 (vect_model_load_cost): Likewise.
3536 (vectorizable_bswap): Add cost vector arg and adjust.
3537 (vectorizable_call): Likewise.
3538 (vectorizable_simd_clone_call): Likewise.
3539 (vectorizable_conversion): Likewise.
3540 (vectorizable_assignment): Likewise.
3541 (vectorizable_shift): Likewise.
3542 (vectorizable_operation): Likewise.
3543 (vectorizable_store): Likewise.
3544 (vectorizable_load): Likewise.
3545 (vectorizable_condition): Likewise.
3546 (vectorizable_comparison): Likewise.
3547 (can_vectorize_live_stmts): Likewise.
3548 (vect_analyze_stmt): Likewise.
3549 (vect_transform_stmt): Adjust calls to vectorizable_*.
3550 * tree-vectorizer.c: Include gimple-pretty-print.h.
3551 (dump_stmt_cost): New function.
3553 2018-05-16 Richard Biener <rguenther@suse.de>
3555 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
3556 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
3557 * tree-ssa-dse.c: Include tree-ssa-loop.h.
3558 (check_name): New callback.
3559 (dse_classify_store): Track cycles via a visited bitmap of PHI
3560 defs and simplify handling of in-loop and across loop dead stores
3561 and properly fail for loop-variant refs. Handle byte-tracking with
3562 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
3565 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
3567 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
3568 (vect_get_mask_type_for_stmt): Likewise.
3569 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
3571 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
3572 to determine the statement's vector type and the vector type that
3573 should be used for calculating nunits. Deal with cases in which
3574 the type has to be deferred.
3575 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
3576 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
3577 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
3578 (vect_determine_vf_for_stmt): New functions, split out from...
3579 (vect_determine_vectorization_factor): ...here.
3580 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
3581 (vect_get_mask_type_for_stmt): New functions, split out from
3582 vect_determine_vectorization_factor.
3584 2018-05-16 Richard Biener <rguenther@suse.de>
3586 * tree-cfg.c (verify_gimple_assign_ternary): Properly
3587 verify the [VEC_]COND_EXPR embedded comparison.
3589 2018-05-15 Martin Sebor <msebor@redhat.com>
3591 PR tree-optimization/85753
3592 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
3593 RECORD_TYPE in addition to ARRAY_TYPE.
3595 2018-05-15 Martin Sebor <msebor@redhat.com>
3598 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
3600 2018-05-15 Richard Biener <rguenther@suse.de>
3602 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
3603 add by_clobber_p one. Change algorithm to collect all defs
3604 representing uses we need to walk and try reducing them to
3605 a single one before failing.
3606 (dse_dom_walker::dse_optimize_stmt): Adjust.
3608 2018-05-13 Mark Wielaard <mark@klomp.org>
3610 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
3611 (size_of_loc_descr): Likewise.
3612 (output_loc_operands): Likewise.
3613 (output_loc_operands_raw): Likewise.
3614 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
3615 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
3616 (hash_loc_operands): Likewise.
3617 (compare_loc_operands): Likewise.
3619 2018-05-14 Mark Wielaard <mark@klomp.org>
3621 * dwarf2out.c (count_index_addrs): New function.
3622 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
3624 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3626 PR tree-optimization/83648
3627 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
3628 return value as malloc candidate.
3630 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3633 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
3634 param as true in call to suggest_attribute.
3636 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
3638 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
3639 -mreadonly-in-sdata.
3641 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3643 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
3645 (aarch64_crypto_aesd_fused): Likewise.
3647 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
3649 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
3650 (movsi_aarch64): Likewise.
3651 (load_pairsi): Likewise.
3652 (load_pairdi): Likewise.
3653 (store_pairsi): Likewise.
3654 (store_pairdi): Likewise.
3655 (load_pairsf): Likewise.
3656 (load_pairdf): Likewise.
3657 (store_pairsf): Likewise.
3658 (store_pairdf): Likewise.
3659 (zero_extend): Likewise.
3660 (trunc): Swap alternatives.
3661 (fcvt_target): Add '?' to prefer w over r.
3663 2018-05-14 Jakub Jelinek <jakub@redhat.com>
3666 * config/i386/i386.md: Disallow non-commutative arithmetics in
3667 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
3668 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
3669 in the peephole2 before it.
3671 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
3673 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
3674 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
3675 (ix86_handle_option): Handle -mcldemote.
3676 * config.gcc: New header.
3677 * config/i386/cldemoteintrin.h: New file.
3678 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
3679 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3681 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3682 OPTION_MASK_ISA_CLDEMOTE.
3683 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
3684 (ix86_valid_target_attribute_inner_p): Ditto.
3685 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
3686 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
3687 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
3688 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
3689 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
3691 * config/i386/i386.opt: Add -mcldemote.
3692 * config/i386/x86intrin.h: New header.
3693 * doc/invoke.texi: Add -mcldemote.
3695 2018-05-14 Richard Biener <rguenther@suse.de>
3697 * doc/match-and-simplify.texi: Adjust :s documentation.
3699 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
3701 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
3702 intended memcpy size.
3703 (REORDER_45): Likewise.
3705 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
3707 * sort.cc: New file.
3708 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
3709 * vec.c (qsort_chk): Use gcc_qsort.
3710 * Makefile.in (OBJS-libcommon): Add sort.o.
3711 (build/sort.o): New target. Use it...
3712 (BUILD_RTL): ... here, and...
3713 (build/gencfn-macros): ... here, and...
3714 (build/genmatch): ... here.
3716 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3717 Chung-Ju Wu <jasonwucj@gmail.com>
3719 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
3720 * config/nds32/nds32-graywolf.md: New file.
3721 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
3722 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
3724 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
3725 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
3726 * config/nds32/nds32.md (pipeline_model): Add graywolf.
3727 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
3728 * config/nds32/pipelines.md: Include n15 settings.
3730 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3731 Chung-Ju Wu <jasonwucj@gmail.com>
3733 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
3734 * config/nds32/nds32-n13.md: New file.
3735 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
3736 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
3738 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
3739 * config/nds32/nds32.md (pipeline_model): Add n13.
3740 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
3741 * config/nds32/pipelines.md: Include n13 settings.
3743 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3744 Chung-Ju Wu <jasonwucj@gmail.com>
3746 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
3747 * config/nds32/nds32-n10.md: New file.
3748 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
3749 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
3751 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
3752 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
3753 * config/nds32/nds32.md (pipeline_model): Add n10.
3754 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
3755 * config/nds32/pipelines.md: Include n10 settings.
3757 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
3758 Kito Cheng <kito.cheng@gmail.com>
3759 Chung-Ju Wu <jasonwucj@gmail.com>
3761 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
3762 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3763 Add enum values for DSP extension instructions.
3764 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
3766 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
3767 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
3769 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
3770 * config/nds32/nds32-dspext.md: New file for DSP implementation.
3771 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
3772 * config/nds32/nds32-intrinsic.md: Likewise.
3773 * config/nds32/nds32_intrinsic.h: Likewise.
3774 * config/nds32/nds32-md-auxiliary.c: Likewise.
3775 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
3776 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
3777 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
3778 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
3779 * config/nds32/nds32-protos.h: New declarations for DSP extension.
3780 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
3781 TYPE_DMAC in switch statement.
3782 * config/nds32/nds32.c: New checking and implementation for DSP
3783 extension instructions.
3784 * config/nds32/nds32.h: Likewise.
3785 * config/nds32/nds32.md: Likewise.
3786 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
3787 * config/nds32/predicates.md: Implement new predicates for DSP
3790 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
3792 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
3793 Reformat alternatives and attributes so it is easier to identify
3794 which constraints/attributes go with which instruction.
3795 (mov<mode>_hardfloat32, FMOVE64): Likewise.
3796 (mov<mode>_softfloat32, FMOVE64): Likewise.
3797 (mov<mode>_hardfloat64, FMOVE64): Likewise.
3798 (mov<mode>_softfloat64, FMOVE64): Likewise.
3800 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
3802 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3804 (Basic PowerPC Built-in Functions): The new name of the
3805 subsection previously known as "PowerPC Built-in Functions".
3806 (Basic PowerPC Built-in Functions Available on all Configurations):
3808 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
3809 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
3810 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
3811 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
3813 2018-05-11 Martin Jambor <mjambor@suse.cz>
3816 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
3819 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
3822 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
3824 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
3826 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
3827 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
3828 (ix86_handle_option): Handle -mwaitpkg.
3829 * config.gcc: New header.
3830 * config/i386/cpuid.h (bit_WAITPKG): New bit.
3831 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
3832 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
3834 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3835 OPTION_MASK_ISA_WAITPKG.
3836 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
3837 (ix86_option_override_internal): Add PTA_WAITPKG.
3838 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
3839 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
3840 IX86_BUILTIN_TPAUSE.
3841 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
3842 __builtin_ia32_umwait and __builtin_ia32_tpause.
3843 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
3844 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
3845 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
3846 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
3847 UNSPECV_TPAUSE): New.
3848 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
3849 * config/i386/i386.opt: Add -mwaitpkg.
3850 * config/i386/waitpkgintrin.h: New file.
3851 * config/i386/x86intrin.h: New header.
3852 * doc/invoke.texi: Add -mwaitpkg.
3854 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
3857 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
3859 (cortex-m0): Use armv6s-m isa.
3860 (cortex-m0plus): Likewise.
3861 (cortex-m1): Likewise.
3862 (cortex-m0.small-multiply): Likewise.
3863 (cortex-m0plus.small-multiply): Likewise.
3864 (cortex-m1.small-multiply): Likewise.
3866 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
3867 Jakub Jelinek <jakub@redhat.com>
3869 PR tree-optimization/85692
3870 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
3871 source permute as well.
3873 2018-05-11 Martin Liska <mliska@suse.cz>
3876 * doc/extend.texi: Document LLVM style format for no_sanitize
3879 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
3881 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
3882 mode_supports_vsx_dform_quad to mode_supports_dq_form.
3883 (mode_supports_vsx_dform_quad): Likewise.
3884 (mode_supports_vmx_dform): Move these functions to be next to the
3885 other mode_supports functions.
3886 (mode_supports_dq_form): Likewise.
3887 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
3888 mode_supports_dq_form.
3889 (reg_offset_addressing_ok_p): Likewise.
3890 (offsettable_ok_by_alignment): Likewise.
3891 (rs6000_legitimate_offset_address_p): Likewise.
3892 (legitimate_lo_sum_address_p): Likewise.
3893 (rs6000_legitimize_address): Likewise.
3894 (rs6000_legitimize_reload_address): Likewise.
3895 (rs6000_secondary_reload_inner): Likewise.
3896 (rs6000_preferred_reload_class): Likewise.
3897 (rs6000_output_move_128bit): Likewise.
3899 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
3901 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
3902 Generate SImode target register for null target.
3903 <case IX86_BUILTIN_XGETBV>: Ditto.
3904 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
3905 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
3907 2018-05-10 Carl Love <cel@us.ibm.com>
3909 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3910 dcbtt and dcbtstt if operands[2] is 0.
3912 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
3915 * config/i386/sse.md (usadv64qi): New expander.
3917 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
3919 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
3920 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
3921 -maltivec=be support.
3922 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
3923 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
3924 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
3925 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
3926 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
3927 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
3928 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
3929 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
3930 altivec_vsumsws): Adjust.
3931 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
3932 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
3933 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
3935 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
3936 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
3937 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
3938 (altivec_lve<VI_char>x): Delete expand.
3939 (*altivec_lve<VI_char>x_internal): Rename to...
3940 (altivec_lve<VI_char>x): ... this.
3941 (altivec_lvxl_<mode>): Delete expand.
3942 (*altivec_lvxl_<mode>_internal): Rename to ...
3943 (altivec_lvxl_<mode>): ... this.
3944 (altivec_stvxl_<mode>): Delete expand.
3945 (*altivec_stvxl_<mode>_internal): Rename to ...
3946 (altivec_stvxl_<mode>): ... this.
3947 (altivec_stve<VI_char>x): Delete expand.
3948 (*altivec_stve<VI_char>x_internal): Rename to ...
3949 (altivec_stve<VI_char>x): ... this.
3950 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
3951 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
3952 reduc_plus_scal_<mode>): Adjust.
3953 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
3955 (rs6000_cpu_cpp_builtins): Adjust.
3956 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
3957 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
3958 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
3959 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
3960 -maltivec=be support.
3961 (rs6000_split_vec_extract_var): Adjust.
3962 (rs6000_split_v4si_init): Adjust.
3963 (swap_selector_for_mode): Delete.
3964 (altivec_expand_lvx_be, altivec_expand_stvx_be,
3965 altivec_expand_stvex_be): Delete.
3966 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
3967 -maltivec=be support.
3968 (rs6000_gimple_fold_builtin): Ditto.
3969 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
3971 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
3972 (TARGET_DIRECT_MOVE_64BIT): Adjust.
3973 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
3974 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
3975 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
3976 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
3977 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
3978 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
3979 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
3980 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
3981 anonymous split): Adjust.
3982 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
3983 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
3985 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
3987 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
3988 when --with-gxx-include-dir is also specified.
3989 * configure: Regenerate.
3991 2018-05-09 Jim Wilson <jimw@sifive.com>
3994 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
3995 * config/riscv/t-withmultilib: New.
3996 * config/riscv/withmultilib.h: New.
3997 * doc/install.texi: Document RISC-V --with-multilib-list support.
3999 2018-05-09 Richard Biener <rguenther@suse.de>
4001 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
4003 (vect_bb_vectorization_profitable_p): Adjust. Compute
4004 actual scalar cost using the cost vector and the add_stmt_cost
4007 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4009 PR rtl-optimization/85645
4010 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
4011 in the REG_CFA_REGISTER note for LR, don't leave it empty.
4013 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4015 PR rtl-optimization/85645
4016 * shrink-wrap.c (spread_components): Return a boolean saying if
4017 anything was changed.
4018 (try_shrink_wrapping_separate): Iterate spread_components until
4019 nothing changes anymore.
4021 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4023 PR rtl-optimization/85645
4024 * regrename.c (build_def_use): Also kill the chains that include the
4025 destination of a REG_CFA_REGISTER note.
4027 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4029 PR rtl-optimization/85645
4030 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
4031 insn that has a REG_CFA_REGISTER note.
4033 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
4035 * cfgexpand.c (expand_clobber): New function.
4036 (expand_gimple_stmt_1): Use it.
4037 * tree-vect-stmts.c (vect_clobber_variable): New function,
4039 (vectorizable_simd_clone_call): ...here.
4040 (vectorizable_store): Emit a clobber either side of an
4041 IFN_STORE_LANES sequence.
4042 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
4044 2018-05-09 Tom de Vries <tom@codesourcery.com>
4047 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
4048 (define_insn "trap_if_false"): Add exit after trap.
4050 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
4052 PR rtl-optimization/85638
4053 * bb-reorder.c: Include common/common-target.h.
4054 (create_forwarder_block): New function extracted from...
4055 (fix_up_crossing_landing_pad): ...here. Rename into...
4056 (dw2_fix_up_crossing_landing_pad): ...this.
4057 (sjlj_fix_up_crossing_landing_pad): New function.
4058 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
4059 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
4060 from both partitions and exit the loop after one iteration.
4062 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4065 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4067 (Basic PowerPC Built-in Functions): The new name of the
4068 subsection previously known as "PowerPC Built-in Functions".
4069 (Basic PowerPC Built-in Functions Available on all Configurations):
4071 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4073 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4075 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4077 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4080 2018-05-08 Jim Wilson <jimw@sifive.com>
4082 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
4083 (LD_EMUL_SUFFIX): New.
4084 (LINK_SPEC): Use it.
4086 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4088 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4090 (Basic PowerPC Built-in Functions): The new name of the
4091 subsection previously known as "PowerPC Built-in Functions".
4092 (Basic PowerPC Built-in Functions Available on all Configurations):
4094 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4096 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4098 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4100 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4103 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4106 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4107 after cmpelim optimization.
4109 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
4111 * config.gcc: Support "goldmont".
4112 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4113 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4115 * config/i386/i386.c (m_GOLDMONT): Define.
4116 (processor_target_table): Add "goldmont".
4117 (PTA_GOLDMONT): Define.
4118 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4119 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4120 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4121 (fold_builtin_cpu): Add "goldmont".
4122 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4123 (ix86_option_override_internal): Add "goldmont".
4124 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4125 (processor_type): Add PROCESSOR_GOLDMONT.
4126 * config/i386/i386.md: Add CPU "glm".
4127 * config/i386/glm.md: New file.
4128 * config/i386/x86-tune.def: Add m_GOLDMONT.
4129 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4131 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4134 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4136 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4137 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
4138 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
4141 * config/i386/i386.c (ix86_fold_builtin): Handle
4142 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4145 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4146 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4148 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4151 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4152 (check_arch): Likewise.
4153 (check_fpu): Return the result rather than printing it.
4154 (end arch): Fix operator precedence.
4155 (end cpu): Likewise.
4156 (END): Print the result from check_fpu.
4158 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4159 Alan Hayward <alan.hayward@arm.com>
4160 David Sherwood <david.sherwood@arm.com>
4162 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4163 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4164 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4165 (*fcmuo<mode>_and): New patterns.
4167 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4169 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4170 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4171 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4172 (cmp_op, sve_imm_con): New code attributes.
4173 (SVE_COND_INT_CMP, imm_con): Delete.
4174 (cmp_op): Remove above unspecs from int attribute.
4175 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4177 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
4178 comparison-specific unspecs.
4179 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4180 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4181 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4182 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4183 (*vec_fcm<cmp_op><mode>): Rename to...
4184 (*fcm<cmp_op><mode>): ...this and adjust likewise.
4185 (*vec_fcmuo<mode>): Rename to...
4186 (*fcmuo<mode>): ...this and adjust likewise.
4187 (*pred_fcm<cmp_op><mode>): New pattern.
4188 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4189 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4191 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4193 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4194 (aarch64_emit_sve_predicated_cond): New function.
4195 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4196 (aarch64_emit_unspec_cond_or): Replace with...
4197 (aarch64_emit_sve_or_conds): ...this new function. Use
4198 aarch64_emit_sve_ptrue_op for the individual comparisons and
4199 aarch64_emit_binop to OR them together.
4200 (aarch64_emit_inverted_unspec_cond): Replace with...
4201 (aarch64_emit_sve_inverted_cond): ...this new function. Use
4202 aarch64_emit_sve_ptrue_op for the comparison and
4203 aarch64_emit_unop to invert the result.
4204 (aarch64_expand_sve_vec_cmp_float): Update after the above
4205 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
4207 2018-05-07 Nathan Sidwell <nathan@acm.org>
4209 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4210 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4211 (Backwards Compatibility): Likewise.
4213 2018-05-07 Luis Machado <luis.machado@linaro.org>
4217 2018-05-07 Luis Machado <luis.machado@linaro.org>
4219 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4220 <prefetch_dynamic_strides>: New const bool field.
4221 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4222 prefetch_dynamic_strides.
4223 (exynosm1_prefetch_tune): Likewise.
4224 (thunderxt88_prefetch_tune): Likewise.
4225 (thunderx_prefetch_tune): Likewise.
4226 (thunderx2t99_prefetch_tune): Likewise.
4227 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4229 (aarch64_override_options_internal): Update to set
4230 PARAM_PREFETCH_DYNAMIC_STRIDES.
4231 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4232 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4233 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4234 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4235 prefetch-dynamic-strides setting.
4237 2018-05-07 Luis Machado <luis.machado@linaro.org>
4239 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4240 <minimum_stride>: New const int field.
4241 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4242 minimum_stride field.
4243 (exynosm1_prefetch_tune): Likewise.
4244 (thunderxt88_prefetch_tune): Likewise.
4245 (thunderx_prefetch_tune): Likewise.
4246 (thunderx2t99_prefetch_tune): Likewise.
4247 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4248 (aarch64_override_options_internal): Update to set
4249 PARAM_PREFETCH_MINIMUM_STRIDE.
4250 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4251 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4252 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4253 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4254 stride is constant and is below the minimum stride threshold.
4256 2018-05-07 Luis Machado <luis.machado@linaro.org>
4258 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
4261 2018-05-07 Luis Machado <luis.machado@linaro.org>
4263 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4264 <prefetch_dynamic_strides>: New const bool field.
4265 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4266 prefetch_dynamic_strides.
4267 (exynosm1_prefetch_tune): Likewise.
4268 (thunderxt88_prefetch_tune): Likewise.
4269 (thunderx_prefetch_tune): Likewise.
4270 (thunderx2t99_prefetch_tune): Likewise.
4271 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4273 (aarch64_override_options_internal): Update to set
4274 PARAM_PREFETCH_DYNAMIC_STRIDES.
4275 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4276 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4277 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4278 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4279 prefetch-dynamic-strides setting.
4281 2018-05-07 Luis Machado <luis.machado@linaro.org>
4283 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4284 <minimum_stride>: New const int field.
4285 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4286 minimum_stride field.
4287 (exynosm1_prefetch_tune): Likewise.
4288 (thunderxt88_prefetch_tune): Likewise.
4289 (thunderx_prefetch_tune): Likewise.
4290 (thunderx2t99_prefetch_tune): Likewise.
4291 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4292 (aarch64_override_options_internal): Update to set
4293 PARAM_PREFETCH_MINIMUM_STRIDE.
4294 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4295 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4296 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4297 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4298 stride is constant and is below the minimum stride threshold.
4300 2018-05-06 Jakub Jelinek <jakub@redhat.com>
4303 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4304 the type is addressable. Don't force op into register if it has
4307 2018-05-05 Roland McGrath <mcgrathr@google.com>
4310 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
4311 any section for which we don't know a specific type it should have,
4312 regardless of name. Previously this was done only for the exact
4313 names ".init_array", ".fini_array", and ".preinit_array".
4314 (default_elf_asm_named_section): Add comment about
4315 relationship with default_section_type_flags and SECTION_NOTYPE.
4316 (get_section): Don't consider it a type conflict if one side has
4317 SECTION_NOTYPE and the other doesn't, as long as neither has the
4318 SECTION_BSS et al used in the default_section_type_flags logic.
4320 2018-05-05 Tom de Vries <tom@codesourcery.com>
4323 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
4324 (workaround_barsyncs): New function.
4325 (nvptx_reorg): Use workaround_barsyncs.
4326 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4327 (define_expand "nvptx_membar_cta"): New define_expand.
4328 (define_insn "*nvptx_membar_cta"): New insn.
4330 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
4332 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
4333 To improve optimization opportunities.
4334 * builtin-types.def: The new needed builtin types for the above.
4336 2018-05-04 Richard Biener <rguenther@suse.de>
4338 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4339 * gimple-ssa-store-merging.c
4340 (imm_store_chain_info::output_merged_store): Remove redundant create,
4341 release split_store vector contents on failure.
4342 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4343 scalar stmt vector on cache hit.
4345 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
4347 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
4349 * config.gcc (powerpc-xilinx-eabi*): Remove.
4350 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
4352 (fusion_addis_mem_combo_load): Ditto.
4353 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
4355 (rs6000_cpu_cpp_builtins): Ditto.
4356 * config/rs6000/rs6000-linux.c
4357 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
4358 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
4359 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
4361 (rs6000_setup_reg_addr_masks): Ditto.
4362 (rs6000_init_hard_regno_mode_ok): Ditto.
4363 (rs6000_option_override_internal): Ditto.
4364 (legitimate_lo_sum_address_p): Ditto.
4365 (rs6000_legitimize_address): Ditto.
4366 (rs6000_legitimize_reload_address): Ditto.
4367 (rs6000_legitimate_address_p): Ditto.
4368 (abi_v4_pass_in_fpr): Ditto.
4369 (setup_incoming_varargs): Ditto.
4370 (rs6000_gimplify_va_arg): Ditto.
4371 (rs6000_split_multireg_move): Ditto.
4372 (rs6000_savres_strategy): Ditto.
4373 (rs6000_emit_prologue_components): Ditto.
4374 (rs6000_emit_epilogue_components): Ditto.
4375 (rs6000_emit_prologue): Ditto.
4376 (rs6000_emit_epilogue): Ditto.
4377 (rs6000_elf_file_end): Ditto.
4378 (rs6000_function_value): Ditto.
4379 (rs6000_libcall_value): Ditto.
4380 * config/rs6000/rs6000.h: Ditto.
4381 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
4382 (TARGET_MINMAX): ... this. New.
4383 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
4384 * config/rs6000/rs6000.md: Remove Xilinx FP support.
4385 (*movsi_internal1_single): Delete.
4386 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
4387 mfpu=, mxilinx-fpu): Delete.
4388 * config/rs6000/singlefp.h: Delete.
4389 * config/rs6000/sysv4.h: Remove Xilinx FP support.
4390 * config/rs6000/t-rs6000: Ditto.
4391 * config/rs6000/t-xilinx: Delete.
4392 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
4393 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
4394 (VStype_simple): Delete.
4395 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
4396 * config/rs6000/xfpu.h: Delete.
4397 * config/rs6000/xfpu.md: Delete.
4398 * config/rs6000/xilinx.h: Delete.
4399 * config/rs6000/xilinx.opt: Delete.
4400 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
4401 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
4403 2018-05-04 Tom de Vries <tom@codesourcery.com>
4406 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
4409 2018-05-04 Richard Biener <rguenther@suse.de>
4412 * tree-complex.c (update_complex_assignment): We are always in SSA form.
4413 (expand_complex_div_wide): Likewise.
4414 (expand_complex_operations_1): Likewise.
4415 (expand_complex_libcall): Preserve EH info of the original stmt.
4416 (tree_lower_complex): Handle removed blocks.
4417 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
4418 on complex multiplication and division libcall builtins.
4420 2018-05-04 Richard Biener <rguenther@suse.de>
4423 * fold-const.c (negate_expr_p): Restrict negation of operand
4424 zero of a division to when we know that can happen without
4426 (fold_negate_expr_1): Likewise.
4428 2018-05-04 Jakub Jelinek <jakub@redhat.com>
4431 * real.h (real_nextafter): Declare.
4432 * real.c (real_nextafter): New function.
4433 * fold-const-call.c (fold_const_nextafter): New function.
4434 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
4435 CASE_CFN_NEXTTOWARD.
4436 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
4437 even when arg1_mode is different from arg0_mode.
4439 2018-05-03 Nathan Sidwell <nathan@acm.org>
4441 * doc/extend.texi (Deprecated Features): Remove
4443 (Backwards Compatibility): Likewise.
4444 * doc/invoke.texi (C++ Language Options): Likewise.
4445 (C++ Dialect Options): Likewise.
4447 2018-05-03 Jakub Jelinek <jakub@redhat.com>
4450 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
4451 _mm512_mask_mullox_epi64): New intrinsics.
4453 2018-05-03 Tom de Vries <tom@codesourcery.com>
4456 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4457 dump files): Add offload-tree.
4459 2018-05-03 Richard Biener <rguenther@suse.de>
4461 PR tree-optimization/85615
4462 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
4463 to loops not nested in BBs loop father to avoid creating multi-entry
4466 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4468 PR tree-optimization/70291
4469 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
4470 arguments. Change return type to tree. Emit libcall as a new
4471 statement rather than replacing existing one when inplace_p is true.
4472 (expand_complex_multiplication_components): New function.
4473 (expand_complex_multiplication): Expand floating-point complex
4474 multiplication using the above.
4475 (expand_complex_division): Rename inner_type parameter to type.
4476 Update expand_complex_libcall call-site.
4477 (expand_complex_operations_1): Update expand_complex_multiplication
4478 and expand_complex_division call-sites.
4480 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4483 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4484 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4485 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
4486 the highest significant bit of the shift count mask is clear. In
4487 check whether and[sq]i3 is needed verify that all significant bits
4488 of the shift count other than the highest are set.
4490 2018-05-02 Tom de Vries <tom@codesourcery.com>
4493 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
4494 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
4495 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
4496 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
4497 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
4498 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
4499 __builtin_goacc_parlevel_size.
4501 2018-05-02 Richard Biener <rguenther@suse.de>
4503 PR tree-optimization/85597
4504 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
4505 do not use split vect_get_vec_defs call but call vect_get_slp_defs
4508 2018-05-02 Tom de Vries <tom@codesourcery.com>
4511 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4512 dump files): Add ltrans-tree.
4514 2018-05-02 Tom de Vries <tom@codesourcery.com>
4517 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4518 dump files): Add wpa-ipa.
4520 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
4522 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
4523 powerpc*-*-linux*paired* target.
4524 * config/rs6000/750cl.h: Delete.
4525 * config/rs6000/paired.h: Delete.
4526 * config/rs6000/paired.md: Delete.
4527 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
4529 * config/rs6000/rs6000-builtin.def: Remove paired float support.
4530 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
4531 comment. Remove paired float support.
4532 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
4533 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4535 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
4536 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
4538 * config/rs6000/rs6000.c: Remove paired float support.
4539 (paired_expand_vector_init, paired_expand_vector_move,
4540 paired_emit_vector_compare, paired_emit_vector_cond_expr,
4541 (paired_expand_lv_builtin, paired_expand_stv_builtin,
4542 paired_expand_builtin, paired_expand_predicate_builtin,
4543 paired_init_builtins): Delete.
4544 * config/rs6000/rs6000.h: Remove paired float support.
4545 * config/rs6000/rs6000.md: Remove paired float support.
4546 (move_from_CR_ov_bit): Delete.
4547 * config/rs6000/rs6000.opt (mpaired): Delete.
4548 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
4549 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
4551 2018-05-02 Richard Biener <rguenther@suse.de>
4554 * gimplify.c (gimplify_save_expr): When in SSA form allow
4555 SAVE_EXPRs to compute to SSA vars.
4557 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4560 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4561 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4562 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
4563 clobber operands[2], instead use a new pseudo. Formatting fixes.
4565 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
4567 PR tree-optimization/85586
4568 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
4569 exit early for statements in the same group if the accesses are
4572 2018-05-02 Tom de Vries <tom@codesourcery.com>
4575 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
4578 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
4580 PR tree-optimization/85143
4581 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
4583 2018-05-01 Tom de Vries <tom@codesourcery.com>
4586 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
4587 not found" error message.
4589 2018-05-01 Tom de Vries <tom@codesourcery.com>
4592 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
4593 * vec.c (test_ordered_remove_if): New function.
4594 (vec_c_tests): Call test_ordered_remove_if.
4595 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
4596 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
4597 * tree-vect-patterns.c (vect_pattern_recog_1): Use
4598 VEC_ORDERED_REMOVE_IF.
4600 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4602 PR tree-optimization/82665
4603 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
4604 pointer subtraction where arguments come from a memchr call.
4606 2018-05-01 Jakub Jelinek <jakub@redhat.com>
4608 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
4609 --push-state --as-needed and --pop-state instead of --as-needed and
4610 --no-as-needed if ld supports it.
4611 * configure: Regenerated.
4614 * doc/install.texi2html: Replace _002d with - and _002a with * in
4615 generated html files using sed.
4617 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4620 * gcc-rich-location.c (blank_line_before_p): New function.
4621 (use_new_line): New function.
4622 (gcc_rich_location::add_fixit_insert_formatted): New function.
4623 * gcc-rich-location.h
4624 (gcc_rich_location::add_fixit_insert_formatted): New function.
4626 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4628 * selftest.c (assert_streq): Rename "expected" and "actual" to
4629 "val1" and "val2". Extend NULL-handling to cover both inputs
4630 symmetrically, while still requiring both to be non-NULL for a pass.
4631 * selftest.h (assert_streq): Rename "expected" and "actual" to
4633 (ASSERT_EQ): Likewise.
4634 (ASSERT_EQ_AT): Likewise.
4635 (ASSERT_KNOWN_EQ): Likewise.
4636 (ASSERT_KNOWN_EQ_AT): Likewise.
4637 (ASSERT_NE): Likewise.
4638 (ASSERT_MAYBE_NE): Likewise.
4639 (ASSERT_MAYBE_NE_AT): Likewise.
4640 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
4641 the assertion to pass.
4642 (ASSERT_STREQ_AT): Likewise.
4644 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4646 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
4647 interaction with -pie.
4649 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4651 * selftest.h: Fix alphabetization of per-source-file selftest
4654 2018-04-30 Jason Merrill <jason@redhat.com>
4656 PR c++/61982 - dead stores to destroyed objects.
4657 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
4660 2018-04-30 Jason Merrill <jason@redhat.com>
4662 * tree.c (build_clobber): New.
4663 * tree.h: Declare it.
4664 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
4666 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4668 * diagnostic-show-locus.c (layout::layout): Update for
4669 location_get_source_line returning a char_span.
4670 (struct char_span): Move to input.h.
4671 (struct correction): Update for fields in char_span becoming
4673 (struct source_line): Update for location_get_source_line
4674 returning a char_span.
4675 (layout::print_line): Likewise.
4676 * edit-context.c (edited_file::print_content): Likewise.
4677 (edited_file::print_diff_hunk): Likewise.
4678 (edited_file::print_run_of_changed_lines): Likewise.
4679 (edited_file::get_num_lines): Likewise.
4680 (edited_line::edited_line): Likewise.
4681 * final.c (asm_show_source): Likewise.
4682 * input.c (location_get_source_line): Convert return type
4683 from const char * to char_span, losing the final "line_len"
4685 (dump_location_info): Update for the above.
4686 (get_substring_ranges_for_loc): Likewise. Use a char_span
4687 when handling the literal within the line.
4688 (test_reading_source_line): Update for location_get_source_line
4689 returning a char_span.
4690 * input.h (class char_span): Move here from
4691 diagnostic-show-locus.c, converting from a struct to a class.
4692 Make data members private.
4693 (char_span::operator bool): New.
4694 (char_span::length): New.
4695 (char_span::get_buffer): New.
4696 (char_span::operator[]): New.
4697 (char_span::subspan): Make const.
4698 (char_span::xstrdup): New.
4699 (location_get_source_line): Convert return type from const char *
4700 to char_span, losing the final "line_size" param.
4702 2018-04-30 Jan Hubicka <jh@suse.cz>
4704 * lto-wrapper.c (ltrans_priorities): New static var.
4705 (cmp_priority): New.
4706 (run_gcc): Read priorities and if doing parallel build order
4707 the Makefile by them.
4709 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4711 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
4713 2018-04-30 Richard Biener <rguenther@suse.de>
4715 * tree-cfg.c (verify_address): Remove base argument, add
4716 flag whether to check TREE_ADDRESSABLE and do that.
4717 (verify_expr): Remove.
4718 (verify_types_in_gimple_reference): Add pieces from verify_expr.
4719 (verify_gimple_assign_single): Likewise.
4720 (verify_gimple_switch): Likewise.
4721 (verify_expr_location_1): Dereference tp once. Add (disabled)
4722 piece from verify_expr.
4723 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
4725 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4727 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
4729 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4731 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
4732 (small_data_pattern): Likewise.
4733 (arc_rewrite_small_data): Likewise.
4734 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
4735 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
4736 (get_symbol_alignment): New function.
4737 (legitimate_small_data_address_p): Likewise.
4738 (legitimate_scaled_address): Update, call
4739 legitimate_small_data_address_p.
4740 (output_sdata): New static variable.
4741 (arc_print_operand): Update how we handle small data operands.
4742 (arc_print_operand_address): Likewise.
4743 (arc_legitimate_address_p): Update, use
4744 legitimate_small_data_address_p.
4745 (arc_rewrite_small_data_p): Remove.
4746 (arc_rewrite_small_data_1): Likewise.
4747 (arc_rewrite_small_data): Likewise.
4748 (small_data_pattern): Likewise.
4749 (compact_sda_memory_operand): Update to use
4750 legitimate_small_data_address_p and get_symbol_alignment.
4751 (prepare_move_operands): Don't rewite sdata pattern.
4752 (prepare_extend_operands): Remove.
4753 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
4755 (zero_extendqisi2): Likewise.
4756 (zero_extendhisi2): Likewise.
4757 (extendqihi2): Likewise.
4758 (extendqisi2): Likewise.
4759 (extendhisi2): Likewise.
4763 * config/arc/constraints.md (Usd): Change it to memory constraint.
4765 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4767 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
4768 as source of std instructions.
4769 * config/arc/arc.md (movsi_insn): Update pattern predicate to
4770 allow 6-bit constants as source for store instructions.
4771 (movdi_insn): Update instruction pattern to allow 6-bit constants
4772 as source for store instructions.
4774 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4776 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
4778 2018-04-30 Nathan Sidwell <nathan@acm.org>
4779 Sandra Loosemore <sandra@codesourcery.com>
4781 * dumpfile.c (dump_open): Allow '-' for stdout.
4782 * doc/invoke.texi (Developer Options): Document dump filename
4783 determination early. Document stdin/stdout selection.
4785 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
4787 Microblaze Target: PIC data text relative
4789 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
4790 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
4792 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
4793 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
4794 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
4795 New addressing mode for data-text relative position indepenedent code.
4796 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
4797 'ADDRESS_SYMBOLIC_TXT_REL'.
4798 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
4799 (microblaze_legitimate_pic_operand): Exclude function calls from
4800 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
4801 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
4803 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4804 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4805 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
4806 for 'address + offset'.
4807 (microblaze_expand_prologue): Add new function prologue call for
4809 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
4810 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
4811 table in case of TARGET_PIC_DATA_TEXT_REL.
4812 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
4813 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
4814 Add new macros 'UNSPEC_TEXT',
4815 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
4816 + exclude function calls from 'UNSPEC_PLT' in case of data text
4818 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4819 new target hook for generating address diff vector tables in case of
4821 * doc/tm.texi : Regenerate.
4822 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
4823 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
4824 of addr diff vector generation.
4825 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4826 target hook definition.
4827 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
4828 Add default function for generate_pic_addr_diff_vec -> flag_pic.
4829 * doc/invoke.texi (Add new pic option): Add new microblaze pic
4830 option for data text relative.
4832 2018-04-30 Richard Biener <rguenther@suse.de>
4834 * tree-chrec.h (evolution_function_is_constant_p): Remove
4836 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
4838 2018-04-30 Richard Biener <rguenther@suse.de>
4841 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
4843 2018-04-30 Richard Biener <rguenther@suse.de>
4845 PR tree-optimization/28364
4846 PR tree-optimization/85275
4847 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
4848 copying first exit test.
4850 2018-04-28 Mark Wielaard <mark@klomp.org>
4852 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
4854 (dwarf_AT): Handle DW_AT_addr_base.
4855 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
4857 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
4860 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
4861 (*ashl<dwi>3_doubleword_mask_1): Ditto.
4862 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
4863 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
4865 2018-04-28 Richard Biener <rguenther@suse.de>
4867 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
4868 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
4869 to reflect use. Only add interesting stmts.
4871 2018-04-27 Martin Jambor <mjambor@suse.cz>
4874 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
4875 the jump function allows for passing through aggregate values.
4877 2018-04-27 David Malcolm <dmalcolm@redhat.com>
4879 * input.h (in_system_header_at): Convert from macro to inline
4881 (from_macro_expansion_at): Likewise.
4882 (from_macro_definition_at): Likewise.
4884 2018-04-27 Jeff Law <law@redhat.com>
4886 * config.gcc: Mark tile* targets as deprecated/obsolete.
4888 2018-04-27 Richard Biener <rguenther@suse.de>
4890 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
4893 2018-04-27 Richard Biener <rguenther@suse.de>
4895 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
4897 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
4899 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
4900 with Yd constraint. Set "preferred_for_speed" attribute from
4901 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
4903 (*movdi_internal): Ditto.
4904 (movti_interunit splitters): Remove
4905 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
4906 (movdi_interunit splitters): Ditto.
4907 * config/i386/constraints.md (Ye): Remove.
4908 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
4910 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4913 * config/aarch64/constraints.md (Usg): Limit to 31.
4916 2018-04-27 Jakub Jelinek <jakub@redhat.com>
4918 PR tree-optimization/85529
4919 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
4920 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
4921 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
4922 zero extension or masking of the MSB bit.
4923 (optimize_range_tests): Add FIRST_BB argument, pass it through
4924 to optimize_range_tests_var_bound.
4925 (maybe_optimize_range_tests, reassociate_bb): Adjust
4926 optimize_range_tests callers.
4928 2018-04-26 Richard Biener <rguenther@suse.de>
4929 Jakub Jelinek <jakub@redhat.com>
4931 * cgraph.h (symbol_table): Just declare debug method here.
4932 * symtab.c (symbol_table::debug): Define.
4934 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
4936 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
4938 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
4940 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
4941 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
4942 (*movdi_internal): Substitute Yi and Yj constraint with x
4943 and Ym and Yn constraint with y constraint. Update "isa"
4944 attribute and set "preferred_for_speed" attribute from
4945 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
4946 (*movsi_internal): Ditto.
4947 (*movdf_internal): Ditto.
4948 (*movsf_internal): Ditto.
4949 (*zero_extendsidi2): Ditto.
4950 * config/i386/sse.md (vec_set<mode>_0): Ditto.
4951 (sse2_loadld): Ditto.
4952 (*vec_extract<ssevecmodelower>_0): Ditto.
4953 (*vec_extractv4si_0_zext_sse4): Ditto.
4954 (vec_concatv2di): Ditto.
4955 (*vec_dup<mode>): Ditto.
4956 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
4957 * config/i386/constraints.md (Yi): Remove.
4962 2018-04-26 Nathan Sidwell <nathan@acm.org>
4964 * dumpfile.c (dump_open): New.
4965 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
4966 (dump_finish): Detect stdio/stderr by value not name.
4968 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
4970 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
4972 2018-04-26 Tom de Vries <tom@codesourcery.com>
4975 * config/nvptx/nvptx.c (verify_neutering_jumps)
4976 (verify_neutering_labels): New function
4977 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
4979 2018-04-26 Tom de Vries <tom@codesourcery.com>
4982 * config/nvptx/nvptx.c (needs_neutering_p): New function.
4983 (nvptx_single): Use needs_neutering_p to skip over insns that do not
4986 2018-04-26 Richard Biener <rguenther@suse.de>
4987 Tom de Vries <tom@codesourcery.com>
4990 * lto-streamer-out.c (output_function): Fixup loops if required to match
4991 discovery done in the reader.
4993 2018-04-26 Richard Biener <rguenther@suse.de>
4995 PR tree-optimization/85116
4996 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
4997 have a loop exit from the single latch predecessor. Remove
4998 case of header with just condition.
4999 (ch_base::copy_headers): Exclude infinite loops from any
5001 (pass_ch::execute): Record exits.
5003 2018-04-26 Richard Biener <rguenther@suse.de>
5005 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5006 prologue cost vector and pass it to vect_get_load_cost.
5007 (vect_get_peeling_costs_all_drs): Likewise.
5008 (vect_peeling_hash_get_lowest_cost): Likewise.
5009 (vect_enhance_data_refs_alignment): Likewise.
5011 2018-04-26 Richard Biener <rguenther@suse.de>
5014 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
5015 checking of integer<->pointer conversions.
5016 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
5017 sign-/zero-extending pointer types.
5018 (expand_omp_for_static_chunk): Likewise.
5020 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
5021 Jean Lee <xiaoyur347@gmail.com>
5023 * config/mips/mips.c (mips_asan_shadow_offset): New function.
5024 (TARGET_ASAN_SHADOW_OFFSET): Define.
5025 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
5026 true for -fsanitize=address.
5028 2018-04-25 Mark Wielaard <mark@klomp.org>
5030 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
5033 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5035 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
5036 than "alu", remove explicit "memory" and "imm_disp" attributes.
5037 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
5040 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
5041 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
5042 gen_lowpart_no_emit.
5044 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
5047 * config/i386/i386.c (ix86_expand_builtin): Change memory
5048 operand to XI, extend p0 to Pmode.
5049 * config/i386/i386.md: Change unspec volatile and operand
5050 1 mode to XI, change operand 0 mode to P.
5052 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5054 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
5055 GET_MODE_MASK before any checking.
5056 (nds32_can_use_bset_p): Likewise.
5057 (nds32_can_use_btgl_p): Likewise.
5059 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5061 * config/nds32/nds32-doubleword.md: New define_split pattern for
5062 illegal register number.
5064 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5066 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
5068 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5070 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
5072 2018-04-25 Richard Biener <rguenther@suse.de>
5074 * lto-streamer.h (LTO_major_version): Bump to 8.
5076 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5078 * BASE-VER: Set to 9.0.0.
5080 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
5082 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
5083 in __abskf2 and __powikf2.
5085 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5088 * config/aarch64/constraints.md (Usg, Usj): New constraints.
5089 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5090 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5091 Use the above on operand 2. Reindent.
5092 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5094 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5097 * common/config/i386/i386-common.c (ix86_handle_option): Don't
5099 * config/i386/i386.opt (mcet): Removed.
5100 * doc/install.texi: Remove -mcet documentation.
5101 * doc/invoke.texi: Likewise.
5103 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5106 * doc/install.texi: Remove -mcet from bootstrap-cet.
5108 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5111 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5112 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5116 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5117 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5118 containing a CONST_VECTOR.
5120 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
5122 * doc/install.texi: Update newlib dependency for nvptx.
5124 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5127 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5128 instead of INTVAL when shifting x left.
5130 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
5132 PR tree-optimization/85478
5133 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5134 vect_grouped_store_supported for single element vectors.
5136 2018-04-24 Richard Biener <rguenther@suse.de>
5139 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5140 load cost increase to the case of non-constant step.
5142 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5145 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5146 destination if any_malformed_asm.
5148 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
5151 * expr.c (store_field): In the bitfield case, if the value comes from
5152 a function call and is returned in registers by means of a PARALLEL,
5153 do not change the mode of the temporary unless BLKmode and VOIDmode.
5155 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
5157 PR rtl-optimization/85423
5158 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5159 dependencies to debug insns when the previous insn is non-debug.
5161 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5163 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5164 enums into a single definition.
5165 (fls): Fix predicates and printing.
5168 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5170 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5171 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5172 and short u6 immediate.
5173 (check_if_valid_sleep_operand): Remove.
5174 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5176 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5178 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5179 flag_always_save_lp condition.
5180 * config/nds32/nds32.opt (malways-save-lp): New option.
5182 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5184 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5185 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5186 * config/nds32/nds32.h
5187 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5188 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5190 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5192 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5193 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5195 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5196 Chung-Ju Wu <jasonwucj@gmail.com>
5198 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5200 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5201 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5203 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5205 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5207 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5209 * config/nds32/nds32-protos.h (nds32_data_alignment,
5210 nds32_local_alignment): Declare.
5211 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5212 nds32_local_alignment): New functions.
5213 (TARGET_CONSTANT_ALIGNMENT): Define.
5214 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5216 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5218 * config/nds32/nds32.c
5219 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5220 (TARGET_MODES_TIEABLE_P): Likewise.
5222 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5224 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5227 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
5228 Chung-Ju Wu <jasonwucj@gmail.com>
5230 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5231 for unaligned access.
5232 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5234 * config/nds32/nds32-intrinsic.md: Likewise.
5235 * config/nds32/nds32_intrinsic.h: Likewise.
5236 * config/nds32/nds32.h (nds32_builtins): Likewise.
5237 * config/nds32/nds32.opt (munaligned-access): New option.
5238 * config/nds32/nds32.c (nds32_asm_file_start): Display
5239 flag_unaligned_access status.
5241 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5243 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5244 -mno-relax is present.
5245 * config/riscv/linux.h (LINK_SPEC): Ditto.
5247 2018-04-20 Martin Sebor <msebor@redhat.com>
5250 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
5252 (gimple_fold_builtin_stxcpy_chk): Same.
5253 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
5255 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
5258 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
5259 __powikf2 when long double is IEEE 128-bit.
5261 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5263 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
5264 step to make sure stack always aligned.
5266 2018-04-20 Carl Love <cel@us.ibm.com>
5269 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
5270 size check for arg0.
5272 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
5273 Tom de Vries <tom@codesourcery.com>
5276 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
5277 Emit insns for calls too.
5278 (nvptx_find_par): Always look for worker-level predecessor insn.
5279 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
5281 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
5282 (nvptx_process_pars): Propagate frames for calls.
5284 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
5287 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
5289 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
5290 (ix86_handle_option): Don't handle OPT_mibt.
5291 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
5293 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
5295 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
5296 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
5297 (ix86_target_macros): Define __CET__ with flag_cf_protection
5298 for -fcf-protection.
5299 * config/i386/i386.c (isa2_opts): Remove -mibt.
5300 * config/i386/i386.h (TARGET_IBT): Removed.
5301 (TARGET_IBT_P): Likewise.
5302 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
5303 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
5304 * config/i386/i386.opt (mcet): Update help message.
5307 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
5308 -mcet as an alias for -mshstk.
5310 2018-04-20 Richard Biener <rguenther@suse.de>
5313 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
5314 complexity by forcing a single use of the multiply operand.
5316 2018-04-20 Martin Jambor <mjambor@suse.cz>
5319 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
5320 recursion dependency to only apply to non-clones.
5322 2018-04-20 Martin Jambor <mjambor@suse.cz>
5325 * ipa-cp.c (create_specialized_node): Check that clones of
5326 self-recursive edges exist during IPA-CP.
5328 2018-04-19 Toon Moene <toon@moene.org>
5330 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
5333 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5335 PR tree-optimization/85467
5336 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
5337 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
5338 VECTOR_CST element to type.
5340 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5343 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
5344 * config/i386/i386.md (builtin_setjmp_setup): Removed.
5345 (builtin_longjmp): Likewise.
5346 (save_stack_nonlocal): New pattern.
5347 (restore_stack_nonlocal): Likewise.
5349 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5352 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5353 Replace ASM_OUTPUT_LABEL with fprintf.
5355 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5358 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5359 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
5360 * config/i386/i386-c.c (ix86_target_macros_internal): Also
5361 define __IBT__ and __SHSTK__ for -fcf-protection.
5362 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
5364 (ix86_trampoline_init): Likewise.
5365 (x86_output_mi_thunk): Likewise.
5366 (ix86_notrack_prefixed_insn_p): Likewise.
5367 (ix86_option_override_internal): Don't disallow -fcf-protection.
5368 * config/i386/i386.md (rdssp<mode>): Also enable for
5370 (incssp<mode>): Likewise.
5371 (nop_endbr): Likewise.
5372 * config/i386/i386.opt (mcet): Change help message to built-in
5376 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
5377 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
5378 enable CET built-in functions.
5380 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
5382 * common/config/i386/i386-common.c
5383 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
5384 OPTION_MASK_ISA_MOVDIRI_UNSET,
5385 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
5386 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
5387 * config.gcc (movdirintrin.h): New header.
5388 * config/i386/cpuid.h (bit_MOVDIRI,
5389 bit_MOVDIR64B): New bits.
5390 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
5392 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
5393 (VOID, PVOID, PCVOID)): New function types.
5394 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
5395 __builtin_ia32_directstoreu_u64,
5396 __builtin_ia32_movdir64b): New builtins.
5397 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
5398 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
5400 (ix86_valid_target_attribute_inner_p): Ditto.
5401 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
5402 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
5403 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
5404 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
5405 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
5406 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
5407 (movdiri<mode>, movdir64b_<mode>): New.
5408 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
5409 * config/i386/immintrin.h: Include movdirintrin.h.
5410 * config/i386/movdirintrin.h: New file.
5411 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
5413 2018-04-19 Richard Biener <rguenther@suse.de>
5416 * cfg.c (clear_bb_flags): When loop state says we have
5417 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
5419 2018-04-19 Richard Biener <rguenther@suse.de>
5421 PR tree-optimization/84737
5422 * tree-vect-data-refs.c (vect_copy_ref_info): New function
5423 copying restrict info.
5424 (vect_setup_realignment): Use it.
5425 * tree-vectorizer.h (vect_copy_ref_info): Declare.
5426 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
5427 the first DR to all generated stores.
5428 (vectorizable_load): Likewise for loads.
5430 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5432 PR tree-optimization/85446
5433 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
5434 the integral and pointer types to have the same precision.
5436 * doc/install.texi: Document --disable-cet being the default and
5439 2018-04-18 Martin Liska <mliska@suse.cz>
5441 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
5444 2018-04-18 Martin Liska <mliska@suse.cz>
5447 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5450 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5451 arguments if they are comparable.
5453 2018-04-18 Martin Liska <mliska@suse.cz>
5456 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5459 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5462 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
5465 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
5466 ENDBR after calling __morestack.
5468 2018-04-18 David Malcolm <dmalcolm@redhat.com>
5471 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
5472 by using gcc_base_ver to generate a gcc_driver_version, and use
5473 it when generating GCC_DRIVER_NAME.
5474 * configure: Regenerate.
5476 2018-04-18 Jakub Jelinek <jakub@redhat.com>
5479 * config.gcc: Obsolete powerpc*-*-*spe*.
5481 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5484 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
5485 (stabstr_D): Change type of unum from unsigned int to
5486 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
5489 2018-04-17 Jim Wilson <jimw@sifive.com>
5492 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
5493 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
5494 Set arg_pointer_offset after using pretend_args_size.
5496 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5498 PR rtl-optimization/85431
5499 * dse.c (record_store): Ignore zero width stores.
5502 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
5503 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
5504 __builtin_stack_restore rather than after it.
5505 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
5506 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
5507 argument instead of virtual_dynamic_stack_rtx.
5509 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
5511 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
5513 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5514 Add note to error message to explain internal mapping of overloaded
5515 built-in function name to non-overloaded built-in function name.
5516 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
5519 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5522 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
5523 where the inputs overlap with the output.
5525 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5527 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
5528 (=v, v) alternative and explicit "memory" attribute.
5529 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
5530 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5532 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
5533 "sselog1" type instead of "sselog".
5534 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
5535 "sselog". Remove explicit "memory" attribute.
5536 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
5537 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5539 (vec_extract_hi_v32hi): Merge all alternatives into one, use
5540 "sselog1" type instead of "sselog". Remove explicit "memory"
5542 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
5543 use "sselog1" type instead of "sselog". Remove explicit "memory"
5545 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
5546 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5548 (vec_extract_hi_v64qi): Merge all alternatives into one, use
5549 "sselog1" type instead of "sselog". Remove explicit "memory"
5551 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
5552 use "sselog1" type instead of "sselog". Remove explicit "memory"
5556 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
5559 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
5562 2018-04-17 Martin Jambor <mjambor@suse.cz>
5565 * ipa-cp.c (create_specialized_node): Call
5566 expand_all_artificial_thunks if necessary.
5568 2018-04-17 Martin Liska <mliska@suse.cz>
5571 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
5572 in message, remote space in between '_G' and '('.
5574 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5577 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
5578 avx512f_vmcmp<mode>3<round_saeonly_name>,
5579 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
5580 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
5581 avx512f_rndscale<mode><round_saeonly_name>,
5582 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
5583 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
5584 Use %<iptr>2 instead of %2 for -masm=intel.
5585 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
5586 avx512f_vcvttss2usi<round_saeonly_name>,
5587 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
5589 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
5590 avx512f_vcvttsd2usi<round_saeonly_name>,
5591 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
5592 Use %q1 instead of %1 for -masm=intel.
5593 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
5594 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
5595 of %3 for -masm=intel.
5596 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
5598 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
5600 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
5602 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
5603 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
5605 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
5607 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
5608 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
5609 %g1 and one with %0 and %1.
5610 (avx512er_vmrcp28<mode><round_saeonly_name>,
5611 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
5613 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
5614 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
5615 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
5616 of %0 and %{%4%} for -masm=intel.
5617 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
5618 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
5619 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
5620 order of %0 and %{%5%}%{z%} for -masm=intel.
5622 2018-04-17 Jan Hubicka <jh@suse.cz>
5625 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
5627 2018-04-17 Martin Liska <mliska@suse.cz>
5630 * multiple_target.c (create_dispatcher_calls): Set apostrophes
5631 for target_clone error message. Make default implementation
5632 clone to be a local declaration.
5633 (separate_attrs): Add new argument and check for an empty
5635 (expand_target_clones): Handle it.
5636 (ipa_target_clone): Make redirection just for target_clones
5639 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
5640 Tom de Vries <tom@codesourcery.com>
5643 * omp-expand.c (expand_oacc_for): Add dummy false branch for
5644 tiled basic blocks without omp continue statements.
5646 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
5649 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
5650 vec_extract expression as having side effects to make sure it gets
5653 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
5656 * config/i386/i386.c (get_builtin_code_for_version): Check
5659 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
5662 * gcc/config.gcc: Support "skylake".
5663 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
5665 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
5666 (processor_target_table): Add "skylake".
5667 (ix86_option_override_internal): Add "skylake".
5668 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
5669 PROCESSOR_CANNONLAKE.
5670 (get_builtin_code_for_version): Fix priority for
5671 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
5672 PROCESSOR_SKYLAKE-AVX512.
5673 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
5674 (processor_type): Add PROCESSOR_SKYLAKE.
5676 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
5677 Jason Merrill <jason@redhat.com>
5680 * convert.c (convert_to_integer_1): Use direct recursion for
5681 enumeral types and types with a precision less than the number
5682 of bits in their mode.
5684 2018-04-16 Julia Koval <julia.koval@intel.com>
5687 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
5688 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
5690 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
5693 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
5694 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
5695 and -mno-direct-move.
5697 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
5700 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
5701 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
5702 Ensure negative shifts result in {0}.
5704 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
5706 PR rtl-optimization/79916
5707 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
5708 regs (if any) to define how to gnerate SD moves when LRA is in
5711 2018-04-13 Jakub Jelinek <jakub@redhat.com>
5713 PR rtl-optimization/85393
5714 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
5715 * except.c (expand_dw2_landing_pad_for_region): Make static.
5716 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
5717 a label and unconditional jump to old_bb, rather than
5718 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
5721 PR rtl-optimization/85376
5722 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
5723 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
5724 instead of a specific value.
5726 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5727 Bin Cheng <bin.cheng@arm.com>
5729 PR tree-optimization/82965
5730 PR tree-optimization/83991
5731 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
5732 by_profile_only parameter.
5733 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
5734 information if the loop was predicted to iterate too many times.
5735 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
5737 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5740 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
5743 2018-04-13 Martin Liska <mliska@suse.cz>
5744 Jakub Jelinek <jakub@redhat.com>
5747 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
5748 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
5749 * builtins.c (expand_builtin_memory_copy_args): Use
5750 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
5751 handle dest_addr == pc_rtx.
5753 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
5756 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
5757 asked to not generate direct moves.
5758 (fix_trunc<mode>si2_stfiwx): Similar.
5759 (fix_trunc<mode>si2_internal): Similar.
5761 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5764 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
5765 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
5766 lookup if dest in some wider mode is known to be const0_rtx and
5767 if so, record permanent equivalence for it to be ZERO_EXTEND of
5768 the narrower mode destination.
5770 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5772 * lto-streamer-out.c (output_function): Revert 259346.
5773 * omp-expand.c (expand_oacc_for): Likewise.
5775 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
5777 PR rtl-optimization/85354
5778 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
5779 * sel-sched.c (sel_global_init): ... here.
5781 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
5784 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
5785 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
5786 mode for PE-COFF targets.
5787 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
5788 (i386_pe_asm_lto_end): Likewise.
5789 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
5790 (TARGET_ASM_LTO_END): Likewise.
5791 * config/i386/winnt.c (saved_debug_info_level): New static variable.
5792 (i386_pe_asm_lto_start): New function.
5793 (i386_pe_asm_lto_end): Likewise.
5795 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5796 Richard Biener <rguenther@suse.de>
5799 * lto-streamer-out.c (output_function): Fix CFG loop state before
5801 * omp-expand.c (expand_oacc_for): Handle calls to internal
5802 functions like regular functions.
5804 2018-04-12 Richard Biener <rguenther@suse.de>
5807 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
5808 for the early LTO debug to properly generate references to it
5809 during DIE emission. Do not re-use that for the skeleton for
5811 (dwarf2out_early_finish): Likewise.
5813 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5816 * config/i386/sse.md
5817 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
5818 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
5819 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
5820 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
5821 and output is a reg, avoid creating invalid lowpart subreg, but
5822 instead split into a 512-bit move. Don't split if not AVX512VL,
5823 input is xmm16+ reg and output is a mem.
5824 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
5825 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
5826 xmm16+ reg and output is a mem.
5828 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5830 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
5831 also for flag_dwarf2_cfi_asm.
5833 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5835 PR rtl-optimization/85342
5836 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
5837 a bool scalar var inside of the loop instead. Don't try to update
5838 recog_data.operand after failed apply_change_group.
5840 2018-04-12 Tom de Vries <tom@codesourcery.com>
5843 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
5844 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
5845 array with flexible array member as array without given dimension.
5846 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
5847 argument for undefined param to true.
5849 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
5852 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5853 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
5854 from PowerPC section.
5855 * config/rs6000/sysv4.opt (mcall-): Improve help text.
5856 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
5857 help text that is too long.
5858 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
5859 help text that is too long.
5860 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
5861 help text that is too long.
5863 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
5865 * config/alpha/alpha.md (stack_probe_internal): Rename
5866 from "probe_stack". Update all callers.
5868 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5870 PR rtl-optimization/84566
5871 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
5872 sched_macro_fuse_insns.
5874 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5877 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
5878 (compute_block_dependences): ... from here.
5880 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5882 PR tree-optimization/85331
5883 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
5884 from int to HOST_WIDE_INT.
5886 2018-04-11 Martin Jambor <mjambor@suse.cz>
5889 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
5890 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
5891 not the same as the source val.
5892 (cgraph_edge_brings_value_p): New parameter.
5893 (gather_edges_for_value): Pass destination value to
5894 cgraph_edge_brings_value_p.
5895 (perhaps_add_new_callers): Likewise.
5896 (get_info_about_necessary_edges): Likewise and exclude values brought
5897 only by self-recursive edges.
5898 (create_specialized_node): Redirect only clones of self-calling edges.
5899 (+self_recursive_pass_through_p): New function.
5900 (find_more_scalar_values_for_callers_subset): Use it.
5901 (find_aggregate_values_for_callers_subset): Likewise.
5902 (known_aggs_to_agg_replacement_list): Removed.
5903 (decide_whether_version_node): Re-calculate known constants for all
5904 remaining context clones.
5906 2018-04-11 Richard Biener <rguenther@suse.de>
5909 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
5910 from early DWARF output.
5911 (dwarf2out_early_finish): Output line info unconditionally into
5912 early DWARF and add reference to it.
5914 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5917 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
5918 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
5919 other than V2DFmode using iptr mode attribute.
5920 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
5922 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5924 PR rtl-optimization/84659
5925 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
5927 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5930 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
5932 (output_loc_list): Pass address of a dummy size variable even in the
5933 locview handling loop.
5934 (index_location_lists): Add comment on why skip_loc_list_entry can't
5937 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
5940 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
5943 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
5946 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5947 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
5948 and -mstring-compare-inline-limit.
5950 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5953 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
5954 for stack clash protection in a register whenever we need it to be in
5957 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5959 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
5960 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
5962 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5965 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
5967 (mlong-double-): Ditto.
5968 * config/rs6000/sysv4.opt (msdata=): Ditto.
5969 (mtls-size=): Ditto.
5971 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5973 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5974 erroneous entries for
5975 "vector int vec_ldl (int, long int *)", and
5976 "vector unsigned int vec_ldl (int, unsigned long int *)".
5977 Add comments and entries for
5978 "vector bool char vec_ldl (int, bool char *)",
5979 "vector bool short vec_ldl (int, bool short *)",
5980 "vector bool int vec_ldl (int, bool int *)",
5981 "vector bool long long vec_ldl (int, bool long long *)",
5982 "vector pixel vec_ldl (int, pixel *)",
5983 "vector long long vec_ldl (int, long long *)",
5984 "vector unsigned long long vec_ldl (int, unsigned long long *)".
5985 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
5986 type tree bool_long_long_type_node and correct definition of
5987 bool_V2DI_type_node to make reference to this new type tree.
5988 (rs6000_mangle_type): Replace erroneous reference to
5989 bool_long_type_node with bool_long_long_type_node.
5990 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
5991 comments to emphasize sign distinctions for char and int types and
5992 replace RS6000_BTI_bool_long constant with
5993 RS6000_BTI_bool_long_long constant. Also add comment to restrict
5994 use of RS6000_BTI_pixel.
5995 (bool_long_type_node): Remove this macro definition.
5996 (bool_long_long_type_node): New macro definition
5998 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6000 PR rtl-optimization/85300
6001 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
6002 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
6003 simplify_unary_operation fails.
6005 2018-04-10 Martin Liska <mliska@suse.cz>
6007 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
6008 cgraph_edge and ipa_ref.
6010 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6014 * config/i386/sse.md
6015 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
6016 computation of the VEC_MERGE selector from mask.
6017 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
6018 Fix decoding of the VEC_MERGE selector into mask.
6020 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6022 PR tree-optimization/85286
6023 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
6025 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6027 * final.c (final_1): Set insn_last_address as well as
6028 insn_current_address.
6030 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6033 * explow.c (emit_stack_probe): Call validize_mem on memory location
6034 before passing it to gen_probe_stack. Create address operand and
6035 legitimize it for the probe_stack_address case.
6037 2018-04-09 Jan Hubicka <jh@suse.cz>
6040 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
6041 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
6042 * tree.c (free_lang_data_in_type): Fix handling of binfos;
6044 (free_lang_data): Rebuild type inheritance graph.
6046 2018-04-09 Martin Sebor <msebor@redhat.com>
6048 * invoke.texi (-finline-small-functions): Mention other optimization
6050 (-findirect-inlining, -fpartial-inlining): Same.
6051 (-finline-functions-called-once): Same.
6052 (-freorder-blocks-and-partition): Same.
6054 2018-04-09 Jan Hubicka <jh@suse.cz>
6057 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
6058 jumps; choose last target that matches the criteria (i.e.
6059 no partition changes for non-crossing jumps).
6060 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
6061 support for redirecting crossing jumps to non-crossing.
6063 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
6065 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
6066 also for naked functions.
6068 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
6070 * config/arc/arc.md (add_shift): New pattern.
6071 (add_shift2): Likewise.
6072 (sub_shift): Likewise.
6073 (sub_shift_cmp0_noout): Likewise.
6074 (compare_si_ashiftsi): Likewise.
6075 (xbfu_cmp0_noout): New combine pattern.
6076 (xbfu_cmp0"): Likewise.
6077 (movsi_set_cc_insn): Place the predicable variant first.
6078 (commutative_binary_cmp0_noout): Remove clobber.
6079 (commutative_binary_cmp0): New pattern.
6080 (noncommutative_binary_cmp0): Likewise.
6081 (noncommutative_binary_cmp0_noout): Likewise.
6082 (noncommutative_binary_comparison_result_used): Removed.
6083 (rsub_cmp0): New pattern.
6084 (rsub_cmp0_noout): Likewise.
6085 (extzvsi): Changed, keep only meaningful variants.
6086 (SQH, SEZ): New iterators.
6087 (SQH_postfix): New mode attribute.
6088 (SEZ_prefix): New code attribute.
6089 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6090 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6091 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6093 (noncommutative_operator): Check the availability of barrel
6096 2018-04-09 Richard Biener <rguenther@suse.de>
6098 PR tree-optimization/85284
6099 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6100 Only use the niter constraining form of simple_iv when the exit
6103 2018-04-09 Tom de Vries <tom@codesourcery.com>
6106 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6107 (define_expand "*memory_barrier"): New define_expand.
6108 (define_insn "memory_barrier"): New insn.
6110 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6112 PR rtl-optimization/80463
6113 PR rtl-optimization/83972
6114 PR rtl-optimization/83480
6116 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6117 correct producer for the insn.
6118 (tidy_control_flow): Fixup seqnos in case of debug insns.
6120 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6122 PR rtl-optimization/83913
6124 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6125 different sched-times when merging exprs.
6127 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6129 PR rtl-optimization/83962
6131 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6132 tidy_fallthru_edge and tidy_control_flow.
6134 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6136 PR rtl-optimization/83530
6138 * sel-sched.c (force_next_insn): New global variable.
6139 (remove_insn_for_debug): When force_next_insn is true, also leave only
6140 next insn in the ready list.
6141 (sel_sched_region): When the region wasn't scheduled, make another pass
6142 over it with force_next_insn set to 1.
6144 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6146 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6148 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6149 for interrupt control.
6150 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6151 functions for interrupt control.
6152 * config/nds32/nds32-intrinsic.md: Likewise.
6153 * config/nds32/nds32_intrinsic.h: Likewise.
6154 * config/nds32/nds32.h (nds32_builtins): Likewise.
6156 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6158 * config/nds32/nds32.c (nds32_init_machine_status,
6159 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6160 strict_aligned_p field.
6161 (nds32_expand_to_rtl_hook): New function.
6162 (TARGET_EXPAND_TO_RTL_HOOK): Define.
6163 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6165 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6166 Chung-Ju Wu <jasonwucj@gmail.com>
6168 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6169 * config/nds32/nds32-n7.md: New file.
6170 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6171 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6173 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6174 * config/nds32/nds32.md (pipeline_model): Add n7.
6175 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6176 * config/nds32/pipelines.md: Include n7 settings.
6178 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6179 Chung-Ju Wu <jasonwucj@gmail.com>
6181 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6182 * config/nds32/nds32-e8.md: New file.
6183 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6184 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6186 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6187 * config/nds32/nds32.md (pipeline_model): Add e8.
6188 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6189 * config/nds32/pipelines.md: Include e8 settings.
6191 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6192 Chung-Ju Wu <jasonwucj@gmail.com>
6194 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6195 * config/nds32/nds32-n8.md: New file.
6196 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6197 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6199 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6200 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6201 * config/nds32/nds32.md (pipeline_model): Add n8.
6202 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6203 * config/nds32/pipelines.md: Include n8 settings.
6205 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6206 Chung-Ju Wu <jasonwucj@gmail.com>
6208 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6209 * config/nds32/nds32-n9-2r1w.md: New file.
6210 * config/nds32/nds32-n9-3r2w.md: New file.
6211 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6212 nds32_register_ports): New or modify for cpu n9.
6213 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6215 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6216 * config/nds32/nds32-utils.c: New file.
6217 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6218 TARGET_MUL_SLOW): Define.
6219 * config/nds32/nds32.md (pipeline_model): New attribute.
6220 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6221 New options that support cpu n9.
6222 * config/nds32/pipelines.md: Include n9 settings.
6223 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6225 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6227 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6228 information if necessary.
6229 (output_cond_branch_compare_zero): Likewise.
6230 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6231 (nds32_target_alignment): Refine for alignment.
6232 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6233 (FUNCTION_BOUNDARY): Modify.
6234 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6236 * config/nds32/nds32.opt (malways-align, malign-functions): New.
6238 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6240 * config/nds32/constants.md (unspec_volatile_element): Add values for
6241 TLB operation and data prefetch.
6242 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6243 functions for TLB operation and data prefetch.
6244 * config/nds32/nds32-intrinsic.md: Likewise.
6245 * config/nds32/nds32_intrinsic.h: Likewise.
6246 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
6247 (nds32_print_operand): Likewise.
6248 * config/nds32/nds32.h (nds32_builtins): Likewise.
6250 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
6251 Andrew Pinski <pinsika@gcc.gnu.org>
6254 * match.pd: Use constant_boolean_node of correct type instead of
6255 boolean_true_node or boolean_false_node for simplifying
6256 pointer comparisons to zero.
6258 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6260 PR tree-optimization/80021
6261 * tree.c (verify_type_variant): Make error call in verify_variant_match
6262 translatable and remove final full stop.
6264 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6266 * config/nds32/constants.md (unspec_volatile_element): Add
6267 UNSPEC_VOLATILE_EH_RETURN.
6268 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
6269 nds32_output_stack_pop): Support dwarf exception handling process.
6270 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
6271 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
6272 exception handling process.
6273 (nds32_compute_stack_frame): Likewise.
6274 (nds32_return_addr_rtx): Likewise.
6275 (nds32_initial_elimination_offset): Likewise.
6276 (nds32_expand_prologue): Likewise.
6277 (nds32_expand_epilogue): Likewise.
6278 (nds32_dynamic_chain_address): New function.
6279 * config/nds32/nds32.h (machine_function): Add fields for dwarf
6281 (DYNAMIC_CHAIN_ADDRESS): Define.
6282 (EH_RETURN_DATA_REGNO): Define.
6283 (EH_RETURN_STACKADJ_RTX): Define.
6284 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
6285 patterns for dwarf exception handling.
6287 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6289 * config/nds32/nds32.h: Clean up obsolete macros.
6291 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6293 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6294 Add enum values for particular instructions.
6295 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6296 particular intrinsic functions.
6297 * config/nds32/nds32-intrinsic.md: Likewise.
6298 * config/nds32/nds32_intrinsic.h: Likewise.
6299 * config/nds32/nds32.h (nds32_builtins): Likewise.
6300 * config/nds32/nds32.md (type): Add pbsad and pbsada.
6301 (btst, ave): New patterns for particular instructions.
6303 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6305 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6306 Add enum values for atomic load/store and memory sync.
6307 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
6309 * config/nds32/nds32-intrinsic.md: Likewise.
6310 * config/nds32/nds32_intrinsic.h: Likewise.
6311 * config/nds32/nds32.h (nds32_builtins): Likewise.
6313 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6315 PR tree-optimization/85257
6316 * fold-const.c (native_encode_vector): If not all elts could fit
6317 and off is -1, return 0 rather than offset.
6318 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
6319 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
6320 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
6321 adjust buffer in native_interpret_expr call.
6323 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6325 * config/nds32/constants.md (unspec_volatile_element): Add cache
6326 control enum values.
6327 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
6328 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
6329 * config/nds32/nds32.c (nds32_cctl_names): New.
6330 (nds32_print_operand): Handle cache control register names.
6331 * config/nds32/nds32.h (nds32_builtins): New enum values.
6332 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
6334 * config/nds32/nds32.md (type): Add mmu.
6335 * config/nds32/pipelines.md (simple_insn): Add mmu.
6337 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6339 * config/nds32/nds32.md (type): Remove call.
6340 * config/nds32/pipelines.md (simple_insn): Likewise.
6342 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6344 * config/nds32/constants.md (unspec_volatile_element): Add
6345 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
6346 UNSPEC_VOLATILE_FMFCFG.
6347 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
6348 description for fmfcfg and fmfcsr.
6349 (bdesc_1arg): Add fmtcsr.
6350 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
6351 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
6352 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
6353 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
6354 unspec_fmfcfg): New patterns.
6355 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
6356 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
6357 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
6358 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
6359 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
6360 __nds32__fmfcfg): Define.
6362 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6364 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
6365 intrinsic register names.
6366 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
6367 intrinsic register enum values and macros.
6369 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6371 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
6372 for load/store addressing form.
6373 (nds32_print_operand_address): Likewise.
6375 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
6378 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
6379 based on LABEL_REF. Remove useless assertion.
6380 (pic_address_needs_scratch): Fix formatting.
6381 (sparc_legitimize_pic_address): Minor tweaks.
6382 (sparc_delegitimize_address): Adjust assertion accordingly.
6383 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
6384 into symbolic_operand.
6385 (movsi_high_pic_label_ref): Likewise.
6386 (movsi_lo_sum_pic_label_ref): Likewise.
6387 (movdi_pic_label_ref): Likewise.
6388 (movdi_high_pic_label_ref): Likewise.
6389 (movdi_lo_sum_pic_label_ref): Likewise.
6391 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
6393 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
6394 custom LIB_SPEC setup.
6396 2018-04-06 Ruslan Bukin <br@bsdpad.com>
6397 Kito Cheng <kito.cheng@gmail.com>
6399 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
6400 * config/riscv/freebsd.h: New.
6402 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6404 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
6405 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
6408 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6409 Kito Cheng <kito.cheng@gmail.com>
6411 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
6412 nds32_output_call, nds32_symbol_binds_local_p): New functions.
6413 * config/nds32/nds32-protos.h (nds32_output_call,
6414 nds32_output_return): Declare.
6415 * config/nds32/nds32.md: Refine all the call and return patterns.
6417 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6420 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
6421 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
6423 PR rtl-optimization/84872
6424 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
6425 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
6428 2018-04-06 Tamar Christina <tamar.christina@arm.com>
6430 * expr.c (copy_blkmode_to_reg): Revert 254862.
6431 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
6433 2018-04-06 Richard Biener <rguenther@suse.de>
6436 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
6437 after seeing a component reference with an adjacent field. Treat
6438 refs to arrays at struct end of external decls similar to
6439 refs to unconstrained commons.
6441 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6444 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
6445 look through SAVE_EXPRs with non-side-effects argument. Adjust
6447 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
6450 2018-04-06 Richard Biener <rguenther@suse.de>
6453 * alias.c (find_base_term): New wrapper around find_base_term
6454 unwinding CSELIB_VAL_PTR changes.
6455 (find_base_term): Do not restore CSELIB_VAL_PTR during the
6458 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6460 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
6462 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
6463 constant definitions.
6464 ("nop"): lr 0,0 -> nopr r0
6465 ("nop_lr0", "nop_lr1"): New insn definitions.
6467 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6469 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
6470 NDS32_V3PUSH_AVAILABLE_P macro.
6472 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
6473 Chung-Ju Wu <jasonwucj@gmail.com>
6475 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
6476 (nds32*-*-*): Add float and fpu_config into supported_defaults.
6477 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
6478 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
6479 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
6480 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
6481 * config/nds32/constraints.md: New constraints and checking for hard
6482 float configuration.
6483 * config/nds32/iterators.md: New mode iterator and attribute for hard
6484 float configuration.
6485 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
6487 * config/nds32/nds32-fpu.md: New file.
6488 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
6489 deal with hard float code generation.
6490 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
6492 (abi_type, float_reg_number): New enum type.
6493 * config/nds32/nds32-predicates.c: New predicates for hard float.
6494 * config/nds32/nds32-protos.h: Declare functions for hard float.
6495 * config/nds32/nds32.c: Implementation for hard float configuration.
6496 * config/nds32/nds32.h: Definitions for hard float configuration.
6497 * config/nds32/nds32.md: Include hard float machine description and
6498 modify patterns for hard float configuration.
6499 * config/nds32/nds32.opt: New options for hard float configuration.
6500 * config/nds32/predicates.md: New predicates for hard float
6503 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6505 * common/config/nds32/nds32-common.c
6506 (nds32_option_optimization_table): Enable -mreleax-hint by default.
6508 2018-04-05 Jakub Jelinek <jakub@redhat.com>
6511 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
6512 CONSTRUCTOR_ELT (ctor, ...)->value.
6514 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
6517 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
6519 2018-04-05 Tom de Vries <tom@codesourcery.com>
6522 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
6525 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6526 Kito Cheng <kito.cheng@gmail.com>
6528 * config/nds32/constraints.md (U33): Fine-tune checking condition.
6529 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
6530 * config/nds32/nds32.h (nds32_16bit_address_type): Add
6531 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
6533 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6534 Kito Cheng <kito.cheng@gmail.com>
6536 * config/nds32/constraints.md (Ufe): New memory constraint.
6537 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
6538 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
6539 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
6541 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
6542 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
6544 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6546 * config/nds32/nds32.md: Use optimize_size in the condition for
6547 alu-shift instructions.
6549 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6551 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
6553 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6555 * config/nds32/nds32.md (negsi2): Refine pattern.
6557 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6558 Chung-Ju Wu <jasonwucj@gmail.com>
6560 * config/nds32/iterators.md (shift_rotate): New code iterator.
6561 (shift): New code attribute.
6562 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
6563 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
6564 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
6565 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
6566 bit-wise operations.
6567 (andsi3, *andsi3): Ditto.
6568 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
6569 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
6570 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
6571 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
6572 nds32_ior_operand, nds32_xor_operand): New predicates.
6574 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6576 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
6577 (addsi3, subsi3): ... this.
6579 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6581 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
6583 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6585 * config/nds32/nds32.md: Adjust indention.
6587 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6589 * config/nds32/nds32.md (feature): New attribute.
6591 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6593 * config/nds32/nds32.md (subtype): New attribute.
6595 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
6598 * config/arm/arm-builtins.c (arm_expand_builtin): Change
6599 expansion to perform a bitwise AND of the argument followed by a
6600 boolean negation of the result.
6602 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
6604 PR rtl-optimization/84878
6605 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
6606 the basic block. Assert the use reference is not artificial and that
6607 it has an associated insn.
6609 2018-04-04 Michael Matz <matz@suse.de>
6611 * builtins.c (compute_objsize): Pass correct operand
6612 to array_at_struct_end_p.
6614 2018-04-04 Richard Biener <rguenther@suse.de>
6617 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
6618 from contexts for DINFO_LEVEL_TERSE and below.
6620 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6622 * config/nds32/nds32-doubleword.md (move_<mode>): Require
6623 resiter_operand condition.
6624 * config/nds32/nds32.md (*move<mode>): Ditto.
6626 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6627 Monk Chiang <sh.chiang04@gmail.com>
6629 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
6631 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6633 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
6635 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6636 Kito Cheng <kito.cheng@gmail.com>
6638 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
6639 nds32_cond_code_str, output_cond_branch,
6640 output_cond_branch_compare_zero, nds32_expand_cbranch,
6641 nds32_expand_cstore, nds32_expand_movcc,
6642 nds32_output_cbranchsi4_equality_zero,
6643 nds32_output_cbranchsi4_equality_reg,
6644 nds32_output_cbranchsi4_equality_reg_or_const_int,
6645 nds32_output_cbranchsi4_greater_less_zero: New functions.
6646 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
6647 nds32_expand_cstore, nds32_expand_movcc,
6648 nds32_output_cbranchsi4_equality_zero,
6649 nds32_output_cbranchsi4_equality_reg,
6650 nds32_output_cbranchsi4_equality_reg_or_const_int,
6651 nds32_output_cbranchsi4_greater_less_zero): Declare.
6652 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
6653 nds32_rimm11s_operand): New predicates.
6654 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
6655 * config/nds32/nds32.md: Rewrite all the branch and conditional move
6658 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6660 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
6661 * config/nds32/nds32.md: Ditto.
6662 * config/nds32/pipelines.md: Ditto.
6664 2018-04-04 Richard Biener <rguenther@suse.de>
6666 PR tree-optimization/85168
6667 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
6668 propagating abnormals.
6670 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6672 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
6674 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6675 Kito Cheng <kito.cheng@gmail.com>
6677 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
6678 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
6679 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
6680 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
6681 * config/nds32/nds32.md (sibcall_internal): New.
6682 (sibcall_register): Remove.
6683 (sibcall_immediate): Remove.
6684 (sibcall_value_internal): New.
6685 (sibcall_value_register): Remove.
6686 (sibcall_value_immediate): Remove.
6687 * config/nds32/predicates.md (nds32_general_register_operand): New.
6688 (nds32_call_address_operand): New.
6690 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6692 PR rtl-optimization/85167
6693 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
6694 bb_defs if *split_p, instead preinitialize it to NULL.
6696 PR tree-optimization/85156
6697 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
6698 evaluating the argument multiple times.
6700 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
6702 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
6704 (_mm_cvtpd_ps): Likewise.
6705 (_mm_cvttpd_epi32): Likewise.
6706 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
6707 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
6708 vector, pixel, and bool following altivec.h include.
6710 2018-04-03 Martin Sebor <msebor@redhat.com>
6712 * doc/extend.texi (Common Function Attributes): Clarify.
6713 (const attribute): Likewise.
6714 (pure attribute): Likewise.
6716 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6719 * config/i386/i386.c (ix86_expand_vector_set): Use
6720 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
6722 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
6724 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
6725 instructions when changing rounding bits to preserve precision bits
6726 in the x87 control word.
6728 2018-04-03 Martin Liska <mliska@suse.cz>
6730 PR tree-optimization/82491
6731 * rtl.h (strip_offset_and_add): Replace += suboffset with
6732 poly_uint64 () + suboffset.
6734 2018-03-29 Martin Liska <mliska@suse.cz>
6735 Martin Jambor <mjambor@suse.cz>
6738 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
6739 param_type is not an integral or pointer type.
6741 2018-04-03 Richard Biener <rguenther@suse.de>
6743 * sese.h (recompute_all_dominators): Remove.
6745 2018-04-02 Martin Sebor <msebor@redhat.com>
6747 * doc/invoke.texi (-Wrestrict): Fix typos.
6749 2018-04-02 Jim Wilson <jimw@sifive.com>
6751 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
6752 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
6753 (<optab>di3, <optab>si3_extend): Likewise.
6754 (<optab>si3_mask, <optab>si3_mask_1): New.
6755 (<optab>di3_mask, <optab>di3_mask_1): New.
6756 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
6757 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
6758 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
6760 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
6762 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
6765 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
6767 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
6768 (nds32_canonicalize_comparison): New function.
6770 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6771 Kito Cheng <kito.cheng@gmail.com>
6772 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6774 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
6775 * config/nds32/constants.md (unspec_volatile_element): Add
6776 UNSPEC_VOLATILE_RELAX_GROUP.
6777 * config/nds32/nds32-relax-opt.c: New file.
6778 * config/nds32/nds32-predicates.c
6779 (nds32_symbol_load_store_p): New function.
6780 * config/nds32/nds32-protos.h
6781 (nds32_symbol_load_store_p): Declare function.
6782 (make_pass_nds32_relax_opt): Declare new rtl pass function.
6783 * config/nds32/nds32.c
6784 (nds32_register_pass): New function to register pass.
6785 (nds32_register_passes): New function to register passes.
6786 * config/nds32/nds32.md (relax_group): New pattern.
6787 * config/nds32/nds32.opt (mrelax-hint): New option.
6788 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
6790 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
6792 * config/nds32/t-nds32: Modify files dependency.
6794 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6796 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
6797 (PROFILE_HOOK): Define its implementation.
6799 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6801 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
6802 type and 32-bit size.
6804 2018-04-01 Jakub Jelinek <jakub@redhat.com>
6807 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
6808 (V_128_256): New mode iterator.
6809 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
6810 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
6811 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
6813 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
6814 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
6816 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
6819 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
6820 NaN inputs correctly.
6822 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
6825 * config/rs6000/vsx.md (??r): New mode attribute.
6826 (*vsx_mov<mode>_64bit): Use it.
6827 (*vsx_mov<mode>_32bit): Likewise.
6829 2018-03-30 Martin Sebor <msebor@redhat.com>
6831 PR tree-optimization/84818
6832 * builtins.c (check_access): Use warning_n.
6834 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6837 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
6839 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
6842 2018-03-30 Julia Koval <julia.koval@intel.com>
6845 * x86-tune.def (movx, partial_reg_dependency): Enable for
6848 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
6851 * lra-constraints.c (process_alt_operands): Move setting
6852 this_alternative_matches below.
6854 2018-03-29 Martin Liska <mliska@suse.cz>
6857 * doc/invoke.texi: Document how LTO works with debug info.
6858 Describe auto-load support of binutils. Mention 'x86-64'
6859 as valid option value of -march option.
6861 2018-03-29 Jakub Jelinek <jakub@redhat.com>
6863 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
6866 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
6867 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
6868 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
6871 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
6874 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
6875 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
6876 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
6877 for RS6000_BTM_POWERPC64.
6878 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
6879 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
6880 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
6885 2018-03-28 Carl Love <cel@us.ibm.com>
6888 2017-09-27 Carl Love <cel@us.ibm.com>
6890 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
6891 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
6892 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
6895 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6897 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
6898 instead of __vector bool.
6899 (_mm_max_pu8): Likewise.
6900 (_mm_min_pi16): Likewise.
6902 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
6905 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
6906 (DIVWEUO): Likewise.
6908 (DIVDEUO): Likewise.
6909 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
6910 DIVWEUO and DIVDEUO.
6911 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
6912 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
6913 (div_extend): Likewise.
6914 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
6916 (__builtin_divweuo): Likewise.
6917 (__builtin_divdeo): Likewise.
6918 (__builtin_divdeuo): Likewise.
6920 2018-03-28 Jakub Jelinek <jakub@redhat.com>
6923 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
6924 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
6926 PR tree-optimization/82004
6927 * gimple-match-head.c (optimize_pow_to_exp): New function.
6928 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
6929 Don't fold to exp if optimize_pow_to_exp is false.
6931 2018-03-28 Martin Liska <mliska@suse.cz>
6934 * calls.c (initialize_argument_information): Fix trailing space.
6935 * common.opt: Fix typo and provide better explanation for
6936 -fsanitize-coverage option.
6937 * config/i386/i386.opt: Fix typo.
6939 2018-03-28 Jakub Jelinek <jakub@redhat.com>
6940 Martin Liska <mliska@suse.cz>
6943 * gimplify.c (asan_poison_variable): Don't do the check for
6944 gimplify_omp_ctxp here.
6945 (gimplify_decl_expr): Do it here.
6946 (gimplify_target_expr): Likewise.
6948 2018-03-28 Martin Liska <mliska@suse.cz>
6951 * config/i386/i386.c (ix86_function_arg_advance): Do not call
6952 chkp_type_bounds_count if MPX is not enabled.
6954 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
6956 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
6958 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
6961 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
6962 function to create the function decl for complex long double
6963 multiply and divide for -mabi=ieeelongdouble.
6964 (init_float128_ieee): Call it.
6966 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
6969 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
6970 -fcf-protection=branch -mibt.
6971 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
6973 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6976 * config/arm/arm.c (arm_valid_symbolic_address): Handle
6977 arm_word_relocations.
6979 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
6982 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
6983 extern array declarations.
6985 2018-03-27 Richard Biener <rguenther@suse.de>
6988 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
6989 explicit single_use checks.
6991 2018-03-27 Richard Biener <rguenther@suse.de>
6993 PR tree-optimization/85082
6994 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
6997 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6999 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
7000 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
7001 Turn on fasynchronous-unwind-tables and funwind-tables.
7003 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
7006 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
7007 (*bmi_blsr_<mode>_ccz): Ditto.
7009 2018-03-26 Tom de Vries <tom@codesourcery.com>
7011 PR tree-optimization/85063
7012 * omp-general.c (offloading_function_p): New function. Factor out
7014 * omp-offload.c (pass_omp_target_link::gate): ... here.
7015 * omp-general.h (offloading_function_p): Declare.
7016 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
7017 with attribute omp declare target for offloading functions.
7019 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
7021 PR tree-optimization/84005
7022 * tree-data-ref.h (get_base_for_alignment): Declare.
7023 * tree-data-ref.c (get_base_for_alignment_1): New function.
7024 (get_base_for_alignment): Likewise.
7025 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
7026 get_base_for_alignment to find a suitable base object, instead
7027 of always using drb->base_address.
7029 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7032 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
7033 known size by default.
7035 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
7038 * lra-constraints.c (process_alt_operands): Don't match BLKmode
7039 and non BLKmode operands.
7041 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7044 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
7045 Clean up attributes.
7047 2018-03-23 Richard Biener <rguenther@suse.de>
7050 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
7051 we are going to emit early debug for LTO.
7053 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7056 * function.c (match_asm_constraints_1): Don't optimize if input
7057 doesn't satisfy general_operand predicate for output's mode.
7060 * alias.c (write_dependence_p): Don't require for x_canonicalized
7061 non-VOIDmode if x has VOIDmode.
7064 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
7065 just don't try to optimize it rather than assert it never happens.
7067 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7069 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
7070 macro expansions for definition of ST_INTERNAL_<mode> and
7071 LD_INTERNAL_<mode> builtins.
7072 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
7074 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
7076 (altivec_expand_st_builtin): Likewise.
7077 (altivec_expand_builtin): Remove calls to deleted functions.
7078 (rs6000_address_for_altivec): Delete this function.
7079 * config/rs6000/vector.md: Remove expands for
7080 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
7082 2018-03-22 Sudakshina Das <sudi.das@arm.com>
7085 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7086 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7087 re-computing once computed.
7088 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7089 (arm_init_machine_status): Initialize
7090 machine->static_chain_stack_bytes.
7092 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7095 * doc/extend.texi: Add four new prototypes for vec_ld.
7096 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7097 definitions for more logical presentation.
7098 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7099 entries for V1TI variants of __builtin_altivec_ld builtin.
7100 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7101 handling of V1TI variant of LVX icode pattern.
7102 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7103 (rs6000_gimple_fold_builtin): Likewise.
7104 (altivec_init_builtins): Add code to define
7105 __builtin_altivec_lvx_v1ti function.
7107 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7110 * function.c (match_asm_constraints_1): Don't do the optimization
7111 if input isn't a REG, SUBREG, MEM or constant.
7113 2018-03-22 Tom de Vries <tom@codesourcery.com>
7115 PR tree-optimization/84956
7116 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7117 bb_has_abnormal_pred.
7119 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7122 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7123 DECL_INITIAL (decl) to decl at the end.
7124 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7127 2018-03-21 Joseph Myers <joseph@codesourcery.com>
7129 * doc/extend.texi (__builtin_tgmath): Document when complex
7130 integer types are treated as _Complex _Float64.
7132 2018-03-21 Tom de Vries <tom@codesourcery.com>
7134 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7136 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7138 PR tree-optimization/84960
7139 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7140 if it is ENTRY block, move them into single succ of ENTRY in that case.
7142 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
7144 PR tree-optimization/84811
7145 * poly-int.h (poly_span_traits): Remove the T3 parameter and
7146 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7147 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7148 (known_subrange_p): Update accordingly. Cast each value involved
7149 in the size comparison, rather than casting the result of the
7152 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7154 PR tree-optimization/84982
7155 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7156 by flipping the least significant bit rather than all bits from
7157 bitpos to bitpos + bitsize - 1.
7159 2018-03-21 Nathan Sidwell <nathan@acm.org>
7161 * doc/extend.texi (Deprecated Features): Remove mention of
7162 long-deleted deprecations.
7164 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7167 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7168 * configure: Regenerate.
7170 2018-03-21 Tom de Vries <tom@codesourcery.com>
7172 PR tree-optimization/83126
7173 * tree-parloops.c (num_phis): New function.
7174 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7176 2018-03-21 Nathan Sidwell <nathan@acm.org>
7178 * doc/extend.texi (Deprecated Features): Update deprecated flags,
7179 mention anon-struct/union members and trailing attributes.
7181 2018-03-21 Bin Cheng <bin.cheng@arm.com>
7183 PR tree-optimization/84969
7184 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7185 builtin memset partitions if they set different rhs values.
7187 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7189 PR rtl-optimization/84989
7190 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7191 VEC_DUPLICATE with scalar result mode.
7193 2018-03-21 Martin Liska <mliska@suse.cz>
7196 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7197 not intended return statement.
7199 2018-03-21 Martin Liska <mliska@suse.cz>
7202 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7203 (chkp_find_bound_slots_1): Limit number of iterations.
7205 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
7208 * Minor grammar fixes for x86 options.
7210 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7213 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7214 holding REG_CFA_RESTORE notes, instead turn them into a USE.
7216 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
7219 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7220 (altivec_lvx_<mode>_1op): Likewise.
7221 (altivec_stvx_<mode>_2op): Likewise.
7222 (altivec_stvx_<mode>_1op): Likewise.
7223 (altivec_lvx_<VM2:mode>): New define_expand.
7224 (altivec_stvx_<VM2:mode>): Likewise.
7225 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7226 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7227 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7228 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7229 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7230 (rs6000_gen_lvx): Likewise.
7231 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7232 (altivec_expand_stv_builtin): Likewise.
7233 (altivec_expand_builtin): Likewise.
7234 * config/rs6000/vector.md: Likewise.
7236 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7239 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7242 2018-03-20 Richard Biener <rguenther@suse.de>
7245 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
7246 sign-conversions as zero, fall back to standard scalar_stmt
7249 2018-03-20 Martin Liska <mliska@suse.cz>
7252 * predict.c (rebuild_frequencies): Handle case when we have
7253 PROFILE_ABSENT, but flag_guess_branch_prob is false.
7255 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7258 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
7259 flag_section_anchors.
7260 * varasm.c (use_blocks_for_decl_p): Remove hack for
7261 dw2_force_const_mem.
7264 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
7266 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
7267 be created, use lowpart_subreg of operands[0] rather than operands[0]
7269 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
7270 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
7271 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
7272 and n constraint instead of aarch64_shift_imm_di and Usd.
7273 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
7274 (*aarch64_<optab>_reg_minus<mode>3): ... this.
7276 2018-03-20 Sudakshina Das <sudi.das@arm.com>
7279 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
7280 to favor GPR over NEON registers.
7281 (<shift>di3_neon): Likewise.
7283 2018-03-20 Tom de Vries <tom@codesourcery.com>
7286 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
7287 (nvptx_process_pars): Emit bar.sync asap and alap.
7289 2018-03-20 Tom de Vries <tom@codesourcery.com>
7292 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
7293 seen_label if seen_label is already set.
7295 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7298 * config/i386/i386.c (fold_builtin_cpu): For features above 31
7299 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
7300 Use 1U instead of 1. Formatting fixes.
7303 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
7304 instead of TREE_TYPE (s1) for the return value.
7306 2018-03-19 Jakub Jelinek <jakub@redhat.com>
7308 PR tree-optimization/84946
7309 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
7310 bitsize + bitsize in poly_uint64 rather than poly_int64.
7313 * dwarf2asm.c: Include fold-const.c.
7314 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
7315 of decl rather than decl itself.
7317 PR rtl-optimization/84643
7318 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
7320 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
7323 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
7324 calling assemble_variable.
7326 2018-03-19 Sudakshina Das <sudi.das@arm.com>
7329 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
7330 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
7332 2018-03-19 Jim Wilson <jimw@sifive.com>
7335 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
7336 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
7337 (riscv_first_stack_step): Likewise.
7338 (riscv_option_override): Use STACK_BOUNDARY instead of
7340 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
7342 (BIGGEST_ALIGNMENT): Set to 128.
7343 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
7344 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
7347 2018-03-19 Richard Biener <rguenther@suse.de>
7349 PR tree-optimization/84933
7350 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
7351 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
7353 2018-03-19 Richard Biener <rguenther@suse.de>
7355 PR tree-optimization/84859
7356 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
7357 (cond_if_else_store_replacement): Perform sinking operation on
7358 single-store BBs regardless of MAX_STORES_TO_SINK setting.
7359 Generalize what a BB with a single eligible store is.
7361 2018-03-19 Richard Biener <rguenther@suse.de>
7363 PR tree-optimization/84929
7364 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
7365 chrec_is_positive against non-chrec arg.
7367 2018-03-19 Tamar Christina <tamar.christina@arm.com>
7370 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
7372 2018-03-18 Martin Liska <mliska@suse.cz>
7374 PR rtl-optimization/84635
7375 * regrename.c (build_def_use): Use matches_mode only when
7378 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
7380 PR tree-optimization/84913
7381 * tree-vect-loop.c (vectorizable_reduction): Don't try to
7382 vectorize chains of COND_EXPRs.
7384 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7386 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
7388 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7390 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
7392 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7394 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
7396 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7397 Kito Cheng <kito.cheng@gmail.com>
7399 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
7400 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
7401 (nds32_adjust_reg_alloc_order): New function.
7402 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
7404 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
7406 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
7407 nds32_print_operand, nds32_print_operand_address): Use
7408 HOST_WIDE_INT_PRINT_DEC instead.
7410 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7412 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
7414 2018-03-17 Jakub Jelinek <jakub@redhat.com>
7417 * config/i386/i386.c (initial_ix86_tune_features,
7418 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
7420 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
7421 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
7422 rather than 1u << ix86_tune. Formatting fix.
7423 (ix86_option_override_internal): Change ix86_arch_mask from
7424 unsigned int to unsigned HOST_WIDE_INT, initialize to
7425 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
7426 (ix86_function_specific_restore): Likewise.
7428 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7431 * postreload.c (reload_combine_recognize_pattern): Perform
7432 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
7433 truncate_int_for_mode the result for the destination's mode.
7436 * hsa-gen.c (mem_type_for_type): Fix comment typo.
7437 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
7439 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7442 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
7445 * lra-assigns.c (lra_split_hard_reg_for): Don't use
7446 regno_allocno_class_array and sorted_pseudos.
7447 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
7448 insns where regno is used.
7450 2018-03-16 Martin Liska <mliska@suse.cz>
7453 * multiple_target.c (create_dispatcher_calls): Redirect
7454 reference in the symbol table.
7456 2018-03-16 Martin Liska <mliska@suse.cz>
7459 * multiple_target.c (create_dispatcher_calls): Redirect also
7462 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7466 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
7467 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
7468 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
7470 2018-03-16 Julia Koval <julia.koval@intel.com>
7472 * doc/invoke.texi (Skylake Server): Add CLWB.
7473 Cannonlake): Remove CLWB.
7475 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7477 PR tree-optimization/84841
7478 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
7480 (FLOAT_ONE_CONST_TYPE): Define.
7481 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
7482 (sort_by_operand_rank): Put entries with higher constant_type last
7483 rather than first to match comments.
7485 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
7487 * config/nios2/nios2.md (movsi_internal): Fix thinko in
7490 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7493 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
7494 check and use address of target always.
7496 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
7499 * config/i386/i386.c (indirect_thunk_needed): Update comments.
7500 (indirect_thunk_bnd_needed): Likewise.
7501 (indirect_thunks_used): Likewise.
7502 (indirect_thunks_bnd_used): Likewise.
7503 (indirect_return_needed): New.
7504 (indirect_return_bnd_needed): Likewise.
7505 (output_indirect_thunk_function): Add a bool argument for
7507 (output_indirect_thunk_function): Don't generate alias for
7508 function return thunk.
7509 (ix86_code_end): Call output_indirect_thunk_function to generate
7510 function return thunks.
7511 (ix86_output_function_return): Set indirect_return_bnd_needed
7512 and indirect_return_needed instead of indirect_thunk_bnd_needed
7513 and indirect_thunk_needed.
7515 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
7517 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
7518 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
7519 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
7521 2018-03-15 David Malcolm <dmalcolm@redhat.com>
7522 Paul Hua <paul.hua.gm@gmail.com>
7525 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
7527 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
7529 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
7530 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
7533 2018-03-15 Tamar Christina <tamar.christina@arm.com>
7536 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7537 instead of GET_MODE_SIZE when comparing Units.
7539 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7542 * varasm.c (hash_section): Return an unchangeble hash value
7543 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
7544 Return !aarch64_can_use_per_function_literal_pools_p ().
7546 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7549 * optabs.c (emit_conditional_move): Pass address of cmode's copy
7550 rather than address of cmode as last argument to prepare_cmp_insn.
7552 2018-03-15 Julia Koval <julia.koval@intel.com>
7554 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
7555 F_AVX512VNNI, F_AVX512BITALG): New.
7557 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
7560 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
7561 insn for floating-point loads and stores.
7563 2018-03-14 Carl Love <cel@us.ibm.com>
7565 * config/rs6000/rs6000-c.c: Add macro definitions for
7566 ALTIVEC_BUILTIN_VEC_PERMXOR.
7567 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
7568 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
7569 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
7570 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
7572 * config/doc/extend.texi: Add prototypes for vec_permxor.
7574 2018-03-14 David Malcolm <dmalcolm@redhat.com>
7577 * diagnostic-show-locus.c (class layout_point): Convert m_line
7578 from int to linenum_type.
7579 (line_span::comparator): Use linenum "compare" function when
7580 comparing line numbers.
7581 (test_line_span): New function.
7582 (layout_range::contains_point): Convert param "row" from int to
7584 (layout_range::intersects_line_p): Likewise.
7585 (layout::will_show_line_p): Likewise.
7586 (layout::print_source_line): Likewise.
7587 (layout::should_print_annotation_line_p): Likewise.
7588 (layout::print_annotation_line): Likewise.
7589 (layout::print_leading_fixits): Likewise.
7590 (layout::annotation_line_showed_range_p): Likewise.
7591 (struct line_corrections): Likewise for field m_row.
7592 (line_corrections::line_corrections): Likewise for param "row".
7593 (layout::print_trailing_fixits): Likewise.
7594 (layout::get_state_at_point): Likewise.
7595 (layout::get_x_bound_for_row): Likewise.
7596 (layout::print_line): Likewise.
7597 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
7598 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
7599 * input.c (selftest::test_linenum_comparisons): New function.
7600 (selftest::input_c_tests): Call it.
7601 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
7602 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
7603 * selftest.h (ASSERT_GT): New macro.
7604 (ASSERT_GT_AT): New macro.
7605 (ASSERT_LT): New macro.
7606 (ASSERT_LT_AT): New macro.
7608 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
7610 PR rtl-optimization/84780
7611 * combine.c (distribute_links): Don't make a link based on pc_rtx.
7613 2018-03-14 Martin Liska <mliska@suse.cz>
7615 * tree.c (record_node_allocation_statistics): Use
7616 get_stats_node_kind.
7617 (get_stats_node_kind): New function extracted from
7618 record_node_allocation_statistics.
7619 (free_node): Use get_stats_node_kind.
7621 2018-03-14 Richard Biener <rguenther@suse.de>
7623 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
7624 that the value-set of ANTIC_IN doesn't grow.
7627 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7629 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7630 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7632 2018-03-14 Julia Koval <julia.koval@intel.com>
7634 * config.gcc (icelake-client, icelake-server): New.
7636 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
7637 (initial_ix86_arch_features): Ditto.
7638 (PTA_SKYLAKE): Add SGX.
7639 (PTA_ICELAKE): Remove.
7640 (PTA_ICELAKE_CLIENT): New.
7641 (PTA_ICELAKE_SERVER): New.
7642 (ix86_option_override_internal): Split up icelake on icelake client and
7644 (get_builtin_code_for_version): Ditto.
7645 (fold_builtin_cpu): Ditto.
7646 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
7647 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
7648 * config/i386/i386.h (processor_type): Ditto.
7649 * doc/invoke.texi: Ditto.
7651 2018-03-14 Jakub Jelinek <jakub@redhat.com>
7654 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
7655 INTEGER_CST offset, add it together with bitpos / 8 and
7656 sign extend based on POINTER_SIZE.
7660 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
7663 * config/i386/constraints.md (Yc): New register constraint.
7664 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
7665 Use Yc constraint for alternative 2 of operand 0. Remove
7666 preferred_for_speed attribute.
7668 2018-03-14 Richard Biener <rguenther@suse.de>
7670 PR tree-optimization/84830
7671 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
7672 with the old one to avoid oscillations.
7674 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
7677 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
7679 (assign_by_spills): Return a flag of reload assignment failure.
7680 Do not process the reload assignment failures. Do not spill other
7681 reload pseudos if they has the same reg class. Update n if
7683 (lra_assign): Add a return arg. Set up from the result of
7684 assign_by_spills call.
7685 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7686 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7687 usage_insns if it is not NULL.
7688 (spill_hard_reg_in_range): New function.
7689 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7690 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7691 function prototypes.
7692 (lra_assign): Change prototype.
7693 * lra.c (lra): Add code to deal with fails by splitting hard reg
7696 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
7698 * config/riscv/riscv.opt (mrelax): New option.
7699 * config/riscv/riscv.c (riscv_file_start): Emit ".option
7700 "norelax" when riscv_mrelax is disabled.
7701 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
7703 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7706 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
7707 reassociation for int modes.
7709 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7711 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
7712 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
7714 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
7715 * config/aarch64/aarch64-sve.md
7716 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
7717 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
7718 (*extend<mode><Vwide>2): Rename to...
7719 (aarch64_sve_extend<mode><Vwide>2): ...this.
7720 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
7721 renaming the old pattern to...
7722 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
7724 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
7725 define_expand, renaming the old pattern to...
7726 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
7727 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
7728 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
7729 account when deciding which SVE instruction the optab should use.
7730 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
7732 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7734 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
7735 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
7736 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
7737 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
7738 (tlsdesc_small_<mode>): Turn a define_expand and use
7739 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
7740 (tlsdesc_small_advsimd_<mode>): ...this.
7741 (tlsdesc_small_sve_<mode>): New pattern.
7743 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7745 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
7746 (UNSPEC_UMUL_HIGHPART): New constants.
7747 (MUL_HIGHPART): New int iteraor.
7748 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
7749 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
7751 (*<su>mul<mode>3_highpart): New define_insn.
7753 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7756 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7759 2018-03-13 Martin Liska <mliska@suse.cz>
7762 * (sem_item_optimizer::sem_item_optimizer): Initialize new
7764 (sem_item_optimizer::~sem_item_optimizer): Release it.
7765 (sem_item_optimizer::merge_classes): Register variable aliases.
7766 (sem_item_optimizer::fixup_pt_set): New function.
7767 (sem_item_optimizer::fixup_points_to_sets): Likewise.
7768 * ipa-icf.h: Declare new variables and functions.
7770 2018-03-13 Jakub Jelinek <jakub@redhat.com>
7773 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
7774 integer_pow2p@2 and test integer_pow2p in condition.
7775 (A < 0 ? C : 0): Similarly for @1.
7778 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
7779 characters starting at p contain '\0' character, don't look beyond
7783 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
7784 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
7787 * reg-stack.c (change_stack): Change update_end var from int to
7788 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
7789 also call set_block_for_insn on the newly added insns and rescan.
7792 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
7793 on the last operand.
7796 * tree.c (stabilize_reference_1): Return save_expr (e) for
7797 STATEMENT_LIST even if it doesn't have side-effects.
7799 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
7801 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
7803 2018-03-12 Renlin Li <renlin.li@arm.com>
7805 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
7806 aarch64_output_scalar_simd_mov_immediate.
7808 2018-03-12 Martin Sebor <msebor@redhat.com>
7810 PR tree-optimization/83456
7811 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
7812 for perfectly overlapping calls to memcpy.
7813 (gimple_fold_builtin_memory_chk): Same.
7814 (gimple_fold_builtin_strcpy): Handle no-warning.
7815 (gimple_fold_builtin_stxcpy_chk): Same.
7816 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
7818 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7820 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
7821 parameter. Use it for SFmode.
7822 (rs6000_function_arg_advance_1): Adjust.
7823 (rs6000_function_arg): Adjust.
7824 (rs6000_gimplify_va_arg): Pass false for that new parameter.
7826 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7828 PR rtl-optimization/84169
7829 PR rtl-optimization/84780
7830 * combine.c (can_combine_p): Check for a 2-insn combination whether
7831 the destination register is used between the two insns, too.
7833 2018-03-12 Richard Biener <rguenther@suse.de>
7835 PR tree-optimization/84803
7836 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
7837 for refs DR analysis didn't process.
7839 2018-03-12 Richard Biener <rguenther@suse.de>
7841 PR tree-optimization/84777
7842 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
7843 force-vectorize loops ignore whether we are optimizing for size.
7845 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
7847 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
7848 (TARGET_MD_ASM_ADJUST): Define.
7850 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
7851 Kito Cheng <kito.cheng@gmail.com>
7852 Chung-Ju Wu <jasonwucj@gmail.com>
7854 * config/nds32/nds32.c (nds32_compute_stack_frame,
7855 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
7856 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
7857 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
7858 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
7859 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
7860 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
7861 * config/nds32/nds32.md (prologue, epilogue): Use macro
7862 NDS32_V3PUSH_AVAILABLE_P to do checking.
7864 2018-03-11 Jakub Jelinek <jakub@redhat.com>
7867 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
7868 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
7869 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
7870 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
7871 addition of most attributes on !orig_type_die or the attribute not
7872 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
7874 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7875 Chung-Ju Wu <jasonwucj@gmail.com>
7877 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
7879 * config/nds32/nds32.opt (mvh): New option.
7881 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7882 Chung-Ju Wu <jasonwucj@gmail.com>
7884 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
7886 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
7887 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
7890 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7891 Chung-Ju Wu <jasonwucj@gmail.com>
7893 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
7895 * config/nds32/nds32-multiple.md (strlensi): New pattern.
7896 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
7898 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
7899 Kito Cheng <kito.cheng@gmail.com>
7900 Chung-Ju Wu <jasonwucj@gmail.com>
7902 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
7903 UNSPEC_FFMISM and UNSPEC_FLMISM.
7904 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
7905 for ffb, ffmism and flmism.
7906 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
7907 (unspec_ffmism): Ditto.
7908 (unspec_flmism): Ditto.
7909 (nds32_expand_builtin_impl): Check if string extension is available.
7910 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
7911 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
7913 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
7916 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
7919 * lra-assigns.c (assign_by_spills): Return a flag of reload
7920 assignment failure. Do not process the reload assignment
7921 failures. Do not spill other reload pseudos if they has the same
7923 (lra_assign): Add a return arg. Set up from the result of
7924 assign_by_spills call.
7925 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7926 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7927 usage_insns if it is not NULL.
7928 (spill_hard_reg_in_range): New function.
7929 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7930 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7931 function prototypes.
7932 (lra_assign): Change prototype.
7933 * lra.c (lra): Add code to deal with fails by splitting hard reg
7936 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
7939 * config/i386/i386.opt: Replace Enforcment with Enforcement.
7941 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
7944 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
7945 (dw_val_node): Add val_symbolic_view.
7946 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
7947 (symview_upper_bound): New.
7948 (new_line_info_table): Initialize symviews_since_reset.
7949 (dwarf2out_source_line): Count symviews_since_reset and set
7950 symview_upper_bound.
7951 (dw_val_equal_p): Handle symview.
7952 (add_AT_symview): New.
7953 (print_dw_val): Handle symview.
7954 (attr_checksum, attr_checksum_ordered): Likewise.
7955 (same_dw_val_p, size_of_die): Likewise.
7956 (value_format, output_die): Likewise.
7957 (add_high_low_attributes): Use add_AT_symview for entry_view.
7958 (dwarf2out_finish): Reset symview_upper_bound, clear
7961 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
7964 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
7965 Add strict argument and use it.
7966 (rs6000_split_multireg_move): Update for new strict argument.
7967 (mem_operand_gpr): Disallow all non-offsettable addresses.
7968 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
7970 2018-03-09 Jakub Jelinek <jakub@redhat.com>
7973 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
7974 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
7975 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
7978 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
7979 decl, use remap_type if we want to use the type.
7981 2018-03-09 Martin Sebor <msebor@redhat.com>
7983 PR tree-optimization/84526
7984 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7986 (builtin_access::generic_overlap): Be prepared to handle non-array
7989 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
7991 PR rtl-optimization/84682
7992 * lra-constraints.c (process_address_1): Check is_address flag
7993 for address constraints.
7994 (process_alt_operands): Likewise.
7995 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
7996 preprocess_constraints.
7997 * recog.h (preprocess_constraints): Add oploc parameter.
7999 * recog.c (preprocess_constraints): Test address_operand for
8000 CT_ADDRESS constraints.
8002 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8005 * lra-assigns.c (assign_by_spills): Return a flag of reload
8006 assignment failure. Do not process the reload assignment
8007 failures. Do not spill other reload pseudos if they has the same
8009 (lra_assign): Add a return arg. Set up from the result of
8010 assign_by_spills call.
8011 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8012 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8013 usage_insns if it is not NULL.
8014 (spill_hard_reg_in_range): New function.
8015 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8016 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8017 function prototypes.
8018 (lra_assign): Change prototype.
8019 * lra.c (lra): Add code to deal with fails by splitting hard reg
8022 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8025 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
8026 Accept complain bool parameter. Only emit errors if it is true.
8027 (arm_parse_cpu_option_name): Likewise.
8028 (arm_target_thumb_only): Adjust callers of the above.
8029 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
8030 prototype to take a default true bool parameter.
8031 (arm_parse_arch_option_name): Likewise.
8033 2018-03-09 David Malcolm <dmalcolm@redhat.com>
8034 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
8038 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
8039 * configure: Regenerate.
8040 * configure.ac ("linker --version-script option"): New.
8041 ("linker soname option"): New.
8043 2018-03-09 Richard Biener <rguenther@suse.de>
8045 PR tree-optimization/84775
8046 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
8047 immediate uses of predicate stmts and mark them modified.
8050 PR tree-optimization/84178
8051 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8053 (version_loop_for_if_conversion): Delay update_ssa call.
8054 (tree_if_conversion): Delay update_ssa until after predicate
8057 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
8060 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
8061 when the function accesses prior frames.
8063 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8066 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
8067 gen_llsym, otherwise call maybe_gen_llsym.
8070 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
8071 has ',' character inside of it.
8073 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8076 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
8077 as clobbering CC_REGNUM.
8079 2018-03-08 Richard Biener <rguenther@suse.de>
8082 * tree-scalar-evolution.c: Include tree-into-ssa.h.
8083 (follow_copies_to_constant): Do not follow SSA names registered
8086 2018-03-08 Richard Biener <rguenther@suse.de>
8088 PR tree-optimization/84178
8089 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8091 (version_loop_for_if_conversion): Delay update_ssa call.
8092 (tree_if_conversion): Delay update_ssa until after predicate
8095 2018-03-08 David Malcolm <dmalcolm@redhat.com>
8097 PR tree-optimization/84178
8098 * tree-if-conv.c (release_bb_predicate): Remove the
8099 the assertion that the stmts have NULL use_ops.
8100 Discard the statements, asserting that they haven't
8101 yet been added to a BB.
8103 2018-03-08 Richard Biener <rguenther@suse.de>
8105 PR tree-optimization/84746
8106 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8107 (phi_translate): Pass in destination ANTIC_OUT set.
8108 (phi_translate_1): Likewise. For a simplified result lookup
8109 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8110 (phi_translate_set): Adjust.
8111 (do_pre_regular_insertion): Likewise.
8112 (do_pre_partial_partial_insertion): Likewise.
8114 2018-03-08 Martin Liska <mliska@suse.cz>
8116 PR gcov-profile/84735
8117 * doc/gcov.texi: Document usage of profile files.
8118 * gcov-io.h: Document changes in the format.
8120 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
8124 * dwarf2out.c (struct dw_line_info_table): Update comments for
8126 (FORCE_RESET_NEXT_VIEW): New.
8127 (FORCE_RESETTING_VIEW_P): New.
8128 (RESETTING_VIEW_P): Check for -1 too.
8129 (ZERO_VIEW_P): Likewise.
8130 (new_line_info_table): Force-reset next view.
8131 (dwarf2out_begin_function): Likewise.
8132 (dwarf2out_source_line): Simplify zero_view_p initialization.
8133 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8134 view directly. Omit view when omitting .loc at line 0.
8136 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8138 PR tree-optimization/84740
8139 * tree-switch-conversion.c (process_switch): Call build_constructors
8140 only if info.phi_count is non-zero.
8142 PR tree-optimization/84739
8143 * tree-tailcall.c (find_tail_calls): Check call arguments against
8144 DECL_ARGUMENTS (current_function_decl) rather than
8145 DECL_ARGUMENTS (func) when checking for tail recursion.
8147 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8149 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8150 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8151 Volker Reichelt's entry and add entries for people that perform
8152 GCC fuzzy testing and report numerous bugs.
8154 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
8157 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8158 readonly data in sdata, if that is disabled.
8159 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8160 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8161 -mreadonly-in-sdata option.
8163 2018-03-07 Martin Sebor <msebor@redhat.com>
8165 PR tree-optimization/84468
8166 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8167 basic block when looking for nul assignment.
8169 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
8172 * except.h (output_function_exception_table): Adjust prototype.
8173 * except.c (output_function_exception_table): Remove FNNAME parameter
8174 and add SECTION parameter. Ouput one part of the table at a time.
8175 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8176 the first part of the exception table and emit unwind directives.
8177 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8178 (i386_pe_seh_cold_init): Likewise.
8179 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8180 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8181 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8182 (ix86_output_call_insn): Emit a nop in one more case for SEH.
8183 * config/i386/winnt.c: Include except.h.
8184 (struct seh_frame_state): Add reg_offset, after_prologue and
8185 in_cold_section fields.
8186 (i386_pe_seh_end_prologue): Set seh->after_prologue.
8187 (i386_pe_seh_cold_init): New function.
8188 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8189 to seh->in_cold_section.
8190 (seh_emit_push): Record the offset of the push.
8191 (seh_emit_save): Record the offet of the save.
8192 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8193 Test seh->after_prologue to disregard the epilogue.
8194 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8195 (i386_pe_end_cold_function): New function.
8197 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8200 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8201 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8204 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8206 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8207 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8210 * multiple_target.c: Include tree-inline.h and intl.h.
8211 (expand_target_clones): Diagnose and fail if node->definition and
8212 !tree_versionable_function_p (node->decl).
8214 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
8216 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8218 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8219 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8220 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8222 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8225 * combine.c (try_combine): Use reg_or_subregno instead of handling
8226 just paradoxical SUBREGs and REGs.
8228 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
8230 * config/arc/arc.c (arc_finalize_pic): Remove function.
8231 (arc_must_save_register): We use single base PIC register, remove
8232 checks to save/restore the PIC register.
8233 (arc_expand_prologue): Likewise.
8234 * config/arc/arc-protos.h (arc_set_default_type_attributes):
8236 (arc_verify_short): Likewise.
8237 (arc_attr_type): Likewise.
8238 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8239 (walk_stores): Likewise.
8240 (arc_address_cost): Make it static.
8241 (arc_verify_short): Likewise.
8242 (branch_dest): Likewise.
8243 (arc_attr_type): Likewise.
8244 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
8245 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
8246 (arc_final_prescan_insn): Remove inserting the nops due to
8247 hardware hazards. It is done in reorg step.
8248 (insn_length_variant_t): Remove.
8249 (insn_length_parameters_t): Likewise.
8250 (arc_insn_length_parameters): Likewise.
8251 (arc_get_insn_variants): Likewise.
8252 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
8254 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8257 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
8260 PR tree-optimization/84687
8261 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
8263 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8265 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8267 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
8268 Rename to ppc_speculation_barrier.
8269 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
8270 __builtin_ppc_speculation_barrier.
8272 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8275 * combine.c (combine_simplify_rtx): Don't try to simplify if
8276 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
8279 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
8281 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
8282 to 32 bytes when compiling for POWER9.
8284 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8287 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
8288 regparm >= 3 with no arg reg available also for calls with
8289 flag_force_indirect_call. Pass decl to ix86_function_regparm.
8292 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
8294 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
8296 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
8299 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
8301 2018-03-05 Richard Biener <rguenther@suse.de>
8303 PR tree-optimization/84486
8304 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
8305 When inserting a __builtin_assume_aligned call set the LHS
8306 SSA name alignment info accordingly.
8308 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
8310 PR tree-optimization/84114
8311 * config/aarch64/aarch64.c (aarch64_reassociation_width)
8312 Avoid reassociation of FLOAT_MODE addition.
8314 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
8316 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
8317 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
8318 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
8319 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
8320 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
8321 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
8322 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
8324 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
8325 __builtin_ia32_wbinvd): New builtins.
8326 (SPECIAL_ARGS2): New.
8327 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
8328 (SPECIAL_ARGS2): New.
8329 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
8330 (ix86_valid_target_attribute_inner_p): Ditto.
8331 (ix86_init_mmx_sse_builtins): Add special_args2.
8332 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
8333 TARGET_WBNOINVD_P): New.
8334 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
8335 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
8336 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
8337 * config/i386/immintrin.h (_wbinvd): New intrinsic.
8338 * config/i386/pconfigintrin.h: New file.
8339 * config/i386/wbnoinvdintrin.h: Ditto.
8340 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
8342 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
8344 2018-03-05 Richard Biener <rguenther@suse.de>
8346 PR tree-optimization/84670
8347 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8349 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8350 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8351 (compute_antic_aux): Only assert the number of values in ANTIC_IN
8352 doesn't grow if all successors (recursively) were visited at least
8355 2018-03-05 Richard Biener <rguenther@suse.de>
8357 PR tree-optimization/84650
8358 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
8359 if executed in the loop pipeline.
8361 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
8363 * doc/configfiles.texi (Configuration Files): Move info about
8364 conditionalizing $target-protos.h to...
8365 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
8366 differs from $target-protos.h.
8368 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
8369 Chung-Ju Wu <jasonwucj@gmail.com>
8371 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
8372 * config/nds32/nds32-multiple.md (setmemsi): Define.
8373 * config/nds32/nds32-memory-manipulation.c
8374 (nds32_gen_dup_4_byte_to_word_value): New.
8375 (emit_setmem_word_loop): New.
8376 (emit_setmem_byte_loop): New.
8377 (nds32_expand_setmem_loop): New.
8378 (nds32_expand_setmem_loop_v3m): New.
8379 (nds32_expand_setmem_unroll): New.
8380 (nds32_expand_setmem): New.
8382 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8383 Chung-Ju Wu <jasonwucj@gmail.com>
8385 * config/nds32/nds32-memory-manipulation.c
8386 (nds32_emit_load_store): New.
8387 (nds32_emit_post_inc_load_store): New.
8388 (nds32_emit_mem_move): New.
8389 (nds32_emit_mem_move_block): New.
8390 (nds32_expand_movmemsi_loop_unknown_size): New.
8391 (nds32_expand_movmemsi_loop_known_size): New.
8392 (nds32_expand_movmemsi_loop): New.
8393 (nds32_expand_movmemsi_unroll): New.
8394 (nds32_expand_movmemqi): Rename ...
8395 (nds32_expand_movmemsi): ... to this.
8396 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
8397 (movmemsi): ... to this.
8398 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
8399 (nds32_expand_movmemsi): ... to this.
8401 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8402 Monk Chiang <sh.chiang04@gmail.com>
8403 Chung-Ju Wu <jasonwucj@gmail.com>
8405 * config/nds32/nds32-protos.h
8406 (nds32_expand_load_multiple): New arguments.
8407 (nds32_expand_store_multiple): Ditto.
8408 (nds32_valid_multiple_load_store): Rename ...
8409 (nds32_valid_multiple_load_store_p): ... to this.
8410 * config/nds32/nds32-memory-manipulation.c
8411 (nds32_expand_load_multiple): Refine implementation.
8412 (nds32_expand_store_multiple): Ditto.
8413 * config/nds32/nds32-multiple.md
8414 (load_multiple): Update nds32_expand_load_multiple interface.
8415 (store_multiple): Update nds32_expand_store_multiple interface.
8416 * config/nds32/nds32-predicates.c
8417 (nds32_valid_multiple_load_store): Rename ...
8418 (nds32_valid_multiple_load_store_p): ... to this and refine
8420 * config/nds32/predicates.md
8421 (nds32_load_multiple_and_update_address_operation): New predicate.
8422 (nds32_store_multiple_and_update_address_operation): New predicate.
8424 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8425 Chung-Ju Wu <jasonwucj@gmail.com>
8427 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
8428 (combo): New attribute.
8429 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
8431 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
8433 * config/nds32/nds32.opt: Change -mcmodel= default value.
8435 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
8436 Monk Chiang <sh.chiang04@gmail.com>
8437 Chung-Ju Wu <jasonwucj@gmail.com>
8439 * config/nds32/constants.md (unspec_element): New enum.
8440 * config/nds32/constraints.md (Umw): New constraint.
8441 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
8442 * config/nds32/nds32-intrinsic.md: Likewise.
8443 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
8444 (nds32_valid_smw_lwm_base_p): New.
8445 (nds32_output_smw_single_word): New.
8446 (nds32_output_lmw_single_word): New.
8447 (nds32_expand_unaligned_load): New.
8448 (nds32_expand_unaligned_store): New.
8449 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
8450 (nds32_output_smw_single_word): Declare.
8451 (nds32_output_lmw_single_word): Declare.
8452 (nds32_expand_unaligned_load): Declare.
8453 (nds32_expand_unaligned_store): Declare.
8454 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
8455 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
8456 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
8457 NDS32_BUILTIN_UASTORE_DW.
8458 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
8461 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
8462 Kito Cheng <kito.cheng@gmail.com>
8463 Chung-Ju Wu <jasonwucj@gmail.com>
8465 * config/nds32/nds32-intrinsic.c
8466 (nds32_expand_builtin_null_ftype_reg): Delete.
8467 (nds32_expand_builtin_reg_ftype_imm): Ditto.
8468 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
8469 (nds32_read_argument): New.
8470 (nds32_legitimize_target): Ditto.
8471 (nds32_legitimize_argument): Ditto.
8472 (nds32_check_constant_argument): Ditto.
8473 (nds32_expand_unop_builtin): Ditto.
8474 (nds32_expand_unopimm_builtin): Ditto.
8475 (nds32_expand_binop_builtin): Ditto.
8476 (nds32_builtin_decl_impl): Ditto.
8477 (builtin_description): Ditto.
8478 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
8479 (nds32_init_builtins_impl): Ditto.
8480 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
8481 (nds32_builtin_decl): New.
8482 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
8483 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
8485 2018-03-02 Jeff Law <law@redhat.com>
8487 * reorg.c (stop_search_p): Handle DEBUG_INSN.
8488 (redundant_insn, fill_simple_delay_slots): Likewise.
8489 (fill_slots_from_thread): Likewise.
8490 * resource.c (mark_referenced_resources): Likewise.
8491 (mark_set_resources, find_dead_or_set_registers): Likewise.
8493 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8495 * substring-locations.h (format_warning_va): Formatting fix for
8497 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
8499 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
8500 * substring-locations.c: Include intl.h.
8501 (format_warning_va): Turned into small wrapper around
8502 format_warning_n_va, renamed to ...
8503 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
8504 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
8506 (format_warning_at_substring_n): New function.
8507 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
8508 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
8509 format_warning_at_substring with just a shorter name instead of
8510 const function pointer.
8511 (fmtwarn_n): New function.
8512 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
8513 appropriate, get rid of all the fmtstr temporaries, move conditionals
8514 with G_() wrapped string literals directly into fmtwarn arguments,
8515 cast dir.len to (int), formatting fixes.
8517 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
8519 * doc/invoke.texi: Remove "Cilk Plus" references.
8521 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8522 Richard Biener <rguenther@suse.de>
8525 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
8526 for error or warning attributes if CALL_FROM_THUNK_P is set.
8529 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8532 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
8533 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
8536 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
8537 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
8539 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
8540 instead of -1U in last predictors element's probability member.
8542 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8545 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8546 arguments if they are comparable.
8548 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8550 PR tree-optimization/84634
8551 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
8552 masks and masked_loop_p with a single loop_masks, making sure it's
8553 null for bb vectorization.
8555 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8557 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8558 (vect_analyze_data_ref_access): Use loop->safe_len rather than
8559 loop->force_vectorize to check whether there is no alias.
8561 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8564 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
8566 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
8568 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
8569 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
8570 instead of a loop around prev_real_insn.
8571 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
8575 * config/i386/i386.c (ix86_print_operand): Use conditional
8576 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
8579 2018-03-02 Richard Biener <rguenther@suse.de>
8581 PR tree-optimization/84427
8582 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
8583 (bitmap_set_subtract_values): Rewrite to handle multiple
8586 (prune_clobbered_mems): Likewise.
8587 (phi_translate): Take edge instead of pred/phiblock.
8588 (phi_translate_1): Likewise.
8589 (phi_translate_set): Likewise. Insert all translated
8590 exprs for a value into the set, keeping possibly multiple
8591 expressions per value.
8592 (compute_antic_aux): Adjust for phi_translate changes.
8593 When intersecting union the expressions and prune those
8594 not in the final value set, keeping possibly multiple
8595 expressions per value. Do not use value-insertion
8596 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
8597 all expressions. Add verification that the value-sets
8598 only shrink during iteration.
8599 (compute_partial_antic_aux): Adjust for the phi_translate changes.
8600 (do_pre_regular_insertion): Likewise.
8601 (do_pre_partial_partial_insertion): Likewise.
8603 2018-03-02 Richard Biener <rguenther@suse.de>
8606 * config/darwin.c (saved_debug_info_level): New static global.
8607 (darwin_asm_lto_start): Disable debug info generation for LTO out.
8608 (darwin_asm_lto_end): Restore debug info generation settings.
8610 2018-03-01 Martin Liska <mliska@suse.cz>
8613 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
8616 2018-03-01 Richard Biener <rguenther@suse.de>
8619 * dwarf2out.c (gen_variable_die): Properly handle late VLA
8620 type annotation with LTO when debug was disabled at compile-time.
8622 2018-03-01 Matthew Fortune <mfortune@gmail.com>
8624 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
8626 (mips_final_postscan_insn): Likewise.
8628 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
8630 PR rtl-optimization/84528
8631 * alias.c (init_alias_target): Add commentary.
8632 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
8633 a unique base value if the frame pointer is not eliminated
8634 to the stack pointer.
8636 2018-03-01 Tom de Vries <tom@codesourcery.com>
8638 PR rtl-optimization/83327
8639 * lra-int.h (hard_regs_spilled_into): Declare.
8640 * lra.c (hard_regs_spilled_into): Define.
8641 (init_reg_info): Init hard_regs_spilled_into.
8642 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
8643 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
8644 (process_bb_lives): Handle hard_regs_spilled_into.
8645 (lra_create_live_ranges_1): Before doing liveness propagation, clear
8646 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
8648 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
8650 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
8651 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
8652 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
8653 * config/rs6000/aix72.h: New file.
8655 2018-02-28 Jakub Jelinek <jakub@redhat.com>
8657 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
8658 instead of warning_at with conditional singular and plural messages
8662 * stor-layout.c (update_alignment_for_field): For
8663 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
8664 && !DECL_PACKED (field), do the alignment update, just use
8665 only desired_align instead of MAX (type_align, desired_align)
8667 (place_field): Don't do known_align < desired_align handling
8668 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
8669 is non-NULL, instead do it after rli->prev_field handling and
8670 only if not within a bitfield word. For DECL_PACKED (field)
8671 use type_align of BITS_PER_UNIT.
8673 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
8675 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
8676 superfluous parentheses and trailing spaces.
8678 2018-02-28 Richard Biener <rguenther@suse.de>
8680 PR tree-optimization/84584
8681 * graphite-scop-detection.c (scop_detection::add_scop): Discard
8682 SCoPs with fake exit edge.
8684 2018-02-28 Martin Liska <mliska@suse.cz>
8687 * timevar.c (timer::print): Fix format to properly print 100%
8690 2018-02-28 Richard Biener <rguenther@suse.de>
8693 * genmatch.c (capture_info::walk_match): Do not mark
8694 captured expressions without operands as expr_p given
8695 they act more like predicates and should be subject to
8696 "lost tail" side-effect preserving.
8698 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
8700 PR rtl-optimization/81611
8701 * auto-inc-dec.c (attempt_change): Move dead note from
8702 mem_insn if it's the next use of regno
8703 (find_address): Take address use of reg holding
8704 non-incremented value. Add parm to limit search to the named
8706 (merge_in_block): Attempt to use a mem insn that is the next
8707 use of the original regno.
8709 2018-02-27 Martin Sebor <msebor@redhat.com>
8712 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
8713 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
8715 2018-02-27 Martin Sebor <msebor@redhat.com>
8717 PR translation/84207
8718 * diagnostic-core.h (warning_n, error_n, inform_n): Change
8719 n argument to unsigned HOST_WIDE_INT.
8720 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
8721 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
8722 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
8723 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
8725 2018-02-27 Richard Biener <rguenther@suse.de>
8727 PR tree-optimization/84512
8728 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
8729 Do not use the estimate returned from record_stmt_cost for
8730 the scalar iteration cost but sum properly using add_stmt_cost.
8732 2018-02-27 Richard Biener <rguenther@suse.de>
8734 PR tree-optimization/84466
8735 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8736 Adjust last change to less strictly validate use operands.
8738 2018-02-27 Martin Liska <mliska@suse.cz>
8740 PR gcov-profile/84548
8741 * gcov.c (process_file): Allow partial overlap and consider it
8742 also as group functions.
8743 (output_lines): Properly calculate range of lines for a group.
8745 2018-02-27 Martin Liska <mliska@suse.cz>
8747 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
8748 'ggc' suffixes. Change first column width.
8749 (timer::print): Fix formatting of the column.
8751 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
8753 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
8754 preserve inline entry blocks for the sake of debug inline
8755 entry point markers alone.
8756 (remove_unused_locals): Suggest in comments a better place to
8757 force the preservation of inline entry blocks that are
8758 otherwise unused, but do not preserve them.
8760 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8762 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
8764 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8767 * config/i386/constraints.md (Bs): Replace
8768 ix86_indirect_branch_register with
8769 TARGET_INDIRECT_BRANCH_REGISTER.
8771 * config/i386/i386.md (indirect_jump): Likewise.
8772 (tablejump): Likewise.
8773 (*sibcall_memory): Likewise.
8774 (*sibcall_value_memory): Likewise.
8775 Peepholes of indirect call and jump via memory: Likewise.
8776 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
8777 (*sibcall_value_GOT_32): Likewise.
8778 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8779 (GOT_memory_operand): Likewise.
8780 (call_insn_operand): Likewise.
8781 (sibcall_insn_operand): Likewise.
8782 (GOT32_symbol_operand): Likewise.
8783 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
8785 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8787 PR rtl-optimization/83496
8788 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
8789 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
8790 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
8791 redundant insn, if any.
8792 (relax_delay_slots): Likewise.
8793 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
8795 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
8797 PR tree-optimization/83965
8798 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
8799 that grouped statements are part of a reduction chain. Return
8800 true if the statement is not marked as a reduction itself but
8802 (vect_recog_dot_prod_pattern): Don't check whether the statement
8803 is part of a group here.
8804 (vect_recog_sad_pattern): Likewise.
8805 (vect_recog_widen_sum_pattern): Likewise.
8807 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8810 * final.c (rest_of_clean_state): Also look for calls inside sequences.
8812 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8815 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
8817 (ix86_output_indirect_function_return): New prototype.
8818 (ix86_split_simple_return_pop_internal): Likewise.
8819 * config/i386/i386.c (indirect_return_via_cx): New.
8820 (indirect_return_via_cx_bnd): Likewise.
8821 (indirect_thunk_name): Handle return va CX_REG.
8822 (output_indirect_thunk_function): Create alias for
8823 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
8824 (ix86_output_indirect_jmp): Remove the bool argument.
8825 (ix86_output_indirect_function_return): New function.
8826 (ix86_split_simple_return_pop_internal): Likewise.
8827 * config/i386/i386.md (*indirect_jump): Don't pass false
8828 to ix86_output_indirect_jmp.
8829 (*tablejump_1): Likewise.
8830 (simple_return_pop_internal): Change it to define_insn_and_split.
8831 Call ix86_split_simple_return_pop_internal to split it for
8833 (simple_return_indirect_internal): Call
8834 ix86_output_indirect_function_return instead of
8835 ix86_output_indirect_jmp.
8837 2018-02-26 Jakub Jelinek <jakub@redhat.com>
8840 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
8841 memset and value initialization afterwards.
8843 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
8845 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
8847 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8850 * common/config/aarch64/aarch64-common.c
8851 (aarch_option_optimization_table[]): Switch
8852 off fomit-frame-pointer
8854 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
8855 Chung-Ju Wu <jasonwucj@gmail.com>
8857 * config/nds32/nds32-multiple.md (load_multiple): Disallow
8859 (store_multiple): Ditto.
8861 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
8863 * config.gcc: Add --with-cpu support for nds32 target.
8864 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
8865 * config/nds32/nds32.opt: Add -mcpu= option.
8867 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
8869 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
8870 isel=yes): Warn for these deprecated options.
8872 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
8874 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
8875 ISA_2_5_MASKS_EMBEDDED.
8877 2018-02-23 Jakub Jelinek <jakub@redhat.com>
8879 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
8880 p->max as pointers rather than using iterative_hash_expr.
8882 2018-02-23 Carl Love <cel@us.ibm.com>
8884 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
8885 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
8887 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
8888 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
8889 P8V_BUILTIN_VEC_VUNSIGNED2.
8891 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
8894 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
8895 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
8897 * lra-constraints.c (curr_insn_transform): Set up
8898 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
8900 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
8901 * lra-eliminations.c (spill_pseudos): Ditto.
8902 (process_insn_for_elimination): Ditto.
8903 * lra-lives.c (reg_early_clobber_p): Use the new macros.
8904 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
8905 LRA_NON_CLOBBERED_ALT.
8907 2018-02-22 Martin Sebor <msebor@redhat.com>
8909 PR tree-optimization/84480
8910 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
8911 to maybe_diag_stxncpy_trunc. Call it.
8912 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
8913 from gimple_fold_builtin_strcpy. Print inlining stack.
8914 (handle_builtin_stxncpy): Print inlining stack.
8915 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
8917 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
8920 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
8921 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
8922 and -fcheck-pointer-bounds are used together.
8923 (indirect_thunk_prefix): New enum.
8924 (indirect_thunk_need_prefix): New function.
8925 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
8926 "_nt" instead of "_bnd" for NOTRACK prefix.
8927 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
8928 (output_indirect_thunk_function): Likewise.
8930 (ix86_code_end): Update output_indirect_thunk_function calls.
8931 (ix86_output_indirect_branch_via_reg): Replace
8932 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
8933 (ix86_output_indirect_branch_via_push): Likewise.
8934 (ix86_output_function_return): Likewise.
8935 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
8936 incompatible with -fcf-protection=branch and
8937 -fcheck-pointer-bounds.
8939 2018-02-22 Steve Ellcey <sellcey@cavium.com>
8942 * config/aarch64/aarch64.c (aarch64_print_address_internal):
8943 Change gcc_assert call to output_operand_lossage.
8945 2018-02-22 Steve Ellcey <sellcey@cavium.com>
8947 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
8949 2018-02-22 DJ Delorie <dj@redhat.com>
8950 Sebastian Perta <sebastian.perta@renesas.com>
8951 Oleg Endo <olegendo@gcc.gnu.org>
8953 * config/rx/rx.c (rx_rtx_costs): New function.
8954 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
8956 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
8958 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
8960 2018-02-22 Martin Liska <mliska@suse.cz>
8963 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
8964 Add "native" as a possible value.
8966 2018-02-22 Martin Liska <mliska@suse.cz>
8969 * config/i386/i386.c (ix86_option_override_internal):
8970 Add "native" as a possible value for -march and -mtune.
8972 2018-02-22 Jakub Jelinek <jakub@redhat.com>
8975 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
8976 to all type variants.
8978 PR tree-optimization/84503
8979 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
8980 width as info->bitpos + info->bitsize - start.
8981 (merged_store_group::merge_overlapping): Simplify width computation.
8982 (check_no_overlap): New function.
8983 (imm_store_chain_info::try_coalesce_bswap): Compute expected
8984 start + width and last_order of the group, fail if check_no_overlap
8986 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
8987 to group if check_no_overlap fails.
8989 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8991 * config/rs6000/altivec.md: Delete contraint arguments to
8992 define_expand, define_split, and define_peephole2, and in
8993 define_insn_and_split if always unused.
8994 * config/rs6000/darwin.md: Ditto.
8995 * config/rs6000/dfp.md: Ditto.
8996 * config/rs6000/rs6000.md: Ditto.
8997 * config/rs6000/sync.md: Ditto.
8998 * config/rs6000/vector.md: Ditto.
8999 * config/rs6000/vsx.md: Ditto.
9001 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9003 * config/rs6000/altivec.md: Write output control strings as braced
9004 blocks instead of double-quoted strings.
9005 * config/rs6000/darwin.md: Ditto.
9006 * config/rs6000/rs6000.md: Ditto.
9007 * config/rs6000/vector.md: Ditto.
9008 * config/rs6000/vsx.md: Ditto.
9010 2018-02-21 Jason Merrill <jason@redhat.com>
9012 PR c++/84314 - ICE with templates and fastcall attribute.
9013 * attribs.c (build_type_attribute_qual_variant): Remove assert.
9015 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9017 * ipa-cp.c (determine_versionability): Fix comment typos.
9019 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9022 * ipa-cp.c (determine_versionability): Do not version functions caling
9025 2018-02-21 Martin Liska <mliska@suse.cz>
9028 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
9029 Add "native" as a possible value.
9030 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
9031 the macro when native cpu detection is available.
9033 2018-02-21 Martin Liska <mliska@suse.cz>
9036 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
9037 Add "native" as a possible value.
9038 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
9039 when native cpu detection is available.
9041 2018-02-21 Jakub Jelinek <jakub@redhat.com>
9042 Martin Sebor <msebor@redhat.com>
9044 PR tree-optimization/84478
9045 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
9047 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
9048 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
9049 support which is conservatively correct, for 2 only stay conservative
9050 for maxlen. Formatting and comment capitalization fixes. Add STRICT
9051 argument to the 2 argument get_range_strlen, adjust 6 arg
9052 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
9054 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
9055 (gimple_fold_builtin_strlen): Pass true as last argument to
9058 2018-02-20 Martin Sebor <msebor@redhat.com>
9061 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
9062 (builtin_memref::set_base_and_offset): Same. Handle inner references.
9063 (builtin_memref::builtin_memref): Factor out parts into
9064 set_base_and_offset and call it.
9066 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9069 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
9070 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
9071 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
9072 search at the associated MODE_INT.
9074 2018-02-20 Jeff Law <law@redhat.com>
9077 PR tree-optimization/81592
9079 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
9080 for range data rather than using global data.
9081 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
9082 range data rather than using global data.
9083 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9084 pass it to children as needed.
9085 (struct directive::fmtresult): Similarly.
9086 (struct directive::set_width): Similarly.
9087 (struct directive::set_precision): Similarly.
9088 (format_integer, format_directive, parse_directive): Similarly.
9089 (format_none): Accept unnamed vr_values parameter.
9090 (format_percent, format_floating, format_character): Similarly.
9091 (format_string, format_plain): Similarly.
9092 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9093 the EVRP range analyzer for range data rather than using global data.
9094 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9095 gimple-ssa-evrp-analyze.h
9096 (class sprintf_dom_walker): Add after_dom_children member function.
9097 Add evrp_range_analyzer member.
9098 (sprintf_dom_walker::before_dom_children): Call into the EVRP
9099 range analyzer as needed.
9100 (sprintf_dom_walker::after_dom_children): New member function.
9101 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9103 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9104 (evrp_range_analyzer::pop_to_marker): Likewise.
9106 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9108 PR tree-optimization/84419
9109 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9110 with the required type if its current type is compatible but
9113 2018-02-20 Jakub Jelinek <jakub@redhat.com>
9116 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9117 after vectorization.
9119 2018-02-20 Martin Liska <mliska@suse.cz>
9122 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9123 possible values if we don't have a hint.
9125 2018-02-20 Martin Liska <mliska@suse.cz>
9129 * final.c (shorten_branches): Build align_tab array with one
9131 * opts.c (finish_options): Add alignment option limit check.
9132 (MAX_CODE_ALIGN): Likewise.
9133 (MAX_CODE_ALIGN_VALUE): Likewise.
9134 * doc/invoke.texi: Document maximum allowed option value for
9135 all -falign-* options.
9137 2018-02-19 Jakub Jelinek <jakub@redhat.com>
9140 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9141 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9142 * var-tracking.c (emit_note_insn_var_location): Remove all references
9143 to NOTE_INSN_CALL_ARG_LOCATION.
9144 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9145 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9146 Use copy_rtx_if_shared.
9147 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9148 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9149 (dwarf2out_var_location): Remove handling of
9150 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9152 * final.c (final_scan_insn): Remove all references to
9153 NOTE_INSN_CALL_ARG_LOCATION.
9154 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
9155 before dumping final insns.
9156 * except.c (emit_note_eh_region_end): Remove all references to
9157 NOTE_INSN_CALL_ARG_LOCATION.
9158 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9159 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9160 * config/arc/arc.c (hwloop_optimize): Likewise.
9161 * config/arm/arm.c (create_fix_barrier): Likewise.
9162 * config/s390/s390.c (s390_chunkify_start): Likewise.
9163 * config/sh/sh.c (find_barrier): Likewise.
9164 * config/i386/i386.c (rest_of_insert_endbranch,
9165 ix86_seh_fixup_eh_fallthru): Likewise.
9166 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9167 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9168 * config/frv/frv.c (frv_function_prologue): Likewise.
9169 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
9171 (note_outside_basic_block_p): Remove all references to
9172 NOTE_INSN_CALL_ARG_LOCATION.
9173 * gengtype.c (adjust_field_rtx_def): Likewise.
9174 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9176 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9177 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9180 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9183 PR tree-optimization/84452
9184 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9185 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9188 2018-02-19 Martin Liska <mliska@suse.cz>
9191 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9193 2018-02-19 Martin Liska <mliska@suse.cz>
9194 Richard Sandiford <richard.sandiford@linaro.org>
9196 PR tree-optimization/82491
9197 * gimple-fold.c (get_base_constructor): Make earlier bail out
9200 2018-02-19 Carl Love <cel@us.ibm.com>
9202 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9203 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9205 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9206 P8V_BUILTIN_VEC_NEG.
9208 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
9210 * config/rl78/rl78.md (movdf): New define expand.
9212 2018-02-19 Martin Liska <mliska@suse.cz>
9215 * doc/invoke.texi: Fix typo.
9216 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9218 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
9220 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9221 handle rs6000_single_float and rs6000_double_float specially for
9224 2018-02-16 Jeff Law <law@redhat.com>
9226 * config/rx/rx.c (add_pop_cfi_notes): New function.;
9229 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9232 * ipa-inline.c (inline_small_functions): Fix a typo.
9234 2018-02-16 Nathan Sidwell <nathan@acm.org>
9236 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9238 2018-02-16 Carl Love <cel@us.ibm.com>
9240 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9241 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9242 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9243 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9244 expansion to P8V_BUILTIN_VEC_FLOAT2.
9246 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
9248 PR rtl-optimization/70023
9249 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
9250 src_regno into account.
9252 2018-02-16 Carl Love <cel@us.ibm.com>
9254 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
9255 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
9256 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
9257 * config/rs6000/rs6000.c: Remove case statements for
9258 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
9259 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
9260 and P9V_BUILTIN_VEC_VINSERT4B.
9261 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
9262 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
9263 * config/rs6000/vsx.md:
9264 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
9267 2018-02-16 Carl Love <cel@us.ibm.com>
9269 * config/rs6000/altivec.h: Add builtin names vec_extract4b
9271 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
9273 * config/rs6000/rs6000-c.c: Add the definitions for
9274 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
9275 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
9276 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
9277 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
9278 definition for insert4b and define insn *insert3b_internal.
9279 * doc/extend.texi: Add documentation for vec_extract4b.
9281 2018-02-16 Nathan Sidwell <nathan@acm.org>
9283 * doc/extend.texi (Backwards Compatibility): Mention friend
9284 injection. Note for-scope is deprecated.
9285 * doc/invoke.texi (-ffriend-injection): Deprecate.
9287 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
9289 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
9290 that moved to I2, also allow destinations that are a paradoxical
9291 subreg (instead of a normal reg).
9293 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
9296 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
9299 2018-02-16 Richard Biener <rguenther@suse.de>
9301 PR tree-optimization/84037
9302 PR tree-optimization/84016
9304 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9305 Adjust vec_construct for the fact we need additional higher latency
9306 128bit inserts for AVX256 and AVX512 vector builds.
9307 (ix86_add_stmt_cost): Scale vector construction cost for
9310 2018-02-16 Richard Biener <rguenther@suse.de>
9312 PR tree-optimization/84417
9313 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
9314 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
9315 (non_rewritable_lvalue_p): Likewise, use poly-ints.
9317 2018-02-16 Martin Liska <mliska@suse.cz>
9320 * internal-fn.def (ASAN_CHECK): Set proper flags.
9321 (ASAN_MARK): Likewise.
9323 2018-02-16 Julia Koval <julia.koval@intel.com>
9325 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
9326 from PTA_CANNONLAKE.
9328 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9331 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
9332 Use ++iter rather than iter++ for std::list iterators.
9333 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
9334 defer deleting them until all nodes in the forest are processed. Do
9335 free even leaf nodes. Change to_process into auto_vec.
9338 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
9339 * vec.h (vec_default_construct): Use memset instead of placement new
9340 if BROKEN_VALUE_INITIALIZATION is defined.
9341 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
9342 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
9345 PR rtl-optimization/83723
9346 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
9347 * lra.c (lra_substitute_pseudo): Likewise. If true, use
9348 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
9350 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
9352 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
9354 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
9356 PR rtl-optimization/81443
9357 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
9358 from inner REGs to paradoxical SUBREGs.
9360 2018-02-16 Richard Biener <rguenther@suse.de>
9362 PR tree-optimization/84399
9363 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9364 For operands we can analyze at their definition make sure we can
9365 analyze them at each use as well.
9367 2018-02-16 Richard Biener <rguenther@suse.de>
9369 PR tree-optimization/84190
9370 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
9371 volatile accesses if the decl isn't volatile.
9373 2018-02-15 Jason Merrill <jason@redhat.com>
9375 PR c++/84314 - ICE with templates and fastcall attribute.
9376 * attribs.c (build_type_attribute_qual_variant): Don't clobber
9377 TYPE_CANONICAL on an existing type.
9379 2018-02-15 Jakub Jelinek <jakub@redhat.com>
9381 PR tree-optimization/84383
9382 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
9383 dstoff nor call operand_equal_p if dstbase is NULL.
9385 PR tree-optimization/84334
9386 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
9387 also a CONSTANT_CLASS_P, punt.
9389 2018-02-14 Jim Wilson <jimw@sifive.com>
9391 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
9392 first SMALL_OPERAND check. New local min_second_step. Move assert
9393 to where locals are set. Add TARGET_RVC support.
9394 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
9396 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
9398 * doc/invoke.texi: Correct -Wformat-overflow code sample.
9400 2018-02-14 Martin Sebor <msebor@redhat.com>
9402 PR tree-optimization/83698
9403 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
9404 arrays constrain the offset range to their bounds.
9405 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
9406 (builtin_access::overlap): Avoid setting the size of overlap if it's
9408 (maybe_diag_overlap): Also consider arrays when deciding what values
9409 of offsets to include in diagnostics.
9411 2018-02-14 Martin Sebor <msebor@redhat.com>
9414 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
9415 that correspond to the kind of a declaration.
9417 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
9420 * config/pa/pa.md: Load address of PIC label using the linkage table
9421 if the label is nonlocal.
9423 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
9425 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
9426 warning message if user requests -maltivec=be.
9427 * doc/invoke.texi: Document deprecation of -maltivec=be.
9429 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
9432 * config/rs6000/rs6000-c.c: Update definitions for
9433 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
9434 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
9436 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9439 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
9440 add _get_ssp intrinsics. Remove argument from
9441 __builtin_ia32_rdssp[d|q].
9442 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
9443 * config/i386/i386-builtin.def: Remove argument from
9444 __builtin_ia32_rdssp[d|q].
9445 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
9446 ix86_expand_special_args_builtin for _rdssp[d|q].
9447 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
9448 Clear register before usage.
9449 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
9450 Add documentation for new _get_ssp and _inc_ssp intrinsics.
9452 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
9454 PR tree-optimization/84357
9455 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
9456 operand 1 of an ARRAY_REF too.
9458 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
9461 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
9462 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9464 (set_of_reg): New struct.
9465 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
9466 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
9467 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9469 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
9470 Split into bitclr, bitset, bitinvert patterns if appropriate.
9471 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
9472 use rx_fuse_in_memory_bitop.
9473 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
9474 to named insn, correct maximum insn length.
9476 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
9479 * machmode.def: Define a complex mode for PARTIAL_INT.
9480 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
9482 * doc/rtl.texi: Document CSPImode.
9483 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
9485 (msp430_hard_regno_nregs_with_padding): Likewise.
9487 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
9490 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
9492 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
9494 PR rtl-optimization/84169
9495 * combine.c (try_combine): New variable split_i2i3. Set it to true if
9496 we generated a parallel as new i3 and we split that to new i2 and i3
9497 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
9498 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
9499 those to i2, not i1. Partially rewrite this scan code.
9501 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9504 * stor-layout.c (place_field): For variable length fields, adjust
9505 offset_align afterwards not just based on the field's alignment,
9506 but also on the size.
9509 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
9510 of exps and logs in the use_exp2 case.
9512 2018-02-13 Jeff Law <law@redhat.com>
9514 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
9517 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
9520 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
9524 * common.opt (gas-loc-support, gas-locview-support): New.
9525 (ginline-points, ginternal-reset-location-views): New.
9526 * doc/invoke.texi: Document them. Use @itemx where intended.
9527 (gvariable-location-views): Adjust.
9528 * target.def (reset_location_view): New.
9529 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
9530 (TARGET_RESET_LOCATION_VIEW): New.
9531 * doc/tm.texi: Rebuilt.
9532 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
9533 (dwarf2out_default_as_locview_support): New.
9534 (output_asm_line_debug_info): Use option variables.
9535 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
9536 (output_loc_list): Likewise.
9537 (add_high_low_attributes): Check option variables.
9538 Don't output entry view attribute in strict mode.
9539 (gen_inlined_subroutine_die): Check option variables.
9540 (dwarf2out_inline_entry): Likewise.
9541 (init_sections_and_labels): Likewise.
9542 (dwarf2out_early_finish): Likewise.
9543 (maybe_reset_location_view): New, from...
9544 (dwarf2out_var_location): ... here. Call it.
9545 * debug.h (dwarf2out_default_as_loc_support): Declare.
9546 (dwarf2out_default_as_locview_support): Declare.
9547 * hooks.c (hook_int_rtx_insn_0): New.
9548 * hooks.h (hook_int_rtx_insn_0): Declare.
9549 * toplev.c (process_options): Take -gas-loc-support and
9550 -gas-locview-support from dwarf2out. Enable
9551 -gvariable-location-views by default only with locview
9552 assembler support. Enable -ginternal-reset-location-views by
9553 default only if the target defines the corresponding hook.
9554 Enable -ginline-points by default if location views are
9555 enabled; force it disabled if statement frontiers are
9557 * tree-inline.c (expand_call_inline): Check option variables.
9558 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9560 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
9562 PR tree-optimization/84321
9563 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
9564 handling. Also check whether the anti-range contains any values
9565 that satisfy the mask; switch to a VR_RANGE if not.
9567 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
9570 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
9572 2018-02-13 Martin Jambor <mjambor@suse.cz>
9575 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
9576 of call statements, also set location of a load to a temporary.
9578 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
9580 * config/rl78/rl78.c (add_vector_labels): New function.
9581 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
9582 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
9583 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
9584 which checks that no arguments are passed.
9585 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
9586 * doc/extend.texi: Documentation for the new attribute.
9588 2018-02-13 Andreas Schwab <schwab@suse.de>
9590 * config/riscv/linux.h (CPP_SPEC): Define.
9592 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9595 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
9596 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
9597 OPTION_MASK_ISA_AES as first argument to def_builtin_const
9598 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
9599 instead of OPTION_MASK_ISA_PCLMUL as first argument to
9600 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
9601 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
9602 temporarily for AES and PCLMUL builtins.
9604 PR tree-optimization/84339
9605 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
9606 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
9610 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
9611 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
9612 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
9614 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
9616 * omp-simd-clone.h: New file.
9617 * omp-simd-clone.c: Include omp-simd-clone.h.
9618 (expand_simd_clones): No longer static.
9619 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
9620 cgraph.h and omp-simd-clone.h.
9621 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
9622 (vect_recog_widen_shift_pattern): Formatting fix.
9623 (vect_pattern_recog_1): Don't check optab for calls.
9626 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
9627 operands[2] into a REG before using gen_lowpart on it.
9629 2018-02-12 Jeff Law <law@redhat.com>
9632 * config/sh/sh.c (find_barrier): Consider a sibling call
9635 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
9636 successfully back substituting a reg.
9638 2018-02-12 Richard Biener <rguenther@suse.de>
9640 PR tree-optimization/84037
9641 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
9642 parameter, move visited init to caller.
9643 (vect_slp_analyze_operations): Separate cost from validity
9644 check, initialize visited once for all instances.
9645 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
9647 * tree-vect-stmts.c (vect_model_simple_cost): Make early
9649 (vect_model_promotion_demotion_cost): Likewise.
9650 (vectorizable_bswap): Guard cost modeling with !slp_node
9651 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
9653 (vectorizable_call): Likewise.
9654 (vectorizable_conversion): Likewise.
9655 (vectorizable_assignment): Likewise.
9656 (vectorizable_shift): Likewise.
9657 (vectorizable_operation): Likewise.
9658 (vectorizable_store): Likewise.
9659 (vectorizable_load): Likewise.
9660 (vectorizable_condition): Likewise.
9661 (vectorizable_comparison): Likewise.
9663 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
9666 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
9667 (ASAN_MARK): Fix fnspec to account for return value, change pointer
9668 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
9670 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9673 * params.def (inline-min-speedup): Increase from 8 to 15.
9674 (max-inline-insns-auto): Decrease from 40 to 30.
9675 * ipa-split.c (consider_split): Add some buffer for function to
9676 be considered inlining candidate.
9677 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
9680 2018-02-12 Richard Biener <rguenther@suse.de>
9682 PR tree-optimization/84037
9683 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
9684 matched stmts if we cannot swap the non-matched ones.
9686 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9688 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
9689 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
9690 _mm_maskz_scalef_round_ss): New intrinsics.
9691 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
9692 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
9693 __builtin_ia32_scalefss_round): Remove.
9694 (__builtin_ia32_scalefsd_mask_round,
9695 __builtin_ia32_scalefss_mask_round): New intrinsics.
9696 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
9697 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
9698 ((match_operand:VF_128 2 "<round_nimm_predicate>"
9699 "<round_constraint>")): Changed to ...
9700 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
9701 "<round_scalar_constraint>")): ... this.
9702 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
9703 %0, %1, %2<round_op3>}"): Changed to ...
9704 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
9705 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
9706 %2<round_scalar_mask_op3>}"): ... this.
9707 * config/i386/subst.md (round_scalar_nimm_predicate): New.
9709 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9711 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
9712 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
9713 (_mm_maskz_sqrt_round_ss): New intrinsics.
9714 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
9715 (__builtin_ia32_sqrtsd_mask_round)
9716 (__builtin_ia32_sqrtss_mask_round): New builtins.
9717 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
9718 (__builtin_ia32_sqrtss_round): Remove.
9719 (__builtin_ia32_sqrtsd_mask_round)
9720 (__builtin_ia32_sqrtss_mask_round): New builtins.
9721 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
9722 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
9723 ((match_operand:VF_128 1 "vector_operand"
9724 "xBm,<round_constraint>")): Changed to ...
9725 ((match_operand:VF_128 1 "vector_operand"
9726 "xBm,<round_scalar_constraint>")): ... this.
9727 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
9728 %0, %2, %<iptr>1<round_op3>}): Changed to ...
9729 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
9730 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
9731 %<iptr>1<round_scalar_mask_op3>}): ... this.
9732 ((set_attr "prefix" "<round_prefix>")): Changed to ...
9733 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
9735 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
9738 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
9739 Cast vec_cmpeq result to correct type.
9740 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
9741 Cast vec_cmpgt result to correct type.
9743 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
9745 * final.c (final_scan_insn_1): Renamed from...
9746 (final_scan_insn): ... this. New wrapper, to recover
9747 seen from the outermost call in recursive ones.
9748 * config/sparc/sparc.c (output_return): Drop seen from call.
9749 (output_sibcall): Likewise.
9750 * config/visium/visium.c (output_branch): Likewise.
9752 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
9754 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
9757 2018-02-10 Alan Modra <amodra@gmail.com>
9760 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
9761 Specify LR as an input.
9763 2018-02-10 Jakub Jelinek <jakub@redhat.com>
9766 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
9767 remove_member_access_dummy_vars): New functions.
9768 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
9769 lower_omp_1, execute_lower_omp): Use them.
9771 PR rtl-optimization/84308
9772 * shrink-wrap.c (spread_components): Release todo vector.
9774 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
9776 PR rtl-optimization/57193
9777 * ira-color.c (struct allocno_color_data): Add member
9778 conflict_allocno_hard_prefs.
9779 (update_conflict_allocno_hard_prefs): New.
9780 (bucket_allocno_compare_func): Add a preference based on
9781 conflict_allocno_hard_prefs.
9782 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
9783 (color_allocnos): Remove a dead code. Initiate
9784 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
9786 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9789 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
9790 constraint from =wa to wa. Avoid a subreg on the output operand,
9791 instead use a pseudo and subreg it in a move.
9792 (p9_xxbrd_<mode>): Changed to ...
9793 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
9794 (p9_xxbrd_v2df): New expander.
9795 (p9_xxbrw_<mode>): Changed to ...
9796 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
9797 (p9_xxbrw_v4sf): New expander.
9799 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9801 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
9803 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
9806 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
9807 multiply in 32-bit mode.
9808 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
9809 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
9812 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9814 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
9815 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
9816 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
9817 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
9819 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
9822 * dwarf2out.c (is_trivial_indirect_ref): New function.
9823 (dwarf2out_late_global_decl): Do not generate a location
9824 attribute for variables that have a non-trivial DECL_VALUE_EXPR
9825 and that are not defined in the current unit.
9827 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
9829 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
9830 instead of a libcall for UNORDERED.
9832 2018-02-09 Tamar Christina <tamar.christina@arm.com>
9835 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
9836 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
9838 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9841 * config/s390/s390.c (s390_set_current_function): Invoke
9842 s390_indirect_branch_settings also if fndecl didn't change.
9844 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
9846 * config/rs6000/rs6000.md (blockage): Set length to zero.
9848 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
9850 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
9852 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9855 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
9856 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
9860 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
9861 PARALLEL incoming that failed vt_get_decl_and_offset check.
9864 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
9865 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
9867 (get_variable_section): For decls in named .bss* sections pass true as
9868 second argument to bss_initializer_p.
9870 2018-02-09 Marek Polacek <polacek@redhat.com>
9871 Jakub Jelinek <jakub@redhat.com>
9874 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
9875 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
9876 Sync some changes from cxx_fold_indirect_ref.
9878 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
9880 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
9882 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
9883 (BLOCK_INLINE_ENTRY_LABEL): New.
9884 (dwarf2out_var_location): Disregard inline entry markers.
9885 (inline_entry_data): New struct.
9886 (inline_entry_data_hasher): New hashtable type.
9887 (inline_entry_data_hasher::hash): New.
9888 (inline_entry_data_hasher::equal): New.
9889 (inline_entry_data_table): New variable.
9890 (add_high_low_attributes): Add DW_AT_entry_pc and
9891 DW_AT_GNU_entry_view attributes if a pending entry is found
9892 in inline_entry_data_table. Add old entry_pc attribute only
9893 if debug nonbinding markers are disabled.
9894 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
9895 markers are enabled.
9896 (block_within_block_p, dwarf2out_inline_entry): New.
9897 (dwarf2out_finish): Check that no entries remained in
9898 inline_entry_data_table.
9899 * final.c (reemit_insn_block_notes): Handle inline entry notes.
9900 (final_scan_insn, notice_source_line): Likewise.
9901 (rest_of_clean_state): Skip inline entry markers.
9902 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
9904 * gimple.c (gimple_build_debug_inline_entry): New.
9905 * gimple.h (enum gimple_debug_subcode): Add
9906 GIMPLE_DEBUG_INLINE_ENTRY.
9907 (gimple_build_debug_inline_entry): Declare.
9908 (gimple_debug_inline_entry_p): New.
9909 (gimple_debug_nonbind_marker_p): Adjust.
9910 * insn-notes.def (INLINE_ENTRY): New.
9911 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
9912 inline entry marker notes.
9913 (print_insn): Likewise.
9914 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
9915 (INSN_DEBUG_MARKER_KIND): Likewise.
9916 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
9917 * tree-inline.c (expand_call_inline): Build and insert
9918 debug_inline_entry stmt.
9919 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
9920 inline entry blocks early, if nonbind markers are enabled.
9921 (dump_scope_block): Dump fragment info.
9922 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
9923 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
9924 (gimple_build_debug_inline_entry): New.
9925 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
9926 Enable/disable inline entry points too.
9927 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
9928 (DEBUG_INSN): Describe inline entry markers.
9930 * common.opt (gvariable-location-views): New.
9931 (gvariable-location-views=incompat5): New.
9932 * config.in: Rebuilt.
9933 * configure: Rebuilt.
9934 * configure.ac: Test assembler for view support.
9935 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
9936 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
9937 * dwarf2out.c (var_loc_view): New typedef.
9938 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
9939 (dwarf2out_locviews_in_attribute): New.
9940 (dwarf2out_locviews_in_loclist): New.
9941 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
9942 (enum dw_line_info_opcode): Add LI_adv_address.
9943 (struct dw_line_info_table): Add view.
9944 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
9945 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
9946 (zero_view_p): New variable.
9947 (ZERO_VIEW_P): New macro.
9948 (output_asm_line_debug_info): New.
9949 (struct var_loc_node): Add view.
9950 (add_AT_view_list, AT_loc_list): New.
9951 (add_var_loc_to_decl): Add view param. Test it against last.
9952 (new_loc_list): Add view params. Record them.
9953 (AT_loc_list_ptr): Handle loc and view lists.
9954 (view_list_to_loc_list_val_node): New.
9955 (print_dw_val): Handle dw_val_class_view_list.
9956 (size_of_die): Likewise.
9957 (value_format): Likewise.
9958 (loc_list_has_views): New.
9959 (gen_llsym): Set vl_symbol too.
9960 (maybe_gen_llsym, skip_loc_list_entry): New.
9961 (dwarf2out_maybe_output_loclist_view_pair): New.
9962 (output_loc_list): Output view list or entries too.
9963 (output_view_list_offset): New.
9964 (output_die): Handle dw_val_class_view_list.
9965 (output_dwarf_version): New.
9966 (output_compilation_unit_header): Use it.
9967 (output_skeleton_debug_sections): Likewise.
9968 (output_rnglists, output_line_info): Likewise.
9969 (output_pubnames, output_aranges): Update version comments.
9970 (output_one_line_info_table): Output view numbers in asm comments.
9971 (dw_loc_list): Determine current endview, pass it to new_loc_list.
9972 Call maybe_gen_llsym.
9973 (loc_list_from_tree_1): Adjust.
9974 (add_AT_location_description): Create view list attribute if
9975 needed, check it's absent otherwise.
9976 (convert_cfa_to_fb_loc_list): Adjust.
9977 (maybe_emit_file): Call output_asm_line_debug_info for test.
9978 (dwarf2out_var_location): Reset views as needed. Precompute
9979 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
9980 attribute. Set view.
9981 (new_line_info_table): Reset next view.
9982 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
9983 (dwarf2out_source_line): Likewise. Output view resets and labels to
9984 the assembler, or select appropriate line info opcodes.
9985 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
9986 (optimize_string_length): Catch it. Adjust.
9987 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
9988 dw_val_class_view_list, and remove it if no longer needed.
9989 (hash_loc_list): Hash view numbers.
9990 (loc_list_hasher::equal): Compare them.
9991 (optimize_location_lists): Check whether a view list symbol is
9992 needed, and whether the locview attribute is present, and
9993 whether they match. Remove the locview attribute if no longer
9995 (index_location_lists): Call skip_loc_list_entry for test.
9996 (dwarf2out_finish): Call output_asm_line_debug_info for test.
9997 Use output_dwarf_version.
9998 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
9999 (struct dw_val_node): Add val_view_list.
10000 * final.c (SEEN_NEXT_VIEW): New.
10001 (set_next_view_needed): New.
10002 (clear_next_view_needed): New.
10003 (maybe_output_next_view): New.
10004 (final_start_function): Rename to...
10005 (final_start_function_1): ... this. Take pointer to FIRST,
10006 add SEEN parameter. Emit param bindings in the initial view.
10007 (final_start_function): Reintroduce SEEN-less interface.
10008 (final): Rename to...
10009 (final_1): ... this. Take SEEN parameter. Output final pending
10010 next view at the end.
10011 (final): Reintroduce seen-less interface.
10012 (final_scan_insn): Output pending next view before switching
10013 sections or ending a block. Mark the next view as needed when
10014 outputting variable locations. Notify debug backend of section
10015 changes, and of location view changes.
10016 (rest_of_handle_final): Adjust.
10017 * toplev.c (process_options): Autodetect value for debug variable
10018 location views option. Warn on incompat5 without -gdwarf-5.
10019 * doc/invoke.texi (gvariable-location-views): New.
10020 (gvariable-location-views=incompat5): New.
10021 (gno-variable-location-views): New.
10023 2018-02-08 David Malcolm <dmalcolm@redhat.com>
10025 PR tree-optimization/84136
10026 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
10027 that the result of find_edge is non-NULL.
10029 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10032 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
10033 storing integer register in SImode. Fix cost of 256 and 512
10034 byte aligned SSE register store.
10036 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10038 * config/i386/i386.c (ix86_multiplication_cost): Fix
10039 multiplication cost for TARGET_AVX512DQ.
10041 2018-02-08 Marek Polacek <polacek@redhat.com>
10043 PR tree-optimization/84238
10044 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
10047 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10049 PR tree-optimization/84265
10050 * tree-vect-stmts.c (vectorizable_store): Don't treat
10051 VMAT_CONTIGUOUS accesses as grouped.
10052 (vectorizable_load): Likewise.
10054 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10056 PR tree-optimization/81635
10057 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
10058 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
10059 (test_round_for_mask): New functions.
10060 (wide_int_cc_tests): Call test_round_for_mask.
10061 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
10062 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
10063 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
10064 range returned by get_range_info.
10066 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10069 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
10070 * symtab.c: Include builtins.h
10071 (symtab_node::output_to_lto_symbol_table_p): Move here
10072 from lto-streamer-out.c:output_symbol_p.
10073 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
10074 (output_symbol_p): Move all logic to symtab.c
10075 (produce_symtab): Update.
10077 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10079 * config/s390/s390-opts.h (enum indirect_branch): Define.
10080 * config/s390/s390-protos.h (s390_return_addr_from_memory)
10081 (s390_indirect_branch_via_thunk)
10082 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
10083 (enum s390_indirect_branch_type): Define.
10084 * config/s390/s390.c (struct s390_frame_layout, struct
10085 machine_function): Remove.
10086 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10087 (indirect_branch_table_label_no, indirect_branch_table_name):
10089 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10090 (enum s390_indirect_branch_option): Define.
10091 (s390_return_addr_from_memory): New function.
10092 (s390_handle_string_attribute): New function.
10093 (s390_attribute_table): Add new attribute handler.
10094 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10095 (s390_indirect_branch_via_thunk): New function.
10096 (s390_indirect_branch_via_inline_thunk): New function.
10097 (s390_function_ok_for_sibcall): When jumping via thunk disallow
10098 sibling call optimization for non z10 compiles.
10099 (s390_emit_call): Force indirect branch target to be a single
10100 register. Add r1 clobber for non-z10 compiles.
10101 (s390_emit_epilogue): Emit return jump via return_use expander.
10102 (s390_reorg): Handle JUMP_INSNs as execute targets.
10103 (s390_option_override_internal): Perform validity checks for the
10104 new command line options.
10105 (s390_indirect_branch_attrvalue): New function.
10106 (s390_indirect_branch_settings): New function.
10107 (s390_set_current_function): Invoke s390_indirect_branch_settings.
10108 (s390_output_indirect_thunk_function): New function.
10109 (s390_code_end): Implement target hook.
10110 (s390_case_values_threshold): Implement target hook.
10111 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10113 * config/s390/s390.h (struct s390_frame_layout)
10114 (struct machine_function): Move here from s390.c.
10115 (TARGET_INDIRECT_BRANCH_NOBP_RET)
10116 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10117 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10118 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10119 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10120 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10121 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10122 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10123 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10124 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10125 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10126 (mnemonic attribute): Add values which aren't recognized
10128 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10129 pattern for branch conversion. Fix mnemonic attribute.
10130 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10131 indirect branch via thunk if requested.
10132 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10133 ("*indirect_jump"): Disable for branch conversion using out of
10135 ("indirect_jump_via_thunk<mode>_z10")
10136 ("indirect_jump_via_thunk<mode>")
10137 ("indirect_jump_via_inlinethunk<mode>_z10")
10138 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10139 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10140 ("casesi_jump_via_inlinethunk<mode>_z10")
10141 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10142 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10143 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10144 ("*indirect2_jump"): Disable for branch conversion.
10145 ("casesi_jump"): Turn into expander and expand patterns for branch
10147 ("return_use"): New expander.
10148 ("*return"): Emit return via thunk and rename it to ...
10149 ("*return<mode>"): ... this one.
10150 * config/s390/s390.opt: Add new options and and enum for the
10153 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10155 * lra-constraints.c (match_reload): Unconditionally use
10156 gen_lowpart_SUBREG, rather than selecting between that
10157 and equivalent gen_rtx_SUBREG code.
10159 2018-02-08 Richard Biener <rguenther@suse.de>
10161 PR tree-optimization/84233
10162 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10163 changed flag instead of boguously re-using phi_inserted.
10165 2018-02-08 Martin Jambor <mjambor@suse.cz>
10167 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10168 static local variables.
10170 2018-02-08 Richard Biener <rguenther@suse.de>
10172 PR tree-optimization/84278
10173 * tree-vect-stmts.c (vectorizable_store): When looking for
10174 smaller vector types to perform grouped strided loads/stores
10175 make sure the mode is supported by the target.
10176 (vectorizable_load): Likewise.
10178 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10180 * config/aarch64/aarch64.c (aarch64_components_for_bb):
10181 Increase LDP/STP opportunities by adding adjacent callee-saves.
10183 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10185 PR rtl-optimization/84068
10186 PR rtl-optimization/83459
10187 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10189 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
10191 PR tree-optimization/84224
10192 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10193 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10194 non-zero arguments.
10196 2018-02-07 Iain Sandoe <iain@codesourcery.com>
10199 * config/rs6000/altivec.md (*restore_world): Remove LR use.
10200 * config/rs6000/predicates.md (restore_world_operation): Adjust op
10201 count, remove one USE.
10203 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
10205 * doc/install.texi (Configuration): Document the
10206 --with-long-double-format={ibm,ieee} PowerPC configuration
10210 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10211 Convert from define_expand to be define_insn_and_split. Rework
10212 float/double/_Float128 conversions to QI/HI/SImode to work with
10213 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
10214 conversions to QI/HImode types did a store and then a load to
10215 truncate the value. For conversions to VSX registers, don't split
10216 the insn, instead emit the code directly. Use the code iterator
10217 any_fix to combine signed and unsigned conversions.
10218 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10219 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10220 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10221 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10222 (fix_<mode>di2_hw): Likewise.
10223 (fixuns_<mode>di2_hw): Likewise.
10224 (fix_<mode>si2_hw): Likewise.
10225 (fixuns_<mode>si2_hw): Likewise.
10226 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10227 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10228 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10229 fix<uns>_trunc<SFDF:mode>si2_p8.
10230 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10231 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10232 (fix<uns>_<mode>_mem): Likewise.
10233 (fctiw<u>z_<mode>_mem): Likewise.
10234 (fix<uns>_<mode>_mem): Likewise.
10235 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10236 the register allocator from doing a direct move to the GPRs to do
10237 a store, and instead use the ISA 3.0 store byte/half-word from
10238 vector register instruction. For IEEE 128-bit floating point,
10239 also optimize stores of 32-bit ints.
10240 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10242 2018-02-07 Alan Hayward <alan.hayward@arm.com>
10244 * genextract.c (push_pathstr_operand): New function to support
10246 (walk_rtx): Call push_pathstr_operand.
10247 (print_path): Support [a-zA-Z].
10249 2018-02-07 Richard Biener <rguenther@suse.de>
10251 PR tree-optimization/84037
10252 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
10253 (cse_and_gimplify_to_preheader): Declare.
10254 (vect_get_place_in_interleaving_chain): Likewise.
10255 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10257 (_loop_vec_info::~_loop_vec_info): Delete it.
10258 (cse_and_gimplify_to_preheader): New function.
10259 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
10260 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
10261 (vectorizable_load): Likewise. For grouped stores always base
10262 the IV on the first element.
10263 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
10264 condition before gimplifying.
10266 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10268 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
10269 *DIV_EXPR and *MOD_EXPR.
10271 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
10274 * config/i386/i386.c (ix86_option_override_internal): Mask out
10275 the CF_SET bit when checking -fcf-protection.
10277 2018-02-07 Tom de Vries <tom@codesourcery.com>
10280 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
10283 2018-02-07 Richard Biener <rguenther@suse.de>
10285 PR tree-optimization/84204
10286 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
10289 PR tree-optimization/84205
10290 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
10291 special-case isl_ast_op_zdiv_r.
10293 PR tree-optimization/84223
10294 * graphite-scop-detection.c (gather_bbs::before_dom_children):
10295 Only add conditions from within the region.
10296 (gather_bbs::after_dom_children): Adjust.
10298 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
10301 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
10302 * config/avr/avr.md: Only post-reload split REG-REG moves if
10303 either register is GENERAL_REG_P.
10305 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10307 PR tree-optimization/84235
10308 * tree-ssa-scopedtables.c
10309 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
10310 if the subtraction is performed in floating point type where NaNs are
10311 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
10312 build 1. Formatting fix.
10314 2018-02-06 Jakub Jelinek <jakub@redhat.com>
10317 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
10318 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
10319 and skip it regardless of bb boundaries. Use CALL_P macro,
10320 don't test INSN_P (insn) together with CALL_P or JUMP_P check
10321 unnecessarily, formatting fix.
10323 2018-02-06 Michael Collison <michael.collison@arm.com>
10325 * config/arm/thumb2.md:
10326 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
10327 (*thumb_mov_notscc): Ditto.
10329 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
10332 * config/rs6000/rs6000.md (su code attribute): Use "u" for
10333 unsigned_fix, not "s".
10335 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10337 * configure.ac (gcc_fn_eh_frame_ro): New function.
10338 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
10339 correct .eh_frame permissions.
10340 * configure: Regenerate.
10342 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
10344 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
10345 irrelevant options.
10347 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10349 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10350 Display warning message for -mno-speculate-indirect-jumps.
10352 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
10354 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
10356 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
10358 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
10360 PR tree-optimization/84225
10361 * tree-eh.c (find_trapping_overflow): Only call
10362 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
10364 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10367 * config/i386/i386.c: Reimplement the check of possible options
10368 -mibt/-mshstk conbination. Change error messages.
10369 * doc/invoke.texi: Fix a typo: remove extra '='.
10371 2018-02-06 Marek Polacek <polacek@redhat.com>
10373 PR tree-optimization/84228
10374 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
10376 2018-02-06 Tamar Christina <tamar.christina@arm.com>
10379 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
10380 emitted arch directives.
10381 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
10382 __ARM_FEATURE_COPROC before changing architectures.
10384 2018-02-06 Richard Biener <rguenther@suse.de>
10386 * config/i386/i386.c (print_reg): Fix typo.
10387 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
10389 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
10391 * configure: Regenerate.
10393 2018-02-05 Martin Sebor <msebor@redhat.com>
10395 PR tree-optimization/83369
10396 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
10399 2018-02-05 Martin Liska <mliska@suse.cz>
10401 * doc/invoke.texi: Cherry-pick upstream r323995.
10403 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
10405 * ira.c (ira_init_register_move_cost): Adjust comment.
10407 2018-02-05 Martin Liska <mliska@suse.cz>
10409 PR gcov-profile/84137
10410 * doc/gcov.texi: Fix typo in documentation.
10412 2018-02-05 Martin Liska <mliska@suse.cz>
10414 PR gcov-profile/83879
10415 * doc/gcov.texi: Document necessity of --dynamic-list-data when
10416 using dlopen functionality.
10418 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
10420 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
10421 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
10422 _mm_maskz_range_ss, _mm_mask_range_round_ss,
10423 _mm_maskz_range_round_ss): New intrinsics.
10424 (__builtin_ia32_rangesd128_round)
10425 (__builtin_ia32_rangess128_round): Remove.
10426 (__builtin_ia32_rangesd128_mask_round,
10427 __builtin_ia32_rangess128_mask_round): New builtins.
10428 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
10429 __builtin_ia32_rangess128_round): Remove.
10430 (__builtin_ia32_rangesd128_mask_round,
10431 __builtin_ia32_rangess128_mask_round): New builtins.
10432 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
10433 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
10434 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
10435 "<round_saeonly_constraint>")): Changed to ...
10436 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
10437 "<round_saeonly_scalar_constraint>")): ... this.
10438 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
10439 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
10440 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
10441 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
10442 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
10444 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
10446 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
10448 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
10449 Remove all values except native, 8540 and 8548.
10451 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
10453 * config/i386/i386.c (ix86_output_function_return): Pass
10454 INVALID_REGNUM, instead of -1, as invalid register number to
10455 indirect_thunk_name and output_indirect_thunk.
10457 2018-02-02 Julia Koval <julia.koval@intel.com>
10459 * config.gcc: Add -march=icelake.
10460 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
10461 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
10462 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
10463 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
10464 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
10465 (processor_target_table): Add icelake.
10466 (ix86_option_override_internal): Handle new PTAs.
10467 (get_builtin_code_for_version): Handle icelake.
10468 (M_INTEL_COREI7_ICELAKE): New.
10469 (fold_builtin_cpu): Handle icelake.
10470 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
10471 * doc/invoke.texi: Add -march=icelake.
10473 2018-02-02 Julia Koval <julia.koval@intel.com>
10475 * config/i386/i386.c (ix86_option_override_internal): Change flags type
10476 to wide_int_bitmask.
10477 * wide-int-bitmask.h: New.
10479 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10482 * config/i386/i386.md: Replace Pmode with word_mode in
10483 builtin_setjmp_setup and builtin_longjmp to support x32.
10485 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
10489 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
10491 (rs6000_supported_cpu_names): New static variable.
10492 (linux_cpu_translation_table): Likewise.
10493 (elf_platform) <cpu>: Define new static variable and use it.
10494 Translate kernel AT_PLATFORM name to canonical name if needed.
10495 Error if platform name is unknown.
10497 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
10500 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
10502 2018-02-01 Jeff Law <law@redhat.com>
10505 * config/i386/i386.c (release_scratch_register_on_entry): Add new
10506 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
10507 the scratch if RELEASE_VIA_POP is false.
10508 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
10509 If we have to save a temporary register, decrement SIZE appropriately.
10510 Pass new arguments to release_scratch_register_on_entry.
10511 (ix86_adjust_stack_and_probe): Likewise.
10512 (ix86_emit_probe_stack_range): Pass new arguments to
10513 release_scratch_register_on_entry.
10515 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
10517 PR rtl-optimization/84157
10518 * combine.c (change_zero_ext): Use REG_P predicate in
10519 front of HARD_REGISTER_P predicate.
10521 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
10523 * config/avr/avr.c (avr_option_override): Move disabling of
10524 -fdelete-null-pointer-checks to...
10525 * common/config/avr/avr-common.c (avr_option_optimization_table):
10528 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10530 PR tree-optimization/81635
10531 * tree-data-ref.c (split_constant_offset_1): For types that
10532 wrap on overflow, try to use range info to prove that wrapping
10535 2018-02-01 Renlin Li <renlin.li@arm.com>
10538 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
10539 TAILCALL_ADDR_REGS.
10540 (aarch64_register_move_cost): Likewise.
10541 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
10542 TAILCALL_ADDR_REGS.
10543 (REG_CLASS_NAMES): Likewise.
10544 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
10545 TAILCALL_ADDR_REGS. Remove IP registers.
10546 * config/aarch64/aarch64.md (Ucs): Update register constraint.
10548 2018-02-01 Richard Biener <rguenther@suse.de>
10550 * domwalk.h (dom_walker::dom_walker): Add additional constructor
10551 for specifying RPO order and allow NULL for that.
10552 * domwalk.c (dom_walker::dom_walker): Likewise.
10553 (dom_walker::walk): Handle NULL RPO order.
10554 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
10556 (rewrite_update_dom_walker): Likewise.
10557 (mark_def_dom_walker): Likewise.
10559 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10561 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
10562 (aarch64_maybe_expand_sve_subreg_move): Declare.
10563 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
10564 * config/aarch64/predicates.md (aarch64_any_register_operand): New
10566 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
10567 that are semantically a reverse operation.
10568 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
10569 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
10570 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
10572 (aarch64_can_change_mode_class): For big-endian, forbid changes
10573 between two SVE modes if they have different element sizes.
10575 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10577 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
10578 the TImode handling for big-endian targets.
10580 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10582 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
10583 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
10585 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
10586 Remove BSWAP handing for big-endian targets and use the form of
10587 LD1RQ appropariate for the mode.
10589 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10591 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
10592 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
10593 duplicated element.
10595 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10598 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
10599 check for operands that need to go through aarch64_sve_reload_be.
10601 2018-02-01 Jakub Jelinek <jakub@redhat.com>
10603 PR tree-optimization/81661
10604 PR tree-optimization/84117
10605 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
10606 * tree-eh.c: Include gimplify.h.
10607 (find_trapping_overflow, replace_trapping_overflow,
10608 rewrite_to_non_trapping_overflow): New functions.
10609 * tree-vect-loop.c: Include tree-eh.h.
10610 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
10611 * tree-data-ref.c: Include tree-eh.h.
10612 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
10614 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
10616 PR rtl-optimization/84123
10617 * combine.c (change_zero_ext): Check if hard register satisfies
10618 can_change_dest_mode before calling gen_lowpart_SUBREG.
10620 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
10623 * ira.c (ira_init_register_move_cost): Remove assert.
10625 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10627 PR rtl-optimization/84071
10628 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
10629 * doc/tm.texi: Regenerate.
10631 2018-01-31 Richard Biener <rguenther@suse.de>
10633 PR tree-optimization/84132
10634 * tree-data-ref.c (analyze_miv_subscript): Properly
10635 check whether evolution_function_is_affine_multivariate_p
10636 before calling gcd_of_steps_may_divide_p.
10638 2018-01-31 Julia Koval <julia.koval@intel.com>
10641 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
10642 * config/i386/i386.md (rdpid_rex64) New.
10643 (rdpid): Make 32bit only.
10645 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
10648 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
10649 an IDENTIFIER_NODE for FUNCTION_TYPE's.
10651 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10654 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10656 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10658 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10660 PR rtl-optimization/84071
10661 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
10662 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
10664 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10666 * config/arc/arc.c (arc_handle_aux_attribute): New function.
10667 (arc_attribute_table): Add 'aux' attribute.
10668 (arc_in_small_data_p): Consider aux like variables.
10669 (arc_is_aux_reg_p): New function.
10670 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
10671 (arc_get_aux_arg): New function.
10672 (prepare_move_operands): Handle aux-register access.
10673 (arc_handle_aux_attribute): New function.
10674 * doc/extend.texi (ARC Variable attributes): Add subsection.
10676 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10678 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
10679 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
10680 (arc_attribute_table): Add 'uncached' attribute.
10681 (arc_print_operand): Print '.di' flag for uncached memory
10683 (arc_in_small_data_p): Do not consider for small data the uncached
10685 (arc_is_uncached_mem_p): New function.
10686 * config/arc/predicates.md (compact_store_memory_operand): Check
10687 for uncached memory accesses.
10688 (nonvol_nonimm_operand): Likewise.
10689 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
10691 2018-01-31 Jakub Jelinek <jakub@redhat.com>
10694 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
10695 falign-loops=): Add Optimization flag.
10697 2018-01-30 Jeff Law <law@redhat.com>
10700 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
10701 INT_REGISTERS_SAVED. Check it prior to calling
10702 get_scratch_register_on_entry.
10703 (ix86_adjust_stack_and_probe): Similarly.
10704 (ix86_emit_probe_stack_range): Similarly.
10705 (ix86_expand_prologue): Corresponding changes.
10707 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10710 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
10711 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
10713 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
10716 * lra-constraints.c (curr_insn_transform): Process AND in the
10719 2018-01-30 Jakub Jelinek <jakub@redhat.com>
10721 PR rtl-optimization/83986
10722 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
10723 dependence against last_pending_memory_flush in addition to
10724 pending_jump_insns.
10726 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
10728 PR tree-optimization/81611
10729 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
10732 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10735 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
10738 2018-01-30 Richard Biener <rguenther@suse.de>
10739 Jakub Jelinek <jakub@redhat.com>
10741 PR tree-optimization/84111
10742 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
10743 inner loops added during recursion, as they don't have up-to-date
10746 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10749 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
10750 (can_inline_edge_by_limits_p): ... here.
10751 (can_early_inline_edge_p, check_callers,
10752 update_caller_keys, update_callee_keys, recursive_inlining,
10753 add_new_edges_to_heap, speculation_useful_p,
10754 inline_small_functions,
10755 inline_small_functions, flatten_function,
10756 inline_to_all_callers_1): Update.
10758 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10760 * profile-count.c (profile_count::combine_with_ipa_count): Handle
10763 2018-01-30 Richard Biener <rguenther@suse.de>
10765 PR tree-optimization/83008
10766 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
10767 invariant and constant vector uses in stmts when they need
10768 more than one stmt.
10770 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10773 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
10774 * configure: Regenerate.
10776 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10778 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
10780 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
10781 Use gen_rtx_REG rather than gen_lowpart.
10783 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10785 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
10786 rather than 0 when creating partial subregs.
10788 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10790 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
10793 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
10796 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
10797 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
10798 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
10799 flags. This restores the settings used before the 2017-07-24.
10800 Turning off pre increment/decrement/modify allows IVOPTS to
10801 optimize DF/SF loops where the index is an int.
10803 2018-01-29 Richard Biener <rguenther@suse.de>
10804 Kelvin Nilsen <kelvin@gcc.gnu.org>
10807 * tree-vect-stmts.c (vectorizable_call): Don't call
10808 targetm.vectorize_builtin_md_vectorized_function if callee is
10811 2018-01-22 Carl Love <cel@us.ibm.com>
10813 * doc/extend.tex: Fix typo in second arg in
10814 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
10816 2018-01-29 Richard Biener <rguenther@suse.de>
10818 PR tree-optimization/84086
10819 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
10820 (flush_ssaname_freelist): When SSA names were released reset
10821 the SCEV hash table.
10823 2018-01-29 Richard Biener <rguenther@suse.de>
10825 PR tree-optimization/84057
10826 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
10827 removed paths when removing edges.
10829 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
10831 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
10832 -mfunction-return=@var{choice}.
10834 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
10836 PR diagnostic/84034
10837 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
10838 Handle CR like TAB.
10839 (layout::print_source_line): Likewise.
10840 (test_get_line_width_without_trailing_whitespace): Add test cases.
10842 2018-01-27 Jakub Jelinek <jakub@redhat.com>
10844 PR middle-end/84040
10845 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
10848 2018-01-26 Jim Wilson <jimw@sifive.com>
10850 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
10852 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
10855 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10857 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
10858 and CMP + SUB-immediate -> SUBS.
10860 2018-01-26 Martin Sebor <msebor@redhat.com>
10862 PR tree-optimization/83896
10863 * tree-ssa-strlen.c (get_string_len): Rename...
10864 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
10865 Avoid assuming length is constant.
10866 (handle_char_store): Use HOST_WIDE_INT for string length.
10868 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
10871 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
10872 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
10874 2018-01-26 Richard Biener <rguenther@suse.de>
10876 PR rtl-optimization/84003
10877 * dse.c (record_store): Only record redundant stores when
10878 the earlier store aliases at least all accesses the later one does.
10880 2018-01-26 Jakub Jelinek <jakub@redhat.com>
10882 PR rtl-optimization/83985
10883 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
10884 REG_CFA_RESTORE insns.
10885 (delete_unmarked_insns): Don't ignore separate shrink wrapping
10886 REG_CFA_RESTORE insns here.
10889 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
10890 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
10892 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10894 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
10895 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
10896 (arc_init): Likewise.
10897 (arc_override_options): Likewise.
10898 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
10900 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
10902 * config/arc/arc.h (TARGET_DBNZ): Define.
10903 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
10904 properly set the tune attribute.
10905 (dbnz): Use TARGET_DBNZ guard.
10906 * config/arc/arc.opt (mtune): Add core3 option.
10908 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10910 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
10911 recognize new pic like addresses.
10912 (arc_delegitimize_address): Clean up.
10914 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10916 * config/arc/arc-arches.def: Option mrf16 valid for all
10918 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
10919 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
10920 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
10921 * config/arc/arc-tables.opt: Regenerate.
10922 * config/arc/arc.c (arc_conditional_register_usage): Handle
10923 reduced register file case.
10924 (arc_file_start): Set must have build attributes.
10925 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
10926 mrf16 option value.
10927 * config/arc/arc.opt (mrf16): Add new option.
10928 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
10929 * config/arc/genmultilib.awk: Handle new mrf16 option.
10930 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
10931 * config/arc/t-multilib: Regenerate.
10932 * doc/invoke.texi (ARC Options): Document mrf16 option.
10934 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10936 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
10937 * config/arc/arc.c (arc_handle_secure_attribute): New function.
10938 (arc_attribute_table): Add 'secure_call' attribute.
10939 (arc_print_operand): Print secure call operand.
10940 (arc_function_ok_for_sibcall): Don't optimize tail calls when
10942 (arc_is_secure_call_p): New function. * config/arc/arc.md
10943 (call_i): Add support for sjli instruction.
10944 (call_value_i): Likewise.
10945 * config/arc/constraints.md (Csc): New constraint.
10947 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10948 John Eric Martin <John.Martin@emmicro-us.com>
10950 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
10951 * config/arc/arc.c (_arc_jli_section): New struct.
10952 (arc_jli_section): New type.
10953 (rc_jli_sections): New static variable.
10954 (arc_handle_jli_attribute): New function.
10955 (arc_attribute_table): Add jli_always and jli_fixed attribute.
10956 (arc_file_end): New function.
10957 (TARGET_ASM_FILE_END): Define.
10958 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
10959 (arc_add_jli_section): New function.
10960 (jli_call_scan): Likewise.
10961 (arc_reorg): Call jli_call_scan.
10962 (arc_output_addsi): Remove 'S' from printing asm operand.
10963 (arc_is_jli_call_p): New function.
10964 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
10966 (movhi_insn): Likewise.
10967 (movsi_insn): Likewise.
10968 (movsi_set_cc_insn): Likewise.
10969 (loadqi_update): Likewise.
10970 (load_zeroextendqisi_update): Likewise.
10971 (load_signextendqisi_update): Likewise.
10972 (loadhi_update): Likewise.
10973 (load_zeroextendhisi_update): Likewise.
10974 (load_signextendhisi_update): Likewise.
10975 (loadsi_update): Likewise.
10976 (loadsf_update): Likewise.
10977 (movsicc_insn): Likewise.
10978 (bset_insn): Likewise.
10979 (bxor_insn): Likewise.
10980 (bclr_insn): Likewise.
10981 (bmsk_insn): Likewise.
10982 (bicsi3_insn): Likewise.
10983 (cmpsi_cc_c_insn): Likewise.
10984 (movsi_ne): Likewise.
10985 (movsi_cond_exec): Likewise.
10986 (clrsbsi2): Likewise.
10987 (norm_f): Likewise.
10997 (call_i): Remove 'S' asm letter, add jli instruction.
10998 (call_value_i): Likewise.
10999 * config/arc/arc.op (mjli-always): New option.
11000 * config/arc/constraints.md (Cji): New constraint.
11001 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
11003 (subsf3_fpx): Likewise.
11004 (mulsf3_fpx): Likewise.
11005 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
11007 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
11008 function attrbutes.
11009 * doc/invoke.texi (ARC): Document mjli-always option.
11011 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
11013 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
11014 avoid addition with 0 and use incw and decw where possible.
11016 2018-01-26 Richard Biener <rguenther@suse.de>
11018 PR tree-optimization/81082
11019 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
11020 association if it requires casting to unsigned.
11021 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
11022 from fold_plusminus_mult_expr to catch important cases late when
11023 range info is available.
11025 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11027 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
11028 * configure.ac (hidden_linkonce): New test.
11029 * configure: Regenerate.
11030 * config.in: Regenerate.
11032 2018-01-26 Julia Koval <julia.koval@intel.com>
11034 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
11035 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
11036 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
11037 _mm_mask_bitshuffle_epi64_mask): Fix type.
11038 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
11039 USI_FTYPE_V4DI_V4DI_USI): Remove.
11040 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
11041 __builtin_ia32_vpshufbitqmb256_mask,
11042 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
11043 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
11044 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
11046 2018-01-26 Alan Modra <amodra@gmail.com>
11049 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
11050 UNSPEC_VBPERMQ. Sort other unspecs.
11052 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
11054 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
11056 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
11058 PR middle-end/83055
11059 * predict.c (drop_profile): Do not push/pop cfun; update also
11061 (handle_missing_profiles): Fix logic looking for zero profiles.
11063 2018-01-25 Jakub Jelinek <jakub@redhat.com>
11065 PR middle-end/83977
11066 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
11067 on functions with #pragma omp declare simd or functions with simd
11069 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
11070 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
11071 Remove trailing \n from warning_at calls.
11073 2018-01-25 Tom de Vries <tom@codesourcery.com>
11076 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11077 for neutered workers.
11079 2018-01-24 Joseph Myers <joseph@codesourcery.com>
11082 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
11083 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11085 2018-01-24 Jeff Law <law@redhat.com>
11088 * i386.c (get_probe_interval): Move to earlier point.
11089 (ix86_compute_frame_layout): If -fstack-clash-protection and
11090 the frame is larger than the probe interval, then use pushes
11091 to save registers rather than reg->mem moves.
11092 (ix86_expand_prologue): Remove conditional for int_registers_saved
11095 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
11098 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11099 min/max for never referenced object.
11101 2018-01-24 Jakub Jelinek <jakub@redhat.com>
11103 PR middle-end/83977
11104 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11106 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11107 attributes from DECL_ATTRIBUTES (decl) without affecting
11108 DECL_ATTRIBUTES (current_function_decl).
11109 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11110 functions with non-NULL DECL_ABSTRACT_ORIGIN.
11112 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
11114 PR tree-optimization/83979
11115 * fold-const.c (fold_comparison): Use constant_boolean_node
11116 instead of boolean_{true,false}_node.
11118 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
11120 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11123 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11125 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11126 Simplify the clause that sets the length attribute.
11127 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11128 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11129 clause that sets the length attribute.
11130 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11132 2018-01-24 Tom de Vries <tom@codesourcery.com>
11135 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11136 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11137 Add strict parameter.
11138 (prevent_branch_around_nothing): Insert dummy insn between branch to
11139 label and label with no ptx insn inbetween.
11140 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11142 2018-01-24 Tom de Vries <tom@codesourcery.com>
11145 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11146 for neutered threads in warp.
11147 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11149 2018-01-24 Richard Biener <rguenther@suse.de>
11151 PR tree-optimization/83176
11152 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11155 2018-01-24 Richard Biener <rguenther@suse.de>
11157 PR tree-optimization/82819
11158 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11159 code generating pluses that are no-ops in the target precision.
11161 2018-01-24 Richard Biener <rguenther@suse.de>
11163 PR middle-end/84000
11164 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11166 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11168 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11169 to merge probabilities.
11170 * predict.c (probably_never_executed): Also mark as cold functions
11171 with global 0 profile and guessed local profile.
11172 * profile-count.c (profile_probability::combine_with_count): New
11174 * profile-count.h (profile_probability::operator*,
11175 profile_probability::operator*=, profile_probability::operator/,
11176 profile_probability::operator/=): Reduce precision to adjusted
11177 and set value to guessed on contradictory divisions.
11178 (profile_probability::combine_with_freq): Remove.
11179 (profile_probability::combine_wiht_count): Declare.
11180 (profile_count::force_nonzero):: Set to adjusted.
11181 (profile_count::probability_in):: Set quality to adjusted.
11182 * tree-ssa-tail-merge.c (replace_block_by): Use
11183 combine_with_count.
11185 2018-01-23 Andrew Waterman <andrew@sifive.com>
11186 Jim Wilson <jimw@sifive.com>
11188 * config/riscv/riscv.c (riscv_stack_boundary): New.
11189 (riscv_option_override): Set riscv_stack_boundary. Handle
11190 riscv_preferred_stack_boundary_arg.
11191 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11192 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11193 (STACK_BOUNDARY): Set to riscv_stack_boundary.
11194 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11195 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11196 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11198 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
11201 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11202 of struct ix86_frame.
11203 (ix86_expand_epilogue): Likewise. Add a local variable for
11204 the reg_save_offset field in struct ix86_frame.
11206 2018-01-23 Bin Cheng <bin.cheng@arm.com>
11208 PR tree-optimization/82604
11209 * tree-loop-distribution.c (enum partition_kind): New enum item
11210 PKIND_PARTIAL_MEMSET.
11211 (partition_builtin_p): Support above new enum item.
11212 (generate_code_for_partition): Ditto.
11213 (compute_access_range): Differentiate cases that equality can be
11214 proven at all loops, the innermost loops or no loops.
11215 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11216 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
11217 (finalize_partitions, distribute_loop): Don't fuse partition of
11218 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11219 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11220 parloop is enabled.
11222 2018-01-23 Martin Liska <mliska@suse.cz>
11224 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11225 order to ignore the predictor.
11226 (PRED_POLYMORPHIC_CALL): Likewise.
11227 (PRED_RECURSIVE_CALL): Likewise.
11229 2018-01-23 Martin Liska <mliska@suse.cz>
11231 * tree-profile.c (tree_profiling): Print function header to
11232 aware reader which function we are working on.
11233 * value-prof.c (gimple_find_values_to_profile): Do not print
11234 not interesting value histograms.
11236 2018-01-23 Martin Liska <mliska@suse.cz>
11238 * profile-count.h (enum profile_quality): Add
11239 profile_uninitialized as the first value. Do not number values
11240 as they are zero based.
11241 (profile_count::verify): Update sanity check.
11242 (profile_probability::verify): Likewise.
11244 2018-01-23 Nathan Sidwell <nathan@acm.org>
11246 * doc/invoke.texi (ffor-scope): Deprecate.
11248 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11250 PR tree-optimization/83510
11251 * domwalk.c (set_all_edges_as_executable): New function.
11252 (dom_walker::dom_walker): Convert bool param
11253 "skip_unreachable_blocks" to enum reachability. Move setup of
11254 edge flags to set_all_edges_as_executable and only do it when
11255 reachability is REACHABLE_BLOCKS.
11256 * domwalk.h (enum dom_walker::reachability): New enum.
11257 (dom_walker::dom_walker): Convert bool param
11258 "skip_unreachable_blocks" to enum reachability.
11259 (set_all_edges_as_executable): New decl.
11260 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
11261 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
11263 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
11264 but converting true to REACHABLE_BLOCKS.
11265 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
11267 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
11268 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
11269 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
11271 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
11272 if check_all_array_refs will be called.
11274 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11276 * tree.c (selftest::test_location_wrappers): Add more test
11279 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11281 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
11282 (selftest::test_bit_in_range): Likewise.
11284 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11287 * doc/sourcebuild.texi (vect_float): Say that the selector
11288 only describes the situation when -funsafe-math-optimizations is on.
11289 (vect_float_strict): Document.
11291 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11293 PR tree-optimization/83965
11294 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11295 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
11296 instead of checking only for a reduction.
11297 (vect_recog_widen_sum_pattern): Likewise.
11299 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11301 * predict.c (probably_never_executed): Only use precise profile info.
11302 (compute_function_frequency): Skip after inlining hack since we now
11303 have quality checking.
11305 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11307 * profile-count.h (profile_probability::very_unlikely,
11308 profile_probability::unlikely, profile_probability::even): Set
11309 precision to guessed.
11311 2018-01-23 Richard Biener <rguenther@suse.de>
11313 PR tree-optimization/83963
11314 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
11315 Properly terminate dominator walk when crossing the exit edge not
11316 when visiting its source block.
11318 2018-01-23 Jakub Jelinek <jakub@redhat.com>
11321 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
11322 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
11324 2018-01-22 Jakub Jelinek <jakub@redhat.com>
11326 PR tree-optimization/83957
11327 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
11328 semicolon after for body surrounded by braces.
11330 PR tree-optimization/83081
11331 * profile-count.h (profile_probability::split): New method.
11332 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
11333 Use profile_probability::split.
11334 (do_compare_rtx_and_jump): Fix adjustment of probabilities
11335 when splitting a single conditional jump into 2.
11337 2018-01-22 David Malcolm <dmalcolm@redhat.com>
11339 PR tree-optimization/69452
11340 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
11343 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11345 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
11346 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
11347 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
11349 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11351 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
11353 * config/rl78/rl78.md (movdi): New define_expand.
11354 * config/rl78/rl78.c (rl78_split_movdi): New function.
11356 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
11359 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
11361 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
11362 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
11363 128-bit to produce an UNSPEC move to get the double word with the
11364 signbit and then a shift directly to do signbit.
11365 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
11366 implementation with a new version that just does either a direct
11367 move or a regular move. Move memory interface to separate insns.
11368 Move insns so they are next to the expander.
11369 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
11370 with signbit move. Split big and little endian case.
11371 (signbit<mode>2_dm_mem_le): Likewise.
11372 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
11373 (signbit<mode>2_dm2): Likewise.
11375 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11377 * config/rl78/rl78.md (anddi3): New define_expand.
11379 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11381 * config/rl78/rl78.md (umindi3): New define_expand.
11383 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11385 * config/rl78/rl78.md (smindi3): New define_expand.
11387 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11389 * config/rl78/rl78.md (smaxdi3): New define_expand.
11391 2018-01-22 Carl Love <cel@us.ibm.com>
11393 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
11394 LVX_V1TI): Add macro expansion.
11395 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
11396 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
11397 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
11398 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
11399 Change check to determine if the instruction is a byte reversing
11400 entry. Fix typo in comment.
11401 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
11402 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
11403 Add def_builtin calls for new builtins.
11404 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
11405 Add define_insn expansion.
11407 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11409 * config/rl78/rl78.md (umaxdi3): New define_expand.
11411 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11413 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
11414 for non-QImode registers.
11416 2018-01-22 Richard Biener <rguenther@suse.de>
11418 PR tree-optimization/83963
11419 * graphite-scop-detection.c (scop_detection::get_sese): Delay
11420 including the loop exit block.
11421 (scop_detection::merge_sese): Likewise.
11422 (scop_detection::add_scop): Do it here instead.
11424 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11426 * doc/sourcebuild.texi (arm_softfloat): Document.
11428 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
11431 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
11432 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
11433 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
11435 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11436 David Edelsohn <dje.gcc@gmail.com>
11439 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11440 Change "crset eq" to "crset 2".
11441 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11442 (*call_indirect_aix<mode>_nospec): Likewise.
11443 (*call_value_indirect_aix<mode>_nospec): Likewise.
11444 (*call_indirect_elfv2<mode>_nospec): Likewise.
11445 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
11446 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
11447 change assembly output from . to $.
11448 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11449 (indirect_jump<mode>_nospec): Change assembly output from . to $.
11450 (*tablejump<mode>_internal1_nospec): Likewise.
11452 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
11455 * config/sh/sh_optimize_sett_clrt.cc:
11456 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
11458 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
11460 PR tree-optimization/83940
11461 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
11462 offset_dt to vect_constant_def rather than vect_unknown_def_type.
11463 (vect_check_load_store_mask): Add a mask_dt_out parameter and
11464 use it to pass back the definition type.
11465 (vect_check_store_rhs): Likewise rhs_dt_out.
11466 (vect_build_gather_load_calls): Add a mask_dt argument and use
11467 it instead of a call to vect_is_simple_use.
11468 (vectorizable_store): Update calls to vect_check_load_store_mask
11469 and vect_check_store_rhs. Use the dt returned by the latter instead
11470 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
11471 instead of calls to vect_is_simple_use. Pass the scalar rather
11472 than the vector operand to vect_is_simple_use when handling
11473 second and subsequent copies of an rhs value.
11474 (vectorizable_load): Update calls to vect_check_load_store_mask
11475 and vect_build_gather_load_calls. Use the cached mask_dt and
11476 gs_info.offset_dt instead of calls to vect_is_simple_use.
11478 2018-01-20 Jakub Jelinek <jakub@redhat.com>
11480 PR middle-end/83945
11481 * tree-emutls.c: Include gimplify.h.
11482 (lower_emutls_2): New function.
11483 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
11484 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
11485 it before further processing.
11488 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
11489 UINTVAL (trueop1) instead of INTVAL (op1).
11491 2018-01-19 Jakub Jelinek <jakub@redhat.com>
11495 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
11496 INCOMING_FRAME_SP_OFFSET if not defined.
11497 (scan_trace): Add ENTRY argument. If true and
11498 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
11499 emit a note to adjust the CFA offset.
11500 (create_cfi_notes): Adjust scan_trace callers.
11501 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
11502 INCOMING_FRAME_SP_OFFSET in the CIE.
11503 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
11504 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
11506 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
11507 * doc/tm.texi: Regenerated.
11509 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11511 PR rtl-optimization/83147
11512 * lra-constraints.c (remove_inheritance_pseudos): Use
11513 lra_substitute_pseudo_within_insn.
11515 2018-01-19 Tom de Vries <tom@codesourcery.com>
11516 Cesar Philippidis <cesar@codesourcery.com>
11519 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
11521 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
11524 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
11525 spaces for function labels.
11527 2018-01-19 Martin Liska <mliska@suse.cz>
11529 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
11530 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
11531 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
11532 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
11533 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
11534 (PRED_CONST_RETURN): Change from 69 to 65.
11535 (PRED_NULL_RETURN): Change from 91 to 71.
11536 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
11537 (PRED_LOOP_GUARD): Change from 66 to 73.
11539 2018-01-19 Martin Liska <mliska@suse.cz>
11541 * predict.c (predict_insn_def): Add new assert.
11542 (struct branch_predictor): Change type to signed integer.
11543 (test_prediction_value_range): Amend test to cover
11544 PROB_UNINITIALIZED.
11545 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
11546 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
11547 (PRED_LOOP_ITERATIONS_MAX): Likewise.
11548 (PRED_LOOP_IV_COMPARE): Likewise.
11549 * predict.h (PROB_UNINITIALIZED): Define new constant.
11551 2018-01-19 Martin Liska <mliska@suse.cz>
11553 * predict.c (dump_prediction): Add new format for
11554 analyze_brprob.py script which is enabled with -details
11556 * profile-count.h (precise_p): New function.
11558 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11560 PR tree-optimization/83922
11561 * tree-vect-loop.c (vect_verify_full_masking): Return false if
11562 there are no statements that need masking.
11563 (vect_active_double_reduction_p): New function.
11564 (vect_analyze_loop_operations): Use it when handling phis that
11565 are not in the loop header.
11567 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11569 PR tree-optimization/83914
11570 * tree-vect-loop.c (vectorizable_induction): Don't convert
11571 init_expr or apply the peeling adjustment for inductions
11572 that are nested within the vectorized loop.
11574 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11576 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
11579 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11583 * function.h (gimplify_parameters): Add gimple_seq * argument.
11584 * function.c: Include gimple.h and options.h.
11585 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
11586 for the added local temporaries if needed.
11587 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
11588 if there are any parameter cleanups, wrap whole body into a
11589 try/finally with the cleanups.
11591 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
11594 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
11595 Use GET_MODE_CLASS for scalar floating point.
11597 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11601 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
11602 Fix call of call_cgraph_insertion_hooks.
11604 2018-01-18 Martin Sebor <msebor@redhat.com>
11606 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
11608 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11611 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
11614 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
11617 * common.opt: (-ffile-prefix-map): New option.
11618 * opts.c (common_handle_option): Defer it.
11619 * opts-global.c (handle_common_deferred_options): Handle it.
11620 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
11621 * file-prefix-map.h: New file.
11622 (remap_debug_filename, add_debug_prefix_map): ...here.
11623 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
11624 * final.c (debug_prefix_map, add_debug_prefix_map
11625 remap_debug_filename): Move to...
11626 * file-prefix-map.c: New file.
11627 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
11628 generalize, get rid of alloca(), use strrchr() instead of strchr().
11629 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
11630 Implement in terms of add_prefix_map().
11631 (remap_macro_filename, remap_debug_filename): Implement in term of
11633 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
11634 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
11635 * dbxout.c: Include file-prefix-map.h.
11636 * varasm.c: Likewise.
11637 * vmsdbgout.c: Likewise.
11638 * xcoffout.c: Likewise.
11639 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
11640 * doc/cppopts.texi (-fmacro-prefix-map): Document.
11641 * doc/invoke.texi (-ffile-prefix-map): Document.
11642 (-fdebug-prefix-map): Update description.
11644 2018-01-18 Martin Liska <mliska@suse.cz>
11646 * config/i386/i386.c (indirect_thunk_name): Document that also
11648 (output_indirect_thunk): Document why both instructions
11649 (pause and lfence) are generated.
11651 2018-01-18 Richard Biener <rguenther@suse.de>
11653 PR tree-optimization/83887
11654 * graphite-scop-detection.c
11655 (scop_detection::get_nearest_dom_with_single_entry): Remove.
11656 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
11657 (scop_detection::merge_sese): Re-implement with a flood-fill
11658 algorithm that properly finds a SESE region if it exists.
11660 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11663 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
11664 pointer_diff optimizations use view_convert instead of convert.
11666 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11668 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11669 Generate different code for -mno-speculate-indirect-jumps.
11670 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11671 (*call_indirect_aix<mode>): Disable for
11672 -mno-speculate-indirect-jumps.
11673 (*call_indirect_aix<mode>_nospec): New define_insn.
11674 (*call_value_indirect_aix<mode>): Disable for
11675 -mno-speculate-indirect-jumps.
11676 (*call_value_indirect_aix<mode>_nospec): New define_insn.
11677 (*sibcall_nonlocal_sysv<mode>): Generate different code for
11678 -mno-speculate-indirect-jumps.
11679 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11681 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
11683 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
11684 long double type, set the flags for noting the default long double
11685 type, even if we don't pass or return a long double type.
11687 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
11690 * ipa-inline.c (flatten_function): Do not overwrite final inlining
11693 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
11695 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
11696 support for merge[hl].
11697 (fold_mergehl_helper): New helper function.
11698 (tree-vector-builder.h): New #include for tree_vector_builder usage.
11699 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
11700 (altivec_vmrglw_direct): Add xxmrglw insn.
11702 2018-01-17 Andrew Waterman <andrew@sifive.com>
11704 * config/riscv/riscv.c (riscv_conditional_register_usage): If
11705 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
11707 2018-01-17 David Malcolm <dmalcolm@redhat.com>
11710 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
11711 call the lto_location_cache before reading the
11712 DECL_SOURCE_LOCATION of the types.
11714 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
11715 Richard Sandiford <richard.sandiford@linaro.org>
11717 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
11718 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
11719 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
11720 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
11721 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
11723 * config/aarch64/constraints.md (aarch64_movti_operand):
11725 * config/aarch64/predicates.md (Uti): Add new constraint.
11727 2018-01-17 Carl Love <cel@us.ibm.com>
11729 * config/rs6000/vsx.md (define_expand xl_len_r,
11730 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
11731 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
11733 (define_expand, define_insn): Move the shift left from the
11734 define_insn to the define_expand for lxvl and stxvl instructions.
11735 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
11736 and XL_LEN_R definitions to PURE.
11738 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
11740 * config/i386/i386.c (indirect_thunk_name): Declare regno
11741 as unsigned int. Compare regno with INVALID_REGNUM.
11742 (output_indirect_thunk): Ditto.
11743 (output_indirect_thunk_function): Ditto.
11744 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
11745 in the call to output_indirect_thunk_function.
11747 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
11749 PR middle-end/83884
11750 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
11751 rather than the size of inner_type to determine the stack slot size
11752 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
11754 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
11757 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
11760 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
11762 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
11763 endian Linux systems to optionally enable multilibs for selecting
11764 the long double type if the user configured an explicit type.
11765 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
11766 have no long double multilibs if not defined.
11767 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
11768 warn if the user used -mabi={ieee,ibm}longdouble and we built
11769 multilibs for long double.
11770 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
11771 appropriate multilib option.
11772 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
11774 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
11775 for building long double multilibs.
11776 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
11778 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
11780 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
11783 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
11785 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
11788 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
11791 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
11794 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11796 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
11797 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
11799 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
11801 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
11802 different rtl trees depending on TARGET_64BIT.
11803 (rs6000_gen_lvx): Likewise.
11805 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11807 * config/visium/visium.md (nop): Tweak comment.
11808 (hazard_nop): Likewise.
11810 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11812 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
11813 -mspeculate-indirect-jumps.
11814 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
11815 for -mno-speculate-indirect-jumps.
11816 (*call_indirect_elfv2<mode>_nospec): New define_insn.
11817 (*call_value_indirect_elfv2<mode>): Disable for
11818 -mno-speculate-indirect-jumps.
11819 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
11820 (indirect_jump): Emit different RTL for
11821 -mno-speculate-indirect-jumps.
11822 (*indirect_jump<mode>): Disable for
11823 -mno-speculate-indirect-jumps.
11824 (*indirect_jump<mode>_nospec): New define_insn.
11825 (tablejump): Emit different RTL for
11826 -mno-speculate-indirect-jumps.
11827 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
11828 (tablejumpsi_nospec): New define_expand.
11829 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
11830 (tablejumpdi_nospec): New define_expand.
11831 (*tablejump<mode>_internal1): Disable for
11832 -mno-speculate-indirect-jumps.
11833 (*tablejump<mode>_internal1_nospec): New define_insn.
11834 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
11837 2018-01-16 Artyom Skrobov tyomitch@gmail.com
11839 * caller-save.c (insert_save): Drop unnecessary parameter. All
11842 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11843 Richard Biener <rguenth@suse.de>
11846 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
11847 return early, inline manually is_gimple_sizepos. Make sure if we
11848 call gimplify_expr we don't end up with a gimple constant.
11849 * tree.c (variably_modified_type_p): Don't return true for
11850 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
11851 * gimplify.h (is_gimple_sizepos): Remove.
11853 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11855 PR tree-optimization/83857
11856 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
11857 vectorizable_live_operation for pure SLP statements.
11858 (vectorizable_live_operation): Handle PHIs.
11860 2018-01-16 Richard Biener <rguenther@suse.de>
11862 PR tree-optimization/83867
11863 * tree-vect-stmts.c (vect_transform_stmt): Precompute
11864 nested_in_vect_loop_p since the scalar stmt may get invalidated.
11866 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11869 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
11870 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
11871 If off is not INTEGER_CST, issue a may not be aligned warning
11872 rather than isn't aligned. Use isn%'t rather than isn't.
11873 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
11875 <case MULT_EXPR>: Improve the case when bottom and one of the
11876 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
11877 operand, in that case check if the other operand is multiple of
11878 bottom divided by the INTEGER_CST operand.
11880 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11883 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
11884 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
11885 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
11886 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
11887 * config/pa/pa.c (pa_function_arg_advance): Likewise.
11888 (pa_function_arg, pa_arg_partial_bytes): Likewise.
11889 (pa_function_arg_size): New function.
11891 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11893 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
11894 in a separate statement.
11896 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11898 PR tree-optimization/83847
11899 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
11900 group gathers and scatters.
11902 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11904 PR rtl-optimization/86620
11905 * params.def (max-sched-ready-insns): Bump minimum value to 1.
11907 PR rtl-optimization/83213
11908 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
11909 to last if both are JUMP_INSNs.
11911 PR tree-optimization/83843
11912 * gimple-ssa-store-merging.c
11913 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
11914 store_immediate_info for bswap/nop orig_stores.
11916 2018-01-15 Andrew Waterman <andrew@sifive.com>
11918 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
11920 <UDIV>: Increase cost if !TARGET_DIV.
11922 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
11924 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
11925 (define_attr "cr_logical_3op"): New.
11926 (cceq_ior_compare): Adjust.
11927 (cceq_ior_compare_complement): Adjust.
11928 (*cceq_rev_compare): Adjust.
11929 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
11930 (is_cracked_insn): Adjust.
11931 (insn_must_be_first_in_group): Adjust.
11932 * config/rs6000/40x.md: Adjust.
11933 * config/rs6000/440.md: Adjust.
11934 * config/rs6000/476.md: Adjust.
11935 * config/rs6000/601.md: Adjust.
11936 * config/rs6000/603.md: Adjust.
11937 * config/rs6000/6xx.md: Adjust.
11938 * config/rs6000/7450.md: Adjust.
11939 * config/rs6000/7xx.md: Adjust.
11940 * config/rs6000/8540.md: Adjust.
11941 * config/rs6000/cell.md: Adjust.
11942 * config/rs6000/e300c2c3.md: Adjust.
11943 * config/rs6000/e500mc.md: Adjust.
11944 * config/rs6000/e500mc64.md: Adjust.
11945 * config/rs6000/e5500.md: Adjust.
11946 * config/rs6000/e6500.md: Adjust.
11947 * config/rs6000/mpc.md: Adjust.
11948 * config/rs6000/power4.md: Adjust.
11949 * config/rs6000/power5.md: Adjust.
11950 * config/rs6000/power6.md: Adjust.
11951 * config/rs6000/power7.md: Adjust.
11952 * config/rs6000/power8.md: Adjust.
11953 * config/rs6000/power9.md: Adjust.
11954 * config/rs6000/rs64.md: Adjust.
11955 * config/rs6000/titan.md: Adjust.
11957 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11959 * config/i386/predicates.md (indirect_branch_operand): Rewrite
11960 ix86_indirect_branch_register logic.
11962 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11964 * config/i386/constraints.md (Bs): Update
11965 ix86_indirect_branch_register check. Don't check
11966 ix86_indirect_branch_register with GOT_memory_operand.
11968 * config/i386/predicates.md (GOT_memory_operand): Don't check
11969 ix86_indirect_branch_register here.
11970 (GOT32_symbol_operand): Likewise.
11972 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11974 * config/i386/predicates.md (constant_call_address_operand):
11975 Rewrite ix86_indirect_branch_register logic.
11976 (sibcall_insn_operand): Likewise.
11978 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11980 * config/i386/constraints.md (Bs): Replace
11981 ix86_indirect_branch_thunk_register with
11982 ix86_indirect_branch_register.
11984 * config/i386/i386.md (indirect_jump): Likewise.
11985 (tablejump): Likewise.
11986 (*sibcall_memory): Likewise.
11987 (*sibcall_value_memory): Likewise.
11988 Peepholes of indirect call and jump via memory: Likewise.
11989 * config/i386/i386.opt: Likewise.
11990 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11991 (GOT_memory_operand): Likewise.
11992 (call_insn_operand): Likewise.
11993 (sibcall_insn_operand): Likewise.
11994 (GOT32_symbol_operand): Likewise.
11996 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11998 PR middle-end/83837
11999 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
12000 type rather than type addr's type points to.
12001 (expand_omp_atomic_mutex): Likewise.
12002 (expand_omp_atomic): Likewise.
12004 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12007 * config/i386/i386.c (output_indirect_thunk_function): Use
12008 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
12009 for __x86_return_thunk.
12011 2018-01-15 Richard Biener <rguenther@suse.de>
12013 PR middle-end/83850
12014 * expmed.c (extract_bit_field_1): Fix typo.
12016 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12019 * config/arm/iterators.md (VF): New mode iterator.
12020 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
12021 Remove integer-related logic from pattern.
12022 (neon_vabd<mode>_3): Likewise.
12024 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12026 PR middle-end/82694
12027 * common.opt (fstrict-overflow): No longer an alias.
12028 (fwrapv-pointer): New option.
12029 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
12030 also for pointer types based on flag_wrapv_pointer.
12031 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
12032 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
12033 opts->x_flag_wrapv got set.
12034 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
12035 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
12036 POINTER_TYPE_OVERFLOW_UNDEFINED.
12037 * match.pd: Likewise in address comparison pattern.
12038 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
12040 2018-01-15 Richard Biener <rguenther@suse.de>
12043 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
12044 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
12045 Reset type names to their identifier if their TYPE_DECL doesn't
12046 have linkage (and thus is used for ODR and devirt).
12047 (save_debug_info_for_decl): Remove.
12048 (save_debug_info_for_type): Likewise.
12049 (add_tree_to_fld_list): Adjust.
12050 * tree-pretty-print.c (dump_generic_node): Make dumping of
12051 type names more robust.
12053 2018-01-15 Richard Biener <rguenther@suse.de>
12055 * BASE-VER: Bump to 8.0.1.
12057 2018-01-14 Martin Sebor <msebor@redhat.com>
12060 * builtins.c (check_access): Avoid warning when the no-warning bit
12063 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
12065 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
12066 * ira-color (allocno_hard_regs_compare): Likewise.
12068 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
12071 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
12072 Use .pushsection/.popsection.
12074 2018-01-14 Martin Sebor <msebor@redhat.com>
12077 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
12079 2018-01-14 Jakub Jelinek <jakub@redhat.com>
12081 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
12082 entry from extra_headers.
12083 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12084 extra_headers, make the list bitwise identical to the i?86-*-* one.
12086 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12088 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12089 -mcmodel=large with -mindirect-branch=thunk,
12090 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12091 -mfunction-return=thunk-extern.
12092 * doc/invoke.texi: Document -mcmodel=large is incompatible with
12093 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12094 -mfunction-return=thunk and -mfunction-return=thunk-extern.
12096 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12098 * config/i386/i386.c (print_reg): Print the name of the full
12099 integer register without '%'.
12100 (ix86_print_operand): Handle 'V'.
12101 * doc/extend.texi: Document 'V' modifier.
12103 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12105 * config/i386/constraints.md (Bs): Disallow memory operand for
12106 -mindirect-branch-register.
12108 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12109 (GOT_memory_operand): Likewise.
12110 (call_insn_operand): Likewise.
12111 (sibcall_insn_operand): Likewise.
12112 (GOT32_symbol_operand): Likewise.
12113 * config/i386/i386.md (indirect_jump): Call convert_memory_address
12114 for -mindirect-branch-register.
12115 (tablejump): Likewise.
12116 (*sibcall_memory): Likewise.
12117 (*sibcall_value_memory): Likewise.
12118 Disallow peepholes of indirect call and jump via memory for
12119 -mindirect-branch-register.
12120 (*call_pop): Replace m with Bw.
12121 (*call_value_pop): Likewise.
12122 (*sibcall_pop_memory): Replace m with Bs.
12123 * config/i386/i386.opt (mindirect-branch-register): New option.
12124 * doc/invoke.texi: Document -mindirect-branch-register option.
12126 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12128 * config/i386/i386-protos.h (ix86_output_function_return): New.
12129 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12130 set function_return_type.
12131 (indirect_thunk_name): Add ret_p to indicate thunk for function
12133 (output_indirect_thunk_function): Pass false to
12134 indirect_thunk_name.
12135 (ix86_output_indirect_branch_via_reg): Likewise.
12136 (ix86_output_indirect_branch_via_push): Likewise.
12137 (output_indirect_thunk_function): Create alias for function
12138 return thunk if regno < 0.
12139 (ix86_output_function_return): New function.
12140 (ix86_handle_fndecl_attribute): Handle function_return.
12141 (ix86_attribute_table): Add function_return.
12142 * config/i386/i386.h (machine_function): Add
12143 function_return_type.
12144 * config/i386/i386.md (simple_return_internal): Use
12145 ix86_output_function_return.
12146 (simple_return_internal_long): Likewise.
12147 * config/i386/i386.opt (mfunction-return=): New option.
12148 (indirect_branch): Mention -mfunction-return=.
12149 * doc/extend.texi: Document function_return function attribute.
12150 * doc/invoke.texi: Document -mfunction-return= option.
12152 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12154 * config/i386/i386-opts.h (indirect_branch): New.
12155 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12156 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12157 with local indirect jump when converting indirect call and jump.
12158 (ix86_set_indirect_branch_type): New.
12159 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12160 (indirectlabelno): New.
12161 (indirect_thunk_needed): Likewise.
12162 (indirect_thunk_bnd_needed): Likewise.
12163 (indirect_thunks_used): Likewise.
12164 (indirect_thunks_bnd_used): Likewise.
12165 (INDIRECT_LABEL): Likewise.
12166 (indirect_thunk_name): Likewise.
12167 (output_indirect_thunk): Likewise.
12168 (output_indirect_thunk_function): Likewise.
12169 (ix86_output_indirect_branch_via_reg): Likewise.
12170 (ix86_output_indirect_branch_via_push): Likewise.
12171 (ix86_output_indirect_branch): Likewise.
12172 (ix86_output_indirect_jmp): Likewise.
12173 (ix86_code_end): Call output_indirect_thunk_function if needed.
12174 (ix86_output_call_insn): Call ix86_output_indirect_branch if
12176 (ix86_handle_fndecl_attribute): Handle indirect_branch.
12177 (ix86_attribute_table): Add indirect_branch.
12178 * config/i386/i386.h (machine_function): Add indirect_branch_type
12179 and has_local_indirect_jump.
12180 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12182 (tablejump): Likewise.
12183 (*indirect_jump): Use ix86_output_indirect_jmp.
12184 (*tablejump_1): Likewise.
12185 (simple_return_indirect_internal): Likewise.
12186 * config/i386/i386.opt (mindirect-branch=): New option.
12187 (indirect_branch): New.
12190 (thunk-inline): Likewise.
12191 (thunk-extern): Likewise.
12192 * doc/extend.texi: Document indirect_branch function attribute.
12193 * doc/invoke.texi: Document -mindirect-branch= option.
12195 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
12198 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12200 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
12202 * ipa-inline.c (want_inline_small_function_p): Return false if
12203 inlining has already failed with CIF_FINAL_ERROR.
12204 (update_caller_keys): Call want_inline_small_function_p before
12206 (update_callee_keys): Likewise.
12208 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12210 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12212 (rs6000_quadword_masked_address_p): Likewise.
12213 (quad_aligned_load_p): Likewise.
12214 (quad_aligned_store_p): Likewise.
12215 (const_load_sequence_p): Add comment to describe the outer-most loop.
12216 (mimic_memory_attributes_and_flags): New function.
12217 (rs6000_gen_stvx): Likewise.
12218 (replace_swapped_aligned_store): Likewise.
12219 (rs6000_gen_lvx): Likewise.
12220 (replace_swapped_aligned_load): Likewise.
12221 (replace_swapped_load_constant): Capitalize argument name in
12222 comment describing this function.
12223 (rs6000_analyze_swaps): Add a third pass to search for vector loads
12224 and stores that access quad-word aligned addresses and replace
12225 with stvx or lvx instructions when appropriate.
12226 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12227 New function prototype.
12228 (rs6000_quadword_masked_address_p): Likewise.
12229 (rs6000_gen_lvx): Likewise.
12230 (rs6000_gen_stvx): Likewise.
12231 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12232 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12233 when memory address is aligned.
12234 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12235 this split to select lvx instruction when memory address is aligned.
12236 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12237 instruction when memory address is aligned.
12238 (*vsx_le_perm_load_v16qi): Likewise.
12239 (four unnamed splitters): Modify to select the stvx instruction
12240 when memory is aligned.
12242 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
12244 * predict.c (determine_unlikely_bbs): Handle correctly BBs
12245 which appears in the queue multiple times.
12247 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12248 Alan Hayward <alan.hayward@arm.com>
12249 David Sherwood <david.sherwood@arm.com>
12251 * tree-vectorizer.h (vec_lower_bound): New structure.
12252 (_loop_vec_info): Add check_nonzero and lower_bounds.
12253 (LOOP_VINFO_CHECK_NONZERO): New macro.
12254 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
12255 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
12256 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
12257 fields. Make seg_len the distance travelled, not including the
12259 (dr_direction_indicator): Declare.
12260 (dr_zero_step_indicator): Likewise.
12261 (dr_known_forward_stride_p): Likewise.
12262 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
12264 (runtime_alias_check_p): Allow runtime alias checks with
12266 (operator ==): Compare access_size and align.
12267 (prune_runtime_alias_test_list): Rework for new distinction between
12268 the access_size and seg_len.
12269 (create_intersect_range_checks_index): Likewise. Cope with polynomial
12271 (get_segment_min_max): New function.
12272 (create_intersect_range_checks): Use it.
12273 (dr_step_indicator): New function.
12274 (dr_direction_indicator): Likewise.
12275 (dr_zero_step_indicator): Likewise.
12276 (dr_known_forward_stride_p): Likewise.
12277 * tree-loop-distribution.c (data_ref_segment_size): Return
12278 DR_STEP * (niters - 1).
12279 (compute_alias_check_pairs): Update call to the dr_with_seg_len
12281 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
12282 (vect_preserves_scalar_order_p): New function, split out from...
12283 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
12284 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
12285 (vect_vfa_access_size): New function.
12286 (vect_vfa_align): Likewise.
12287 (vect_compile_time_alias): Take access_size_a and access_b arguments.
12288 (dump_lower_bound): New function.
12289 (vect_check_lower_bound): Likewise.
12290 (vect_small_gap_p): Likewise.
12291 (vectorizable_with_step_bound_p): Likewise.
12292 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
12293 depencies if the vectorization factor is 1. Convert the checks
12294 for nonzero steps into checks on the bounds of DR_STEP. Try using
12295 a bunds check for variable steps if the minimum required step is
12296 relatively small. Update calls to the dr_with_seg_len
12297 constructor and to vect_compile_time_alias.
12298 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
12300 (vect_loop_versioning): Call it.
12301 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
12303 (vect_estimate_min_profitable_iters): Account for any bounds checks.
12305 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12306 Alan Hayward <alan.hayward@arm.com>
12307 David Sherwood <david.sherwood@arm.com>
12309 * doc/sourcebuild.texi (vect_scatter_store): Document.
12310 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
12312 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
12314 * genopinit.c (main): Add supports_vec_scatter_store and
12315 supports_vec_scatter_store_cached to target_optabs.
12316 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
12317 IFN_MASK_SCATTER_STORE.
12318 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
12320 * internal-fn.h (internal_store_fn_p): Declare.
12321 (internal_fn_stored_value_index): Likewise.
12322 * internal-fn.c (scatter_store_direct): New macro.
12323 (expand_scatter_store_optab_fn): New function.
12324 (direct_scatter_store_optab_supported_p): New macro.
12325 (internal_store_fn_p): New function.
12326 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
12327 IFN_MASK_SCATTER_STORE.
12328 (internal_fn_mask_index): Likewise.
12329 (internal_fn_stored_value_index): New function.
12330 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
12331 for scatter stores.
12332 * optabs-query.h (supports_vec_scatter_store_p): Declare.
12333 * optabs-query.c (supports_vec_scatter_store_p): New function.
12334 * tree-vectorizer.h (vect_get_store_rhs): Declare.
12335 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
12336 true for scatter stores.
12337 (vect_gather_scatter_fn_p): Handle scatter stores too.
12338 (vect_check_gather_scatter): Consider using scatter stores if
12339 supports_vec_scatter_store_p.
12340 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
12341 scatter stores too.
12342 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12343 internal_fn_stored_value_index.
12344 (check_load_store_masking): Handle scatter stores too.
12345 (vect_get_store_rhs): Make public.
12346 (vectorizable_call): Use internal_store_fn_p.
12347 (vectorizable_store): Handle scatter store internal functions.
12348 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
12349 when deciding whether the end of the group has been reached.
12350 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
12351 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
12352 (mask_scatter_store<mode>): New insns.
12354 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12355 Alan Hayward <alan.hayward@arm.com>
12356 David Sherwood <david.sherwood@arm.com>
12358 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
12359 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
12360 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
12362 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
12363 Use vect_truncate_gather_scatter_offset if we can't treat the
12364 operation as a normal gather load or scatter store.
12365 (get_group_load_store_type): Take the gather_scatter_info
12366 as argument. Try using a gather load or scatter store for
12367 single-element groups.
12368 (get_load_store_type): Update calls to get_group_load_store_type
12369 and vect_use_strided_gather_scatters_p.
12371 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12372 Alan Hayward <alan.hayward@arm.com>
12373 David Sherwood <david.sherwood@arm.com>
12375 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
12376 optional tree argument.
12377 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
12379 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
12380 but continue to use the current value as a fallback.
12381 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
12382 to compare the updates.
12383 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
12384 (get_load_store_type): Use it when handling a strided access.
12385 (vect_get_strided_load_store_ops): New function.
12386 (vect_get_data_ptr_increment): Likewise.
12387 (vectorizable_load): Handle strided gather loads. Always pass
12388 a step to vect_create_data_ref_ptr and bump_vector_ptr.
12390 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12391 Alan Hayward <alan.hayward@arm.com>
12392 David Sherwood <david.sherwood@arm.com>
12394 * doc/md.texi (gather_load@var{m}): Document.
12395 (mask_gather_load@var{m}): Likewise.
12396 * genopinit.c (main): Add supports_vec_gather_load and
12397 supports_vec_gather_load_cached to target_optabs.
12398 * optabs-tree.c (init_tree_optimization_optabs): Use
12399 ggc_cleared_alloc to allocate target_optabs.
12400 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
12401 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
12403 * internal-fn.h (internal_load_fn_p): Declare.
12404 (internal_gather_scatter_fn_p): Likewise.
12405 (internal_fn_mask_index): Likewise.
12406 (internal_gather_scatter_fn_supported_p): Likewise.
12407 * internal-fn.c (gather_load_direct): New macro.
12408 (expand_gather_load_optab_fn): New function.
12409 (direct_gather_load_optab_supported_p): New macro.
12410 (direct_internal_fn_optab): New function.
12411 (internal_load_fn_p): Likewise.
12412 (internal_gather_scatter_fn_p): Likewise.
12413 (internal_fn_mask_index): Likewise.
12414 (internal_gather_scatter_fn_supported_p): Likewise.
12415 * optabs-query.c (supports_at_least_one_mode_p): New function.
12416 (supports_vec_gather_load_p): Likewise.
12417 * optabs-query.h (supports_vec_gather_load_p): Declare.
12418 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
12419 and memory_type field.
12420 (NUM_PATTERNS): Bump to 15.
12421 * tree-vect-data-refs.c: Include internal-fn.h.
12422 (vect_gather_scatter_fn_p): New function.
12423 (vect_describe_gather_scatter_call): Likewise.
12424 (vect_check_gather_scatter): Try using internal functions for
12425 gather loads. Recognize existing calls to a gather load function.
12426 (vect_analyze_data_refs): Consider using gather loads if
12427 supports_vec_gather_load_p.
12428 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
12429 (vect_get_gather_scatter_offset_type): Likewise.
12430 (vect_convert_mask_for_vectype): Likewise.
12431 (vect_add_conversion_to_patterm): Likewise.
12432 (vect_try_gather_scatter_pattern): Likewise.
12433 (vect_recog_gather_scatter_pattern): New pattern recognizer.
12434 (vect_vect_recog_func_ptrs): Add it.
12435 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12436 internal_fn_mask_index and internal_gather_scatter_fn_p.
12437 (check_load_store_masking): Take the gather_scatter_info as an
12438 argument and handle gather loads.
12439 (vect_get_gather_scatter_ops): New function.
12440 (vectorizable_call): Check internal_load_fn_p.
12441 (vectorizable_load): Likewise. Handle gather load internal
12443 (vectorizable_store): Update call to check_load_store_masking.
12444 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
12445 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
12446 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
12447 (aarch64_gather_scale_operand_d): New predicates.
12448 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
12449 (mask_gather_load<mode>): New insns.
12451 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12452 Alan Hayward <alan.hayward@arm.com>
12453 David Sherwood <david.sherwood@arm.com>
12455 * optabs.def (fold_left_plus_optab): New optab.
12456 * doc/md.texi (fold_left_plus_@var{m}): Document.
12457 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
12458 * internal-fn.c (fold_left_direct): Define.
12459 (expand_fold_left_optab_fn): Likewise.
12460 (direct_fold_left_optab_supported_p): Likewise.
12461 * fold-const-call.c (fold_const_fold_left): New function.
12462 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
12463 * tree-parloops.c (valid_reduction_p): New function.
12464 (gather_scalar_reductions): Use it.
12465 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
12466 (vect_finish_replace_stmt): Declare.
12467 * tree-vect-loop.c (fold_left_reduction_fn): New function.
12468 (needs_fold_left_reduction_p): New function, split out from...
12469 (vect_is_simple_reduction): ...here. Accept reductions that
12470 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
12471 (vect_force_simple_reduction): Also store the reduction type in
12472 the assignment's STMT_VINFO_REDUC_TYPE.
12473 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
12474 (merge_with_identity): New function.
12475 (vect_expand_fold_left): Likewise.
12476 (vectorize_fold_left_reduction): Likewise.
12477 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
12478 scalar phi in place for it. Check for target support and reject
12479 cases that would reassociate the operation. Defer the transform
12480 phase to vectorize_fold_left_reduction.
12481 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
12482 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
12483 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
12485 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12487 * tree-if-conv.c (predicate_mem_writes): Remove redundant
12488 call to ifc_temp_var.
12490 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12491 Alan Hayward <alan.hayward@arm.com>
12492 David Sherwood <david.sherwood@arm.com>
12494 * target.def (legitimize_address_displacement): Take the original
12495 offset as a poly_int.
12496 * targhooks.h (default_legitimize_address_displacement): Update
12498 * targhooks.c (default_legitimize_address_displacement): Likewise.
12499 * doc/tm.texi: Regenerate.
12500 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
12501 as an argument, moving assert of ad->disp == ad->disp_term to...
12502 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
12503 Try calling targetm.legitimize_address_displacement before expanding
12504 the address rather than afterwards, and adjust for the new interface.
12505 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
12506 Match the new hook interface. Handle SVE addresses.
12507 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
12508 new hook interface.
12510 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12512 * Makefile.in (OBJS): Add early-remat.o.
12513 * target.def (select_early_remat_modes): New hook.
12514 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
12515 * doc/tm.texi: Regenerate.
12516 * targhooks.h (default_select_early_remat_modes): Declare.
12517 * targhooks.c (default_select_early_remat_modes): New function.
12518 * timevar.def (TV_EARLY_REMAT): New timevar.
12519 * passes.def (pass_early_remat): New pass.
12520 * tree-pass.h (make_pass_early_remat): Declare.
12521 * early-remat.c: New file.
12522 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
12524 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
12526 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12527 Alan Hayward <alan.hayward@arm.com>
12528 David Sherwood <david.sherwood@arm.com>
12530 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
12531 vfm1 with a bound_epilog parameter.
12532 (vect_do_peeling): Update calls accordingly, and move the prologue
12533 call earlier in the function. Treat the base bound_epilog as 0 for
12534 fully-masked loops and retain vf - 1 for other loops. Add 1 to
12535 this base when peeling for gaps.
12536 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
12537 with fully-masked loops.
12538 (vect_estimate_min_profitable_iters): Handle the single peeled
12539 iteration in that case.
12541 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12542 Alan Hayward <alan.hayward@arm.com>
12543 David Sherwood <david.sherwood@arm.com>
12545 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
12546 single-element interleaving even if the size is not a power of 2.
12547 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
12548 accesses for single-element interleaving if the group size is
12551 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12552 Alan Hayward <alan.hayward@arm.com>
12553 David Sherwood <david.sherwood@arm.com>
12555 * doc/md.texi (fold_extract_last_@var{m}): Document.
12556 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
12557 * optabs.def (fold_extract_last_optab): New optab.
12558 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
12559 * internal-fn.c (fold_extract_direct): New macro.
12560 (expand_fold_extract_optab_fn): Likewise.
12561 (direct_fold_extract_optab_supported_p): Likewise.
12562 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
12563 * tree-vect-loop.c (vect_model_reduction_cost): Handle
12564 EXTRACT_LAST_REDUCTION.
12565 (get_initial_def_for_reduction): Do not create an initial vector
12566 for EXTRACT_LAST_REDUCTION reductions.
12567 (vectorizable_reduction): Leave the scalar phi in place for
12568 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
12569 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
12570 epilogue code for EXTRACT_LAST_REDUCTION and defer the
12571 transform phase to vectorizable_condition.
12572 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
12574 (vect_finish_stmt_generation): ...here.
12575 (vect_finish_replace_stmt): New function.
12576 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
12577 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
12579 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
12581 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12582 Alan Hayward <alan.hayward@arm.com>
12583 David Sherwood <david.sherwood@arm.com>
12585 * doc/md.texi (extract_last_@var{m}): Document.
12586 * optabs.def (extract_last_optab): New optab.
12587 * internal-fn.def (EXTRACT_LAST): New internal function.
12588 * internal-fn.c (cond_unary_direct): New macro.
12589 (expand_cond_unary_optab_fn): Likewise.
12590 (direct_cond_unary_optab_supported_p): Likewise.
12591 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
12592 loops using EXTRACT_LAST.
12593 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
12594 (extract_last_<mode>): ...this optab.
12595 (vec_extract<mode><Vel>): Update accordingly.
12597 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12598 Alan Hayward <alan.hayward@arm.com>
12599 David Sherwood <david.sherwood@arm.com>
12601 * target.def (empty_mask_is_expensive): New hook.
12602 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
12603 * doc/tm.texi: Regenerate.
12604 * targhooks.h (default_empty_mask_is_expensive): Declare.
12605 * targhooks.c (default_empty_mask_is_expensive): New function.
12606 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
12607 if the target says that empty masks are expensive.
12608 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
12610 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
12612 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12613 Alan Hayward <alan.hayward@arm.com>
12614 David Sherwood <david.sherwood@arm.com>
12616 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
12617 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
12618 (vect_use_loop_mask_for_alignment_p): New function.
12619 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
12620 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
12621 niters_skip argument. Make sure that the first niters_skip elements
12622 of the first iteration are inactive.
12623 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
12624 Update call to vect_set_loop_masks_directly.
12625 (get_misalign_in_elems): New function, split out from...
12626 (vect_gen_prolog_loop_niters): ...here.
12627 (vect_update_init_of_dr): Take a code argument that specifies whether
12628 the adjustment should be added or subtracted.
12629 (vect_update_init_of_drs): Likewise.
12630 (vect_prepare_for_masked_peels): New function.
12631 (vect_do_peeling): Skip prologue peeling if we're using a mask
12632 instead. Update call to vect_update_inits_of_drs.
12633 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12635 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
12636 alignment. Do not include the number of peeled iterations in
12637 the minimum threshold in that case.
12638 (vectorizable_induction): Adjust the start value down by
12639 LOOP_VINFO_MASK_SKIP_NITERS iterations.
12640 (vect_transform_loop): Call vect_prepare_for_masked_peels.
12641 Take the number of skipped iterations into account when calculating
12643 * tree-vect-stmts.c (vect_gen_while_not): New function.
12645 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12646 Alan Hayward <alan.hayward@arm.com>
12647 David Sherwood <david.sherwood@arm.com>
12649 * doc/sourcebuild.texi (vect_fully_masked): Document.
12650 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
12651 default value to 0.
12652 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
12654 (vect_analyze_loop_2): ...here. Don't check the vectorization
12655 factor against the number of loop iterations if the loop is
12658 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12659 Alan Hayward <alan.hayward@arm.com>
12660 David Sherwood <david.sherwood@arm.com>
12662 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
12663 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
12664 (dump_groups): Update accordingly.
12665 (iv_use::mem_type): New member variable.
12666 (address_p): New function.
12667 (record_use): Add a mem_type argument and initialize the new
12669 (record_group_use): Add a mem_type argument. Use address_p.
12670 Remove obsolete null checks of base_object. Update call to record_use.
12671 (find_interesting_uses_op): Update call to record_group_use.
12672 (find_interesting_uses_cond): Likewise.
12673 (find_interesting_uses_address): Likewise.
12674 (get_mem_type_for_internal_fn): New function.
12675 (find_address_like_use): Likewise.
12676 (find_interesting_uses_stmt): Try find_address_like_use before
12677 calling find_interesting_uses_op.
12678 (addr_offset_valid_p): Use the iv mem_type field as the type
12679 of the addressed memory.
12680 (add_autoinc_candidates): Likewise.
12681 (get_address_cost): Likewise.
12682 (split_small_address_groups_p): Use address_p.
12683 (split_address_groups): Likewise.
12684 (add_iv_candidate_for_use): Likewise.
12685 (autoinc_possible_for_pair): Likewise.
12686 (rewrite_groups): Likewise.
12687 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
12688 (determine_group_iv_cost): Update after split of USE_ADDRESS.
12689 (get_alias_ptr_type_for_ptr_address): New function.
12690 (rewrite_use_address): Rewrite address uses in calls that were
12691 identified by find_address_like_use.
12693 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12694 Alan Hayward <alan.hayward@arm.com>
12695 David Sherwood <david.sherwood@arm.com>
12697 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
12699 * gimple-expr.h (is_gimple_addressable: Likewise.
12700 * gimple-expr.c (is_gimple_address): Likewise.
12701 * internal-fn.c (expand_call_mem_ref): New function.
12702 (expand_mask_load_optab_fn): Use it.
12703 (expand_mask_store_optab_fn): Likewise.
12705 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12706 Alan Hayward <alan.hayward@arm.com>
12707 David Sherwood <david.sherwood@arm.com>
12709 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
12710 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
12711 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
12712 (cond_umax@var{mode}): Document.
12713 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
12714 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
12715 (cond_umin_optab, cond_umax_optab): New optabs.
12716 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
12717 (COND_IOR, COND_XOR): New internal functions.
12718 * internal-fn.h (get_conditional_internal_fn): Declare.
12719 * internal-fn.c (cond_binary_direct): New macro.
12720 (expand_cond_binary_optab_fn): Likewise.
12721 (direct_cond_binary_optab_supported_p): Likewise.
12722 (get_conditional_internal_fn): New function.
12723 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
12724 Cope with reduction statements that are vectorized as calls rather
12726 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
12727 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
12728 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
12729 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12730 (UNSPEC_COND_EOR): New unspecs.
12731 (optab): Add mappings for them.
12732 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
12733 (sve_int_op, sve_fp_op): New int attributes.
12735 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12736 Alan Hayward <alan.hayward@arm.com>
12737 David Sherwood <david.sherwood@arm.com>
12739 * optabs.def (while_ult_optab): New optab.
12740 * doc/md.texi (while_ult@var{m}@var{n}): Document.
12741 * internal-fn.def (WHILE_ULT): New internal function.
12742 * internal-fn.h (direct_internal_fn_supported_p): New override
12743 that takes two types as argument.
12744 * internal-fn.c (while_direct): New macro.
12745 (expand_while_optab_fn): New function.
12746 (convert_optab_supported_p): Likewise.
12747 (direct_while_optab_supported_p): New macro.
12748 * wide-int.h (wi::udiv_ceil): New function.
12749 * tree-vectorizer.h (rgroup_masks): New structure.
12750 (vec_loop_masks): New typedef.
12751 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
12752 and fully_masked_p.
12753 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
12754 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
12755 (vect_max_vf): New function.
12756 (slpeel_make_loop_iterate_ntimes): Delete.
12757 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
12758 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
12759 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
12760 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
12761 internal-fn.h, stor-layout.h and optabs-query.h.
12762 (vect_set_loop_mask): New function.
12763 (add_preheader_seq): Likewise.
12764 (add_header_seq): Likewise.
12765 (interleave_supported_p): Likewise.
12766 (vect_maybe_permute_loop_masks): Likewise.
12767 (vect_set_loop_masks_directly): Likewise.
12768 (vect_set_loop_condition_masked): Likewise.
12769 (vect_set_loop_condition_unmasked): New function, split out from
12770 slpeel_make_loop_iterate_ntimes.
12771 (slpeel_make_loop_iterate_ntimes): Rename to..
12772 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
12773 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
12774 (vect_do_peeling): Update call accordingly.
12775 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
12777 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12778 mask_compare_type, can_fully_mask_p and fully_masked_p.
12779 (release_vec_loop_masks): New function.
12780 (_loop_vec_info): Use it to free the loop masks.
12781 (can_produce_all_loop_masks_p): New function.
12782 (vect_get_max_nscalars_per_iter): Likewise.
12783 (vect_verify_full_masking): Likewise.
12784 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
12785 retries, and free the mask rgroups before retrying. Check loop-wide
12786 reasons for disallowing fully-masked loops. Make the final decision
12787 about whether use a fully-masked loop or not.
12788 (vect_estimate_min_profitable_iters): Do not assume that peeling
12789 for the number of iterations will be needed for fully-masked loops.
12790 (vectorizable_reduction): Disable fully-masked loops.
12791 (vectorizable_live_operation): Likewise.
12792 (vect_halve_mask_nunits): New function.
12793 (vect_double_mask_nunits): Likewise.
12794 (vect_record_loop_mask): Likewise.
12795 (vect_get_loop_mask): Likewise.
12796 (vect_transform_loop): Handle the case in which the final loop
12797 iteration might handle a partial vector. Call vect_set_loop_condition
12798 instead of slpeel_make_loop_iterate_ntimes.
12799 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
12800 (check_load_store_masking): New function.
12801 (prepare_load_store_mask): Likewise.
12802 (vectorizable_store): Handle fully-masked loops.
12803 (vectorizable_load): Likewise.
12804 (supportable_widening_operation): Use vect_halve_mask_nunits for
12806 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
12807 (vect_gen_while): New function.
12808 * config/aarch64/aarch64.md (umax<mode>3): New expander.
12809 (aarch64_uqdec<mode>): New insn.
12811 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12812 Alan Hayward <alan.hayward@arm.com>
12813 David Sherwood <david.sherwood@arm.com>
12815 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
12816 (reduc_xor_scal_optab): New optabs.
12817 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
12818 (reduc_xor_scal_@var{m}): Document.
12819 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
12820 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
12821 internal functions.
12822 * fold-const-call.c (fold_const_call): Handle them.
12823 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
12824 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
12825 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
12826 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
12827 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
12828 (UNSPEC_XORV): New unspecs.
12829 (optab): Add entries for them.
12830 (BITWISEV): New int iterator.
12831 (bit_reduc_op): New int attributes.
12833 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12834 Alan Hayward <alan.hayward@arm.com>
12835 David Sherwood <david.sherwood@arm.com>
12837 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
12838 * internal-fn.def (VEC_SHL_INSERT): New internal function.
12839 * optabs.def (vec_shl_insert_optab): New optab.
12840 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
12841 (duplicate_and_interleave): Likewise.
12842 * tree-vect-loop.c: Include internal-fn.h.
12843 (neutral_op_for_slp_reduction): New function, split out from
12844 get_initial_defs_for_reduction.
12845 (get_initial_def_for_reduction): Handle option 2 for variable-length
12846 vectors by loading the neutral value into a vector and then shifting
12847 the initial value into element 0.
12848 (get_initial_defs_for_reduction): Replace the code argument with
12849 the neutral value calculated by neutral_op_for_slp_reduction.
12850 Use gimple_build_vector for constant-length vectors.
12851 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
12852 but the first group_size elements have a neutral value.
12853 Use duplicate_and_interleave otherwise.
12854 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
12855 Update call to get_initial_defs_for_reduction. Handle SLP
12856 reductions for variable-length vectors by creating one vector
12857 result for each scalar result, with the elements associated
12858 with other scalar results stubbed out with the neutral value.
12859 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
12860 Require IFN_VEC_SHL_INSERT for double reductions on
12861 variable-length vectors, or SLP reductions that have
12862 a neutral value. Require can_duplicate_and_interleave_p
12863 support for variable-length unchained SLP reductions if there
12864 is no neutral value, such as for MIN/MAX reductions. Also require
12865 the number of vector elements to be a multiple of the number of
12866 SLP statements when doing variable-length unchained SLP reductions.
12867 Update call to vect_create_epilog_for_reduction.
12868 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
12869 and remove initial values.
12870 (duplicate_and_interleave): Make public.
12871 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
12872 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
12874 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12875 Alan Hayward <alan.hayward@arm.com>
12876 David Sherwood <david.sherwood@arm.com>
12878 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
12879 (can_duplicate_and_interleave_p): New function.
12880 (vect_get_and_check_slp_defs): Take the vector of statements
12881 rather than just the current one. Remove excess parentheses.
12882 Restriction rejectinon of vect_constant_def and vect_external_def
12883 for variable-length vectors to boolean types, or types for which
12884 can_duplicate_and_interleave_p is false.
12885 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
12886 (duplicate_and_interleave): New function.
12887 (vect_get_constant_vectors): Use gimple_build_vector for
12888 constant-length vectors and suitable variable-length constant
12889 vectors. Use duplicate_and_interleave for other variable-length
12890 vectors. Don't defer the update when inserting new statements.
12892 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12893 Alan Hayward <alan.hayward@arm.com>
12894 David Sherwood <david.sherwood@arm.com>
12896 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
12897 min_profitable_iters doesn't go negative.
12899 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12900 Alan Hayward <alan.hayward@arm.com>
12901 David Sherwood <david.sherwood@arm.com>
12903 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
12904 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
12905 * optabs.def (vec_mask_load_lanes_optab): New optab.
12906 (vec_mask_store_lanes_optab): Likewise.
12907 * internal-fn.def (MASK_LOAD_LANES): New internal function.
12908 (MASK_STORE_LANES): Likewise.
12909 * internal-fn.c (mask_load_lanes_direct): New macro.
12910 (mask_store_lanes_direct): Likewise.
12911 (expand_mask_load_optab_fn): Handle masked operations.
12912 (expand_mask_load_lanes_optab_fn): New macro.
12913 (expand_mask_store_optab_fn): Handle masked operations.
12914 (expand_mask_store_lanes_optab_fn): New macro.
12915 (direct_mask_load_lanes_optab_supported_p): Likewise.
12916 (direct_mask_store_lanes_optab_supported_p): Likewise.
12917 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
12919 (vect_load_lanes_supported): Likewise.
12920 * tree-vect-data-refs.c (strip_conversion): New function.
12921 (can_group_stmts_p): Likewise.
12922 (vect_analyze_data_ref_accesses): Use it instead of checking
12923 for a pair of assignments.
12924 (vect_store_lanes_supported): Take a masked_p parameter.
12925 (vect_load_lanes_supported): Likewise.
12926 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
12927 vect_store_lanes_supported and vect_load_lanes_supported.
12928 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12929 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
12930 parameter. Don't allow gaps for masked accesses.
12931 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
12932 and vect_load_lanes_supported.
12933 (get_load_store_type): Take a masked_p parameter and update
12934 call to get_group_load_store_type.
12935 (vectorizable_store): Update call to get_load_store_type.
12936 Handle IFN_MASK_STORE_LANES.
12937 (vectorizable_load): Update call to get_load_store_type.
12938 Handle IFN_MASK_LOAD_LANES.
12940 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12941 Alan Hayward <alan.hayward@arm.com>
12942 David Sherwood <david.sherwood@arm.com>
12944 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
12946 * config/aarch64/aarch64-protos.h
12947 (aarch64_sve_struct_memory_operand_p): Declare.
12948 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
12949 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
12950 (VPRED, vpred): Handle SVE structure modes.
12951 * config/aarch64/constraints.md (Utx): New constraint.
12952 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
12953 (aarch64_sve_struct_nonimmediate_operand): New predicates.
12954 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
12955 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
12956 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
12957 structure modes. Split into pieces after RA.
12958 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
12959 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
12961 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
12962 SVE structure modes.
12963 (aarch64_classify_address): Likewise.
12964 (sizetochar): Move earlier in file.
12965 (aarch64_print_operand): Handle SVE register lists.
12966 (aarch64_array_mode): New function.
12967 (aarch64_sve_struct_memory_operand_p): Likewise.
12968 (TARGET_ARRAY_MODE): Redefine.
12970 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12971 Alan Hayward <alan.hayward@arm.com>
12972 David Sherwood <david.sherwood@arm.com>
12974 * target.def (array_mode): New target hook.
12975 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
12976 * doc/tm.texi: Regenerate.
12977 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
12978 * hooks.c (hook_optmode_mode_uhwi_none): New function.
12979 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
12980 targetm.array_mode.
12981 * stor-layout.c (mode_for_array): Likewise. Support polynomial
12984 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12985 Alan Hayward <alan.hayward@arm.com>
12986 David Sherwood <david.sherwood@arm.com>
12988 * fold-const.c (fold_binary_loc): Check the argument types
12989 rather than the result type when testing for a vector operation.
12991 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12993 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
12994 * doc/tm.texi: Regenerate.
12996 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12997 Alan Hayward <alan.hayward@arm.com>
12998 David Sherwood <david.sherwood@arm.com>
13000 * doc/invoke.texi (-msve-vector-bits=): Document new option.
13001 (sve): Document new AArch64 extension.
13002 * doc/md.texi (w): Extend the description of the AArch64
13003 constraint to include SVE vectors.
13004 (Upl, Upa): Document new AArch64 predicate constraints.
13005 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
13007 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
13008 (msve-vector-bits=): New option.
13009 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
13010 SVE when these are disabled.
13011 (sve): New extension.
13012 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
13013 modes. Adjust their number of units based on aarch64_sve_vg.
13014 (MAX_BITSIZE_MODE_ANY_MODE): Define.
13015 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
13016 aarch64_addr_query_type.
13017 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
13018 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
13019 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
13020 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
13021 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
13022 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
13023 (aarch64_simd_imm_zero_p): Delete.
13024 (aarch64_check_zero_based_sve_index_immediate): Declare.
13025 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13026 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13027 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13028 (aarch64_sve_float_mul_immediate_p): Likewise.
13029 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13030 rather than an rtx.
13031 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
13032 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
13033 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
13034 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
13035 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
13036 (aarch64_regmode_natural_size): Likewise.
13037 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
13038 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
13040 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
13041 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
13042 for VG and the SVE predicate registers.
13043 (V_ALIASES): Add a "z"-prefixed alias.
13044 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
13045 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
13046 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
13047 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
13048 (REG_CLASS_NAMES): Add entries for them.
13049 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
13050 and the predicate registers.
13051 (aarch64_sve_vg): Declare.
13052 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
13053 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
13054 (REGMODE_NATURAL_SIZE): Define.
13055 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
13057 * config/aarch64/aarch64.c: Include cfgrtl.h.
13058 (simd_immediate_info): Add a constructor for series vectors,
13059 and an associated step field.
13060 (aarch64_sve_vg): New variable.
13061 (aarch64_dbx_register_number): Handle VG and the predicate registers.
13062 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
13063 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
13064 (VEC_ANY_DATA, VEC_STRUCT): New constants.
13065 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
13066 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
13067 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
13068 (aarch64_get_mask_mode): New functions.
13069 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
13070 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
13071 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
13072 predicate modes and predicate registers. Explicitly restrict
13073 GPRs to modes of 16 bytes or smaller. Only allow FP registers
13074 to store a vector mode if it is recognized by
13075 aarch64_classify_vector_mode.
13076 (aarch64_regmode_natural_size): New function.
13077 (aarch64_hard_regno_caller_save_mode): Return the original mode
13079 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
13080 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
13081 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
13082 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
13084 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
13085 does not overlap dest if the function is frame-related. Handle
13087 (aarch64_split_add_offset): New function.
13088 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13089 them aarch64_add_offset.
13090 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13091 and update call to aarch64_sub_sp.
13092 (aarch64_add_cfa_expression): New function.
13093 (aarch64_expand_prologue): Pass extra temporary registers to the
13094 functions above. Handle the case in which we need to emit new
13095 DW_CFA_expressions for registers that were originally saved
13096 relative to the stack pointer, but now have to be expressed
13097 relative to the frame pointer.
13098 (aarch64_output_mi_thunk): Pass extra temporary registers to the
13100 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
13101 IP0 and IP1 values for SVE frames.
13102 (aarch64_expand_vec_series): New function.
13103 (aarch64_expand_sve_widened_duplicate): Likewise.
13104 (aarch64_expand_sve_const_vector): Likewise.
13105 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13106 Handle SVE constants. Use emit_move_insn to move a force_const_mem
13107 into the register, rather than emitting a SET directly.
13108 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13109 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13110 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13111 (offset_9bit_signed_scaled_p): New functions.
13112 (aarch64_replicate_bitmask_imm): New function.
13113 (aarch64_bitmask_imm): Use it.
13114 (aarch64_cannot_force_const_mem): Reject expressions involving
13115 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
13116 (aarch64_classify_index): Handle SVE indices, by requiring
13117 a plain register index with a scale that matches the element size.
13118 (aarch64_classify_address): Handle SVE addresses. Assert that
13119 the mode of the address is VOIDmode or an integer mode.
13120 Update call to aarch64_classify_symbol.
13121 (aarch64_classify_symbolic_expression): Update call to
13122 aarch64_classify_symbol.
13123 (aarch64_const_vec_all_in_range_p): New function.
13124 (aarch64_print_vector_float_operand): Likewise.
13125 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
13126 "vN" for FP registers with SVE modes. Handle (const ...) vectors
13127 and the FP immediates 1.0 and 0.5.
13128 (aarch64_print_address_internal): Handle SVE addresses.
13129 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13130 (aarch64_regno_regclass): Handle predicate registers.
13131 (aarch64_secondary_reload): Handle big-endian reloads of SVE
13133 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13134 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13135 (aarch64_convert_sve_vector_bits): New function.
13136 (aarch64_override_options): Use it to handle -msve-vector-bits=.
13137 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13138 rather than an rtx.
13139 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13140 Handle SVE vector and predicate modes. Accept VL-based constants
13141 that need only one temporary register, and VL offsets that require
13142 no temporary registers.
13143 (aarch64_conditional_register_usage): Mark the predicate registers
13144 as fixed if SVE isn't available.
13145 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13146 Return true for SVE vector and predicate modes.
13147 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13148 rather than an unsigned int. Handle SVE modes.
13149 (aarch64_preferred_simd_mode): Update call accordingly. Handle
13151 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13153 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13154 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13155 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13156 (aarch64_sve_float_mul_immediate_p): New functions.
13157 (aarch64_sve_valid_immediate): New function.
13158 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13159 Explicitly reject structure modes. Check for INDEX constants.
13160 Handle PTRUE and PFALSE constants.
13161 (aarch64_check_zero_based_sve_index_immediate): New function.
13162 (aarch64_simd_imm_zero_p): Delete.
13163 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13164 vector modes. Accept constants in the range of CNT[BHWD].
13165 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13166 ask for an Advanced SIMD mode.
13167 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13168 (aarch64_simd_vector_alignment): Handle SVE predicates.
13169 (aarch64_vectorize_preferred_vector_alignment): New function.
13170 (aarch64_simd_vector_alignment_reachable): Use it instead of
13172 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13173 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13175 (MAX_VECT_LEN): Delete.
13176 (expand_vec_perm_d): Add a vec_flags field.
13177 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13178 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13179 (aarch64_evpc_ext): Don't apply a big-endian lane correction
13181 (aarch64_evpc_rev): Rename to...
13182 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
13183 (aarch64_evpc_rev_global): New function.
13184 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13185 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13187 (aarch64_evpc_sve_tbl): New function.
13188 (aarch64_expand_vec_perm_const_1): Update after rename of
13189 aarch64_evpc_rev. Handle SVE permutes too, trying
13190 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13191 than aarch64_evpc_tbl.
13192 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13193 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13194 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13195 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13196 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13197 (aarch64_expand_sve_vcond): New functions.
13198 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13199 of aarch64_vector_mode_p.
13200 (aarch64_dwarf_poly_indeterminate_value): New function.
13201 (aarch64_compute_pressure_classes): Likewise.
13202 (aarch64_can_change_mode_class): Likewise.
13203 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13204 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13205 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13206 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13207 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13208 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13209 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13210 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13212 (Dn, Dl, Dr): Accept const as well as const_vector.
13213 (Dz): Likewise. Compare against CONST0_RTX.
13214 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13215 of "vector" where appropriate.
13216 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13217 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13218 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13219 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13220 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13221 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13222 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13223 (v_int_equiv): Extend to SVE modes.
13224 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13226 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13227 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13228 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13229 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13230 (SVE_COND_FP_CMP): New int iterators.
13231 (perm_hilo): Handle the new unpack unspecs.
13232 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13234 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13235 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13236 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13237 (aarch64_equality_operator, aarch64_constant_vector_operand)
13238 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13239 (aarch64_sve_nonimmediate_operand): Likewise.
13240 (aarch64_sve_general_operand): Likewise.
13241 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13242 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13243 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13244 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
13245 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
13246 (aarch64_sve_float_arith_immediate): Likewise.
13247 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
13248 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
13249 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
13250 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
13251 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
13252 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
13253 (aarch64_sve_float_arith_operand): Likewise.
13254 (aarch64_sve_float_arith_with_sub_operand): Likewise.
13255 (aarch64_sve_float_mul_operand): Likewise.
13256 (aarch64_sve_vec_perm_operand): Likewise.
13257 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
13258 (aarch64_mov_operand): Accept const_poly_int and const_vector.
13259 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
13260 as well as const_vector.
13261 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
13262 in file. Use CONST0_RTX and CONSTM1_RTX.
13263 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
13264 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
13265 Use aarch64_simd_imm_zero.
13266 * config/aarch64/aarch64-sve.md: New file.
13267 * config/aarch64/aarch64.md: Include it.
13268 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
13269 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
13270 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
13271 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
13272 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
13273 (sve): New attribute.
13274 (enabled): Disable instructions with the sve attribute unless
13276 (movqi, movhi): Pass CONST_POLY_INT operaneds through
13277 aarch64_expand_mov_immediate.
13278 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
13279 CNT[BHSD] immediates.
13280 (movti): Split CONST_POLY_INT moves into two halves.
13281 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
13282 Split additions that need a temporary here if the destination
13283 is the stack pointer.
13284 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
13285 (*add<mode>3_poly_1): New instruction.
13286 (set_clobber_cc): New expander.
13288 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13290 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
13291 parameter and use it instead of GET_MODE_SIZE (innermode). Use
13292 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
13293 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
13294 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
13295 Change innermode from fixed_mode_size to machine_mode.
13296 (simplify_subreg): Update call accordingly. Handle a constant-sized
13297 subreg of a variable-length CONST_VECTOR.
13299 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13300 Alan Hayward <alan.hayward@arm.com>
13301 David Sherwood <david.sherwood@arm.com>
13303 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
13304 (add_offset_to_base): New function, split out from...
13305 (create_mem_ref): ...here. When handling a scale other than 1,
13306 check first whether the address is valid without the offset.
13307 Add it into the base if so, leaving the index and scale as-is.
13309 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13312 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
13313 fold_for_warn before checking if arg2 is INTEGER_CST.
13315 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
13317 * config/rs6000/predicates.md (load_multiple_operation): Delete.
13318 (store_multiple_operation): Delete.
13319 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
13320 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
13321 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
13322 guarded by TARGET_STRING.
13323 (rs6000_output_load_multiple): Delete.
13324 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
13325 OPTION_MASK_STRING / TARGET_STRING handling.
13326 (print_operand) <'N', 'O'>: Add comment that these are unused now.
13327 (const rs6000_opt_masks) <"string">: Change mask to 0.
13328 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
13329 (MASK_STRING): Delete.
13330 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
13332 (load_multiple): Delete.
13339 (store_multiple): Delete.
13346 (movmemsi_8reg): Delete.
13347 (corresponding unnamed define_insn): Delete.
13348 (movmemsi_6reg): Delete.
13349 (corresponding unnamed define_insn): Delete.
13350 (movmemsi_4reg): Delete.
13351 (corresponding unnamed define_insn): Delete.
13352 (movmemsi_2reg): Delete.
13353 (corresponding unnamed define_insn): Delete.
13354 (movmemsi_1reg): Delete.
13355 (corresponding unnamed define_insn): Delete.
13356 * config/rs6000/rs6000.opt (mno-string): New.
13357 (mstring): Replace by deprecation warning stub.
13358 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
13360 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13362 * regrename.c (regrename_do_replace): If replacing the same
13363 reg multiple times, try to reuse last created gen_raw_REG.
13366 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
13367 main to workaround a bug in GDB.
13369 2018-01-12 Tom de Vries <tom@codesourcery.com>
13372 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
13374 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
13376 PR rtl-optimization/80481
13377 * ira-color.c (get_cap_member): New function.
13378 (allocnos_conflict_by_live_ranges_p): Use it.
13379 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
13380 (setup_slot_coalesced_allocno_live_ranges): Ditto.
13382 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
13385 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
13386 (*saddl_se_1): Ditto.
13387 (*ssubsi_1): Ditto.
13388 (*ssubl_se_1): Ditto.
13390 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13392 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
13393 rather than wi::to_widest for DR_INITs.
13394 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
13395 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
13396 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
13398 (vect_analyze_group_access_1): Note that here.
13400 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13402 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
13403 polynomial type sizes.
13405 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13407 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
13408 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
13409 (gimple_add_tmp_var): Likewise.
13411 2018-01-12 Martin Liska <mliska@suse.cz>
13413 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
13414 (gimple_alloc_sizes): Likewise.
13415 (dump_gimple_statistics): Use PRIu64 in printf format.
13416 * gimple.h: Change uint64_t to int.
13418 2018-01-12 Martin Liska <mliska@suse.cz>
13420 * tree-core.h: Use uint64_t instead of int.
13421 * tree.c (tree_node_counts): Likewise.
13422 (tree_node_sizes): Likewise.
13423 (dump_tree_statistics): Use PRIu64 in printf format.
13425 2018-01-12 Martin Liska <mliska@suse.cz>
13427 * Makefile.in: As qsort_chk is implemented in vec.c, add
13428 vec.o to linkage of gencfn-macros.
13429 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
13430 passing the info to record_node_allocation_statistics.
13431 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
13433 * ggc-common.c (struct ggc_usage): Add operator== and use
13434 it in operator< and compare function.
13435 * mem-stats.h (struct mem_usage): Likewise.
13436 * vec.c (struct vec_usage): Remove operator< and compare
13437 function. Can be simply inherited.
13439 2018-01-12 Martin Jambor <mjambor@suse.cz>
13442 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
13443 * tree-ssa-math-opts.c: Include domwalk.h.
13444 (convert_mult_to_fma_1): New function.
13445 (fma_transformation_info): New type.
13446 (fma_deferring_state): Likewise.
13447 (cancel_fma_deferring): New function.
13448 (result_of_phi): Likewise.
13449 (last_fma_candidate_feeds_initial_phi): Likewise.
13450 (convert_mult_to_fma): Added deferring logic, split actual
13451 transformation to convert_mult_to_fma_1.
13452 (math_opts_dom_walker): New type.
13453 (math_opts_dom_walker::after_dom_children): New method, body moved
13454 here from pass_optimize_widening_mul::execute, added deferring logic
13456 (pass_optimize_widening_mul::execute): Moved most of code to
13457 math_opts_dom_walker::after_dom_children.
13458 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
13459 * config/i386/i386.c (ix86_option_override_internal): Added
13460 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
13462 2018-01-12 Richard Biener <rguenther@suse.de>
13465 * dwarf2out.c (gen_variable_die): Do not reset old_die for
13466 inline instance vars.
13468 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
13471 * config/rx/rx.c (rx_is_restricted_memory_address):
13472 Handle SUBREG case.
13474 2018-01-12 Richard Biener <rguenther@suse.de>
13476 PR tree-optimization/80846
13477 * target.def (split_reduction): New target hook.
13478 * targhooks.c (default_split_reduction): New function.
13479 * targhooks.h (default_split_reduction): Declare.
13480 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
13481 target requests first reduce vectors by combining low and high
13483 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
13484 (get_vectype_for_scalar_type_and_size): Export.
13485 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
13486 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
13487 * doc/tm.texi: Regenerate.
13488 * config/i386/i386.c (ix86_split_reduction): Implement
13489 TARGET_VECTORIZE_SPLIT_REDUCTION.
13491 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13494 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
13495 in PIC mode except for TARGET_VXWORKS_RTP.
13496 * config/sparc/sparc.c: Include cfgrtl.h.
13497 (TARGET_INIT_PIC_REG): Define.
13498 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
13499 (sparc_pic_register_p): New predicate.
13500 (sparc_legitimate_address_p): Use it.
13501 (sparc_legitimize_pic_address): Likewise.
13502 (sparc_delegitimize_address): Likewise.
13503 (sparc_mode_dependent_address_p): Likewise.
13504 (gen_load_pcrel_sym): Remove 4th parameter.
13505 (load_got_register): Adjust call to above. Remove obsolete stuff.
13506 (sparc_expand_prologue): Do not call load_got_register here.
13507 (sparc_flat_expand_prologue): Likewise.
13508 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
13509 (sparc_use_pseudo_pic_reg): New function.
13510 (sparc_init_pic_reg): Likewise.
13511 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13512 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
13514 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
13516 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
13517 Add item for branch_cost.
13519 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13521 PR rtl-optimization/83565
13522 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
13523 not extend the result to a larger mode for rotate operations.
13524 (num_sign_bit_copies1): Likewise.
13526 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13529 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
13531 Use values-Xc.o for -pedantic.
13532 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
13534 2018-01-12 Martin Liska <mliska@suse.cz>
13537 * ipa-devirt.c (final_warning_record::grow_type_warnings):
13539 (possible_polymorphic_call_targets): Use it.
13540 (ipa_devirt): Likewise.
13542 2018-01-12 Martin Liska <mliska@suse.cz>
13544 * profile-count.h (enum profile_quality): Use 0 as invalid
13545 enum value of profile_quality.
13547 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
13549 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
13550 -mext-string options.
13552 2018-01-12 Richard Biener <rguenther@suse.de>
13554 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
13555 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
13556 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
13558 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
13560 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13562 * configure.ac (--with-long-double-format): Add support for the
13563 configuration option to change the default long double format on
13565 * config.gcc (powerpc*-linux*-*): Likewise.
13566 * configure: Regenerate.
13567 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
13568 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
13569 used without modification.
13571 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13573 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
13574 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
13575 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
13576 MISC_BUILTIN_SPEC_BARRIER.
13577 (rs6000_init_builtins): Likewise.
13578 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
13580 (speculation_barrier): New define_insn.
13581 * doc/extend.texi: Document __builtin_speculation_barrier.
13583 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13586 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
13587 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
13588 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
13590 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
13591 integral modes instead of "ss" and "sd".
13592 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
13593 vectors with 32-bit and 64-bit elements.
13594 (vecdupssescalarmodesuffix): New mode attribute.
13595 (vec_dup<mode>): Use it.
13597 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
13600 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
13601 frame if argument is passed on stack.
13603 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13606 * ree.c (combine_reaching_defs): Optimize also
13607 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
13608 reg2=any_extend(exp); reg1=reg2;, formatting fix.
13610 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13612 PR middle-end/83189
13613 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
13615 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13617 PR middle-end/83718
13618 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
13619 after they are computed.
13621 2018-01-11 Bin Cheng <bin.cheng@arm.com>
13623 PR tree-optimization/83695
13624 * gimple-loop-linterchange.cc
13625 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
13626 reset cached scev information after interchange.
13627 (pass_linterchange::execute): Remove call to scev_reset_htab.
13629 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13631 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
13632 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
13633 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
13634 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
13635 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
13636 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
13637 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
13638 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
13639 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
13640 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
13641 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
13642 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
13643 (V_lane_reg): Likewise.
13644 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
13646 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
13647 (vfmal_lane_low<mode>_intrinsic,
13648 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
13649 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
13650 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
13651 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
13652 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
13653 vfmsl_lane_high<mode>_intrinsic): New define_insns.
13655 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13657 * config/arm/arm-cpus.in (fp16fml): New feature.
13658 (ALL_SIMD): Add fp16fml.
13659 (armv8.2-a): Add fp16fml as an option.
13660 (armv8.3-a): Likewise.
13661 (armv8.4-a): Add fp16fml as part of fp16.
13662 * config/arm/arm.h (TARGET_FP16FML): Define.
13663 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
13665 * config/arm/arm-modes.def (V2HF): Define.
13666 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
13667 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
13668 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
13669 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
13670 vfmsl_low, vfmsl_high): New set of builtins.
13671 * config/arm/iterators.md (PLUSMINUS): New code iterator.
13672 (vfml_op): New code attribute.
13673 (VFMLHALVES): New int iterator.
13674 (VFML, VFMLSEL): New mode attributes.
13675 (V_reg): Define mapping for V2HF.
13676 (V_hi, V_lo): New mode attributes.
13677 (VF_constraint): Likewise.
13678 (vfml_half, vfml_half_selector): New int attributes.
13679 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
13681 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
13682 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
13684 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
13685 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
13686 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
13687 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
13689 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
13690 Document new effective target and option set.
13692 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13694 * config/arm/arm-cpus.in (armv8_4): New feature.
13695 (ARMv8_4a): New fgroup.
13696 (armv8.4-a): New arch.
13697 * config/arm/arm-tables.opt: Regenerate.
13698 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
13699 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
13700 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
13701 Add matching rules for -march=armv8.4-a and extensions.
13702 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
13704 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
13707 * config/rx/rx.md (BW): New mode attribute.
13708 (sync_lock_test_and_setsi): Add mode suffix to insn output.
13710 2018-01-11 Richard Biener <rguenther@suse.de>
13712 PR tree-optimization/83435
13713 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
13714 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
13715 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
13717 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13718 Alan Hayward <alan.hayward@arm.com>
13719 David Sherwood <david.sherwood@arm.com>
13721 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
13723 (aarch64_classify_address): Initialize it. Track polynomial offsets.
13724 (aarch64_print_address_internal): Use it to check for a zero offset.
13726 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13727 Alan Hayward <alan.hayward@arm.com>
13728 David Sherwood <david.sherwood@arm.com>
13730 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
13731 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
13732 Return a poly_int64 rather than a HOST_WIDE_INT.
13733 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
13734 rather than a HOST_WIDE_INT.
13735 * config/aarch64/aarch64.h (aarch64_frame): Protect with
13736 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
13737 hard_fp_offset, frame_size, initial_adjust, callee_offset and
13738 final_offset from HOST_WIDE_INT to poly_int64.
13739 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
13740 to_constant when getting the number of units in an Advanced SIMD
13742 (aarch64_builtin_vectorized_function): Check for a constant number
13744 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
13746 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
13747 attribute instead of GET_MODE_NUNITS.
13748 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
13749 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
13750 GET_MODE_SIZE for fixed-size registers.
13751 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
13752 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
13753 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
13754 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
13755 (aarch64_print_operand, aarch64_print_address_internal)
13756 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
13757 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
13758 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
13759 Handle polynomial GET_MODE_SIZE.
13760 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
13761 wider than SImode without modification.
13762 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
13763 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
13764 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
13765 passing and returning SVE modes.
13766 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
13767 rather than GEN_INT.
13768 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
13769 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
13770 (aarch64_allocate_and_probe_stack_space): Likewise.
13771 (aarch64_layout_frame): Cope with polynomial offsets.
13772 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
13773 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
13774 polynomial offsets.
13775 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
13776 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
13777 poly_int64 rather than a HOST_WIDE_INT.
13778 (aarch64_get_separate_components, aarch64_process_components)
13779 (aarch64_expand_prologue, aarch64_expand_epilogue)
13780 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
13781 (aarch64_anchor_offset): New function, split out from...
13782 (aarch64_legitimize_address): ...here.
13783 (aarch64_builtin_vectorization_cost): Handle polynomial
13784 TYPE_VECTOR_SUBPARTS.
13785 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
13787 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
13788 number of elements from the PARALLEL rather than the mode.
13789 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
13790 rather than GET_MODE_BITSIZE.
13791 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
13792 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
13793 (aarch64_expand_vec_perm_const_1): Handle polynomial
13794 d->perm.length () and d->perm elements.
13795 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
13796 Apply to_constant to d->perm elements.
13797 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
13798 polynomial CONST_VECTOR_NUNITS.
13799 (aarch64_move_pointer): Take amount as a poly_int64 rather
13801 (aarch64_progress_pointer): Avoid temporary variable.
13802 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
13803 the mode attribute instead of GET_MODE.
13805 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13806 Alan Hayward <alan.hayward@arm.com>
13807 David Sherwood <david.sherwood@arm.com>
13809 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
13810 x exists before using it.
13811 (aarch64_add_constant_internal): Rename to...
13812 (aarch64_add_offset_1): ...this. Replace regnum with separate
13813 src and dest rtxes. Handle the case in which they're different,
13814 including when the offset is zero. Replace scratchreg with an rtx.
13815 Use 2 additions if there is no spare register into which we can
13816 move a 16-bit constant.
13817 (aarch64_add_constant): Delete.
13818 (aarch64_add_offset): Replace reg with separate src and dest
13819 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
13820 Use aarch64_add_offset_1.
13821 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
13822 an rtx rather than an int. Take the delta as a poly_int64
13823 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
13824 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
13825 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
13826 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
13827 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
13828 and aarch64_add_sp.
13829 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
13830 aarch64_add_constant.
13832 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13834 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
13835 Use scalar_float_mode.
13837 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13839 * config/aarch64/aarch64-simd.md
13840 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
13841 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
13842 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
13843 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
13844 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
13845 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
13846 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
13847 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
13848 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
13849 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
13851 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13854 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
13855 targ_options->x_arm_arch_string is non NULL.
13857 2018-01-11 Tamar Christina <tamar.christina@arm.com>
13859 * config/aarch64/aarch64.h
13860 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
13862 2018-01-11 Sudakshina Das <sudi.das@arm.com>
13865 * expmed.c (emit_store_flag_force): Swap if const op0
13866 and change VOIDmode to mode of op0.
13868 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13870 PR rtl-optimization/83761
13871 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
13872 than bytes to mode_for_size.
13874 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13876 PR middle-end/83189
13877 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
13878 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
13881 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13883 PR middle-end/83575
13884 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
13885 when in layout mode.
13886 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
13887 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
13890 2018-01-10 Michael Collison <michael.collison@arm.com>
13892 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
13893 * config/aarch64/aarch64-option-extension.def: Add
13894 AARCH64_OPT_EXTENSION of 'fp16fml'.
13895 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13896 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
13897 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
13898 * config/aarch64/constraints.md (Ui7): New constraint.
13899 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
13900 (VFMLA_SEL_W): Ditto.
13903 (VFMLA16_LOW): New int iterator.
13904 (VFMLA16_HIGH): Ditto.
13905 (UNSPEC_FMLAL): New unspec.
13906 (UNSPEC_FMLSL): Ditto.
13907 (UNSPEC_FMLAL2): Ditto.
13908 (UNSPEC_FMLSL2): Ditto.
13909 (f16mac): New code attribute.
13910 * config/aarch64/aarch64-simd-builtins.def
13911 (aarch64_fmlal_lowv2sf): Ditto.
13912 (aarch64_fmlsl_lowv2sf): Ditto.
13913 (aarch64_fmlalq_lowv4sf): Ditto.
13914 (aarch64_fmlslq_lowv4sf): Ditto.
13915 (aarch64_fmlal_highv2sf): Ditto.
13916 (aarch64_fmlsl_highv2sf): Ditto.
13917 (aarch64_fmlalq_highv4sf): Ditto.
13918 (aarch64_fmlslq_highv4sf): Ditto.
13919 (aarch64_fmlal_lane_lowv2sf): Ditto.
13920 (aarch64_fmlsl_lane_lowv2sf): Ditto.
13921 (aarch64_fmlal_laneq_lowv2sf): Ditto.
13922 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
13923 (aarch64_fmlalq_lane_lowv4sf): Ditto.
13924 (aarch64_fmlsl_lane_lowv4sf): Ditto.
13925 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
13926 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
13927 (aarch64_fmlal_lane_highv2sf): Ditto.
13928 (aarch64_fmlsl_lane_highv2sf): Ditto.
13929 (aarch64_fmlal_laneq_highv2sf): Ditto.
13930 (aarch64_fmlsl_laneq_highv2sf): Ditto.
13931 (aarch64_fmlalq_lane_highv4sf): Ditto.
13932 (aarch64_fmlsl_lane_highv4sf): Ditto.
13933 (aarch64_fmlalq_laneq_highv4sf): Ditto.
13934 (aarch64_fmlsl_laneq_highv4sf): Ditto.
13935 * config/aarch64/aarch64-simd.md:
13936 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
13937 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13938 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
13939 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13940 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
13941 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
13942 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
13943 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
13944 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
13945 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
13946 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
13947 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
13948 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
13949 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
13950 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
13951 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
13952 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
13953 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
13954 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
13955 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
13956 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
13957 (vfmlsl_low_u32): Ditto.
13958 (vfmlalq_low_u32): Ditto.
13959 (vfmlslq_low_u32): Ditto.
13960 (vfmlal_high_u32): Ditto.
13961 (vfmlsl_high_u32): Ditto.
13962 (vfmlalq_high_u32): Ditto.
13963 (vfmlslq_high_u32): Ditto.
13964 (vfmlal_lane_low_u32): Ditto.
13965 (vfmlsl_lane_low_u32): Ditto.
13966 (vfmlal_laneq_low_u32): Ditto.
13967 (vfmlsl_laneq_low_u32): Ditto.
13968 (vfmlalq_lane_low_u32): Ditto.
13969 (vfmlslq_lane_low_u32): Ditto.
13970 (vfmlalq_laneq_low_u32): Ditto.
13971 (vfmlslq_laneq_low_u32): Ditto.
13972 (vfmlal_lane_high_u32): Ditto.
13973 (vfmlsl_lane_high_u32): Ditto.
13974 (vfmlal_laneq_high_u32): Ditto.
13975 (vfmlsl_laneq_high_u32): Ditto.
13976 (vfmlalq_lane_high_u32): Ditto.
13977 (vfmlslq_lane_high_u32): Ditto.
13978 (vfmlalq_laneq_high_u32): Ditto.
13979 (vfmlslq_laneq_high_u32): Ditto.
13980 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
13981 (AARCH64_FL_FOR_ARCH8_4): New.
13982 (AARCH64_ISA_F16FML): New ISA flag.
13983 (TARGET_F16FML): New feature flag for fp16fml.
13984 (doc/invoke.texi): Document new fp16fml option.
13986 2018-01-10 Michael Collison <michael.collison@arm.com>
13988 * config/aarch64/aarch64-builtins.c:
13989 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
13990 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13991 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
13992 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
13993 (AARCH64_ISA_SHA3): New ISA flag.
13994 (TARGET_SHA3): New feature flag for sha3.
13995 * config/aarch64/iterators.md (sha512_op): New int attribute.
13996 (CRYPTO_SHA512): New int iterator.
13997 (UNSPEC_SHA512H): New unspec.
13998 (UNSPEC_SHA512H2): Ditto.
13999 (UNSPEC_SHA512SU0): Ditto.
14000 (UNSPEC_SHA512SU1): Ditto.
14001 * config/aarch64/aarch64-simd-builtins.def
14002 (aarch64_crypto_sha512hqv2di): New builtin.
14003 (aarch64_crypto_sha512h2qv2di): Ditto.
14004 (aarch64_crypto_sha512su0qv2di): Ditto.
14005 (aarch64_crypto_sha512su1qv2di): Ditto.
14006 (aarch64_eor3qv8hi): Ditto.
14007 (aarch64_rax1qv2di): Ditto.
14008 (aarch64_xarqv2di): Ditto.
14009 (aarch64_bcaxqv8hi): Ditto.
14010 * config/aarch64/aarch64-simd.md:
14011 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
14012 (aarch64_crypto_sha512su0qv2di): Ditto.
14013 (aarch64_crypto_sha512su1qv2di): Ditto.
14014 (aarch64_eor3qv8hi): Ditto.
14015 (aarch64_rax1qv2di): Ditto.
14016 (aarch64_xarqv2di): Ditto.
14017 (aarch64_bcaxqv8hi): Ditto.
14018 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
14019 (vsha512h2q_u64): Ditto.
14020 (vsha512su0q_u64): Ditto.
14021 (vsha512su1q_u64): Ditto.
14022 (veor3q_u16): Ditto.
14023 (vrax1q_u64): Ditto.
14024 (vxarq_u64): Ditto.
14025 (vbcaxq_u16): Ditto.
14026 * config/arm/types.md (crypto_sha512): New type attribute.
14027 (crypto_sha3): Ditto.
14028 (doc/invoke.texi): Document new sha3 option.
14030 2018-01-10 Michael Collison <michael.collison@arm.com>
14032 * config/aarch64/aarch64-builtins.c:
14033 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
14034 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14035 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
14036 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
14037 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
14038 (AARCH64_ISA_SM4): New ISA flag.
14039 (TARGET_SM4): New feature flag for sm4.
14040 * config/aarch64/aarch64-simd-builtins.def
14041 (aarch64_sm3ss1qv4si): Ditto.
14042 (aarch64_sm3tt1aq4si): Ditto.
14043 (aarch64_sm3tt1bq4si): Ditto.
14044 (aarch64_sm3tt2aq4si): Ditto.
14045 (aarch64_sm3tt2bq4si): Ditto.
14046 (aarch64_sm3partw1qv4si): Ditto.
14047 (aarch64_sm3partw2qv4si): Ditto.
14048 (aarch64_sm4eqv4si): Ditto.
14049 (aarch64_sm4ekeyqv4si): Ditto.
14050 * config/aarch64/aarch64-simd.md:
14051 (aarch64_sm3ss1qv4si): Ditto.
14052 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
14053 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
14054 (aarch64_sm4eqv4si): Ditto.
14055 (aarch64_sm4ekeyqv4si): Ditto.
14056 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
14057 (sm3part_op): Ditto.
14058 (CRYPTO_SM3TT): Ditto.
14059 (CRYPTO_SM3PART): Ditto.
14060 (UNSPEC_SM3SS1): New unspec.
14061 (UNSPEC_SM3TT1A): Ditto.
14062 (UNSPEC_SM3TT1B): Ditto.
14063 (UNSPEC_SM3TT2A): Ditto.
14064 (UNSPEC_SM3TT2B): Ditto.
14065 (UNSPEC_SM3PARTW1): Ditto.
14066 (UNSPEC_SM3PARTW2): Ditto.
14067 (UNSPEC_SM4E): Ditto.
14068 (UNSPEC_SM4EKEY): Ditto.
14069 * config/aarch64/constraints.md (Ui2): New constraint.
14070 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
14071 * config/arm/types.md (crypto_sm3): New type attribute.
14072 (crypto_sm4): Ditto.
14073 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
14074 (vsm3tt1aq_u32): Ditto.
14075 (vsm3tt1bq_u32): Ditto.
14076 (vsm3tt2aq_u32): Ditto.
14077 (vsm3tt2bq_u32): Ditto.
14078 (vsm3partw1q_u32): Ditto.
14079 (vsm3partw2q_u32): Ditto.
14080 (vsm4eq_u32): Ditto.
14081 (vsm4ekeyq_u32): Ditto.
14082 (doc/invoke.texi): Document new sm4 option.
14084 2018-01-10 Michael Collison <michael.collison@arm.com>
14086 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14087 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14088 (AARCH64_FL_FOR_ARCH8_4): New.
14089 (AARCH64_FL_V8_4): New flag.
14090 (doc/invoke.texi): Document new armv8.4-a option.
14092 2018-01-10 Michael Collison <michael.collison@arm.com>
14094 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14095 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14096 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14097 * config/aarch64/aarch64-option-extension.def: Add
14098 AARCH64_OPT_EXTENSION of 'sha2'.
14099 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14100 (crypto): Disable sha2 and aes if crypto disabled.
14101 (crypto): Enable aes and sha2 if enabled.
14102 (simd): Disable sha2 and aes if simd disabled.
14103 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14105 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14106 (TARGET_SHA2): New feature flag for sha2.
14107 (TARGET_AES): New feature flag for aes.
14108 * config/aarch64/aarch64-simd.md:
14109 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14110 conditional on TARGET_AES.
14111 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14112 (aarch64_crypto_sha1hsi): Make pattern conditional
14114 (aarch64_crypto_sha1hv4si): Ditto.
14115 (aarch64_be_crypto_sha1hv4si): Ditto.
14116 (aarch64_crypto_sha1su1v4si): Ditto.
14117 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14118 (aarch64_crypto_sha1su0v4si): Ditto.
14119 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14120 (aarch64_crypto_sha256su0v4si): Ditto.
14121 (aarch64_crypto_sha256su1v4si): Ditto.
14122 (doc/invoke.texi): Document new aes and sha2 options.
14124 2018-01-10 Martin Sebor <msebor@redhat.com>
14126 PR tree-optimization/83781
14127 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14130 2018-01-11 Martin Sebor <msebor@gmail.com>
14131 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14133 PR tree-optimization/83501
14134 PR tree-optimization/81703
14136 * tree-ssa-strlen.c (get_string_cst): Rename...
14137 (get_string_len): ...to this. Handle global constants.
14138 (handle_char_store): Adjust.
14140 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
14141 Jim Wilson <jimw@sifive.com>
14143 * config/riscv/riscv-protos.h (riscv_output_return): New.
14144 * config/riscv/riscv.c (struct machine_function): New naked_p field.
14145 (riscv_attribute_table, riscv_output_return),
14146 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14147 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14148 (riscv_compute_frame_info): Only compute frame->mask if not a naked
14150 (riscv_expand_prologue): Add early return for naked function.
14151 (riscv_expand_epilogue): Likewise.
14152 (riscv_function_ok_for_sibcall): Return false for naked function.
14153 (riscv_set_current_function): New.
14154 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14155 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14156 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14157 * doc/extend.texi (RISC-V Function Attributes): New.
14159 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14161 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14162 check for 128-bit long double before checking TCmode.
14163 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14164 128-bit long doubles before checking TFmode or TCmode.
14165 (FLOAT128_IBM_P): Likewise.
14167 2018-01-10 Martin Sebor <msebor@redhat.com>
14169 PR tree-optimization/83671
14170 * builtins.c (c_strlen): Unconditionally return zero for the empty
14172 Use -Warray-bounds for warnings.
14173 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14174 for non-constant array indices with COMPONENT_REF, arrays of
14175 arrays, and pointers to arrays.
14176 (gimple_fold_builtin_strlen): Determine and set length range for
14177 non-constant character arrays.
14179 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
14181 PR middle-end/81897
14182 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14185 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
14187 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14189 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14192 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14193 VECTOR_MEM_ALTIVEC_OR_VSX_P.
14194 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14195 indexed_or_indirect_operand predicate.
14196 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14197 (*vsx_le_perm_load_v8hi): Likewise.
14198 (*vsx_le_perm_load_v16qi): Likewise.
14199 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14200 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14201 (*vsx_le_perm_store_v8hi): Likewise.
14202 (*vsx_le_perm_store_v16qi): Likewise.
14203 (eight unnamed splitters): Likewise.
14205 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14207 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14208 * config/rs6000/emmintrin.h: Likewise.
14209 * config/rs6000/mmintrin.h: Likewise.
14210 * config/rs6000/xmmintrin.h: Likewise.
14212 2018-01-10 David Malcolm <dmalcolm@redhat.com>
14215 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14217 * tree.c (tree_nop_conversion): Return true for location wrapper
14219 (maybe_wrap_with_location): New function.
14220 (selftest::check_strip_nops): New function.
14221 (selftest::test_location_wrappers): New function.
14222 (selftest::tree_c_tests): Call it.
14223 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14224 (maybe_wrap_with_location): New decl.
14225 (EXPR_LOCATION_WRAPPER_P): New macro.
14226 (location_wrapper_p): New inline function.
14227 (tree_strip_any_location_wrapper): New inline function.
14229 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
14232 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14233 stack_realign_offset for the largest alignment of stack slot
14235 (ix86_find_max_used_stack_alignment): New function.
14236 (ix86_finalize_stack_frame_flags): Use it. Set
14237 max_used_stack_alignment if we don't realign stack.
14238 * config/i386/i386.h (machine_function): Add
14239 max_used_stack_alignment.
14241 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
14243 * config/arm/arm.opt (-mbranch-cost): New option.
14244 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
14247 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
14250 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
14251 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
14253 2018-01-10 Richard Biener <rguenther@suse.de>
14256 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
14257 early out so it also covers the case where we have a non-NULL
14260 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14262 PR tree-optimization/83753
14263 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
14264 for non-strided grouped accesses if the number of elements is 1.
14266 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14269 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
14270 * i386.h (TARGET_USE_GATHER): Define.
14271 * x86-tune.def (X86_TUNE_USE_GATHER): New.
14273 2018-01-10 Martin Liska <mliska@suse.cz>
14276 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
14277 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
14279 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
14280 CLEANUP_NO_PARTITIONING is not set.
14282 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14284 * doc/rtl.texi: Remove documentation of (const ...) wrappers
14285 for vectors, as a partial revert of r254296.
14286 * rtl.h (const_vec_p): Delete.
14287 (const_vec_duplicate_p): Don't test for vector CONSTs.
14288 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
14289 * expmed.c (make_tree): Likewise.
14292 * common.md (E, F): Use CONSTANT_P instead of checking for
14294 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
14295 checking for CONST_VECTOR.
14297 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14299 PR middle-end/83575
14300 * predict.c (force_edge_cold): Handle in more sane way edges
14301 with no prediction.
14303 2018-01-09 Carl Love <cel@us.ibm.com>
14305 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
14307 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
14308 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
14309 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
14310 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
14311 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
14312 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
14313 * config/rs6000/rs6000-protos.h: Add extern defition for
14314 rs6000_generate_float2_double_code.
14315 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
14317 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
14318 (float2_v2df): Add define_expand.
14320 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
14323 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
14324 op_mode in the force_to_mode call.
14326 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14328 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
14329 instead of checking each element individually.
14330 (aarch64_evpc_uzp): Likewise.
14331 (aarch64_evpc_zip): Likewise.
14332 (aarch64_evpc_ext): Likewise.
14333 (aarch64_evpc_rev): Likewise.
14334 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
14335 instead of checking each element individually. Return true without
14337 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
14338 whether all selected elements come from the same input, instead of
14339 checking each element individually. Remove calls to gen_rtx_REG,
14340 start_sequence and end_sequence and instead assert that no rtl is
14343 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14345 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
14346 order of HIGH and CONST checks.
14348 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14350 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
14351 if the destination isn't an SSA_NAME.
14353 2018-01-09 Richard Biener <rguenther@suse.de>
14355 PR tree-optimization/83668
14356 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
14358 (canonicalize_loop_form): ... here, renamed from ...
14359 (canonicalize_loop_closed_ssa_form): ... this and amended to
14360 swap successor edges for loop exit blocks to make us use
14361 the RPO order we need for initial schedule generation.
14363 2018-01-09 Joseph Myers <joseph@codesourcery.com>
14365 PR tree-optimization/64811
14366 * match.pd: When optimizing comparisons with Inf, avoid
14367 introducing or losing exceptions from comparisons with NaN.
14369 2018-01-09 Martin Liska <mliska@suse.cz>
14372 * asan.c (shadow_mem_size): Add gcc_assert.
14374 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
14376 Don't save registers in main().
14379 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
14380 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
14381 * config/avr/avr.c (avr_set_current_function): Don't error if
14382 naked, OS_task or OS_main are specified at the same time.
14383 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
14385 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
14387 * common/config/avr/avr-common.c (avr_option_optimization_table):
14388 Switch on -mmain-is-OS_task for optimizing compilations.
14390 2018-01-09 Richard Biener <rguenther@suse.de>
14392 PR tree-optimization/83572
14393 * graphite.c: Include cfganal.h.
14394 (graphite_transform_loops): Connect infinite loops to exit
14395 and remove fake edges at the end.
14397 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14399 * ipa-inline.c (edge_badness): Revert accidental checkin.
14401 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14404 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
14405 symbols; not inline clones.
14407 2018-01-09 Jakub Jelinek <jakub@redhat.com>
14410 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
14411 hard registers. Formatting fixes.
14413 PR preprocessor/83722
14414 * gcc.c (try_generate_repro): Pass
14415 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
14416 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
14419 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
14420 Kito Cheng <kito.cheng@gmail.com>
14422 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
14423 (riscv_leaf_function_p): Delete.
14424 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
14426 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14428 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
14430 (do_ifelse): New function.
14431 (do_isel): New function.
14432 (do_sub3): New function.
14433 (do_add3): New function.
14434 (do_load_mask_compare): New function.
14435 (do_overlap_load_compare): New function.
14436 (expand_compare_loop): New function.
14437 (expand_block_compare): Call expand_compare_loop() when appropriate.
14438 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
14439 option description.
14440 (-mblock-compare-inline-loop-limit): New option.
14442 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14445 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
14446 Reverse order of second and third operands in first alternative.
14447 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
14448 of first and second elements in UNSPEC_VPERMR vector.
14449 (altivec_expand_vec_perm_le): Likewise.
14451 2018-01-08 Jeff Law <law@redhat.com>
14453 PR rtl-optimizatin/81308
14454 * tree-switch-conversion.c (cfg_altered): New file scoped static.
14455 (process_switch): If group_case_labels makes a change, then set
14457 (pass_convert_switch::execute): If a switch is converted, then
14458 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
14460 PR rtl-optimization/81308
14461 * recog.c (split_all_insns): Conditionally cleanup the CFG after
14464 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
14466 PR target/83663 - Revert r255946
14467 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
14468 generation for cases where splatting a value is not useful.
14469 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
14470 across a vec_duplicate and a paradoxical subreg forming a vector
14471 mode to a vec_concat.
14473 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14475 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
14476 -march=armv8.3-a variants.
14477 * config/arm/t-multilib: Likewise.
14478 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
14480 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14482 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
14484 (cceq_ior_compare_complement): Give it a name so I can use it, and
14485 change boolean_or_operator predicate to boolean_operator so it can
14486 be used to generate a crand.
14487 (eqne): New code iterator.
14488 (bd/bd_neg): New code_attrs.
14489 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
14490 a single define_insn.
14491 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
14492 decrement (bdnzt/bdnzf/bdzt/bdzf).
14493 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
14494 with the new names of the branch decrement patterns, and added the
14495 names of the branch decrement conditional patterns.
14497 2018-01-08 Richard Biener <rguenther@suse.de>
14499 PR tree-optimization/83563
14500 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
14503 2018-01-08 Richard Biener <rguenther@suse.de>
14505 PR middle-end/83713
14506 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
14508 2018-01-08 Richard Biener <rguenther@suse.de>
14510 PR tree-optimization/83685
14511 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
14512 references to abnormals.
14514 2018-01-08 Richard Biener <rguenther@suse.de>
14517 * dwarf2out.c (output_indirect_strings): Handle empty
14518 skeleton_debug_str_hash.
14519 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
14521 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14523 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
14524 (emit_store_direct): Likewise.
14525 (arc_trampoline_adjust_address): Likewise.
14526 (arc_asm_trampoline_template): New function.
14527 (arc_initialize_trampoline): Use asm_trampoline_template.
14528 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
14529 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
14530 * config/arc/arc.md (flush_icache): Delete pattern.
14532 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14534 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
14535 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
14538 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14541 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
14542 by not USED_FOR_TARGET.
14543 (make_pass_resolve_sw_modes): Likewise.
14545 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14547 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
14550 2018-01-08 Richard Biener <rguenther@suse.de>
14552 PR middle-end/83580
14553 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
14555 2018-01-08 Richard Biener <rguenther@suse.de>
14557 PR middle-end/83517
14558 * match.pd ((t * 2) / 2) -> t): Add missing :c.
14560 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
14562 PR middle-end/81897
14563 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
14564 basic blocks with a small number of successors.
14565 (convert_control_dep_chain_into_preds): Improve handling of
14567 (dump_predicates): Split apart into...
14568 (dump_pred_chain): ...here...
14569 (dump_pred_info): ...and here.
14570 (can_one_predicate_be_invalidated_p): Add debugging printfs.
14571 (can_chain_union_be_invalidated_p): Improve check for invalidation
14573 (uninit_uses_cannot_happen): Avoid unnecessary if
14574 convert_control_dep_chain_into_preds yielded nothing.
14576 2018-01-06 Martin Sebor <msebor@redhat.com>
14578 PR tree-optimization/83640
14579 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
14580 subtracting negative offset from size.
14581 (builtin_access::overlap): Adjust offset bounds of the access to fall
14582 within the size of the object if possible.
14584 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
14586 PR rtl-optimization/83699
14587 * expmed.c (extract_bit_field_1): Restrict the vector usage of
14588 extract_bit_field_as_subreg to cases in which the extracted
14589 value is also a vector.
14591 * lra-constraints.c (process_alt_operands): Test for the equivalence
14592 substitutions when detecting a possible reload cycle.
14594 2018-01-06 Jakub Jelinek <jakub@redhat.com>
14597 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
14598 by default if flag_selective_schedling{,2}. Formatting fixes.
14600 PR rtl-optimization/83682
14601 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
14602 if it has non-VECTOR_MODE element mode.
14603 (vec_duplicate_p): Likewise.
14605 PR middle-end/83694
14606 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
14607 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
14609 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14612 * config/i386/i386-builtin.def
14613 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
14614 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
14615 Require also OPTION_MASK_ISA_AVX512F in addition to
14616 OPTION_MASK_ISA_GFNI.
14617 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
14618 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
14619 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
14620 to OPTION_MASK_ISA_GFNI.
14621 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
14622 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
14623 OPTION_MASK_ISA_AVX512BW.
14624 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
14625 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
14626 addition to OPTION_MASK_ISA_GFNI.
14627 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
14628 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
14629 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
14630 to OPTION_MASK_ISA_GFNI.
14631 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
14632 a requirement for all ISAs rather than any of them with a few
14634 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
14636 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
14637 bitmasks to be enabled with 3 exceptions, instead of requiring any
14638 enabled ISA with lots of exceptions.
14639 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
14640 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
14641 Change avx512bw in isa attribute to avx512f.
14642 * config/i386/sgxintrin.h: Add license boilerplate.
14643 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
14644 to __AVX512F__ and __AVX512VL to __AVX512VL__.
14645 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
14646 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
14648 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
14649 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
14650 temporarily sse2 rather than sse if not enabled already.
14653 * config/i386/sse.md (VI248_VLBW): Rename to ...
14654 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
14655 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
14656 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
14657 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
14658 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
14659 mode iterator instead of VI248_VLBW.
14661 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
14663 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
14664 (record_modified): Skip clobbers; add debug output.
14665 (param_change_prob): Use sreal frequencies.
14667 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14669 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
14670 punt for user-aligned variables.
14672 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14674 * tree-chrec.c (chrec_contains_symbols): Return true for
14677 2018-01-05 Sudakshina Das <sudi.das@arm.com>
14680 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
14681 of (x|y) == x for BICS pattern.
14683 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14685 PR tree-optimization/83605
14686 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
14687 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
14690 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
14692 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
14693 * config/epiphany/rtems.h: New file.
14695 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14696 Uros Bizjak <ubizjak@gmail.com>
14699 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
14700 QIreg_operand instead of register_operand predicate.
14701 * config/i386/i386.c (ix86_rop_should_change_byte_p,
14702 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
14703 comments instead of -fmitigate[-_]rop.
14705 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14708 * cgraphunit.c (symbol_table::compile): Switch to text_section
14709 before calling assembly_start debug hook.
14710 * run-rtl-passes.c (run_rtl_passes): Likewise.
14713 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14715 * tree-vrp.c (extract_range_from_binary_expr_1): Check
14716 range_int_cst_p rather than !symbolic_range_p before calling
14717 extract_range_from_multiplicative_op_1.
14719 2018-01-04 Jeff Law <law@redhat.com>
14721 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
14722 redundant test in assertion.
14724 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14726 * doc/rtl.texi: Document machine_mode wrapper classes.
14728 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14730 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
14731 using tree_to_uhwi.
14733 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14735 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
14736 the VEC_PERM_EXPR fold to fail.
14738 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14741 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
14742 to switched_sections.
14744 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14747 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
14748 test for d.testing.
14750 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
14753 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
14754 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
14756 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14759 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
14760 is BLKmode and bitpos not zero or mode change is needed.
14762 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14765 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
14768 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
14771 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
14772 instead of MULT rtx. Update all corresponding splitters.
14773 (*saddl_se): Ditto.
14774 (*ssub<modesuffix>): Ditto.
14775 (*ssubl_se): Ditto.
14776 (*cmp_sadd_di): Update split patterns.
14777 (*cmp_sadd_si): Ditto.
14778 (*cmp_sadd_sidi): Ditto.
14779 (*cmp_ssub_di): Ditto.
14780 (*cmp_ssub_si): Ditto.
14781 (*cmp_ssub_sidi): Ditto.
14782 * config/alpha/predicates.md (const23_operand): New predicate.
14783 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
14784 Look for ASHIFT, not MULT inner operand.
14785 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
14787 2018-01-04 Martin Liska <mliska@suse.cz>
14789 PR gcov-profile/83669
14790 * gcov.c (output_intermediate_file): Add version to intermediate
14792 * doc/gcov.texi: Document new field 'version' in intermediate
14793 file format. Fix location of '-k' option of gcov command.
14795 2018-01-04 Martin Liska <mliska@suse.cz>
14798 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
14800 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14802 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
14804 2018-01-03 Martin Sebor <msebor@redhat.com>
14806 PR tree-optimization/83655
14807 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
14808 checking calls with invalid arguments.
14810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14812 * tree-vect-stmts.c (vect_get_store_rhs): New function.
14813 (vectorizable_mask_load_store): Delete.
14814 (vectorizable_call): Return false for masked loads and stores.
14815 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
14816 instead of gimple_assign_rhs1.
14817 (vectorizable_load): Handle IFN_MASK_LOAD.
14818 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
14820 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14822 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
14824 (vectorizable_mask_load_store): ...here.
14825 (vectorizable_load): ...and here.
14827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14829 * tree-vect-stmts.c (vect_build_all_ones_mask)
14830 (vect_build_zero_merge_argument): New functions, split out from...
14831 (vectorizable_load): ...here.
14833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14835 * tree-vect-stmts.c (vect_check_store_rhs): New function,
14837 (vectorizable_mask_load_store): ...here.
14838 (vectorizable_store): ...and here.
14840 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14842 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
14844 (vectorizable_mask_load_store): ...here.
14846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14848 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
14849 (vect_model_store_cost): Take a vec_load_store_type instead of a
14851 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
14852 (vect_model_store_cost): Take a vec_load_store_type instead of a
14854 (vectorizable_mask_load_store): Update accordingly.
14855 (vectorizable_store): Likewise.
14856 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
14858 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14860 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
14861 IFN_MASK_LOAD calls here rather than...
14862 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
14864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14865 Alan Hayward <alan.hayward@arm.com>
14866 David Sherwood <david.sherwood@arm.com>
14868 * expmed.c (extract_bit_field_1): For vector extracts,
14869 fall back to extract_bit_field_as_subreg if vec_extract
14872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14873 Alan Hayward <alan.hayward@arm.com>
14874 David Sherwood <david.sherwood@arm.com>
14876 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
14877 they are variable or constant sized.
14878 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
14879 slots for constant-sized data.
14881 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14882 Alan Hayward <alan.hayward@arm.com>
14883 David Sherwood <david.sherwood@arm.com>
14885 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
14886 handling COND_EXPRs with boolean comparisons, try to find a better
14887 basis for the mask type than the boolean itself.
14889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14891 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
14892 is calculated and how it can be overridden.
14893 * genmodes.c (max_bitsize_mode_any_mode): New variable.
14894 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
14896 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
14899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14900 Alan Hayward <alan.hayward@arm.com>
14901 David Sherwood <david.sherwood@arm.com>
14903 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
14904 Remove the mode argument.
14905 (aarch64_simd_valid_immediate): Remove the mode and inverse
14907 * config/aarch64/iterators.md (bitsize): New iterator.
14908 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
14909 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
14910 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
14911 aarch64_simd_valid_immediate.
14912 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
14913 (aarch64_reg_or_bic_imm): Likewise.
14914 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
14915 with an insn_type enum and msl with a modifier_type enum.
14916 Replace element_width with a scalar_mode. Change the shift
14917 to unsigned int. Add constructors for scalar_float_mode and
14918 scalar_int_mode elements.
14919 (aarch64_vect_float_const_representable_p): Delete.
14920 (aarch64_can_const_movi_rtx_p)
14921 (aarch64_simd_scalar_immediate_valid_for_move)
14922 (aarch64_simd_make_constant): Update call to
14923 aarch64_simd_valid_immediate.
14924 (aarch64_advsimd_valid_immediate_hs): New function.
14925 (aarch64_advsimd_valid_immediate): Likewise.
14926 (aarch64_simd_valid_immediate): Remove mode and inverse
14927 arguments. Rewrite to use the above. Use const_vec_duplicate_p
14928 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
14929 and aarch64_float_const_representable_p on the result.
14930 (aarch64_output_simd_mov_immediate): Remove mode argument.
14931 Update call to aarch64_simd_valid_immediate and use of
14932 simd_immediate_info.
14933 (aarch64_output_scalar_simd_mov_immediate): Update call
14936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14937 Alan Hayward <alan.hayward@arm.com>
14938 David Sherwood <david.sherwood@arm.com>
14940 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
14941 (mode_nunits): Likewise CONST_MODE_NUNITS.
14942 * machmode.def (ADJUST_NUNITS): Document.
14943 * genmodes.c (mode_data::need_nunits_adj): New field.
14944 (blank_mode): Update accordingly.
14945 (adj_nunits): New variable.
14946 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
14948 (emit_mode_size_inline): Set need_bytesize_adj for all modes
14949 listed in adj_nunits.
14950 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
14951 listed in adj_nunits. Don't emit case statements for such modes.
14952 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
14953 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
14954 nothing if adj_nunits is nonnull.
14955 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
14956 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
14957 (emit_mode_fbit): Update use of print_maybe_const_decl.
14958 (emit_move_size): Likewise. Treat the array as non-const
14960 (emit_mode_adjustments): Handle adj_nunits.
14962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14964 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
14965 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
14966 (VECTOR_MODES): Use it.
14967 (make_vector_modes): Take the prefix as an argument.
14969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14970 Alan Hayward <alan.hayward@arm.com>
14971 David Sherwood <david.sherwood@arm.com>
14973 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
14974 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
14975 for MODE_VECTOR_BOOL.
14976 * machmode.def (VECTOR_BOOL_MODE): Document.
14977 * genmodes.c (VECTOR_BOOL_MODE): New macro.
14978 (make_vector_bool_mode): New function.
14979 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
14981 * lto-streamer-in.c (lto_input_mode_table): Likewise.
14982 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
14984 * stor-layout.c (int_mode_for_mode): Likewise.
14985 * tree.c (build_vector_type_for_mode): Likewise.
14986 * varasm.c (output_constant_pool_2): Likewise.
14987 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
14988 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
14989 for MODE_VECTOR_BOOL.
14990 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
14991 of mode class checks.
14992 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
14993 instead of a list of mode class checks.
14994 (expand_vector_scalar_condition): Likewise.
14995 (type_for_widest_vector_mode): Handle BImode as an inner mode.
14997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14998 Alan Hayward <alan.hayward@arm.com>
14999 David Sherwood <david.sherwood@arm.com>
15001 * machmode.h (mode_size): Change from unsigned short to
15003 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
15004 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15005 or if measurement_type is not polynomial.
15006 (fixed_size_mode::includes_p): Check for constant-sized modes.
15007 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
15008 return a poly_uint16 rather than an unsigned short.
15009 (emit_mode_size): Change the type of mode_size from unsigned short
15010 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
15011 (emit_mode_adjustments): Cope with polynomial vector sizes.
15012 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15014 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15016 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
15017 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
15018 * caller-save.c (setup_save_areas): Likewise.
15019 (replace_reg_with_saved_mem): Likewise.
15020 * calls.c (emit_library_call_value_1): Likewise.
15021 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
15022 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
15023 (gen_lowpart_for_combine): Likewise.
15024 * convert.c (convert_to_integer_1): Likewise.
15025 * cse.c (equiv_constant, cse_insn): Likewise.
15026 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
15027 (cselib_subst_to_values): Likewise.
15028 * dce.c (word_dce_process_block): Likewise.
15029 * df-problems.c (df_word_lr_mark_ref): Likewise.
15030 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
15031 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
15032 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
15033 (rtl_for_decl_location): Likewise.
15034 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
15035 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
15036 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
15037 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
15038 (expand_expr_real_1): Likewise.
15039 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
15040 (pad_below): Likewise.
15041 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15042 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
15043 * ira.c (get_subreg_tracking_sizes): Likewise.
15044 * ira-build.c (ira_create_allocno_objects): Likewise.
15045 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
15046 (ira_sort_regnos_for_alter_reg): Likewise.
15047 * ira-costs.c (record_operand_costs): Likewise.
15048 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
15049 (resolve_simple_move): Likewise.
15050 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
15051 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
15052 (lra_constraints): Likewise.
15053 (CONST_POOL_OK_P): Reject variable-sized modes.
15054 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
15055 (add_pseudo_to_slot, lra_spill): Likewise.
15056 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15057 * optabs-query.c (get_best_extraction_insn): Likewise.
15058 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15059 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
15060 (expand_mult_highpart, valid_multiword_target_p): Likewise.
15061 * recog.c (offsettable_address_addr_space_p): Likewise.
15062 * regcprop.c (maybe_mode_change): Likewise.
15063 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
15064 * regrename.c (build_def_use): Likewise.
15065 * regstat.c (dump_reg_info): Likewise.
15066 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
15067 (find_reloads, find_reloads_subreg_address): Likewise.
15068 * reload1.c (eliminate_regs_1): Likewise.
15069 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
15070 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
15071 (simplify_binary_operation_1, simplify_subreg): Likewise.
15072 * targhooks.c (default_function_arg_padding): Likewise.
15073 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
15074 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
15075 (verify_gimple_assign_ternary): Likewise.
15076 * tree-inline.c (estimate_move_cost): Likewise.
15077 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15078 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
15079 (get_address_cost_ainc): Likewise.
15080 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
15081 (vect_supportable_dr_alignment): Likewise.
15082 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15083 (vectorizable_reduction): Likewise.
15084 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15085 (vectorizable_operation, vectorizable_load): Likewise.
15086 * tree.c (build_same_sized_truth_vector_type): Likewise.
15087 * valtrack.c (cleanup_auto_inc_dec): Likewise.
15088 * var-tracking.c (emit_note_insn_var_location): Likewise.
15089 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15090 (ADDR_VEC_ALIGN): Likewise.
15092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15093 Alan Hayward <alan.hayward@arm.com>
15094 David Sherwood <david.sherwood@arm.com>
15096 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
15098 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15099 or if measurement_type is polynomial.
15100 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15101 * combine.c (make_extraction): Likewise.
15102 * dse.c (find_shift_sequence): Likewise.
15103 * dwarf2out.c (mem_loc_descriptor): Likewise.
15104 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15105 (extract_bit_field, extract_low_bits): Likewise.
15106 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15107 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15108 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15109 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15110 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15111 * reload.c (find_reloads): Likewise.
15112 * reload1.c (alter_reg): Likewise.
15113 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15114 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15115 * tree-if-conv.c (predicate_mem_writes): Likewise.
15116 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15117 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15118 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15119 * valtrack.c (dead_debug_insert_temp): Likewise.
15120 * varasm.c (mergeable_constant_section): Likewise.
15121 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15123 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15124 Alan Hayward <alan.hayward@arm.com>
15125 David Sherwood <david.sherwood@arm.com>
15127 * expr.c (expand_assignment): Cope with polynomial mode sizes
15128 when assigning to a CONCAT.
15130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15131 Alan Hayward <alan.hayward@arm.com>
15132 David Sherwood <david.sherwood@arm.com>
15134 * machmode.h (mode_precision): Change from unsigned short to
15136 (mode_to_precision): Return a poly_uint16 rather than an unsigned
15138 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15139 or if measurement_type is not polynomial.
15140 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
15141 in which the mode is already known to be a scalar_int_mode.
15142 * genmodes.c (emit_mode_precision): Change the type of mode_precision
15143 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
15145 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15146 for GET_MODE_PRECISION.
15147 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15148 for GET_MODE_PRECISION.
15149 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15151 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15152 (expand_field_assignment, make_extraction): Likewise.
15153 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15154 (get_last_value): Likewise.
15155 * convert.c (convert_to_integer_1): Likewise.
15156 * cse.c (cse_insn): Likewise.
15157 * expr.c (expand_expr_real_1): Likewise.
15158 * lra-constraints.c (simplify_operand_subreg): Likewise.
15159 * optabs-query.c (can_atomic_load_p): Likewise.
15160 * optabs.c (expand_atomic_load): Likewise.
15161 (expand_atomic_store): Likewise.
15162 * ree.c (combine_reaching_defs): Likewise.
15163 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15164 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15165 * tree.h (type_has_mode_precision_p): Likewise.
15166 * ubsan.c (instrument_si_overflow): Likewise.
15168 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15169 Alan Hayward <alan.hayward@arm.com>
15170 David Sherwood <david.sherwood@arm.com>
15172 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15173 polynomial numbers of units.
15174 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15175 (valid_vector_subparts_p): New function.
15176 (build_vector_type): Remove temporary shim and take the number
15177 of units as a poly_uint64 rather than an int.
15178 (build_opaque_vector_type): Take the number of units as a
15179 poly_uint64 rather than an int.
15180 * tree.c (build_vector_from_ctor): Handle polynomial
15181 TYPE_VECTOR_SUBPARTS.
15182 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15183 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15184 (build_vector_from_val): If the number of units is variable,
15185 use build_vec_duplicate_cst for constant operands and
15186 VEC_DUPLICATE_EXPR otherwise.
15187 (make_vector_type): Remove temporary is_constant ().
15188 (build_vector_type, build_opaque_vector_type): Take the number of
15189 units as a poly_uint64 rather than an int.
15190 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15192 * cfgexpand.c (expand_debug_expr): Likewise.
15193 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15194 (store_constructor, expand_expr_real_1): Likewise.
15195 (const_scalar_mask_from_tree): Likewise.
15196 * fold-const-call.c (fold_const_reduction): Likewise.
15197 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15198 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15199 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15200 (fold_relational_const): Likewise.
15201 (native_interpret_vector): Likewise. Change the size from an
15202 int to an unsigned int.
15203 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15204 TYPE_VECTOR_SUBPARTS.
15205 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15206 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15207 duplicating a non-constant operand into a variable-length vector.
15208 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15209 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15210 * ipa-icf.c (sem_variable::equals): Likewise.
15211 * match.pd: Likewise.
15212 * omp-simd-clone.c (simd_clone_subparts): Likewise.
15213 * print-tree.c (print_node): Likewise.
15214 * stor-layout.c (layout_type): Likewise.
15215 * targhooks.c (default_builtin_vectorization_cost): Likewise.
15216 * tree-cfg.c (verify_gimple_comparison): Likewise.
15217 (verify_gimple_assign_binary): Likewise.
15218 (verify_gimple_assign_ternary): Likewise.
15219 (verify_gimple_assign_single): Likewise.
15220 * tree-pretty-print.c (dump_generic_node): Likewise.
15221 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15222 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15223 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15224 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15225 (vect_shift_permute_load_chain): Likewise.
15226 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15227 (expand_vector_condition, optimize_vector_constructor): Likewise.
15228 (lower_vec_perm, get_compute_type): Likewise.
15229 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15230 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15231 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15232 (vect_recog_mask_conversion_pattern): Likewise.
15233 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15234 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15235 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15236 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15237 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15238 (vectorizable_shift, vectorizable_operation, vectorizable_store)
15239 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15240 (supportable_widening_operation): Likewise.
15241 (supportable_narrowing_operation): Likewise.
15242 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15244 * varasm.c (output_constant): Likewise.
15246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15247 Alan Hayward <alan.hayward@arm.com>
15248 David Sherwood <david.sherwood@arm.com>
15250 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
15251 so that both the length == 3 and length != 3 cases set up their
15252 own permute vectors. Add comments explaining why we know the
15253 number of elements is constant.
15254 (vect_permute_load_chain): Likewise.
15256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15257 Alan Hayward <alan.hayward@arm.com>
15258 David Sherwood <david.sherwood@arm.com>
15260 * machmode.h (mode_nunits): Change from unsigned char to
15262 (ONLY_FIXED_SIZE_MODES): New macro.
15263 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
15264 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
15265 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
15267 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
15268 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
15269 or if measurement_type is not polynomial.
15270 * genmodes.c (ZERO_COEFFS): New macro.
15271 (emit_mode_nunits_inline): Make mode_nunits_inline return a
15273 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
15274 Use ZERO_COEFFS when emitting initializers.
15275 * data-streamer.h (bp_pack_poly_value): New function.
15276 (bp_unpack_poly_value): Likewise.
15277 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15278 for GET_MODE_NUNITS.
15279 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15280 for GET_MODE_NUNITS.
15281 * tree.c (make_vector_type): Remove temporary shim and make
15282 the real function take the number of units as a poly_uint64
15283 rather than an int.
15284 (build_vector_type_for_mode): Handle polynomial nunits.
15285 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
15286 * emit-rtl.c (const_vec_series_p_1): Likewise.
15287 (gen_rtx_CONST_VECTOR): Likewise.
15288 * fold-const.c (test_vec_duplicate_folding): Likewise.
15289 * genrecog.c (validate_pattern): Likewise.
15290 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
15291 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15292 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
15293 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
15294 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
15295 * rtlanal.c (subreg_get_info): Likewise.
15296 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15297 (vect_grouped_load_supported): Likewise.
15298 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
15299 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
15300 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15301 (simplify_const_unary_operation, simplify_binary_operation_1)
15302 (simplify_const_binary_operation, simplify_ternary_operation)
15303 (test_vector_ops_duplicate, test_vector_ops): Likewise.
15304 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
15305 instead of CONST_VECTOR_NUNITS.
15306 * varasm.c (output_constant_pool_2): Likewise.
15307 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
15308 explicit-encoded elements in the XVEC for variable-length vectors.
15310 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15312 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
15314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15315 Alan Hayward <alan.hayward@arm.com>
15316 David Sherwood <david.sherwood@arm.com>
15318 * coretypes.h (fixed_size_mode): Declare.
15319 (fixed_size_mode_pod): New typedef.
15320 * builtins.h (target_builtins::x_apply_args_mode)
15321 (target_builtins::x_apply_result_mode): Change type to
15322 fixed_size_mode_pod.
15323 * builtins.c (apply_args_size, apply_result_size, result_vector)
15324 (expand_builtin_apply_args_1, expand_builtin_apply)
15325 (expand_builtin_return): Update accordingly.
15327 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15329 * cse.c (hash_rtx_cb): Hash only the encoded elements.
15330 * cselib.c (cselib_hash_rtx): Likewise.
15331 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
15332 CONST_VECTOR encoding.
15334 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15335 Jeff Law <law@redhat.com>
15338 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
15339 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
15340 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
15341 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
15344 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
15345 explicitly probe *sp in a noreturn function if there were any callee
15346 register saves or frame pointer is needed.
15348 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15351 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
15352 BLKmode for ternary, binary or unary expressions.
15355 * var-tracking.c (delete_vta_debug_insn): New inline function.
15356 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
15357 insns from get_insns () to NULL instead of each bb separately.
15358 Use delete_vta_debug_insn. No longer static.
15359 (vt_debug_insns_local, variable_tracking_main_1): Adjust
15360 delete_vta_debug_insns callers.
15361 * rtl.h (delete_vta_debug_insns): Declare.
15362 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
15363 instead of variable_tracking_main.
15365 2018-01-03 Martin Sebor <msebor@redhat.com>
15367 PR tree-optimization/83603
15368 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
15369 arguments past the endof the argument list in functions declared
15370 without a prototype.
15371 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
15372 Avoid checking when arguments are null.
15374 2018-01-03 Martin Sebor <msebor@redhat.com>
15377 * doc/extend.texi (attribute const): Fix a typo.
15378 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
15379 issuing -Wsuggest-attribute for void functions.
15381 2018-01-03 Martin Sebor <msebor@redhat.com>
15383 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
15384 offset_int::from instead of wide_int::to_shwi.
15385 (maybe_diag_overlap): Remove assertion.
15386 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
15387 * gimple-ssa-sprintf.c (format_directive): Same.
15388 (parse_directive): Same.
15389 (sprintf_dom_walker::compute_format_length): Same.
15390 (try_substitute_return_value): Same.
15392 2018-01-03 Jeff Law <law@redhat.com>
15394 PR middle-end/83654
15395 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
15396 non-constant residual for zero at runtime and avoid probing in
15397 that case. Reorganize code for trailing problem to mirror handling
15400 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15402 PR tree-optimization/83501
15403 * tree-ssa-strlen.c (get_string_cst): New.
15404 (handle_char_store): Call get_string_cst.
15406 2018-01-03 Martin Liska <mliska@suse.cz>
15408 PR tree-optimization/83593
15409 * tree-ssa-strlen.c: Include tree-cfg.h.
15410 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
15411 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
15412 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
15414 (strlen_dom_walker::before_dom_children): Call
15415 gimple_purge_dead_eh_edges. Dump tranformation with details
15417 (strlen_dom_walker::before_dom_children): Update call by adding
15418 new argument cleanup_eh.
15419 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
15421 2018-01-03 Martin Liska <mliska@suse.cz>
15424 * cif-code.def (VARIADIC_THUNK): New enum value.
15425 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
15428 2018-01-03 Jan Beulich <jbeulich@suse.com>
15430 * sse.md (mov<mode>_internal): Tighten condition for when to use
15431 vmovdqu<ssescalarsize> for TI and OI modes.
15433 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15435 Update copyright years.
15437 2018-01-03 Martin Liska <mliska@suse.cz>
15440 * ipa-visibility.c (function_and_variable_visibility): Skip
15441 functions with noipa attribure.
15443 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15445 * gcc.c (process_command): Update copyright notice dates.
15446 * gcov-dump.c (print_version): Ditto.
15447 * gcov.c (print_version): Ditto.
15448 * gcov-tool.c (print_version): Ditto.
15449 * gengtype.c (create_file): Ditto.
15450 * doc/cpp.texi: Bump @copying's copyright year.
15451 * doc/cppinternals.texi: Ditto.
15452 * doc/gcc.texi: Ditto.
15453 * doc/gccint.texi: Ditto.
15454 * doc/gcov.texi: Ditto.
15455 * doc/install.texi: Ditto.
15456 * doc/invoke.texi: Ditto.
15458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15460 * vector-builder.h (vector_builder::m_full_nelts): Change from
15461 unsigned int to poly_uint64.
15462 (vector_builder::full_nelts): Update prototype accordingly.
15463 (vector_builder::new_vector): Likewise.
15464 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
15465 (vector_builder::operator ==): Likewise.
15466 (vector_builder::finalize): Likewise.
15467 * int-vector-builder.h (int_vector_builder::int_vector_builder):
15468 Take the number of elements as a poly_uint64 rather than an
15470 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
15471 from unsigned int to poly_uint64.
15472 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
15473 (vec_perm_indices::new_vector): Likewise.
15474 (vec_perm_indices::length): Likewise.
15475 (vec_perm_indices::nelts_per_input): Likewise.
15476 (vec_perm_indices::input_nelts): Likewise.
15477 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
15478 number of elements per input as a poly_uint64 rather than an
15479 unsigned int. Use the original encoding for variable-length
15480 vectors, rather than clamping each individual element.
15481 For the second and subsequent elements in each pattern,
15482 clamp the step and base before clamping their sum.
15483 (vec_perm_indices::series_p): Handle polynomial element counts.
15484 (vec_perm_indices::all_in_range_p): Likewise.
15485 (vec_perm_indices_to_tree): Likewise.
15486 (vec_perm_indices_to_rtx): Likewise.
15487 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
15488 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
15489 (tree_vector_builder::new_binary_operation): Handle polynomial
15490 element counts. Return false if we need to know the number
15491 of elements at compile time.
15492 * fold-const.c (fold_vec_perm): Punt if the number of elements
15493 isn't known at compile time.
15495 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15497 * vec-perm-indices.h (vec_perm_builder): Change element type
15498 from HOST_WIDE_INT to poly_int64.
15499 (vec_perm_indices::element_type): Update accordingly.
15500 (vec_perm_indices::clamp): Handle polynomial element_types.
15501 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15502 (vec_perm_indices::all_in_range_p): Likewise.
15503 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
15505 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
15506 polynomial vec_perm_indices element types.
15507 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
15508 * fold-const.c (fold_vec_perm): Likewise.
15509 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
15510 * tree-vect-generic.c (lower_vec_perm): Likewise.
15511 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15512 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
15513 element type to HOST_WIDE_INT.
15515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15516 Alan Hayward <alan.hayward@arm.com>
15517 David Sherwood <david.sherwood@arm.com>
15519 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
15520 rather than an int. Use plus_constant.
15521 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
15522 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
15524 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15525 Alan Hayward <alan.hayward@arm.com>
15526 David Sherwood <david.sherwood@arm.com>
15528 * calls.c (emit_call_1, expand_call): Change struct_value_size from
15529 a HOST_WIDE_INT to a poly_int64.
15531 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15532 Alan Hayward <alan.hayward@arm.com>
15533 David Sherwood <david.sherwood@arm.com>
15535 * calls.c (load_register_parameters): Cope with polynomial
15536 mode sizes. Require a constant size for BLKmode parameters
15537 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
15538 forces a parameter to be padded at the lsb end in order to
15539 fill a complete number of words, require the parameter size
15540 to be ordered wrt UNITS_PER_WORD.
15542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15543 Alan Hayward <alan.hayward@arm.com>
15544 David Sherwood <david.sherwood@arm.com>
15546 * reload1.c (spill_stack_slot_width): Change element type
15547 from unsigned int to poly_uint64_pod.
15548 (alter_reg): Treat mode sizes as polynomial.
15550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15551 Alan Hayward <alan.hayward@arm.com>
15552 David Sherwood <david.sherwood@arm.com>
15554 * reload.c (complex_word_subreg_p): New function.
15555 (reload_inner_reg_of_subreg, push_reload): Use it.
15557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15558 Alan Hayward <alan.hayward@arm.com>
15559 David Sherwood <david.sherwood@arm.com>
15561 * lra-constraints.c (process_alt_operands): Reject matched
15562 operands whose sizes aren't ordered.
15563 (match_reload): Refer to this check here.
15565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15566 Alan Hayward <alan.hayward@arm.com>
15567 David Sherwood <david.sherwood@arm.com>
15569 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
15570 that the mode size is in the set {1, 2, 4, 8, 16}.
15572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15573 Alan Hayward <alan.hayward@arm.com>
15574 David Sherwood <david.sherwood@arm.com>
15576 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
15577 Use plus_constant instead of gen_rtx_PLUS.
15579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15580 Alan Hayward <alan.hayward@arm.com>
15581 David Sherwood <david.sherwood@arm.com>
15583 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
15584 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
15585 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
15586 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
15587 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
15588 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
15589 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
15590 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
15591 * config/i386/i386.c (ix86_push_rounding): ...this new function.
15592 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
15594 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
15595 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
15596 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
15597 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
15598 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
15599 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
15600 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
15601 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
15602 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
15603 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
15605 * expr.c (emit_move_resolve_push): Treat the input and result
15606 of PUSH_ROUNDING as a poly_int64.
15607 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
15608 (emit_push_insn): Likewise.
15609 * lra-eliminations.c (mark_not_eliminable): Likewise.
15610 * recog.c (push_operand): Likewise.
15611 * reload1.c (elimination_effects): Likewise.
15612 * rtlanal.c (nonzero_bits1): Likewise.
15613 * calls.c (store_one_arg): Likewise. Require the padding to be
15614 known at compile time.
15616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15617 Alan Hayward <alan.hayward@arm.com>
15618 David Sherwood <david.sherwood@arm.com>
15620 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
15621 Use plus_constant instead of gen_rtx_PLUS.
15623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15624 Alan Hayward <alan.hayward@arm.com>
15625 David Sherwood <david.sherwood@arm.com>
15627 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
15628 rather than an int.
15630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15631 Alan Hayward <alan.hayward@arm.com>
15632 David Sherwood <david.sherwood@arm.com>
15634 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
15635 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
15636 via stack temporaries. Treat the mode size as polynomial too.
15638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15639 Alan Hayward <alan.hayward@arm.com>
15640 David Sherwood <david.sherwood@arm.com>
15642 * expr.c (expand_expr_real_2): When handling conversions involving
15643 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
15644 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
15645 as a poly_uint64 too.
15647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15648 Alan Hayward <alan.hayward@arm.com>
15649 David Sherwood <david.sherwood@arm.com>
15651 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
15653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15654 Alan Hayward <alan.hayward@arm.com>
15655 David Sherwood <david.sherwood@arm.com>
15657 * combine.c (can_change_dest_mode): Handle polynomial
15658 REGMODE_NATURAL_SIZE.
15659 * expmed.c (store_bit_field_1): Likewise.
15660 * expr.c (store_constructor): Likewise.
15661 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
15662 and polynomial REGMODE_NATURAL_SIZE.
15663 (gen_lowpart_common): Likewise.
15664 * reginfo.c (record_subregs_of_mode): Likewise.
15665 * rtlanal.c (read_modify_subreg_p): Likewise.
15667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15668 Alan Hayward <alan.hayward@arm.com>
15669 David Sherwood <david.sherwood@arm.com>
15671 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
15672 numbers of elements.
15674 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15675 Alan Hayward <alan.hayward@arm.com>
15676 David Sherwood <david.sherwood@arm.com>
15678 * match.pd: Cope with polynomial numbers of vector elements.
15680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15681 Alan Hayward <alan.hayward@arm.com>
15682 David Sherwood <david.sherwood@arm.com>
15684 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
15685 in a POINTER_PLUS_EXPR.
15687 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15688 Alan Hayward <alan.hayward@arm.com>
15689 David Sherwood <david.sherwood@arm.com>
15691 * omp-simd-clone.c (simd_clone_subparts): New function.
15692 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
15693 (ipa_simd_modify_function_body): Likewise.
15695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15696 Alan Hayward <alan.hayward@arm.com>
15697 David Sherwood <david.sherwood@arm.com>
15699 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
15700 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
15701 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
15702 (expand_vector_condition, vector_element): Likewise.
15703 (subparts_gt): New function.
15704 (get_compute_type): Use subparts_gt.
15705 (count_type_subparts): Delete.
15706 (expand_vector_operations_1): Use subparts_gt instead of
15707 count_type_subparts.
15709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15710 Alan Hayward <alan.hayward@arm.com>
15711 David Sherwood <david.sherwood@arm.com>
15713 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
15714 (vect_compile_time_alias): ...this new function. Do the calculation
15715 on poly_ints rather than trees.
15716 (vect_prune_runtime_alias_test_list): Update call accordingly.
15718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15719 Alan Hayward <alan.hayward@arm.com>
15720 David Sherwood <david.sherwood@arm.com>
15722 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
15724 (vect_schedule_slp_instance): Likewise.
15726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15727 Alan Hayward <alan.hayward@arm.com>
15728 David Sherwood <david.sherwood@arm.com>
15730 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
15731 constant and extern definitions for variable-length vectors.
15732 (vect_get_constant_vectors): Note that the number of units
15733 is known to be constant.
15735 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15736 Alan Hayward <alan.hayward@arm.com>
15737 David Sherwood <david.sherwood@arm.com>
15739 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
15740 of units as polynomial. Choose between WIDE and NARROW based
15743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15744 Alan Hayward <alan.hayward@arm.com>
15745 David Sherwood <david.sherwood@arm.com>
15747 * tree-vect-stmts.c (simd_clone_subparts): New function.
15748 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
15750 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15751 Alan Hayward <alan.hayward@arm.com>
15752 David Sherwood <david.sherwood@arm.com>
15754 * tree-vect-stmts.c (vectorizable_call): Treat the number of
15755 vectors as polynomial. Use build_index_vector for
15756 IFN_GOMP_SIMD_LANE.
15758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15759 Alan Hayward <alan.hayward@arm.com>
15760 David Sherwood <david.sherwood@arm.com>
15762 * tree-vect-stmts.c (get_load_store_type): Treat the number of
15763 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
15764 for variable-length vectors.
15765 (vectorizable_mask_load_store): Treat the number of units as
15766 polynomial, asserting that it is constant if the condition has
15767 already been enforced.
15768 (vectorizable_store, vectorizable_load): Likewise.
15770 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15771 Alan Hayward <alan.hayward@arm.com>
15772 David Sherwood <david.sherwood@arm.com>
15774 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
15775 of units as polynomial. Punt if we can't tell at compile time
15776 which vector contains the final result.
15778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15779 Alan Hayward <alan.hayward@arm.com>
15780 David Sherwood <david.sherwood@arm.com>
15782 * tree-vect-loop.c (vectorizable_induction): Treat the number
15783 of units as polynomial. Punt on SLP inductions. Use an integer
15784 VEC_SERIES_EXPR for variable-length integer reductions. Use a
15785 cast of such a series for variable-length floating-point
15788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15789 Alan Hayward <alan.hayward@arm.com>
15790 David Sherwood <david.sherwood@arm.com>
15792 * tree.h (build_index_vector): Declare.
15793 * tree.c (build_index_vector): New function.
15794 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
15795 of units as polynomial, forcibly converting it to a constant if
15796 vectorizable_reduction has already enforced the condition.
15797 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
15798 to create a {1,2,3,...} vector.
15799 (vectorizable_reduction): Treat the number of units as polynomial.
15800 Choose vectype_in based on the largest scalar element size rather
15801 than the smallest number of units. Enforce the restrictions
15804 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15805 Alan Hayward <alan.hayward@arm.com>
15806 David Sherwood <david.sherwood@arm.com>
15808 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
15809 number of units as polynomial.
15811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15812 Alan Hayward <alan.hayward@arm.com>
15813 David Sherwood <david.sherwood@arm.com>
15815 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
15816 * target.def (autovectorize_vector_sizes): Return the vector sizes
15817 by pointer, using vector_sizes rather than a bitmask.
15818 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
15819 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
15820 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
15822 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
15823 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
15824 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
15825 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
15826 * omp-general.c (omp_max_vf): Likewise.
15827 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15828 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
15829 * tree-vect-loop.c (vect_analyze_loop): Likewise.
15830 * tree-vect-slp.c (vect_slp_bb): Likewise.
15831 * doc/tm.texi: Regenerate.
15832 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
15834 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
15835 the vector size as a poly_uint64 rather than an unsigned int.
15836 (current_vector_size): Change from an unsigned int to a poly_uint64.
15837 (get_vectype_for_scalar_type): Update accordingly.
15838 * tree.h (build_truth_vector_type): Take the size and number of
15839 units as a poly_uint64 rather than an unsigned int.
15840 (build_vector_type): Add a temporary overload that takes
15841 the number of units as a poly_uint64 rather than an unsigned int.
15842 * tree.c (make_vector_type): Likewise.
15843 (build_truth_vector_type): Take the number of units as a poly_uint64
15844 rather than an unsigned int.
15846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15847 Alan Hayward <alan.hayward@arm.com>
15848 David Sherwood <david.sherwood@arm.com>
15850 * target.def (get_mask_mode): Take the number of units and length
15851 as poly_uint64s rather than unsigned ints.
15852 * targhooks.h (default_get_mask_mode): Update accordingly.
15853 * targhooks.c (default_get_mask_mode): Likewise.
15854 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
15855 * doc/tm.texi: Regenerate.
15857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15858 Alan Hayward <alan.hayward@arm.com>
15859 David Sherwood <david.sherwood@arm.com>
15861 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
15862 * omp-general.c (omp_max_vf): Likewise.
15863 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
15864 (expand_omp_simd): Handle polynomial safelen.
15865 * omp-low.c (omplow_simd_context): Add a default constructor.
15866 (omplow_simd_context::max_vf): Change from int to poly_uint64.
15867 (lower_rec_simd_input_clauses): Update accordingly.
15868 (lower_rec_input_clauses): Likewise.
15870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15871 Alan Hayward <alan.hayward@arm.com>
15872 David Sherwood <david.sherwood@arm.com>
15874 * tree-vectorizer.h (vect_nunits_for_cost): New function.
15875 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
15876 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
15877 (vect_analyze_slp_cost): Likewise.
15878 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
15879 (vect_model_load_cost): Likewise.
15881 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15882 Alan Hayward <alan.hayward@arm.com>
15883 David Sherwood <david.sherwood@arm.com>
15885 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
15886 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
15887 from an unsigned int * to a poly_uint64_pod *.
15888 (calculate_unrolling_factor): New function.
15889 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
15891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15892 Alan Hayward <alan.hayward@arm.com>
15893 David Sherwood <david.sherwood@arm.com>
15895 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
15896 from an unsigned int to a poly_uint64.
15897 (_loop_vec_info::slp_unrolling_factor): Likewise.
15898 (_loop_vec_info::vectorization_factor): Change from an int
15900 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
15901 (vect_get_num_vectors): New function.
15902 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
15903 (vect_get_num_copies): Use vect_get_num_vectors.
15904 (vect_analyze_data_ref_dependences): Change max_vf from an int *
15905 to an unsigned int *.
15906 (vect_analyze_data_refs): Change min_vf from an int * to a
15908 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15909 than an unsigned HOST_WIDE_INT.
15910 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
15911 (vect_analyze_data_ref_dependence): Change max_vf from an int *
15912 to an unsigned int *.
15913 (vect_analyze_data_ref_dependences): Likewise.
15914 (vect_compute_data_ref_alignment): Handle polynomial vf.
15915 (vect_enhance_data_refs_alignment): Likewise.
15916 (vect_prune_runtime_alias_test_list): Likewise.
15917 (vect_shift_permute_load_chain): Likewise.
15918 (vect_supportable_dr_alignment): Likewise.
15919 (dependence_distance_ge_vf): Take the vectorization factor as a
15920 poly_uint64 rather than an unsigned HOST_WIDE_INT.
15921 (vect_analyze_data_refs): Change min_vf from an int * to a
15923 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
15924 vfm1 as a poly_uint64 rather than an int. Make the same change
15925 for the returned bound_scalar.
15926 (vect_gen_vector_loop_niters): Handle polynomial vf.
15927 (vect_do_peeling): Likewise. Update call to
15928 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
15929 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
15931 * tree-vect-loop.c (vect_determine_vectorization_factor)
15932 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
15933 (vect_get_known_peeling_cost): Likewise.
15934 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
15935 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
15936 (vect_transform_loop): Likewise. Use the lowest possible VF when
15937 updating the upper bounds of the loop.
15938 (vect_min_worthwhile_factor): Make static. Return an unsigned int
15939 rather than an int.
15940 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
15941 polynomial unroll factors.
15942 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
15943 (vect_make_slp_decision): Likewise.
15944 (vect_supported_load_permutation_p): Likewise, and polynomial
15946 (vect_analyze_slp_cost): Handle polynomial vf.
15947 (vect_slp_analyze_node_operations): Likewise.
15948 (vect_slp_analyze_bb_1): Likewise.
15949 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15950 than an unsigned HOST_WIDE_INT.
15951 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
15952 (vectorizable_load): Handle polynomial vf.
15953 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
15955 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
15957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15958 Alan Hayward <alan.hayward@arm.com>
15959 David Sherwood <david.sherwood@arm.com>
15961 * match.pd: Handle bit operations involving three constants
15962 and try to fold one pair.
15964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15966 * tree-vect-loop-manip.c: Include gimple-fold.h.
15967 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
15968 niters_maybe_zero parameters. Handle other cases besides a step of 1.
15969 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
15970 Add a path that uses a step of VF instead of 1, but disable it
15972 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
15973 and niters_no_overflow parameters. Update calls to
15974 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
15975 Create a new SSA name if the latter choses to use a ste other
15976 than zero, and return it via niters_vector_mult_vf_var.
15977 * tree-vect-loop.c (vect_transform_loop): Update calls to
15978 vect_do_peeling, vect_gen_vector_loop_niters and
15979 slpeel_make_loop_iterate_ntimes.
15980 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
15981 (vect_gen_vector_loop_niters): Update declarations after above changes.
15983 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
15985 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
15986 128-bit round to integer instructions.
15987 (ceil<mode>2): Likewise.
15988 (btrunc<mode>2): Likewise.
15989 (round<mode>2): Likewise.
15991 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15993 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
15994 unaligned VSX load/store on P8/P9.
15995 (expand_block_clear): Allow the use of unaligned VSX
15996 load/store on P8/P9.
15998 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16000 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
16002 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
16003 swap associated with both a load and a store.
16005 2018-01-02 Andrew Waterman <andrew@sifive.com>
16007 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
16008 * config/riscv/riscv.md (clear_cache): Use it.
16010 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
16012 * web.c: Remove out-of-date comment.
16014 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16016 * expr.c (fixup_args_size_notes): Check that any existing
16017 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
16018 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
16019 (emit_single_push_insn): ...here.
16021 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16023 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
16024 (const_vector_encoded_nelts): New function.
16025 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
16026 (const_vector_int_elt, const_vector_elt): Declare.
16027 * emit-rtl.c (const_vector_int_elt_1): New function.
16028 (const_vector_elt): Likewise.
16029 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
16030 of CONST_VECTOR_ELT.
16032 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16034 * expr.c: Include rtx-vector-builder.h.
16035 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
16036 directly on the tree encoding.
16037 (const_vector_from_tree): Likewise.
16038 * optabs.c: Include rtx-vector-builder.h.
16039 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
16040 sequence of "u" values.
16041 * vec-perm-indices.c: Include rtx-vector-builder.h.
16042 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
16043 directly on the vec_perm_indices encoding.
16045 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16047 * doc/rtl.texi (const_vector): Describe new encoding scheme.
16048 * Makefile.in (OBJS): Add rtx-vector-builder.o.
16049 * rtx-vector-builder.h: New file.
16050 * rtx-vector-builder.c: Likewise.
16051 * rtl.h (rtx_def::u2): Add a const_vector field.
16052 (CONST_VECTOR_NPATTERNS): New macro.
16053 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
16054 (CONST_VECTOR_DUPLICATE_P): Likewise.
16055 (CONST_VECTOR_STEPPED_P): Likewise.
16056 (CONST_VECTOR_ENCODED_ELT): Likewise.
16057 (const_vec_duplicate_p): Check for a duplicated vector encoding.
16058 (unwrap_const_vec_duplicate): Likewise.
16059 (const_vec_series_p): Check for a non-duplicated vector encoding.
16060 Say that the function only returns true for integer vectors.
16061 * emit-rtl.c: Include rtx-vector-builder.h.
16062 (gen_const_vec_duplicate_1): Delete.
16063 (gen_const_vector): Call gen_const_vec_duplicate instead of
16064 gen_const_vec_duplicate_1.
16065 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
16066 (gen_const_vec_duplicate): Use rtx_vector_builder.
16067 (gen_const_vec_series): Likewise.
16068 (gen_rtx_CONST_VECTOR): Likewise.
16069 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
16070 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16071 Build a new vector rather than modifying a CONST_VECTOR in-place.
16072 (handle_special_swappables): Update call accordingly.
16073 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
16074 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16075 Build a new vector rather than modifying a CONST_VECTOR in-place.
16076 (handle_special_swappables): Update call accordingly.
16078 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16080 * simplify-rtx.c (simplify_const_binary_operation): Use
16081 CONST_VECTOR_ELT instead of XVECEXP.
16083 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16085 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16086 the selector elements to be different from the data elements
16087 if the selector is a VECTOR_CST.
16088 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16089 ssizetype for the selector.
16091 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16093 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16094 before testing each element individually.
16095 * tree-vect-generic.c (lower_vec_perm): Likewise.
16097 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16099 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16100 * selftest-run-tests.c (selftest::run_tests): Call it.
16101 * vector-builder.h (vector_builder::operator ==): New function.
16102 (vector_builder::operator !=): Likewise.
16103 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16104 (vec_perm_indices::all_from_input_p): New function.
16105 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16106 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16107 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16108 instead of reading the VECTOR_CST directly. Detect whether both
16109 vector inputs are the same before constructing the vec_perm_indices,
16110 and update the number of inputs argument accordingly. Use the
16111 utility functions added above. Only construct sel2 if we need to.
16113 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16115 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16116 the broadcast of the low byte.
16117 (expand_mult_highpart): Use an explicit encoding for the permutes.
16118 * optabs-query.c (can_mult_highpart_p): Likewise.
16119 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16120 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16121 (vectorizable_bswap): Likewise.
16122 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16123 explicit encoding for the power-of-2 permutes.
16124 (vect_permute_store_chain): Likewise.
16125 (vect_grouped_load_supported): Likewise.
16126 (vect_permute_load_chain): Likewise.
16128 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16130 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16131 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16132 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16133 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16134 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16135 (vect_gen_perm_mask_any): Likewise.
16137 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16139 * int-vector-builder.h: New file.
16140 * vec-perm-indices.h: Include int-vector-builder.h.
16141 (vec_perm_indices): Redefine as an int_vector_builder.
16142 (auto_vec_perm_indices): Delete.
16143 (vec_perm_builder): Redefine as a stand-alone class.
16144 (vec_perm_indices::vec_perm_indices): New function.
16145 (vec_perm_indices::clamp): Likewise.
16146 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16147 (vec_perm_indices::new_vector): New function.
16148 (vec_perm_indices::new_expanded_vector): Update for new
16149 vec_perm_indices class.
16150 (vec_perm_indices::rotate_inputs): New function.
16151 (vec_perm_indices::all_in_range_p): Operate directly on the
16152 encoded form, without computing elided elements.
16153 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16154 encoding. Update for new vec_perm_indices class.
16155 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16156 the given vec_perm_builder.
16157 (expand_vec_perm_var): Update vec_perm_builder constructor.
16158 (expand_mult_highpart): Use vec_perm_builder instead of
16159 auto_vec_perm_indices.
16160 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16161 vec_perm_indices instead of auto_vec_perm_indices. Use a single
16162 or double series encoding as appropriate.
16163 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16164 vec_perm_indices instead of auto_vec_perm_indices.
16165 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16166 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16167 (vect_permute_store_chain): Likewise.
16168 (vect_grouped_load_supported): Likewise.
16169 (vect_permute_load_chain): Likewise.
16170 (vect_shift_permute_load_chain): Likewise.
16171 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16172 (vect_transform_slp_perm_load): Likewise.
16173 (vect_schedule_slp_instance): Likewise.
16174 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16175 (vectorizable_mask_load_store): Likewise.
16176 (vectorizable_bswap): Likewise.
16177 (vectorizable_store): Likewise.
16178 (vectorizable_load): Likewise.
16179 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16180 vec_perm_indices instead of auto_vec_perm_indices. Use
16181 tree_to_vec_perm_builder to read the vector from a tree.
16182 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16183 vec_perm_builder instead of a vec_perm_indices.
16184 (have_whole_vector_shift): Use vec_perm_builder and
16185 vec_perm_indices instead of auto_vec_perm_indices. Leave the
16186 truncation to calc_vec_perm_mask_for_shift.
16187 (vect_create_epilog_for_reduction): Likewise.
16188 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16189 from auto_vec_perm_indices to vec_perm_indices.
16190 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16191 instead of changing individual elements.
16192 (aarch64_vectorize_vec_perm_const): Use new_vector to install
16193 the vector in d.perm.
16194 * config/arm/arm.c (expand_vec_perm_d::perm): Change
16195 from auto_vec_perm_indices to vec_perm_indices.
16196 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16197 instead of changing individual elements.
16198 (arm_vectorize_vec_perm_const): Use new_vector to install
16199 the vector in d.perm.
16200 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16201 Update vec_perm_builder constructor.
16202 (rs6000_expand_interleave): Likewise.
16203 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16204 (rs6000_expand_interleave): Likewise.
16206 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16208 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16209 to qimode could truncate the indices.
16210 * optabs.c (expand_vec_perm_var): Likewise.
16212 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16214 * Makefile.in (OBJS): Add vec-perm-indices.o.
16215 * vec-perm-indices.h: New file.
16216 * vec-perm-indices.c: Likewise.
16217 * target.h (vec_perm_indices): Replace with a forward class
16219 (auto_vec_perm_indices): Move to vec-perm-indices.h.
16220 * optabs.h: Include vec-perm-indices.h.
16221 (expand_vec_perm): Delete.
16222 (selector_fits_mode_p, expand_vec_perm_var): Declare.
16223 (expand_vec_perm_const): Declare.
16224 * target.def (vec_perm_const_ok): Replace with...
16225 (vec_perm_const): ...this new hook.
16226 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16227 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16228 * doc/tm.texi: Regenerate.
16229 * optabs.def (vec_perm_const): Delete.
16230 * doc/md.texi (vec_perm_const): Likewise.
16231 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16232 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16233 expand_vec_perm for constant permutation vectors. Assert that
16234 the mode of variable permutation vectors is the integer equivalent
16235 of the mode that is being permuted.
16236 * optabs-query.h (selector_fits_mode_p): Declare.
16237 * optabs-query.c: Include vec-perm-indices.h.
16238 (selector_fits_mode_p): New function.
16239 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16240 is defined, instead of checking whether the vec_perm_const_optab
16241 exists. Use targetm.vectorize.vec_perm_const instead of
16242 targetm.vectorize.vec_perm_const_ok. Check whether the indices
16243 fit in the vector mode before using a variable permute.
16244 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
16245 vec_perm_indices instead of an rtx.
16246 (expand_vec_perm): Replace with...
16247 (expand_vec_perm_const): ...this new function. Take the selector
16248 as a vec_perm_indices rather than an rtx. Also take the mode of
16249 the selector. Update call to shift_amt_for_vec_perm_mask.
16250 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
16251 Use vec_perm_indices::new_expanded_vector to expand the original
16252 selector into bytes. Check whether the indices fit in the vector
16253 mode before using a variable permute.
16254 (expand_vec_perm_var): Make global.
16255 (expand_mult_highpart): Use expand_vec_perm_const.
16256 * fold-const.c: Includes vec-perm-indices.h.
16257 * tree-ssa-forwprop.c: Likewise.
16258 * tree-vect-data-refs.c: Likewise.
16259 * tree-vect-generic.c: Likewise.
16260 * tree-vect-loop.c: Likewise.
16261 * tree-vect-slp.c: Likewise.
16262 * tree-vect-stmts.c: Likewise.
16263 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
16265 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
16266 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
16267 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
16268 (aarch64_vectorize_vec_perm_const): ...this new function.
16269 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16270 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16271 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
16272 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
16273 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16274 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16275 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
16277 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
16278 check for NEON modes.
16279 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
16280 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
16281 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
16282 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
16284 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
16285 the old VEC_PERM_CONST conditions.
16286 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
16287 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
16288 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
16289 (ia64_vectorize_vec_perm_const_ok): Merge into...
16290 (ia64_vectorize_vec_perm_const): ...this new function.
16291 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
16292 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
16293 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
16294 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
16295 * config/mips/mips.c (mips_expand_vec_perm_const)
16296 (mips_vectorize_vec_perm_const_ok): Merge into...
16297 (mips_vectorize_vec_perm_const): ...this new function.
16298 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
16299 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
16300 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
16301 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
16302 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
16303 (rs6000_expand_vec_perm_const): Delete.
16304 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
16306 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16307 (altivec_expand_vec_perm_const_le): Take each operand individually.
16308 Operate on constant selectors rather than rtxes.
16309 (altivec_expand_vec_perm_const): Likewise. Update call to
16310 altivec_expand_vec_perm_const_le.
16311 (rs6000_expand_vec_perm_const): Delete.
16312 (rs6000_vectorize_vec_perm_const_ok): Delete.
16313 (rs6000_vectorize_vec_perm_const): New function.
16314 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16315 an element count and rtx array.
16316 (rs6000_expand_extract_even): Update call accordingly.
16317 (rs6000_expand_interleave): Likewise.
16318 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
16319 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
16320 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
16321 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
16322 (rs6000_expand_vec_perm_const): Delete.
16323 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16324 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16325 (altivec_expand_vec_perm_const_le): Take each operand individually.
16326 Operate on constant selectors rather than rtxes.
16327 (altivec_expand_vec_perm_const): Likewise. Update call to
16328 altivec_expand_vec_perm_const_le.
16329 (rs6000_expand_vec_perm_const): Delete.
16330 (rs6000_vectorize_vec_perm_const_ok): Delete.
16331 (rs6000_vectorize_vec_perm_const): New function. Remove stray
16332 reference to the SPE evmerge intructions.
16333 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16334 an element count and rtx array.
16335 (rs6000_expand_extract_even): Update call accordingly.
16336 (rs6000_expand_interleave): Likewise.
16337 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
16338 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
16340 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16342 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16344 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
16345 vector mode and that that mode matches the mode of the data
16347 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
16348 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
16349 directly using expand_vec_perm_1 when forcing selectors into
16351 (expand_vec_perm_var): New function, split out from expand_vec_perm.
16353 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16355 * optabs-query.h (can_vec_perm_p): Delete.
16356 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
16357 * optabs-query.c (can_vec_perm_p): Split into...
16358 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
16359 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
16360 particular selector is valid.
16361 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16362 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16363 (vect_grouped_load_supported): Likewise.
16364 (vect_shift_permute_load_chain): Likewise.
16365 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16366 (vect_transform_slp_perm_load): Likewise.
16367 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16368 (vectorizable_bswap): Likewise.
16369 (vect_gen_perm_mask_checked): Likewise.
16370 * fold-const.c (fold_ternary_loc): Likewise. Don't take
16371 implementations of variable permutation vectors into account
16372 when deciding which selector to use.
16373 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
16374 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
16375 with a false third argument.
16376 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
16377 to test whether the constant selector is valid and can_vec_perm_var_p
16378 to test whether a variable selector is valid.
16380 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16382 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
16383 * optabs-query.c (can_vec_perm_p): Likewise.
16384 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
16385 instead of vec_perm_indices.
16386 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
16387 (vect_gen_perm_mask_checked): Likewise,
16388 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
16389 (vect_gen_perm_mask_checked): Likewise,
16391 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16393 * optabs-query.h (qimode_for_vec_perm): Declare.
16394 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
16395 (qimode_for_vec_perm): ...this new function.
16396 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
16398 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16400 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
16401 does not have a conditional at the top.
16403 2018-01-02 Richard Biener <rguenther@suse.de>
16405 * ipa-inline.c (big_speedup_p): Fix expression.
16407 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16410 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
16413 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16417 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
16418 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
16419 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
16420 cond_taken_branch_cost 3->4.
16422 2018-01-01 Jakub Jelinek <jakub@redhat.com>
16424 PR tree-optimization/83581
16425 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
16426 TODO_cleanup_cfg if any changes have been made.
16428 PR middle-end/83608
16429 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
16430 convert_modes if target mode has the right side, but different mode
16433 PR middle-end/83609
16434 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
16435 last argument when extracting from CONCAT. If either from_real or
16436 from_imag is NULL, use expansion through memory. If result is not
16437 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
16438 the parts directly to inner mode, if even that fails, use expansion
16441 PR middle-end/83623
16442 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
16443 check for bswap in mode rather than HImode and use that in expand_unop
16446 Copyright (C) 2018 Free Software Foundation, Inc.
16448 Copying and distribution of this file, with or without modification,
16449 are permitted in any medium without royalty provided the copyright
16450 notice and this notice are preserved.