1 2018-06-12 Martin Sebor <msebor@redhat.com>
3 PR tree-optimization/85259
4 * builtins.c (compute_objsize): Handle constant offsets.
5 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
6 true iff a warning has been issued.
7 * gimple.h (gimple_nonartificial_location): New function.
8 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
9 gimple_nonartificial_location and handle -Wno-system-headers.
10 (handle_builtin_stxncpy): Same.
12 2018-06-12 Martin Sebor <msebor@redhat.com>
15 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
17 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
19 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
20 BUILTIN_VEC_XST entries for pointer to double and long long.
22 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
25 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
27 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
30 2018-06-12 Martin Liska <mliska@suse.cz>
32 * doc/options.texi: Document IntegerRange.
34 2018-06-12 Martin Liska <mliska@suse.cz>
36 * config/i386/i386.opt: Make MPX-related options as Deprecated.
37 * opt-functions.awk: Handle Deprecated flag.
38 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
40 (read_cmdline_option): Report warning for a deprecated option.
41 * opts.h (struct cl_option): Add new field cl_deprecated.
42 (CL_ERR_DEPRECATED): New.
44 2018-06-12 Martin Liska <mliska@suse.cz>
46 * doc/options.texi: Document Deprecated option flag.
48 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
50 * config/arc/arc-arch.h (arc_extras): New enum.
51 (arc_cpu_t):Add field extra.
52 (arc_cpu_types): Consider the extras.
53 * config/arc/arc-cpus.def: Add extras info.
54 * config/arc/arc-opts.h (processor_type): Consider extra field.
55 * config/arc/arc.c (arc_override_options): Handle extra field.
57 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
59 * config/arc/arc-arch.h: Update ARC_OPTX macro.
60 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
62 * config/arc/arc.c (arc_init): Update pic warning.
63 (irq_range): Update irq range parsing warnings.
64 (arc_override_options): Update various warning messages.
65 (arc_handle_aux_attribute): Likewise.
67 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
69 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
71 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
73 * doc/sourcebuild.texi: Document usage of line number 0 in verify
74 compiler messages directives.
76 2018-06-12 Matthew Fortune <mfortune@gmail.com>
78 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
79 * config/mips/mips-tables.opt: Regenerate.
80 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
82 * doc/invoke.texi: Document -march=i6500.
84 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
86 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
87 (i6400_gpmul): Add cpu_unit.
88 (i6400_gpdiv): Likewise.
89 (i6400_msa_add_d): Update reservations.
90 (i6400_msa_int_add) Likewise.
91 (i6400_msa_short_logic3) Likewise.
92 (i6400_msa_short_logic2) Likewise.
93 (i6400_msa_short_logic) Likewise.
94 (i6400_msa_move) Likewise.
95 (i6400_msa_cmp) Likewise.
96 (i6400_msa_short_float2) Likewise.
97 (i6400_msa_div_d) Likewise.
98 (i6400_msa_long_logic1) Likewise.
99 (i6400_msa_long_logic2) Likewise.
100 (i6400_msa_mult) Likewise.
101 (i6400_msa_long_float2) Likewise.
102 (i6400_msa_long_float4) Likewise.
103 (i6400_msa_long_float5) Likewise.
104 (i6400_msa_long_float8) Likewise.
105 (i6400_fpu_fadd): Include frint type.
106 (i6400_fpu_store): New define_insn_reservation.
107 (i6400_fpu_load): Likewise.
108 (i6400_fpu_move): Likewise.
109 (i6400_fpu_fcmp): Likewise.
110 (i6400_fpu_fmadd): Likewise.
111 (i6400_int_mult): Include imul3nc type and update reservation.
112 (i6400_int_div): Include idiv3 type and update reservation.
113 (i6400_int_load): Update to check type not move_type.
114 (i6400_int_store): Likewise.
115 (i6400_int_prefetch): Set zero latency.
117 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
119 * gcc.c: Document new %@{...} sequence.
120 (LINK_COMMAND_SPEC): Use it for the -L switches.
121 (cpp_unique_options): Use it for the -I switches.
122 (at_file_argbuf): New global variable.
123 (in_at_file): Likewise.
124 (alloc_args): Create at_file_argbuf.
125 (clear_args): Truncate at_file_argbuf.
126 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
127 (open_at_file): New function.
128 (close_at_file): Likewise.
129 (create_at_file): Delete.
130 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
133 (validate_switches_from_spec): Deal with %@{...} sequence.
134 (validate_switches): Likewise.
135 (driver::finalize): Call clear_args.
137 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
139 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
141 2018-06-11 Martin Sebor <msebor@redhat.com>
143 * doc/invoke.texi (-Wall): List -Wc++17-compat.
144 (Wno-class-memaccess): Add @opindex.
145 (Wno-templates, Wno-multiple-inheritance): Same.
146 (Wno-virtual-inheritance, Wno-namespaces): Same.
147 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
148 (Wno-format-overflow, Wno-format-truncation): Same.
149 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
150 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
151 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
152 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
153 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
154 (Wno-misspelled-isr): Same.
156 2018-06-11 Martin Sebor <msebor@redhat.com>
158 * PR tree-optimization/86083
159 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
161 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
163 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
165 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
168 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
169 on the correct operand.
170 (*movdi_internal64): Ditto.
172 2018-06-11 Martin Liska <mliska@suse.cz>
174 PR tree-optimization/86089
175 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
177 2018-06-11 Julia Koval <julia.koval@intel.com>
179 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
180 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
181 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
183 2018-06-11 Olivier Hainque <hainque@adacore.com>
185 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
186 for Ada with strict dwarf2.
188 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
191 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
194 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
196 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
198 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
200 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
203 2018-06-08 Martin Liska <mliska@suse.cz>
205 * tree-cfg.h (debug_function): Fix argument type to match
208 2018-06-08 Martin Liska <mliska@suse.cz>
210 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
211 Remove usage of MPX-related (and removed) fields.
212 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
214 2018-06-08 David Malcolm <dmalcolm@redhat.com>
216 * cfg.c (debug): Use TDF_NONE rather than 0.
217 * cfghooks.c (debug): Likewise.
218 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
219 (struct dump_option_value_info): Convert to...
220 (struct kv_pair): ...this template type.
221 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
223 (optinfo_verbosity_options): Likewise.
224 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
226 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
227 than int for "optgroup_flags" param.
228 (dump_generic_expr_loc): Use dump_flags_t rather than int for
230 (dump_dec): Likewise.
231 (dump_finish): Use TDF_NONE rather than 0.
232 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
233 rather than int for "optgroup_flags" param. Use TDF_NONE rather
234 than 0. Update for change to option_ptr.
235 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
236 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
237 0. Update for changes to optinfo_verbosity_options and
239 (opt_info_switch_p): Convert optgroup_flags from int to
241 (dump_basic_block): Use dump_flags_t rather than int
242 for "dump_kind" param.
243 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
244 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
245 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
246 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
247 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
248 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
249 TDF_NONE): Convert from macros to...
250 (enum dump_flag): ...this new enum.
251 (dump_flags_t): Update to use enum.
252 (operator|, operator&, operator~, operator|=, operator&=):
253 Implement for dump_flags_t.
254 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
255 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
256 Convert from macros to...
257 (enum optgroup_flag): ...this new enum.
258 (optgroup_flags_t): New typedef.
259 (operator|, operator|=): Implement for optgroup_flags_t.
260 (struct dump_file_info): Convert field "alt_flags" to
261 dump_flags_t. Convert field "optgroup_flags" to
263 (dump_basic_block): Use dump_flags_t rather than int for param.
264 (dump_generic_expr_loc): Likewise.
265 (dump_dec): Likewise.
266 (dump_register): Convert param "optgroup_flags" to
268 (opt_info_enable_passes): Likewise.
269 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
271 * gimple-pretty-print.c (debug): Likewise.
272 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
273 (merged_store_group::apply_stores): Likewise.
274 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
275 * gimple.c (verify_gimple_pp): Likewise.
276 * graphite-poly.c (print_pbb_body): Likewise.
277 * passes.c (pass_manager::register_one_dump_file): Convert
278 local "optgroup_flags" to optgroup_flags_t.
279 * print-tree.c (print_node): Use TDF_NONE rather than 0.
281 (debug_body): Likewise.
282 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
284 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
286 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
287 (convert_mult_to_fma): Likewise.
288 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
289 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
290 * tree-vect-data-refs.c (dump_lower_bound): Convert param
291 "dump_kind" to dump_flags_t.
293 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
295 * config/rs6000/rs6000.c (min, max): Delete.
297 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
299 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
300 -mabi=spe and -mabi=no-spe.
302 2018-06-08 Martin Liska <mliska@suse.cz>
304 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
305 where we expect an existing summary.
307 2018-06-08 Martin Liska <mliska@suse.cz>
309 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
310 * ipa-inline.h (estimate_edge_growth): Likewise.
312 2018-06-08 Martin Liska <mliska@suse.cz>
314 * cgraph.c (function_version_hasher::hash): Use
315 cgraph_node::get_uid ().
316 (function_version_hasher::equal):
317 * cgraph.h (cgraph_node::get_uid): New method.
318 * ipa-inline.c (update_caller_keys): Use
319 cgraph_node::get_uid ().
320 (update_callee_keys): Likewise.
321 * ipa-utils.c (searchc): Likewise.
322 (ipa_reduced_postorder): Likewise.
323 * lto-cgraph.c (input_node): Likewise.
324 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
325 * symbol-summary.h (symtab_insertion): Likewise.
326 (symtab_removal): Likewise.
327 (symtab_duplication): Likewise.
328 * tree-pretty-print.c (dump_function_header): Likewise.
329 * tree-sra.c (convert_callers_for_node): Likewise.
331 2018-06-08 Martin Liska <mliska@suse.cz>
333 * cgraph.c (symbol_table::create_edge): Always assign a new
335 (symbol_table::free_edge): Do not recycle numbers.
336 * cgraph.h (cgraph_edge::get): New method.
337 * symbol-summary.h (symtab_removal): Use it.
338 (symtab_duplication): Likewise.
339 (call_summary::hashable_uid): Remove.
341 2018-06-08 Martin Liska <mliska@suse.cz>
343 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
344 (initialize_growth_caches): Remove.
345 (free_growth_caches): Likewise.
346 (do_estimate_edge_time): Use edge_growth_cache.
347 (do_estimate_edge_size): Likewise.
348 (do_estimate_edge_hints): Likewise.
349 * ipa-inline.c (reset_edge_caches): Likewise.
350 (recursive_inlining): Likewise.
351 (inline_small_functions): Likewise.
352 * ipa-inline.h (initialize_growth_caches): Remove.
353 (estimate_edge_size): Likewise.
354 (estimate_edge_time): Likewise.
355 (estimate_edge_hints): Likewise.
356 (reset_edge_growth_cache): Likewise.
357 * symbol-summary.h (call_summary::remove): New method.
359 2018-06-08 Martin Liska <mliska@suse.cz>
361 * ipa-cp.c (class edge_clone_summary): New summary.
362 (grow_edge_clone_vectors): Remove.
363 (ipcp_edge_duplication_hook): Remove.
364 (class edge_clone_summary_t): New call_summary class.
365 (ipcp_edge_removal_hook): Remove.
366 (edge_clone_summary_t::duplicate): New function.
367 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
368 (create_specialized_node): Likewise.
369 (ipcp_driver): Initialize edge_clone_summaries and do not
372 2018-06-08 Martin Liska <mliska@suse.cz>
374 * symbol-summary.h (get): New function.
375 (call_summary::m_initialize_when_cloning): New class member.
377 2018-06-08 Martin Liska <mliska@suse.cz>
379 * cgraph.c (cgraph_node::remove): Do not recycle uid.
380 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
381 (symbol_table::allocate_cgraph_symbol): Do not set uid.
382 * passes.c (uid_hash_t): Record removed_nodes by their uids.
383 (remove_cgraph_node_from_order): Use the removed_nodes set.
384 (do_per_function_toporder): Likwise.
385 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
386 instead of summary_uid.
387 (symtab_removal): Likewise.
388 (symtab_duplication): Likewise.
390 2018-06-08 Martin Liska <mliska@suse.cz>
392 * ipa-cp.c (ipcp_store_bits_results): Use
393 ipcp_transformation_sum.
394 (ipcp_store_vr_results): Likewise.
395 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
397 (ipcp_transformation_initialize): ... this.
398 (ipa_set_node_agg_value_chain):
399 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
400 (write_ipcp_transformation_info): Likewise.
401 (read_ipcp_transformation_info): Likewise.
402 (ipcp_update_bits): Likewise.
403 (ipcp_update_vr): Likewise.
404 (ipcp_transform_function): Likewise.
405 * ipa-prop.h: Rename ipcp_transformation_summary to
407 (class ipcp_transformation_t): New function summary.
408 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
409 (ipa_get_agg_replacements_for_node): Likewise.
411 2018-06-08 Martin Liska <mliska@suse.cz>
413 * ipa-pure-const.c (struct funct_state_d): Do it class instead
415 (class funct_state_summary_t): New function_summary class.
416 (has_function_state): Remove.
417 (get_function_state): Likewise.
418 (set_function_state): Likewise.
419 (add_new_function): Likewise.
420 (funct_state_summary_t::insert): New function.
421 (duplicate_node_data): Remove.
422 (remove_node_data): Remove.
423 (funct_state_summary_t::duplicate): New function.
424 (register_hooks): Create new funct_state_summaries.
425 (pure_const_generate_summary): Use it.
426 (pure_const_write_summary): Likewise.
427 (pure_const_read_summary): Likewise.
428 (propagate_pure_const): Likewise.
429 (propagate_nothrow): Likewise.
430 (dump_malloc_lattice): Likewise.
431 (propagate_malloc): Likewise.
432 (execute): Do not register hooks, just remove summary
434 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
437 2018-06-08 Martin Liska <mliska@suse.cz>
439 * ipa-reference.c (remove_node_data): Remove.
440 (duplicate_node_data): Likewise.
441 (class ipa_ref_var_info_summary_t): New class.
442 (class ipa_ref_opt_summary_t): Likewise.
443 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
444 (get_reference_optimization_summary): Use
445 ipa_ref_opt_sum_summaries.
446 (set_reference_vars_info): Remove.
447 (set_reference_optimization_summary): Likewise.
448 (ipa_init): Create summaries.
449 (init_function_info): Use function summary.
450 (ipa_ref_opt_summary_t::duplicate): New function.
451 (ipa_ref_opt_summary_t::remove): New function.
452 (get_read_write_all_from_node): Fix GNU coding style.
453 (propagate): Use function summary.
454 (write_node_summary_p): Fix GNU coding style.
455 (stream_out_bitmap): Likewise.
456 (ipa_reference_read_optimization_summary): Use function summary.
457 (ipa_reference_c_finalize): Do not release hooks.
459 2018-06-08 Martin Liska <mliska@suse.cz>
461 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
462 (analyze_function_body): Extract multiple calls of get_create.
463 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
464 * ipa-inline.c (recursive_inlining): Use ::get method.
465 * ipa-inline.h (estimate_edge_growth): Likewise.
467 2018-06-08 Martin Liska <mliska@suse.cz>
469 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
471 (hsa_function_summary::hsa_function_summary): Use the new enum
473 (hsa_gpu_implementation_p): Use hsa_summaries::get.
474 * hsa-gen.c (hsa_get_host_function): Likewise.
475 (get_brig_function_name): Likewise.
476 * ipa-hsa.c (process_hsa_functions): Likewise.
477 (ipa_hsa_write_summary): Likewise.
478 * symbol-summary.h (symtab_duplication): Use ::get function/
481 2018-06-08 Martin Liska <mliska@suse.cz>
483 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
485 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
486 (hsa_register_kernel): Likewise.
487 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
488 * hsa-gen.c (hsa_get_host_function): Likewise.
489 (get_brig_function_name): Likewise.
490 (generate_hsa): Likewise.
491 (pass_gen_hsail::execute): Likewise.
492 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
493 (devirtualization_time_bonus): Likewise.
494 (ipcp_propagate_stage): Likewise.
495 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
496 (edge_set_predicate): Likewise.
497 (evaluate_conditions_for_known_args): Likewise.
498 (evaluate_properties_for_edge): Likewise.
499 (ipa_fn_summary::reset): Likewise.
500 (ipa_fn_summary_t::duplicate): Likewise.
501 (dump_ipa_call_summary): Likewise.
502 (ipa_dump_fn_summary): Likewise.
503 (analyze_function_body): Likewise.
504 (compute_fn_summary): Likewise.
505 (estimate_edge_devirt_benefit): Likewise.
506 (estimate_edge_size_and_time): Likewise.
507 (estimate_calls_size_and_time): Likewise.
508 (estimate_node_size_and_time): Likewise.
509 (inline_update_callee_summaries): Likewise.
510 (remap_edge_change_prob): Likewise.
511 (remap_edge_summaries): Likewise.
512 (ipa_merge_fn_summary_after_inlining): Likewise.
513 (ipa_update_overall_fn_summary): Likewise.
514 (read_ipa_call_summary): Likewise.
515 (inline_read_section): Likewise.
516 (write_ipa_call_summary): Likewise.
517 (ipa_fn_summary_write): Likewise.
518 (ipa_free_fn_summary): Likewise.
519 * ipa-hsa.c (process_hsa_functions): Likewise.
520 (ipa_hsa_write_summary): Likewise.
521 (ipa_hsa_read_section): Likewise.
522 * ipa-icf.c (sem_function::merge): Likewise.
523 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
524 (do_estimate_edge_time): Likewise.
525 (estimate_size_after_inlining): Likewise.
526 (estimate_growth): Likewise.
527 (growth_likely_positive): Likewise.
528 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
529 (inline_call): Likewise.
530 * ipa-inline.c (caller_growth_limits): Likewise.
531 (can_inline_edge_p): Likewise.
532 (can_inline_edge_by_limits_p): Likewise.
533 (compute_uninlined_call_time): Likewise.
534 (compute_inlined_call_time): Likewise.
535 (want_inline_small_function_p): Likewise.
536 (edge_badness): Likewise.
537 (update_caller_keys): Likewise.
538 (update_callee_keys): Likewise.
539 (recursive_inlining): Likewise.
540 (inline_small_functions): Likewise.
541 (inline_to_all_callers_1): Likewise.
542 (dump_overall_stats): Likewise.
543 (early_inline_small_functions): Likewise.
544 (early_inliner): Likewise.
545 * ipa-inline.h (estimate_edge_growth): Likewise.
546 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
547 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
548 * ipa-prop.h (IPA_NODE_REF): Likewise.
549 (IPA_EDGE_REF): Likewise.
550 * ipa-pure-const.c (malloc_candidate_p): Likewise.
551 (propagate_malloc): Likewise.
552 * ipa-split.c (execute_split_functions): Likewise.
553 * symbol-summary.h: Rename get to get_create.
555 (get_create): Likewise.
556 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
558 2018-06-08 Martin Liska <mliska@suse.cz>
560 * symbol-summary.h (release): Move definition out of class
562 (symtab_removal): Likewise.
563 (symtab_duplication): Likewise.
565 2018-06-08 Martin Liska <mliska@suse.cz>
567 * symbol-summary.h (function_summary): Move constructor
568 implementation out of class declaration.
570 (symtab_insertion): Likewise.
571 (symtab_removal): Likewise.
572 (symtab_duplication): Likewise.
575 2018-06-08 Martin Liska <mliska@suse.cz>
577 * Makefile.in: Remove support for MPX (macros, related functions,
578 fields in cgraph_node, ...).
579 * builtin-types.def (BT_BND): Likewise.
580 (BT_FN_BND_CONST_PTR): Likewise.
581 (BT_FN_CONST_PTR_BND): Likewise.
582 (BT_FN_VOID_PTR_BND): Likewise.
583 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
584 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
585 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
586 (expand_builtin_mempcpy_with_bounds): Likewise.
587 (expand_builtin_memset_with_bounds): Likewise.
588 (expand_builtin_memset_args): Likewise.
589 (std_expand_builtin_va_start): Likewise.
590 (expand_builtin): Likewise.
591 (expand_builtin_with_bounds): Likewise.
592 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
593 (DEF_LIB_BUILTIN_CHKP): Likewise.
594 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
595 (DEF_CHKP_BUILTIN): Likewise.
596 (BUILT_IN_MEMCPY): Likewise.
597 (BUILT_IN_MEMMOVE): Likewise.
598 (BUILT_IN_MEMPCPY): Likewise.
599 (BUILT_IN_MEMSET): Likewise.
600 (BUILT_IN_STPCPY): Likewise.
601 (BUILT_IN_STRCAT): Likewise.
602 (BUILT_IN_STRCHR): Likewise.
603 (BUILT_IN_STRCPY): Likewise.
604 (BUILT_IN_STRLEN): Likewise.
605 (BUILT_IN_MEMCPY_CHK): Likewise.
606 (BUILT_IN_MEMMOVE_CHK): Likewise.
607 (BUILT_IN_MEMPCPY_CHK): Likewise.
608 (BUILT_IN_MEMSET_CHK): Likewise.
609 (BUILT_IN_STPCPY_CHK): Likewise.
610 (BUILT_IN_STRCAT_CHK): Likewise.
611 (BUILT_IN_STRCPY_CHK): Likewise.
612 * calls.c (store_bounds): Likewise.
613 (emit_call_1): Likewise.
614 (special_function_p): Likewise.
615 (maybe_warn_nonstring_arg): Likewise.
616 (initialize_argument_information): Likewise.
617 (finalize_must_preallocate): Likewise.
618 (compute_argument_addresses): Likewise.
619 (expand_call): Likewise.
620 * cfgexpand.c (expand_call_stmt): Likewise.
621 (expand_return): Likewise.
622 (expand_gimple_stmt_1): Likewise.
623 (pass_expand::execute): Likewise.
624 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
625 (cgraph_node::remove): Likewise.
626 (cgraph_node::dump): Likewise.
627 (cgraph_node::verify_node): Likewise.
628 * cgraph.h (chkp_function_instrumented_p): Likewise.
629 (symtab_node::get_alias_target): Likewise.
630 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
631 (cgraph_local_p): Likewise.
632 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
633 (cgraph_edge::rebuild_references): Likewise.
634 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
635 (walk_polymorphic_call_targets): Likewise.
636 (cgraph_node::expand_thunk): Likewise.
637 (symbol_table::output_weakrefs): Likewise.
638 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
639 (ix86_handle_option): Likewise.
640 * config/i386/constraints.md: Likewise.
641 * config/i386/i386-builtin-types.def (BND): Likewise.
645 * config/i386/i386-builtin.def (BDESC_END): Likewise.
646 (BDESC_FIRST): Likewise.
648 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
649 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
650 * config/i386/i386.c (enum reg_class): Likewise.
651 (ix86_target_string): Likewise.
652 (ix86_option_override_internal): Likewise.
653 (ix86_conditional_register_usage): Likewise.
654 (ix86_valid_target_attribute_inner_p): Likewise.
655 (ix86_set_indirect_branch_type): Likewise.
656 (ix86_set_current_function): Likewise.
657 (ix86_function_arg_regno_p): Likewise.
658 (init_cumulative_args): Likewise.
659 (ix86_function_arg_advance): Likewise.
660 (ix86_function_arg): Likewise.
661 (ix86_pass_by_reference): Likewise.
662 (ix86_function_value_regno_p): Likewise.
663 (ix86_function_value_1): Likewise.
664 (ix86_function_value_bounds): Likewise.
665 (ix86_return_in_memory): Likewise.
666 (ix86_setup_incoming_vararg_bounds): Likewise.
667 (ix86_va_start): Likewise.
668 (indirect_thunk_need_prefix): Likewise.
669 (print_reg): Likewise.
670 (ix86_print_operand): Likewise.
671 (ix86_expand_call): Likewise.
672 (ix86_output_function_return): Likewise.
673 (reg_encoded_number): Likewise.
674 (BDESC_VERIFYS): Likewise.
675 (ix86_init_mpx_builtins): Likewise.
676 (ix86_init_builtins): Likewise.
677 (ix86_emit_cmove): Likewise.
678 (ix86_emit_move_max): Likewise.
679 (ix86_expand_builtin): Likewise.
680 (ix86_builtin_mpx_function): Likewise.
681 (ix86_get_arg_address_for_bt): Likewise.
682 (ix86_load_bounds): Likewise.
683 (ix86_store_bounds): Likewise.
684 (ix86_load_returned_bounds): Likewise.
685 (ix86_store_returned_bounds): Likewise.
686 (ix86_class_likely_spilled_p): Likewise.
687 (ix86_hard_regno_mode_ok): Likewise.
688 (x86_order_regs_for_local_alloc): Likewise.
689 (ix86_mitigate_rop): Likewise.
690 (ix86_bnd_prefixed_insn_p): Likewise.
691 (ix86_mpx_bound_mode): Likewise.
692 (ix86_make_bounds_constant): Likewise.
693 (ix86_initialize_bounds): Likewise.
694 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
695 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
696 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
697 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
698 (TARGET_CHKP_BOUND_MODE): Likewise.
699 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
700 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
701 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
702 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
703 * config/i386/i386.h (TARGET_MPX): Likewise.
704 (TARGET_MPX_P): Likewise.
705 (VALID_BND_REG_MODE): Likewise.
706 (FIRST_BND_REG): Likewise.
707 (LAST_BND_REG): Likewise.
708 (enum reg_class): Likewise.
709 (BND_REG_P): Likewise.
710 (BND_REGNO_P): Likewise.
712 (ADJUST_INSN_LENGTH): Likewise.
713 * config/i386/i386.md: Likewise.
714 * config/i386/i386.opt: Likewise.
715 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
717 (LINK_MPX): Likewise.
718 (MPX_SPEC): Likewise.
719 (LIBMPX_SPEC): Likewise.
720 (LIBMPXWRAPPERS_SPEC): Likewise.
721 (CHKP_SPEC): Likewise.
722 * config/i386/predicates.md: Likewise.
723 * dbxout.c (dbxout_type): Likewise.
724 * doc/extend.texi: Likewise.
725 * doc/invoke.texi: Likewise.
726 * doc/md.texi: Likewise.
727 * doc/tm.texi: Likewise.
728 * doc/tm.texi.in: Likewise.
729 * dwarf2out.c (is_base_type): Likewise.
730 (gen_formal_types_die): Likewise.
731 (gen_subprogram_die): Likewise.
732 (gen_type_die_with_usage): Likewise.
733 (gen_decl_die): Likewise.
734 (dwarf2out_late_global_decl): Likewise.
735 * expr.c (expand_assignment): Likewise.
736 (emit_storent_insn): Likewise.
737 (store_expr_with_bounds): Likewise.
738 (store_expr): Likewise.
739 (expand_expr_real_1): Likewise.
740 * expr.h (store_expr_with_bounds): Likewise.
741 * function.c (use_register_for_decl): Likewise.
742 (struct bounds_parm_data): Likewise.
743 (assign_parms_augmented_arg_list): Likewise.
744 (assign_parm_find_entry_rtl): Likewise.
745 (assign_parm_is_stack_parm): Likewise.
746 (assign_parm_load_bounds): Likewise.
747 (assign_bounds): Likewise.
748 (assign_parms): Likewise.
749 (expand_function_start): Likewise.
750 * gcc.c (CHKP_SPEC): Likewise.
751 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
752 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
753 (wrestrict_dom_walker::check_call): Likewise.
754 * gimple.c (gimple_build_call_from_tree): Likewise.
755 * gimple.h (enum gf_mask): Likewise.
756 (gimple_call_with_bounds_p): Likewise.
757 (gimple_call_set_with_bounds): Likewise.
758 * gimplify.c (gimplify_init_constructor): Likewise.
759 * ipa-cp.c (initialize_node_lattices): Likewise.
760 (propagate_constants_across_call): Likewise.
761 (find_more_scalar_values_for_callers_subset): Likewise.
762 * ipa-hsa.c (process_hsa_functions): Likewise.
763 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
764 * ipa-icf.c (sem_function::merge): Likewise.
765 * ipa-inline.c (early_inliner): Likewise.
766 * ipa-pure-const.c (warn_function_noreturn): Likewise.
767 (warn_function_cold): Likewise.
768 (propagate_pure_const): Likewise.
769 * ipa-ref.h (enum GTY): Likewise.
770 * ipa-split.c (find_retbnd): Likewise.
771 (consider_split): Likewise.
772 (split_function): Likewise.
773 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
774 * ipa.c (walk_polymorphic_call_targets): Likewise.
775 (symbol_table::remove_unreachable_nodes): Likewise.
776 (process_references): Likewise.
777 (cgraph_build_static_cdtor_1): Likewise.
778 * lto-cgraph.c (lto_output_node): Likewise.
779 (output_refs): Likewise.
780 (compute_ltrans_boundary): Likewise.
781 (input_overwrite_node): Likewise.
782 (input_node): Likewise.
783 (input_cgraph_1): Likewise.
784 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
785 * passes.c (pass_manager::execute_early_local_passes): Likewise.
786 (class pass_chkp_instrumentation_passes): Likewise.
787 (make_pass_chkp_instrumentation_passes): Likewise.
788 * passes.def: Likewise.
789 * rtl.h (struct GTY): Likewise.
790 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
791 * stor-layout.c (layout_type): Likewise.
792 * symtab.c: Likewise.
793 * target.def: Likewise.
794 * targhooks.c (default_chkp_bound_type): Likewise.
795 (default_chkp_bound_mode): Likewise.
796 (default_builtin_chkp_function): Likewise.
797 (default_chkp_function_value_bounds): Likewise.
798 (default_chkp_make_bounds_constant): Likewise.
799 (default_chkp_initialize_bounds): Likewise.
800 * targhooks.h (default_chkp_bound_type): Likewise.
801 (default_chkp_bound_mode): Likewise.
802 (default_builtin_chkp_function): Likewise.
803 (default_chkp_function_value_bounds): Likewise.
804 (default_chkp_make_bounds_constant): Likewise.
805 (default_chkp_initialize_bounds): Likewise.
806 * toplev.c (compile_file): Likewise.
807 (process_options): Likewise.
808 * tree-core.h (DEF_BUILTIN): Likewise.
809 (DEF_BUILTIN_CHKP): Likewise.
810 * tree-inline.c (declare_return_variable): Likewise.
811 (remap_gimple_stmt): Likewise.
813 (initialize_inlined_parameters): Likewise.
814 (expand_call_inline): Likewise.
815 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
816 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
817 (make_pass_ipa_chkp_produce_thunks): Likewise.
818 (make_pass_chkp): Likewise.
819 (make_pass_chkp_opt): Likewise.
820 (make_pass_chkp_instrumentation_passes): Likewise.
821 * tree-pretty-print.c (dump_generic_node): Likewise.
822 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
823 * tree-ssa-dce.c (propagate_necessity): Likewise.
824 (eliminate_unnecessary_stmts): Likewise.
825 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
826 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
827 * tree-ssa-sccvn.h: Likewise.
828 * tree-ssa-strlen.c (get_string_length): Likewise.
829 (valid_builtin_call): Likewise.
830 (adjust_last_stmt): Likewise.
831 (handle_builtin_strchr): Likewise.
832 (handle_builtin_strcpy): Likewise.
833 (handle_builtin_stxncpy): Likewise.
834 (handle_builtin_memcpy): Likewise.
835 (handle_builtin_strcat): Likewise.
836 (strlen_check_and_optimize_stmt): Likewise.
837 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
838 * tree-streamer-in.c: Likewise.
839 * tree-streamer.c (record_common_node): Likewise.
840 * tree.c (tree_code_size): Likewise.
841 (wide_int_to_tree_1): Likewise.
842 (type_contains_placeholder_1): Likewise.
843 (build_common_tree_nodes): Likewise.
844 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
845 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
846 (POINTER_BOUNDS_P): Likewise.
847 (BOUNDED_TYPE_P): Likewise.
848 (BOUNDED_P): Likewise.
849 (CALL_WITH_BOUNDS_P): Likewise.
850 (pointer_bounds_type_node): Likewise.
851 * value-prof.c (gimple_ic): Likewise.
852 * var-tracking.c (vt_add_function_parameters): Likewise.
853 * varasm.c (make_decl_rtl): Likewise.
854 (assemble_start_function): Likewise.
855 (output_constant): Likewise.
856 (maybe_assemble_visibility): Likewise.
857 * varpool.c (ctor_for_folding): Likewise.
858 * chkp-builtins.def: Remove.
859 * ipa-chkp.c: Remove.
860 * ipa-chkp.h: Remove.
861 * rtl-chkp.c: Remove.
862 * rtl-chkp.h: Remove.
863 * tree-chkp-opt.c: Remove.
864 * tree-chkp.c: Remove.
865 * tree-chkp.h: Remove.
867 2018-06-07 Carl Love <cel@us.ibm.com>
869 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
870 vextract_fp_from_shortl): Add BE support.
872 2018-06-07 Paul Koning <ni1d@arrl.net>
874 * compare-elim.c (try_merge_compare): Don't merge compare if
875 address contains a side effect.
876 (try_eliminate_compare): Likewise.
878 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
880 * config.gcc: Support "tremont".
881 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
882 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
884 * config/i386/i386.c (m_TREMONT): Define.
885 (processor_target_table): Add "tremont".
886 (PTA_TREMONT): Define.
887 (ix86_lea_outperforms): Add TARGET_TREMONT.
888 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
889 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
890 and M_INTEL_GOLDMONT_PLUS.
891 (fold_builtin_cpu): Add "tremont".
892 (ix86_add_stmt_cost): Add TARGET_TREMONT.
893 (ix86_option_override_internal): Add "tremont".
894 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
895 (processor_type): Add PROCESSOR_TREMONT.
896 * config/i386/x86-tune.def: Add m_TREMONT.
897 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
899 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
901 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
902 symbol defined for msp430i* devices to be lower case.
904 gcc/testsuite/gcc.target/msp430/
905 * msp430i-device-symbol.c: New test.
906 * msp430f-device-symbol.c: New test.
907 * msp430.h: New test header file.
909 2018-06-07 Richard Biener <rguenther@suse.de>
911 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
912 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
913 Properly wrap signed arithmetic if overflow wraps.
915 2018-06-07 Jakub Jelinek <jakub@redhat.com>
917 PR tree-optimization/69615
918 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
919 of a cast from a same precision integral SSA_NAME in a bb dominated
920 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
921 cast to utype if rhs2 has already a compatible type.
923 2018-06-07 Richard Biener <rguenther@suse.de>
925 PR tree-optimization/85935
926 * graphite-scop-detection.c (find_params_in_bb): Analyze
927 condition operands with respect to the correct loop. Assert
928 the analysis doesn't fail.
930 2018-06-04 Carl Love <cel@us.ibm.com>
932 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
933 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
934 as it is slightly cheaper.
935 (first_match_or_eos_index_<mode>):
936 Calculate index using natural element order.
937 (first_match_index_<mode>):
938 Calculate index using natural element order.
939 (first_match_or_eos_index_<mode>):
940 Calculate index using natural order.
941 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
943 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
944 P9V_BUILTIN_VCLZLSBB_V16QI.
945 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
948 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
950 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
951 indentation and line wrap for many prototypes. Add missing
952 @smallexample directives around block of prototypes for vec_xl and
955 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
957 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
958 track if we pass or return IEEE 128-bit floating point.
959 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
960 C++ mangling that is compatible with GCC 8.1.
961 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
962 (init_cumulative_args): Note if we pass or return IEEE 128-bit
963 floating point types.
964 (rs6000_function_arg_advance_1): Likewise.
965 (rs6000_mangle_type): Optionally generate mangled names that match
966 what GCC 8.1 generated for IEEE 128-bit floating point types.
967 (rs6000_globalize_decl_name): If we have an external function that
968 passes or returns IEEE 128-bit types, generate a weak reference
969 from the mangled name used in GCC 8.1 to the current mangled
971 (rs6000_init_builtins): Make __ibm128 use the long double type if
972 long double is IBM extended double. Make __float128 use the long
973 double type if long double is IEEE 128-bit.
976 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
977 macro for __ibm128 built-in functions.
978 (PACK_IF): Add __ibm128 pack/unpack functions.
979 (UNPACK_IF): Likewise.
980 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
981 enable long double built-in functions if long double is IEEE
982 128-bit floating point.
983 (rs6000_invalid_builtin): Update long double built-in function
985 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
986 functions, adjust the built-in function to use the long double
987 built-in function if __ibm128 and long double are the same type.
988 * doc/extend.texi (PowerPC builtins): Update documention for
989 __builtin_{,un}pack_longdouble. Add documentation for
990 __builtin_{,un}pack_ibm128.
992 2018-06-06 Jim Wilson <jimw@sifive.com>
994 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
995 (struct machine_function): New field interrupt_mode.
996 (riscv_handle_type_attribute): New function. Add forward declaration.
997 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
998 (riscv_expand_epilogue): Check interrupt_mode field.
999 (riscv_set_current_function): Check interrupt attribute args and
1000 set interrupt_mode field.
1001 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
1002 (riscv_sret, riscv_uret): New.
1003 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
1004 new arguments to interrupt attribute.
1006 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
1009 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
1010 Don't handle -mcpu=power8 if -mpower9-vector is also used.
1012 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1014 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
1015 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1016 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1017 several redundant entries.
1019 2018-06-06 David Malcolm <dmalcolm@redhat.com>
1021 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1022 type from "rtx" to "rtx_insn *".
1023 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1024 for local "call_insn", removing cast.
1025 (ix86_expand_call): Likewise, introducing a "call_insn" local.
1027 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
1029 PR tree-optimization/86066
1030 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1031 for BIT_INSERT_EXPR stores.
1033 2018-06-06 Richard Biener <rguenther@suse.de>
1035 PR tree-optimization/86062
1036 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1037 component refs ontop
1038 of to be offsetted base.
1040 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1042 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1043 to be static and remove check on interrupt attribute name.
1045 gcc/testsuite/gcc.target/msp430/
1046 * function-attributes-4.c: New test.
1047 * static-interrupts.c: New test.
1049 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1051 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1052 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1054 2018-06-05 Steve Ellcey <sellcey@cavium.com>
1057 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1059 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1060 Remove second argument, change how error is called.
1061 (aarch64_layout_arg): Remove second argument from
1062 aarch64_err_no_fpadvsimd call.
1063 (aarch64_init_cumulative_args): Ditto.
1064 (aarch64_gimplify_va_arg_expr): Ditto.
1065 * config/aarch64/aarch64.md (mov<mode>): Ditto.
1067 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
1069 * config/i386/i386.md (simple_return_indirect_internal): New expander.
1070 (*simple_return_indirect_internal<mode>): Rename from
1071 simple_return_indirect_internal. Use W mode iterator.
1072 (rstorssp): New expander.
1073 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
1074 (clrssbsy): New expander.
1075 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
1077 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
1079 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1081 (cmse_check_pointed_object): Likewise.
1083 2018-06-05 Martin Liska <mliska@suse.cz>
1085 PR gcov-profile/47618
1086 * doc/invoke.texi: Document how -fprofile-dir format
1089 2018-06-05 Richard Biener <rguenther@suse.de>
1091 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1092 removal pretend DOM info isn't available so we do not update
1093 it and only remove edges, not dominated blocks. Actually free
1094 DOM info in case we removed something. Remove unreachable blocks.
1095 (mfb_keep_latches): Work with either DOM info or marked backedges.
1096 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1097 first. Mark backedges if DOM info isn't available.
1098 (Re-)compute DOM info after cleanup_control_flow_pre.
1100 2018-06-05 Richard Biener <rguenther@suse.de>
1102 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1103 (locus_discrim_hasher::hash): Adjust.
1104 (locus_discrim_hasher::equal): Likewise.
1105 (next_discriminator_for_locus): Work on line directly.
1106 (same_line_p): Pass in expanded locus1 as well.
1107 (assign_discriminators): Avoid redundant location expansions.
1109 2018-06-05 Richard Biener <rguenther@suse.de>
1111 PR tree-optimization/86046
1112 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1113 if required after clearing TREE_ADDRESSABLE.
1115 2018-06-05 Richard Biener <rguenther@suse.de>
1117 PR tree-optimization/86047
1118 * tree-ssa-loop.c (for_each_index): Glob handling of all
1119 decls and constants and really handle all of them.
1121 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1124 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1125 qualifier_void_pointer and qualifier_const_void_pointer.
1126 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1127 (arm_init_builtins): Handle the above.
1128 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1129 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1130 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1133 2018-06-05 Martin Liska <mliska@suse.cz>
1135 * auto-profile.c (read_autofdo_file): Do not use
1136 gcov_ctr_summary struct.
1137 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1138 * coverage.c (struct counts_entry): Likewise.
1139 (read_counts_file): Read just single summary entry.
1140 (get_coverage_counts): Use gcov_summary struct.
1141 * coverage.h (get_coverage_counts): Likewise.
1142 * gcov-dump.c (dump_working_sets): Likewise.
1143 (tag_summary): Dump just single summary.
1144 * gcov-io.c (gcov_write_summary): Write just histogram
1146 (gcov_read_summary): Read just single summary.
1147 (compute_working_sets): Use gcov_summary struct.
1148 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1149 of GCOV_COUNTERS_SUMMABLE.
1150 (GCOV_COUNTERS_SUMMABLE): Remove.
1151 (GCOV_FIRST_VALUE_COUNTER): Replace with
1152 GCOV_COUNTER_V_INTERVAL.
1153 (struct gcov_ctr_summary): Remove.
1154 (struct gcov_summary): Directly use fields of former
1156 (compute_working_sets): Use gcov_summary struct.
1157 * gcov.c (read_count_file): Do not use ctrs fields.
1158 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1160 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1162 * profile.c: Likewise.
1163 * profile.h: Likewise.
1165 2018-06-05 Martin Liska <mliska@suse.cz>
1167 PR gcov-profile/84846
1168 * gcov.c (output_lines): Print working directory only
1169 in intermediate format.
1171 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
1173 * config/s390/s390-builtin-types.def: Add void function type.
1174 * config/s390/s390-builtins.def: Use the function type for the
1177 2018-06-04 Jim Wilson <jimw@sifive.com>
1179 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
1181 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
1182 and maybe_eh_return. Change regno to unsigned int. Use new args to
1183 handle EH_RETURN_DATA_REGNO registers properly.
1184 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
1185 (riscv_expand_epilogue): Update comment. Change argument name and
1186 type. Update code to use new name and type. Pass new args to
1187 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
1189 * config/riscv/riscv.md (NORMAL_RETURN): New.
1190 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
1191 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
1192 (eh_return): Call gen_eh_return_internal and emit barrier.
1193 (eh_return_internal): Call riscv_expand_epilogue.
1195 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
1197 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
1198 bit_insertion field and declare can_be_merged_into method.
1199 (merged_store_group::can_be_merged_into): New method.
1200 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
1201 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
1202 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
1204 2018-06-04 Richard Biener <rguenther@suse.de>
1206 PR tree-optimization/85955
1207 * builtins.c (fold_builtin_sincos): Convert pointers to
1208 destination to appropriate type before dereferencing.
1210 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
1212 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
1214 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
1216 * expr.c (expand_expr_real_1): Force the operand into memory if
1217 its TYPE_MODE is BLKmode and if there is no integer mode for
1218 the number of bits being extracted.
1220 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1224 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1225 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
1227 2018-06-04 Richard Biener <rguenther@suse.de>
1229 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
1230 (cleanup_tree_cfg_noloop): ... single caller. Do
1231 start_recording_case_labels later.
1233 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
1235 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
1236 to _IMMINTRIN_H_INCLUDED.
1237 * config/i386/pconfigintrin.h: Ditto.
1238 * config/i386/waitpkgintrin.h: Ditto.
1239 * config/i386/immintrin.h: Add includes for sgxintrin.h,
1240 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
1241 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
1242 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
1243 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
1244 waitpkgintrin.h and cldemoteintrin.h.
1246 2018-06-04 Richard Biener <rguenther@suse.de>
1248 PR tree-optimization/86038
1249 * tracer.c (find_best_successor): Check probability for
1250 being initialized, bail out if not.
1252 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
1255 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
1256 of bits to ignore when comparing architectures.
1258 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1260 PR tree-optimization/69615
1261 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
1262 maximum or minimum of the type, try to merge it also as if
1263 range1 is + [-, x - 1] or + [x + 1, -].
1266 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
1268 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
1270 PR tree-optimization/86034
1271 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
1272 the unsigned bitfield type in a bit insertion sequence if it does not
1273 have a larger precision than the bitfield size.
1274 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
1276 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
1278 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
1280 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1282 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
1283 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
1284 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
1285 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
1287 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1289 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
1290 Disable -fdelete-null-pointer-checks for ELF toolchain.
1292 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1293 Kito Cheng <kito.cheng@gmail.com>
1295 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
1296 (nds32le-*-*, nds32be-*-*): Integrate checking process.
1297 (nds32*-*-*): Add glibc and uclibc conditions.
1298 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
1299 (TARGET_EXCEPT_UNWIND_INFO): Define.
1300 * config/nds32/elf.h: New file.
1301 * config/nds32/linux.h: New file.
1302 * config/nds32/nds32-elf.opt: New file.
1303 * config/nds32/nds32-linux.opt: New file.
1304 * config/nds32/nds32-fp-as-gp.c
1305 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
1306 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
1308 (nds32_asm_file_end): Ditto.
1309 (nds32_print_operand): Ditto.
1310 (nds32_insert_attributes): Ditto.
1311 (nds32_init_libfuncs): New function.
1312 (TARGET_HAVE_TLS): Define.
1313 (TARGET_INIT_LIBFUNCS): Define.
1314 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
1316 (TARGET_ELF): Apply different mcmodel setting.
1317 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
1318 been migrated into elf.h and linux.h files.
1319 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
1320 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
1321 (mcmodel): The content has been migrated into nds32-elf.opt and
1322 nds32-linux.opt files.
1323 * config/nds32/t-elf: New file.
1324 * config/nds32/t-linux: New file.
1326 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1327 Shiva Chen <shiva0217@gmail.com>
1329 * config/nds32/constants.md (unspec_volatile_element): Add
1330 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
1331 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
1333 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
1334 (make_pass_nds32_fp_as_gp): Declare.
1335 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
1337 (nds32_asm_function_end_prologue): Remove unused asm output.
1338 (nds32_asm_function_begin_epilogue): Remove unused asm output.
1339 (nds32_asm_file_start): Output necessary fp_as_gp information.
1340 (nds32_option_override): Adjust register usage.
1341 (nds32_expand_prologue): Consider fp_as_gp situation.
1342 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
1343 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
1346 (simple_return): Ditto.
1347 (omit_fp_begin): Output special directive for fp_as_gp.
1348 (omit_fp_end): Output special directive for fp_as_gp.
1349 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
1350 mforbid-fp-as-gp): New options.
1352 2018-06-01 Mark Wielaard <mark@klomp.org>
1354 * dwarf2out.c (dwarf2out_finish): Remove generation of
1355 DW_AT_loclists_base.
1357 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
1359 * gimple-ssa-store-merging.c: Include gimple-fold.h.
1360 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
1361 (struct merged_store_group): Add bit_insertion field.
1362 (dump_char_array): Use standard hexadecimal format.
1363 (merged_store_group::merged_store_group): Set bit_insertion to false.
1364 (merged_store_group::apply_stores): Use optimal buffer size. Deal
1365 with BIT_INSERT_EXPR stores. Move up code updating the mask and
1366 also print the mask in the dump file.
1367 (pass_store_merging::gate): Minor tweak.
1368 (imm_store_chain_info::coalesce_immediate): Fix wrong association
1369 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
1370 stores with INTEGER_CST stores.
1371 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
1372 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
1373 and use it throughout. Generate bit insertion sequences if need be.
1374 (pass_store_merging::process_store): Remove redundant condition.
1375 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
1377 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
1379 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
1380 the 128-bit floating point types. Fix function comment.
1382 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1384 * config/aarch64/aarch64-simd.md
1385 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
1387 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
1390 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1392 PR tree-optimization/85989
1393 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
1395 (backprop::intersect_uses): Check it when deciding whether this
1396 is a backedge reference.
1397 (backprop::process_block): Add each phi to m_visited_phis
1398 after visiting it, then clear it at the end.
1400 2018-06-01 Richard Biener <rguenther@suse.de>
1402 * tree-vectorizer.h (vect_dr_stmt): New function.
1403 (vect_get_load_cost): Adjust.
1404 (vect_get_store_cost): Likewise.
1405 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
1406 Use vect_dr_stmt instead of DR_SMTT.
1407 (vect_record_base_alignments): Likewise.
1408 (vect_calculate_target_alignment): Likewise.
1409 (vect_compute_data_ref_alignment): Likewise and make static.
1410 (vect_update_misalignment_for_peel): Likewise.
1411 (vect_verify_datarefs_alignment): Likewise.
1412 (vector_alignment_reachable_p): Likewise.
1413 (vect_get_data_access_cost): Likewise. Pass down
1414 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
1415 (vect_get_peeling_costs_all_drs): Likewise.
1416 (vect_peeling_hash_get_lowest_cost): Likewise.
1417 (vect_enhance_data_refs_alignment): Likewise.
1418 (vect_find_same_alignment_drs): Likewise.
1419 (vect_analyze_data_refs_alignment): Likewise.
1420 (vect_analyze_group_access_1): Likewise.
1421 (vect_analyze_group_access): Likewise.
1422 (vect_analyze_data_ref_access): Likewise.
1423 (vect_analyze_data_ref_accesses): Likewise.
1424 (vect_vfa_segment_size): Likewise.
1425 (vect_small_gap_p): Likewise.
1426 (vectorizable_with_step_bound_p): Likewise.
1427 (vect_prune_runtime_alias_test_list): Likewise.
1428 (vect_analyze_data_refs): Likewise.
1429 (vect_supportable_dr_alignment): Likewise.
1430 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1431 (vect_gen_prolog_loop_niters): Likewise.
1432 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1433 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
1435 (vect_recog_mask_conversion_pattern): Likewise.
1436 (vect_try_gather_scatter_pattern): Likewise.
1437 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
1438 to vect_get_store_cost.
1439 (vect_get_store_cost): Get stmt_info instead of DR.
1440 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
1441 (vect_get_load_cost): Get stmt_info instead of DR.
1443 2018-06-01 Richard Biener <rguenther@suse.de>
1446 * gimple-fold.c (var_decl_component_p): Also allow offsetted
1447 vars wrapped in MEM_REFs.
1449 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1451 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1452 Fix subreg tests so that we only return a choice between
1453 GENERAL_REGS and FP_REGS if the original classes included both.
1455 2018-06-01 Richard Biener <rguenther@suse.de>
1458 * tree-ssa-structalias.c (get_function_part_constraint):
1459 Handle NULL fi->decl.
1460 (find_func_aliases_for_call): Properly handle indirect
1461 fi from direct call.
1462 (find_func_clobbers): Likewise.
1463 (ipa_pta_execute): Likewise.
1464 (create_variable_info_for): For functions that are ifunc_resolver
1465 resolve to a varinfo that contains the result of the resolver call.
1466 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
1469 2018-05-31 Michael Collison <michael.collison@arm.com>
1471 * config/aarch64/aarch64.md:
1472 (*fix_to_zero_extenddfdi2): New pattern.
1473 * gcc.target/aarch64/fix_extend1.c: New testcase.
1475 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
1479 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
1480 and BUILT_IN_STRNCMP_EQ.
1481 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
1482 BUILT_IN_STRNCMP_EQ.
1483 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
1484 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1485 (gimple_fold_builtin): Likewise.
1486 * tree-ssa-strlen.c (compute_string_length): New function.
1487 (determine_min_obsize): New function.
1488 (handle_builtin_string_cmp): New function to handle calls to
1489 string compare functions.
1490 (strlen_optimize_stmt): Add handling to builtin string compare
1492 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1493 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1494 * tree.c (build_common_builtin_nodes): Add new defines of
1495 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
1497 2018-05-31 Jakub Jelinek <jakub@redhat.com>
1500 * bb-reorder.c (pass_partition_blocks::gate): Return false for
1501 functions with naked attribute.
1503 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
1505 * config/i386/sse.md (avx_vec_concat<mode>):
1506 Substitute concat_tg_mode mode attribute with xtg_mode.
1507 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
1508 (concat_tg_mode): Remove mode attribute.
1510 2018-05-31 Martin Sebor <msebor@redhat.com>
1513 * calls.c (alloc_max_size): Correct a logic error/typo.
1514 Treat excessive arguments as infinite. Warn for invalid arguments.
1515 * doc/invoke.texi (-Walloc-size-larger-than): Update.
1517 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
1520 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
1521 and movx for Haswell.
1523 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
1524 Cesar Philippidis <cesar@codesourcery.com>
1527 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
1528 when emitting error on private/firstprivate reductions.
1529 * omp-low.c (lower_omp_target): Avoid reference-type processing
1530 on pointers for firstprivate clause.
1532 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1534 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
1537 * config/aarch64/aarch64-simd.md
1538 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
1539 (aarch64_ld1_x3_<mode>): Likewise
1540 (aarch64_st1x2<VALLDIF:mode>): Likewise
1541 (aarch64_st1_x2_<mode>): Likewise
1542 (aarch64_st1x3<VALLDIF:mode>): Likewise
1543 (aarch64_st1_x3_<mode>): Likewise
1544 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
1545 (vld1_s8_x3): Likewise.
1546 (vld1_u16_x3): Likewise.
1547 (vld1_s16_x3): Likewise.
1548 (vld1_u32_x3): Likewise.
1549 (vld1_s32_x3): Likewise.
1550 (vld1_u64_x3): Likewise.
1551 (vld1_s64_x3): Likewise.
1552 (vld1_f16_x3): Likewise.
1553 (vld1_f32_x3): Likewise.
1554 (vld1_f64_x3): Likewise.
1555 (vld1_p8_x3): Likewise.
1556 (vld1_p16_x3): Likewise.
1557 (vld1_p64_x3): Likewise.
1558 (vld1q_u8_x3): Likewise.
1559 (vld1q_s8_x3): Likewise.
1560 (vld1q_u16_x3): Likewise.
1561 (vld1q_s16_x3): Likewise.
1562 (vld1q_u32_x3): Likewise.
1563 (vld1q_s32_x3): Likewise.
1564 (vld1q_u64_x3): Likewise.
1565 (vld1q_s64_x3): Likewise.
1566 (vld1q_f16_x3): Likewise.
1567 (vld1q_f32_x3): Likewise.
1568 (vld1q_f64_x3): Likewise.
1569 (vld1q_p8_x3): Likewise.
1570 (vld1q_p16_x3): Likewise.
1571 (vld1q_p64_x3): Likewise.
1572 (vst1_s64_x2): Likewise.
1573 (vst1_u64_x2): Likewise.
1574 (vst1_f64_x2): Likewise.
1575 (vst1_s8_x2): Likewise.
1576 (vst1_p8_x2): Likewise.
1577 (vst1_s16_x2): Likewise.
1578 (vst1_p16_x2): Likewise.
1579 (vst1_s32_x2): Likewise.
1580 (vst1_u8_x2): Likewise.
1581 (vst1_u16_x2): Likewise.
1582 (vst1_u32_x2): Likewise.
1583 (vst1_f16_x2): Likewise.
1584 (vst1_f32_x2): Likewise.
1585 (vst1_p64_x2): Likewise.
1586 (vst1q_s8_x2): Likewise.
1587 (vst1q_p8_x2): Likewise.
1588 (vst1q_s16_x2): Likewise.
1589 (vst1q_p16_x2): Likewise.
1590 (vst1q_s32_x2): Likewise.
1591 (vst1q_s64_x2): Likewise.
1592 (vst1q_u8_x2): Likewise.
1593 (vst1q_u16_x2): Likewise.
1594 (vst1q_u32_x2): Likewise.
1595 (vst1q_u64_x2): Likewise.
1596 (vst1q_f16_x2): Likewise.
1597 (vst1q_f32_x2): Likewise.
1598 (vst1q_f64_x2): Likewise.
1599 (vst1q_p64_x2): Likewise.
1600 (vst1_s64_x3): Likewise.
1601 (vst1_u64_x3): Likewise.
1602 (vst1_f64_x3): Likewise.
1603 (vst1_s8_x3): Likewise.
1604 (vst1_p8_x3): Likewise.
1605 (vst1_s16_x3): Likewise.
1606 (vst1_p16_x3): Likewise.
1607 (vst1_s32_x3): Likewise.
1608 (vst1_u8_x3): Likewise.
1609 (vst1_u16_x3): Likewise.
1610 (vst1_u32_x3): Likewise.
1611 (vst1_f16_x3): Likewise.
1612 (vst1_f32_x3): Likewise.
1613 (vst1_p64_x3): Likewise.
1614 (vst1q_s8_x3): Likewise.
1615 (vst1q_p8_x3): Likewise.
1616 (vst1q_s16_x3): Likewise.
1617 (vst1q_p16_x3): Likewise.
1618 (vst1q_s32_x3): Likewise.
1619 (vst1q_s64_x3): Likewise.
1620 (vst1q_u8_x3): Likewise.
1621 (vst1q_u16_x3): Likewise.
1622 (vst1q_u32_x3): Likewise.
1623 (vst1q_u64_x3): Likewise.
1624 (vst1q_f16_x3): Likewise.
1625 (vst1q_f32_x3): Likewise.
1626 (vst1q_f64_x3): Likewise.
1627 (vst1q_p64_x3): Likewise.
1629 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1631 * config/msp430/msp430.c (msp430_output_labelref): Prepend
1632 user_label_prefix to name.
1634 * tree-core.h: Update comment about the format of NAME string
1635 passed to handler in attribute_spec.
1637 * config/msp430/msp430.md: Remove erroneous subreg expression from
1638 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
1639 zero_extend{q,h}isi2.
1641 2018-05-30 Borislav Petkov <bp@suse.de>
1643 * doc/extend.texi: Document some architecture specific
1644 constraints and sort entries.
1646 2018-05-30 Martin Sebor <msebor@redhat.com>
1649 * builtins.c (expand_builtin_stpcpy_1): New function.
1650 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
1651 only if the former succeeds.
1653 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1655 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
1658 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1660 * doc/invoke.texi (-flinker-output): Document
1662 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1664 * passes.c (ipa_write_summaries): Only modify statements if body
1666 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
1667 incrementally linking.
1668 (ipa_passes): Likewise.
1669 * lto-cgraph.c (lto_output_node): When incrementally linking do not
1670 pass down resolution info.
1671 * common.opt (flag_incremental_link): Update info.
1672 * gcc.c (plugin specs): Turn flinker-output=* to
1673 -plugin-opt=-linker-output-known
1674 * toplev.c (compile_file): Also cut compilation when doing incremental
1676 * flag-types. (enum lto_partition_model): Add
1677 LTO_LINKER_OUTPUT_NOLTOREL.
1678 (invoke.texi): Add -flinker-output docs.
1679 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
1680 link same way as WPA; do not stream in dead initializers.
1682 * dwarf2out.c (dwarf2out_die_ref_for_decl,
1683 darf2out_register_external_decl): Support incremental link.
1685 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1687 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
1689 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1691 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
1692 it down to simple_object_copy_lto_debug_sections.
1693 (run_gcc): Determine incremental LTO link time and configure
1694 lto1 into non-wpa mode, disable renaming of debug sections.
1696 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
1698 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1699 descriptions of various incorrectly documented functions.
1701 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
1704 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1705 address check not strict.
1707 2018-05-30 Richard Biener <rguenther@suse.de>
1709 PR tree-optimization/85964
1710 * tracer.c (better_p): Drop initialized count check, we only
1711 call the function with initialized counts now.
1712 (find_best_successor): Do find a best edge if one
1713 has uninitialized count.
1714 (find_best_predecessor): Likewise. Do BB frequency check only
1715 if count is initialized.
1717 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
1719 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
1720 (aarch64_ldrstr_offset_compare): New.
1721 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
1722 load/store orderings.
1723 (aarch64_gen_adjusted_ldpstp): Likewise.
1725 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
1727 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1728 Check for subset of GENERAL_REGS and FP_REGS.
1729 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
1732 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
1734 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
1735 and wi::to_poly_offset. Add the current offset and then check
1736 whether the sum fits, rather than using an unchecked addition of
1737 a checked term. Check for a shwi rather than a uhwi.
1738 * expr.c (get_bit_range): Use tree_to_poly_uint64.
1739 (store_constructor): Use poly_int_tree_p.
1740 (expand_expr_real_1): Likewise.
1741 * function.c (assign_temp): Likewise.
1742 * fold-const.c (const_binop): Use poly_int_tree_p and
1744 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
1746 * ipa-icf-gimple.c (func_checker::compare_operand): Use
1747 to_poly_offset for MEM offsets.
1748 * ipa-icf.c (sem_variable::equals): Likewise.
1749 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
1750 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
1751 wi::to_poly_offset for BIT_FIELD_REF offsets.
1752 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
1754 * var-tracking.c (emit_note_insn_var_location): Use
1755 tree_to_poly_uint64.
1757 2018-05-29 Jim Wilson <jimw@sifive.com>
1759 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
1761 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
1764 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
1765 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
1767 (sse4_1_round<mode>2): Use nonimmediate_operand
1768 for operand 1 predicate.
1770 2018-05-29 Martin Sebor <msebor@redhat.com>
1771 Richard Biener <rguenther@suse.de>
1774 * calls.c (get_size_range): Call determine_value_range instead
1775 of get_value_range..
1776 * tree-vrp.h (determine_value_range): Declared new function.
1777 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
1779 2018-05-29 Richard Biener <rguenther@suse.de>
1781 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
1782 sure to use non-pattern stmts for get_earlier_stmt arguments.
1783 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
1784 called on pattern stmts.
1785 (get_later_stmt): Likewise.
1787 2018-05-29 Martin Liska <mliska@suse.cz>
1789 PR gcov-profile/85759
1790 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
1793 2018-05-29 Jakub Jelinek <jakub@redhat.com>
1795 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
1797 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
1801 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
1802 VEC_PACK_FLOAT_EXPR): New tree codes.
1803 * tree-pretty-print.c (op_code_prio): Handle
1804 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
1805 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
1806 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
1807 * tree-inline.c (estimate_operator_cost): Likewise.
1808 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
1809 * fold-const.c (const_binop): Likewise.
1810 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
1811 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
1812 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1813 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
1814 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
1815 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
1816 * expr.c (expand_expr_real_2): Likewise.
1817 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
1818 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
1819 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
1821 * optabs.c (expand_widen_pattern_expr): For
1822 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
1823 sign from result type rather than operand's type.
1824 (expand_binop_directly): For vec_packu_float_optab and
1825 vec_packs_float_optab allow result type to be different from operand's
1827 * optabs-tree.c (optab_for_tree_code): Handle
1828 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
1829 VEC_PACK_FLOAT_EXPR. Formatting fixes.
1830 * tree-vect-generic.c (expand_vector_operations_1): Handle
1831 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
1832 VEC_PACK_FLOAT_EXPR.
1833 * tree-vect-stmts.c (supportable_widening_operation): Handle
1835 (supportable_narrowing_operation): Handle FLOAT_EXPR.
1836 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
1837 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
1838 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
1839 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
1841 (vec_pack<floatprefix>_float_<mode>): New expander.
1842 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
1844 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
1845 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
1846 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
1847 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
1848 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
1850 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
1851 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
1852 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
1853 VEC_PACK_FLOAT_EXPR): Document.
1855 2018-05-29 Richard Biener <rguenther@suse.de>
1857 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
1859 (stmt_vec_info_vec): Make pointer.
1860 (init_stmt_vec_info_vec): Remove.
1861 (free_stmt_vec_info_vec): Likewise.
1862 (set_stmt_vec_info_vec): New function.
1863 (free_stmt_vec_infos): Likewise.
1864 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
1865 (set_vinfo_for_stmt): Likewise.
1866 (get_earlier_stmt): Likewise.
1867 (get_later_stmt): Likewise.
1868 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
1869 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
1870 (vec_info::~vec_info): Free stmt_vec_infos.
1871 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
1872 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
1873 (pass_slp_vectorize::execute): Likewise.
1874 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
1875 (free_stmt_vec_info_vec): Likewise.
1876 (set_stmt_vec_info_vec): New function.
1877 (free_stmt_vec_infos): Likewise.
1878 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
1879 the global stmt_vec_info_vec.
1880 * tree-parloops.c (gather_scalar_reductions): Use
1881 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
1884 2018-05-29 Richard Biener <rguenther@suse.de>
1886 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
1888 2018-05-29 Martin Liska <mliska@suse.cz>
1889 David Malcolm <dmalcolm@redhat.com>
1891 * vec.c (test_reverse): New.
1892 (vec_c_tests): Add new test.
1893 * vec.h (vl_ptr>::reverse): New function.
1895 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
1897 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
1899 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
1902 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
1904 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
1906 2018-05-28 Richard Biener <rguenther@suse.de>
1908 PR tree-optimization/85933
1909 * tree-vect-data-refs.c (vect_record_base_alignments): Only
1910 look at stmts marked as vectorizable.
1912 2018-05-28 Richard Biener <rguenther@suse.de>
1914 PR tree-optimization/85934
1915 * tree-vect-generic.c (expand_vector_operations_1): Hoist
1916 vector boolean check before scalar optimization.
1918 2018-05-28 Jakub Jelinek <jakub@redhat.com>
1920 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
1923 2018-05-28 Mark Wielaard <mark@klomp.org>
1925 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
1926 if it is an expression containing a minus sign.
1928 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
1930 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1932 2018-05-27 Paul Koning <ni1d@arrl.net>
1934 * config/pdp11/pdp11.md (truncsihi2): Remove.
1936 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
1937 Chung-Ju Wu <jasonwucj@gmail.com>
1939 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
1941 (unaligned_store_dw): Ditto.
1942 * config/nds32/nds32-memory-manipulation.c
1943 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
1944 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
1945 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
1946 (emit_setmem_word_loop): Rename to ...
1947 (emit_setmem_doubleword_loop): ... this.
1948 (nds32_gen_dup_4_byte_to_word_value): New function.
1949 (nds32_gen_dup_8_byte_to_double_word_value): New function.
1950 (nds32_expand_setmem_loop): Refine implementation.
1951 (nds32_expand_setmem_loop_v3m): Ditto.
1952 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
1955 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
1957 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
1959 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
1961 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
1962 (nds32_init_machine_status): Initialize machine->attr_naked_p and
1963 machine->attr_no_prologue_p.
1964 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
1965 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
1966 (nds32_expand_epilogue): Consider attr_naked_p.
1967 (nds32_expand_epilogue_v3pop): Likewise.
1968 (nds32_can_use_return_insn): Likewise.
1969 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
1970 attr_no_prologue_p fields.
1971 * config/nds32/nds32.opt (mret-in-naked-func): New option.
1973 2018-05-27 Jakub Jelinek <jakub@redhat.com>
1976 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
1978 * config/i386/sse.md
1979 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
1981 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
1983 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
1985 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
1987 (*<floatsuffix>floatv2div2sf2): Rename to ...
1988 (*float<floatunssuffix>v2div2sf2): ... this.
1989 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
1990 (float<floatunssuffix>v2div2sf2_mask): ... this.
1991 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
1992 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
1993 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
1995 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
1997 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1999 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2001 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2003 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2005 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
2006 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
2007 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
2008 gen_ufix_truncv8dfv8si2.
2009 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
2010 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
2011 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
2012 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
2013 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
2014 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
2015 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2016 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2018 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2022 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2024 2018-05-25 Jim Wilson <jimw@sifive.com>
2026 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2027 * config/riscv/riscv.c (struct machine_function): Add
2028 interrupt_handler_p and attribute_checked_p fields.
2029 (riscv_attribute_table): Add interrupt.
2030 (riscv_interrupt_type_p): New.
2031 (riscv_save_reg_p): Save extra regs for interrupt handler.
2032 (riscv_use_save_libcall): Return false for interrupt handler.
2033 (riscv_first_stack_step): Add forward declaration.
2034 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
2035 for interrupt handler with large frame. Use it for saved reg list.
2036 (riscv_expand_prologue): Move flag_stack_usage_info support to
2037 eliminate duplication.
2038 (riscv_expand_epilogue): Generate mret for interrupt handler.
2039 (riscv_epilogue_uses): New.
2040 (riscv_can_use_return_insn): Return false for interrupt handler.
2041 (riscv_function_ok_for_sibcall): Likewise.
2042 (riscv_set_current_function): Add interrupt handler support.
2043 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2044 * config/riscv/riscv.md (UNSPECV_MRET): New.
2046 (riscv_frflags, riscv_fsflags): Use tab after opcode.
2048 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2050 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
2052 PR tree-optimization/85712
2053 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2054 this candidate has already been replaced in-situ by a copy.
2056 2018-05-25 Jason Merrill <jason@redhat.com>
2058 PR c++/80485 - inline function non-zero address.
2059 * symtab.c (nonzero_address): Check DECL_COMDAT.
2061 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
2064 * config/alpha/alpha.md (ashlsi3): New insn pattern.
2065 (*ashlsi_se): Rename from *ashldi_se. Define as sign
2066 extension of SImode operation. Use const123_operand predicate.
2067 (*saddsi_1): Remove.
2068 (*saddl_se_1): Ditto.
2070 (*ssubl_se_1): Ditto.
2071 * config/alpha/predicates.md (const123_operand): New predicate.
2072 * config/alpha/constraints.md (P): Use IN_RANGE.
2074 2018-05-25 Richard Biener <rguenther@suse.de>
2076 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2078 (ref_maybe_used_by_stmt_p): Likewise.
2079 (stmt_may_clobber_ref_p): Likewise.
2080 (stmt_may_clobber_ref_p_1): Likewise.
2081 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2083 (ref_maybe_used_by_stmt_p): Likewise.
2084 (stmt_may_clobber_ref_p): Likewise.
2085 (stmt_may_clobber_ref_p_1): Likewise.
2086 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2087 the alias oracle to disambiguate DRs with stmts DR analysis
2089 (vect_analyze_data_refs): Do not give up on not analyzable
2090 DRs for BB vectorization. Remove code truncating the dataref
2093 2018-05-25 Jakub Jelinek <jakub@redhat.com>
2096 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2097 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
2098 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2100 2018-05-25 Richard Biener <rguenther@suse.de>
2102 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2103 function, combining stmt data ref gathering and fatal analysis
2105 (vect_analyze_data_refs): Remove now redudnant code and simplify.
2106 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2107 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2108 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2109 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2111 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2113 PR tree-optimization/85720
2114 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2115 SCC if all partitions are builtins.
2116 (version_loop_by_alias_check): New parameter. Generate cancelable
2117 runtime alias check if all partitions are builtins.
2118 (distribute_loop): Update call to above function.
2120 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2122 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2123 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2124 (parm_default_def_partition_arg): Ditto.
2125 (set_parm_default_def_partition): Ditto.
2126 (get_parm_default_def_partitions): Ditto and make it static.
2127 (get_undefined_value_partitions): Ditto and make it static.
2128 (remove_ssa_form): Refactor call to init_var_map here.
2129 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2130 computation for loop region.
2131 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2132 (register_default_def): Delete.
2133 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2134 (parm_default_def_partition_arg): Ditto.
2135 (set_parm_default_def_partition): Ditto.
2136 (get_parm_default_def_partitions): Ditto and make it static.
2137 (get_undefined_value_partitions): Ditto and make it static.
2138 (coalesce_with_default, coalesce_with_default): Update comment.
2139 (create_coalesce_list_for_region): New func factored out from
2140 create_outofssa_var_map.
2141 (populate_coalesce_list_for_outofssa): New func factored out from
2142 create_outofssa_var_map and coalesce_ssa_name.
2143 (create_outofssa_var_map): Delete.
2144 (coalesce_ssa_name): Refactor to support live range computation.
2145 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2146 (get_parm_default_def_partitions): Delete.
2147 (get_undefined_value_partitions): Ditto.
2148 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2149 computation for loop region.
2150 (new_tree_live_info, loe_visit_block): Ditto.
2151 (live_worklist, set_var_live_on_entry): Ditto.
2152 (calculate_live_on_exit, verify_live_on_entry): Ditto.
2153 * tree-ssa-live.h (struct _var_map): New fields.
2154 (init_var_map): Change decl.
2155 (region_contains_p): New.
2157 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2159 * tree-ssa-live.h (live_merge_and_clear): Delete.
2161 2018-05-25 Richard Biener <rguenther@suse.de>
2164 * tree-dump.c (dequeue_and_dump): Remove access to removed
2165 operand 2 of a SWITCH_EXPR.
2167 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2169 * doc/sourcebuild.texi (vect_double_cond_arith): Include
2170 multiplication and division.
2171 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2172 (cond_udiv@var{m}, cond_umod@var{m}): Document.
2173 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2174 (cond_udiv_optab, cond_umod_optab): New optabs.
2175 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2176 (IFN_COND_RDIV): New internal functions.
2177 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
2178 TRUNC_MOD_EXPR and RDIV_EXPR.
2179 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
2180 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
2182 (SVE_INT_BINARY): Include mult.
2183 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
2184 (optab, sve_int_op): Handle mult.
2185 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
2187 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
2188 for SVE_INT_BINARY_SD.
2190 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2192 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
2193 (optab, sve_int_op): Handle div and udiv.
2194 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
2195 for SVE_INT_BINARY_SD.
2196 (*<optab><mode>3): New insn for the same.
2198 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2200 * tree-vect-patterns.c: Include predict.h.
2201 (vect_recog_divmod_pattern): Restrict check for division support
2202 to when optimizing for size.
2204 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2206 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
2207 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
2208 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
2209 (gimple_match_op::set_op): Likewise.
2210 (gimple_resimplify4): Declare.
2211 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
2212 (expr::gen_transform): Likewise.
2213 (decision_tree::gen): Generate a simplification routine for 4 operands.
2214 * gimple-match-head.c (gimple_simplify): Add an overload for
2215 4 operands. In the top-level function, handle up to 4 call
2216 arguments and call gimple_resimplify4.
2217 (gimple_resimplify4): New function.
2218 (build_call_internal): Pass a fourth operand.
2219 (maybe_push_to_seq): Likewise.
2220 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
2221 Fold VEC_COND_EXPRs of an operation and a default value into
2222 an IFN_COND_* function if possible.
2223 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
2225 (SVE_COND_FP_BINARY): Include them.
2226 (optab, sve_fp_op): Handle them.
2227 (SVE_INT_BINARY_REV): New code iterator.
2228 (SVE_COND_FP_BINARY_REV): New int iterator.
2229 (commutative): New int attribute.
2230 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
2232 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
2234 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
2235 (*cond_<optab><mode>): New patterns for reversed operands.
2237 2018-05-25 Richard Biener <rguenther@suse.de>
2239 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
2240 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
2241 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
2242 (STMT_VINFO_GROUPED_ACCESS): Adjust.
2243 * tree-vect-data-refs.c (everywhere): Adjust users.
2244 * tree-vect-loop.c (everywhere): Likewise.
2245 * tree-vect-slp.c (everywhere): Likewise.
2246 * tree-vect-stmts.c (everywhere): Likewise.
2247 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2249 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2251 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
2253 (gcc_cv_as_section_exclude): ... this.
2254 Try Solaris as #exclude syntax.
2255 * configure: Regenerate.
2256 * config.in: Regenerate.
2257 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
2259 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
2260 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
2262 * varasm.c (default_elf_asm_named_section): Don't check if
2263 HAVE_GAS_SECTION_EXCLUDE is defined.
2265 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2267 * doc/md.texi: Update the documentation of the cond_* optabs
2268 to mention the new final operand. Fix GET_MODE_NUNITS call.
2269 Describe the scalar case too.
2270 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
2271 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
2273 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
2274 (get_conditional_internal_fn): Update comment.
2275 * tree-vect-loop.c (vectorizable_reduction): Pass the original
2276 accumulator value as a final argument to conditional functions.
2277 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
2278 a define_expand and add an "else" operand. Assert for now that
2279 the else operand is equal to operand 2. Use SVE_INT_BINARY and
2280 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
2281 (*cond_<optab><mode>): New patterns.
2282 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
2283 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
2284 (UNSPEC_COND_EOR): Delete.
2285 (optab): Remove associated mappings.
2286 (SVE_INT_BINARY): New code iterator.
2287 (sve_int_op): Remove int attribute and add "minus" to the code
2289 (SVE_COND_INT_OP): Delete.
2290 (SVE_COND_FP_OP): Rename to...
2291 (SVE_COND_FP_BINARY): ...this.
2293 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2295 * optabs.c (can_reuse_operands_p): New function.
2296 (maybe_legitimize_operands): Try to reuse the results for
2299 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2301 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
2302 Add {q} suffix to insn mnemonic.
2304 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2306 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
2307 (msp430_warn_func_return): New.
2309 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
2311 * fold-const.c (tree_nonzero_bits): New function.
2312 * fold-const.h (tree_nonzero_bits): Likewise.
2313 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
2314 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
2316 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2320 * varasm.c (assemble_alias): Check ifunc_resolver only on
2323 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2326 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
2327 when memory input operand is handled.
2329 2018-05-24 Luis Machado <luis.machado@linaro.org>
2331 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
2333 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
2335 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2337 * match.pd: Delay FMA folds until after vectorization.
2339 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
2342 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2343 address check not strict.
2345 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2347 * gimple-match.h (gimple_match_op): New class.
2348 (mprts_hook): Replace parameters with a gimple_match_op *.
2349 (maybe_build_generic_op): Likewise.
2350 (gimple_simplified_result_is_gimple_val): Replace parameters with
2351 a const gimple_match_op *.
2352 (gimple_simplify): Replace code_helper * and tree * parameters with
2353 a gimple_match_op * parameter.
2354 (gimple_resimplify1): Replace code_helper *, tree and tree *
2355 parameters with a gimple_match_op * parameter.
2356 (gimple_resimplify2): Likewise.
2357 (gimple_resimplify3): Likewise.
2358 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
2359 parameters with a gimple_match_op * parameter.
2360 * gimple-match-head.c (gimple_simplify): Change prototypes of
2361 auto-generated functions to take a gimple_match_op * instead of
2362 separate code_helper * and tree * parameters. Make the same
2363 change in the top-level overload and update calls to the
2364 gimple_resimplify routines. Update calls to the auto-generated
2365 functions and to maybe_push_res_to_seq in the publicly-facing
2366 operation-specific gimple_simplify overloads.
2367 (gimple_match_op::MAX_NUM_OPS): Define.
2368 (gimple_resimplify1): Replace rcode and ops with a single res_op
2369 parameter. Update call to gimple_simplify.
2370 (gimple_resimplify2): Likewise.
2371 (gimple_resimplify3): Likewise.
2372 (mprts_hook): Replace parameters with a gimple_match_op *.
2373 (maybe_build_generic_op): Likewise.
2374 (build_call_internal): Replace type, nargs and ops with
2375 a gimple_match_op *.
2376 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
2377 with a single gimple_match_op *. Update calls to mprts_hook,
2378 build_call_internal and gimple_simplified_result_is_gimple_val.
2379 Factor out code that is common to the tree_code and combined_fn cases.
2380 * genmatch.c (expr::gen_transform): Replace tem_code and
2381 tem_ops with a gimple_match_op called tem_op. Update calls
2382 to the gimple_resimplify functions and maybe_push_res_to_seq.
2383 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
2384 res_ops. Update call to the gimple_resimplify functions.
2385 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
2386 (decision_tree::gen): Make the functions take a gimple_match_op *
2387 called res_op instead of separate res_code and res_ops parameters.
2388 Update call accordingly.
2389 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
2390 and ops with a single res_op parameter. Update calls to
2391 maybe_build_generic_op and maybe_push_res_to_seq.
2392 (fold_stmt_1): Update calls to gimple_simplify and
2393 replace_stmt_with_simplification.
2394 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
2395 and gimple_simplified_result_is_gimple_val.
2396 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
2398 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
2399 with a gimple_match_op *.
2400 (vn_nary_build_or_lookup): Likewise. Update call to
2401 vn_nary_build_or_lookup_1.
2402 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
2403 gimple_match_op *. Update calls to the gimple_resimplify routines
2404 and to gimple_simplified_result_is_gimple_val.
2405 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
2406 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
2407 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
2408 (visit_nary_op): Likewise.
2409 (visit_reference_op_load): Likewise.
2411 2018-05-23 Luis Machado <luis.machado@linaro.org>
2413 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
2414 modifier for printing the step amount.
2416 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
2419 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
2422 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
2424 * doc/sourcebuild.texi (Endianness): New subsubsection.
2426 2018-05-23 Luis Machado <luis.machado@linaro.org>
2428 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2429 <prefetch_dynamic_strides>: New const bool field.
2430 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2431 prefetch_dynamic_strides.
2432 (exynosm1_prefetch_tune): Likewise.
2433 (thunderxt88_prefetch_tune): Likewise.
2434 (thunderx_prefetch_tune): Likewise.
2435 (thunderx2t99_prefetch_tune): Likewise.
2436 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
2438 (aarch64_override_options_internal): Update to set
2439 PARAM_PREFETCH_DYNAMIC_STRIDES.
2440 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2441 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2442 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2443 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2444 prefetch-dynamic-strides setting.
2446 2018-05-23 Luis Machado <luis.machado@linaro.org>
2448 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2449 <minimum_stride>: New const int field.
2450 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2451 minimum_stride field defaulting to -1.
2452 (exynosm1_prefetch_tune): Likewise.
2453 (thunderxt88_prefetch_tune): Likewise.
2454 (thunderx_prefetch_tune): Likewise.
2455 (thunderx2t99_prefetch_tune): Likewise.
2456 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
2457 <default_opt_level>: Set to 3.
2458 (aarch64_override_options_internal): Update to set
2459 PARAM_PREFETCH_MINIMUM_STRIDE.
2460 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2461 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2462 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2463 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2464 stride is constant and is below the minimum stride threshold.
2466 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2468 * config/arm/arm-cpus.in (mode26): Delete.
2469 (armv4): Delete mode26 reference.
2470 * config/arm/arm.c (arm_configure_build_target): Delete use of
2473 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2475 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
2477 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
2478 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
2480 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
2481 (floatunsdidf2): Ditto.
2483 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2485 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
2486 (fixuns_trunc<mode>si2_avx512f): Ditto.
2487 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
2488 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
2489 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
2491 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2493 PR rtl-optimization/79985
2494 * df-scan.c (df_insn_refs_collect): Remove special case for
2495 global registers and asm statements.
2497 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2499 * extend.texi (Global Register Variables): Rewrite the bullet list.
2500 Note that the register is available for allocation. Note that access
2501 via inline asm must use constraints. Add note about async-signal
2502 handlers. Remove paragraph about automagic register selection.
2504 2018-05-23 Richard Biener <rguenther@suse.de>
2506 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
2507 of fixed offset from memset VN.
2509 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
2511 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
2513 (alloc_cand_and_find_basis): Initialize first_interp field.
2514 (slsr_process_mul): Modify first_interp field.
2515 (slsr_process_add): Likewise.
2516 (slsr_process_cast): Modify first_interp field for each new
2518 (slsr_process_copy): Likewise.
2519 (dump_candidate): Dump first_interp field.
2520 (replace_mult_candidate): Process all interpretations, not just
2522 (replace_rhs_if_not_dup): Likewise.
2523 (replace_one_candidate): Likewise.
2525 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
2527 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
2529 (aarch64_needs_frame_chain): New function.
2530 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
2532 2018-05-23 Sudakshina Das <sudi.das@arm.com>
2535 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
2536 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
2537 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
2538 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
2539 as true for strict-align.
2540 (aarch64_can_inline_p): Perform checks even when callee has no
2541 attributes to check for strict alignment.
2542 * doc/extend.texi (AArch64 Function Attributes): Document
2544 * doc/invoke.texi: (AArch64 Options): Likewise.
2546 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
2548 PR tree-optimization/85853
2549 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
2550 the handling of the root of the node to...
2551 (vect_slp_analyze_node_operations_1): ...this new function,
2552 and run the whole thing with the child nodes' def types
2553 set according to their SLP node's def type.
2555 2018-05-23 Richard Biener <rguenther@suse.de>
2558 * tree-data-ref.c (create_runtime_alias_checks): Defer
2559 and ignore overflow warnings.
2561 2018-05-23 Yury Gribov <tetra2005@gmail.com>
2563 PR tree-optimization/85822
2564 * tree-vrp.c (is_masked_range_test): Fix handling of negative
2567 2018-05-23 Richard Biener <rguenther@suse.de>
2569 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2570 memset constants via native_interpret_expr.
2572 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
2575 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
2577 (cgraph_node::create_alias): Likewise.
2578 (cgraph_node::get_availability): Check ifunc_resolver instead
2579 of looking up ifunc attribute.
2580 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2581 * varasm.c (do_assemble_alias): Likewise.
2582 (assemble_alias): Likewise.
2583 (default_binds_local_p_3): Likewise.
2584 * cgraph.h (cgraph_node): Add ifunc_resolver.
2585 (cgraph_node::only_called_directly_or_aliased_p): Return false
2587 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
2589 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
2590 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
2591 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
2592 instead of looking up ifunc attribute.
2594 2018-05-22 Luis Machado <luis.machado@linaro.org>
2596 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
2598 2018-05-22 Martin Sebor <msebor@redhat.com>
2601 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
2602 only when expasion succeeds.
2603 (expand_builtin_strcmp): Same.
2604 (expand_builtin_strncmp): Same.
2606 2018-05-22 Martin Sebor <msebor@redhat.com>
2608 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
2610 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2611 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2613 * config/aarch64/aarch64-ldpstp.md: Replace uses of
2614 aarch64_mem_pair_operand with memory_operand and delete operand swapping
2616 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2617 Add check for legitimate_address.
2618 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
2619 (aarch64_swap_ldrstr_operands): New.
2620 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
2623 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2624 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2626 * config/aarch64/aarch64.md: New patterns to generate stp
2628 (store_pair_sw, store_pair_dw): New patterns to generate stp for
2629 single words and double words.
2630 (load_pair_sw, load_pair_dw): Likewise.
2631 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
2633 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
2635 * config/aarch64/aarch64-ldpstp.md: Modify peephole
2636 for different mode ldpstp and add peephole for merged zero stores.
2638 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2640 (aarch64_gen_store_pair): Rename calls to match new patterns.
2641 (aarch64_gen_load_pair): Rename calls to match new patterns.
2642 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
2643 (load_pair<DREG:mode><DREG2:mode>): ... This.
2644 (store_pair<mode>): Rename to...
2645 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
2646 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
2648 (V_INT_EQUIV): Handle SImode.
2649 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
2652 2018-05-22 Martin Sebor <msebor@redhat.com>
2655 * calls.c (maybe_warn_nonstring_arg): Use string length to set
2656 or ajust the presumed bound on an operation to avoid unnecessary
2659 2018-05-22 Martin Sebor <msebor@redhat.com>
2661 PR tree-optimization/85826
2662 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
2663 assuming that a DECL necesarily has a constant size.
2665 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
2668 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
2670 2018-05-22 Richard Biener <rguenther@suse.de>
2672 PR tree-optimization/85834
2673 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
2674 non-constant and non-zero memset arguments.
2676 2018-05-22 Martin Liska <mliska@suse.cz>
2679 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
2681 2018-05-22 Richard Biener <rguenther@suse.de>
2683 PR tree-optimization/85863
2684 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
2685 comparisons when vectype is specified.
2686 (vectorizable_condition): Do not specify vectype for
2687 vect_is_simple_cond when SLP vectorizing.
2689 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
2692 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
2693 define __ibm128 as long double.
2694 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
2695 as a distinct type when IEEE 128-bit support is enabled.
2696 (init_float128_ieee): Fix up conversions between IFmode and IEEE
2697 128-bit types to use the correct functions.
2698 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
2699 convert between 128-bit floating point types that have different
2700 modes but the same representation, instead of using gen_lowpart to
2702 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
2704 (IFKF_reg): New attributes to give the register constraints for
2706 (extend<mode>tf2_internal): New insns to mark an explicit
2707 conversion between 128-bit floating point types that have a
2708 different mode but share the same representation.
2710 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
2712 PR tree-optimization/85814
2713 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
2714 a null return from get_strinfo when unsharing the next
2715 strinfo in the chain.
2717 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2720 * varasm.c (weak_finish): Clean up weak_decls.
2722 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2724 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
2725 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
2726 UNSPEC_UADALP values.
2727 * config/aarch64/iterators.md (ABAL): New int iterator.
2730 (sur): Add mappings for the above.
2731 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
2733 (aarch64_<sur>abal<mode>_4): Likewise.
2734 (aarch64_<sur>adalp<mode>_3): Likewise.
2735 (<sur>sadv16qi): New define_expand.
2737 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
2739 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
2740 (*movdf_internal): Ditto.
2741 (*rcpsf2_sse): Ditto.
2742 (*rsqrtsf2_sse): Ditto.
2743 (*sqrt<mode>2_sse): Ditto.
2745 2018-05-21 Tamar Christina <tamar.christina@arm.com>
2747 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
2749 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
2750 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
2751 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
2752 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
2754 * config/aarch64/arm_neon.h: Likewise.
2755 * config/aarch64/iterators.md (VQ_I): New.
2757 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
2759 * config.gcc: Add arc/t-multilib-linux to tmake_file for
2761 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
2764 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
2766 * config/nds32/constraints.md (S): New constraint.
2767 * config/nds32/nds32.md (call_internal): Use constraint S.
2768 (call_value_internal): Likewise.
2769 (sibcall_internal): Likewise.
2770 (sibcall_value_internal): Likewise.
2772 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
2773 Chung-Ju Wu <jasonwucj@gmail.com>
2775 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
2778 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
2779 Chung-Ju Wu <jasonwucj@gmail.com>
2781 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
2782 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
2783 (nds32_rtx_costs_impl): Simplify.
2784 (nds32_address_cost_impl): Simplify.
2785 (nds32_init_rtx_costs): New function.
2786 (nds32_rtx_costs_speed_prefer): Likewise.
2787 (nds32_rtx_costs_size_prefer): Likewise.
2788 (nds32_address_cost_speed_prefer): Likewise.
2789 (nds32_address_cost_speed_fwprop): Likewise.
2790 (nds32_address_cost_size_prefer): Likewise.
2791 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
2792 * config/nds32/nds32.c (nds32_option_override): Use
2793 nds32_init_rtx_costs function.
2795 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
2797 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
2798 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
2799 (TARGET_PIPELINE_N8): Likewise.
2800 (TARGET_PIPELINE_N10): Likewise.
2801 (TARGET_PIPELINE_N13): Likewise.
2802 (TARGET_PIPELINE_GRAYWOLF): Likewise.
2804 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
2806 * config/nds32/nds32-fpu.md: Update copyright year.
2808 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2810 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
2812 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2814 * config/nds32/nds32.c
2815 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
2816 * config/nds32/nds32.opt (minline-asm-r15): New option.
2818 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2820 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
2822 * config/nds32/nds32.md (abssi2): New pattern.
2824 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
2826 * config/i386/i386.md (rex64namesuffix): New mode attribute.
2827 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
2828 Merge insn pattern from sse_cvtsi2ss<round_name> and
2829 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
2830 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
2831 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
2832 using SWI48 mode iterator.
2833 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
2834 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
2835 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
2836 pattern from sse_cvttss2si<round_saeonly_name>
2837 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
2838 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
2839 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
2840 using SWI48 mode iterator.
2841 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
2842 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
2843 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
2844 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
2845 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
2846 using SWI48 mode iterator.
2847 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
2848 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
2849 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
2850 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
2851 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
2852 SWI48 mode iterator.
2853 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
2854 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
2855 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
2856 pattern from sse_cvttsd2si<round_saeonly_name>
2857 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
2859 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2861 * config/nds32/nds32-md-auxiliary.c
2862 (nds32_valid_smw_lwm_base_p): Refine.
2863 (nds32_output_smw_single_word): Refine.
2864 (nds32_output_smw_double_word): New.
2865 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
2867 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2869 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
2870 (nds32_output_stack_pop): Refine.
2871 (nds32_expand_unaligned_load): Refine.
2872 (nds32_expand_unaligned_store): Refine.
2874 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2875 Chung-Ju Wu <jasonwucj@gmail.com>
2877 * config/nds32/constants.md: Add TP_REGNUM constant.
2878 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
2879 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
2881 * config/nds32/nds32-doubleword.md: Consider flag_pic.
2882 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
2883 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
2884 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
2885 and PIC code generation.
2886 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
2888 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
2890 * config/nds32/nds32.md: Support TLS and PIC.
2891 * config/nds32/nds32.c: Support TLS and PIC.
2892 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
2893 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
2896 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2898 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
2899 mode with E_ prefix.
2901 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2902 Chung-Ju Wu <jasonwucj@gmail.com>
2904 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
2905 * config/nds32/nds32-md-auxiliary.c
2906 (symbolic_reference_mentioned_p): New.
2907 (nds32_legitimize_ict_address): New.
2908 (nds32_expand_ict_move): New.
2909 (nds32_indirect_call_referenced_p): New.
2910 (nds32_symbol_binds_local_p): Delete.
2911 (nds32_long_call_p): Modify.
2912 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
2913 * config/nds32/nds32-protos.h
2914 (symbolic_reference_mentioned_p): Declare.
2915 (nds32_legitimize_ict_address): Declare.
2916 (nds32_expand_ict_move): Declare.
2917 (nds32_indirect_call_referenced_p): Declare.
2918 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
2919 (nds32_relax_group): Use nds32_ict_const_p as condition.
2920 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
2921 (nds32_asm_file_start): Output ict_model directive in asm code.
2922 (nds32_legitimate_address_p): Consider indirect call.
2923 (nds32_print_operand): Consider indirect call.
2924 (nds32_print_operand_address): Consider indirect call.
2925 (nds32_insert_attributes): Handle "indirect_call" attribute.
2926 (TARGET_LEGITIMATE_ADDRESS_P): Define.
2927 (TARGET_LEGITIMATE_CONSTANT_P): Define.
2928 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
2929 (TARGET_DELEGITIMIZE_ADDRESS): Define.
2930 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
2931 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
2932 (TARGET_ICT_MODEL_SMALL): Define.
2933 (TARGET_ICT_MODEL_LARGE): Define.
2934 * config/nds32/nds32.md (movsi): Consider ict model.
2935 (call, call_value): Consider ict model.
2936 (sibcall, sibcall_value): Consider ict model.
2937 * config/nds32/nds32.opt (mict-model): New option.
2938 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
2941 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
2942 Monk Chiang <sh.chiang04@gmail.com>
2943 Jim Wilson <jimw@sifive.com>
2945 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
2946 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
2947 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
2948 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
2949 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
2950 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
2951 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
2952 compute save_libcall_adjustment properly.
2953 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
2954 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
2955 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
2956 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
2957 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
2958 (ABI_SPEC): Handle mabi=ilp32e.
2959 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
2960 (RVE): Add RVE mask.
2961 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
2962 <-march>: Add rv32e as an example.
2964 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
2967 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
2968 (intra_create_variable_infos): Handle C++ constructors.
2970 2018-05-18 Martin Liska <mliska@suse.cz>
2972 * passes.def: Remove a redundant pass.
2974 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
2977 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
2979 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2981 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
2983 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
2985 (armv2, armv2a, armv3, armv3m): Delete architectures.
2986 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
2987 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
2988 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
2990 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
2991 (*mulsidi3adddi): Likewise.
2992 (mulsidi3): Likewise.
2993 (*mulsidi3_nov6): Likewise.
2994 (umulsidi3): Likewise.
2995 (umulsidi3_nov6): Likewise.
2996 (umaddsidi4): Likewise.
2997 (*umulsidi3adddi): Likewise.
2998 (smulsi3_highpart): Likewise.
2999 (*smulsi3_highpart_nov6): Likewise.
3000 (umulsi3_highpart): Likewise.
3001 (*umulsi3_highpart_nov6): Likewise.
3002 * config/arm/arm.h (arm_arch3m): Delete.
3003 * config/arm/arm.c (arm_arch3m): Delete.
3004 (arm_option_override_internal): Update armv3-related comment.
3005 (arm_configure_build_target): Delete use of isa_bit_mode32.
3006 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
3007 (arm_rtx_costs_internal): Delete check of arm_arch3m.
3008 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
3010 (mulusa3): Likewise.
3011 * config/arm/arm-protos.h (arm_arch3m): Delete.
3012 * config/arm/arm-tables.opt: Regenerate.
3013 * config/arm/arm-tune.md: Likewise.
3014 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
3015 deleted architectures.
3017 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3019 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3020 (armv5t, armv5te): New features.
3021 (ARMv5, ARMv5e): Delete fgroups.
3022 (ARMv5t, ARMv5te): Adjust for above changes.
3024 (armv5, armv5e): Delete arches.
3025 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3027 (*call_reg_arm): Likewise.
3028 (*call_value_reg_armv5): Likewise.
3029 (*call_value_reg_arm): Likewise.
3030 (*call_symbol): Likewise.
3031 (*call_value_symbol): Likewise.
3032 (*sibcall_insn): Likewise.
3033 (*sibcall_value_insn): Likewise.
3035 (prefetch): Likewise.
3036 (define_split and define_peephole2 dependent on arm_arch5):
3038 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3040 (TARGET_ARM_QBIT): Likewise.
3041 (TARGET_DSP_MULTIPLY): Likewise.
3042 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3043 (arm_arch5, arm_arch5e): Delete.
3044 (arm_arch5t, arm_arch5te): Declare.
3045 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3046 (arm_arch5t): Declare.
3047 (arm_option_reconfigure_globals): Update for the above.
3048 (arm_options_perform_arch_sanity_checks): Update comment, replace
3049 use of arm_arch5 with arm_arch5t.
3050 (use_return_insn): Likewise.
3051 (arm_emit_call_insn): Likewise.
3052 (output_return_instruction): Likewise.
3053 (arm_final_prescan_insn): Likewise.
3054 (arm_coproc_builtin_available): Likewise.
3055 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3056 arm_arch5e with arm_arch5t and arm_arch5te.
3057 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3058 (arm_arch5t, arm_arch5te): Declare.
3059 * config/arm/arm-tables.opt: Regenerate.
3060 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3061 * config/arm/t-multilib: Likewise.
3062 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3063 instead of arm_arch5.
3064 (*call_reg_thumb1): Likewise.
3065 (*call_value_reg_thumb1_v5): Likewise.
3066 (*call_value_reg_thumb1): Likewise.
3067 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3069 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3071 2018-05-18 Martin Liska <mliska@suse.cz>
3073 PR gcov-profile/84846
3074 * doc/gcov.texi: Document -t option of gcov tool.
3076 2018-05-18 Martin Liska <mliska@suse.cz>
3078 PR gcov-profile/84846
3079 * gcov.c (print_usage): Add new -t option.
3080 (process_args): Handle the option.
3081 (generate_results): Use stdout as output when requested by
3084 2018-05-18 Martin Liska <mliska@suse.cz>
3086 PR gcov-profile/84846
3087 * coverage.c (coverage_init): Write PWD to .gcno file.
3088 * doc/gcov.texi: Document how working directory is printed.
3089 * gcov-dump.c (dump_gcov_file): Print PWD.
3090 * gcov.c (output_intermediate_file): Likewise.
3091 (read_graph_file): Read PWD string.
3092 (output_lines): Print PWD.
3094 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3097 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3098 for retval and return false if all args to phi are zero.
3100 2018-05-18 Richard Biener <rguenther@suse.de>
3102 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3104 (evrp_dom_walker::before_dom_children): Call it.
3106 2018-05-18 Richard Biener <rguenther@suse.de>
3108 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3109 results when processing array refs with variable index.
3111 2018-05-18 Toon Moene <toon@moene.org>
3113 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3114 directly after that of -floop-interchange. Indicate that both
3115 options are enabled by default when specifying -O3.
3117 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3119 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3120 iterator. Delete separate integer-mode vec_set<mode> expander.
3121 (aarch64_simd_vec_setv2di): Delete.
3122 (vec_setv2di): Delete.
3123 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3124 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
3125 the "w, r" alternative.
3127 2018-05-18 Martin Liska <mliska@suse.cz>
3129 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3130 * tree-pass.h (make_pass_lower_switch_O0): New function.
3131 * tree-switch-conversion.c (node_has_low_bound): Remove.
3132 (node_has_high_bound): Likewise.
3133 (node_is_bounded): Likewise.
3134 (class pass_lower_switch): Make it a template type and create
3136 (pass_lower_switch::execute): Add template argument.
3137 (make_pass_lower_switch): New function.
3138 (make_pass_lower_switch_O0): New function.
3139 (do_jump_if_equal): Remove.
3140 (emit_case_nodes): Simplify to just handle all 3 cases and leave
3141 all the hard work to tree optimization passes.
3143 2018-05-18 Martin Liska <mliska@suse.cz>
3145 * dbgcnt.c (limit_low): Renamed from limit.
3146 (limit_high): New variable.
3147 (dbg_cnt_is_enabled): Check for upper limit.
3148 (dbg_cnt): Adjust dumping.
3149 (dbg_cnt_set_limit_by_index): Add new argument for high
3151 (dbg_cnt_set_limit_by_name): Likewise.
3152 (dbg_cnt_process_single_pair): Parse new format.
3153 (dbg_cnt_process_opt): Use strtok.
3154 (dbg_cnt_list_all_counters): Remove 'value' and add
3156 * doc/invoke.texi: Document changes.
3158 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
3160 * doc/sourcebuild.texi (scalar_all_fma): Document.
3161 * tree.def (FMA_EXPR): Delete.
3162 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3163 * internal-fn.c (ternary_direct): New macro.
3164 (expand_ternary_optab_fn): Likewise.
3165 (direct_ternary_optab_supported_p): Likewise.
3166 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3167 * builtins.c (fold_builtin_fma): Delete.
3168 (fold_builtin_3): Don't call it.
3169 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3170 * expr.c (expand_expr_real_2): Likewise.
3171 * fold-const.c (operand_equal_p): Likewise.
3172 (fold_ternary_loc): Likewise.
3173 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3174 * gimple.c (DEFTREECODE): Likewise.
3175 * gimplify.c (gimplify_expr): Likewise.
3176 * optabs-tree.c (optab_for_tree_code): Likewise.
3177 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
3178 * tree-eh.c (operation_could_trap_p): Likewise.
3179 (stmt_could_throw_1_p): Likewise.
3180 * tree-inline.c (estimate_operator_cost): Likewise.
3181 * tree-pretty-print.c (dump_generic_node): Likewise.
3182 (op_code_prio): Likewise.
3183 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3184 * tree-ssa-operands.c (get_expr_operands): Likewise.
3185 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
3186 * fold-const-call.h (fold_fma): Delete.
3187 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
3188 CFN_FNMA and CFN_FNMS.
3190 * genmatch.c (combined_fn): New enum.
3191 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
3192 (commutative_op): New function.
3193 (commutate): Use it. Handle more than 2 operands.
3194 (dt_operand::gen_gimple_expr): Use commutative_op.
3195 (parser::parse_expr): Allow :c to be used with non-binary
3196 operators if the commutative operand is known.
3197 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
3198 CFN_FMS, CFN_FNMA and CFN_FNMS.
3199 (backprop::process_assign_use): Remove FMA_EXPR handling.
3200 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
3201 (gen_hsa_fma): New function.
3202 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
3203 IFN_FNMA and IFN_FNMS.
3204 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
3205 * gimple-fold.h (follow_all_ssa_edges): Declare.
3206 * gimple-fold.c (follow_all_ssa_edges): New function.
3207 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
3208 gimple_build interface and use follow_all_ssa_edges to fold the result.
3209 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
3210 instead of checking for optabs directly.
3211 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
3212 rather than FMA_EXPRs.
3213 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
3214 call to IFN_FMA instead of an FMA_EXPR.
3216 2018-05-17 Jim Wilson <jimw@sifive.com>
3218 * expr.c (do_tablejump): When converting index to Pmode, if we have a
3219 sign extended promoted subreg, and the range does not have the sign bit
3220 set, then do a sign extend.
3222 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
3223 test, check for sign extended subreg and/or constant operands, and
3224 do a sign extend in that case.
3226 2018-05-17 Steve Ellcey <sellcey@cavium.com>
3228 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
3229 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
3231 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
3232 Change logics_shift_reg to logics_shift_imm.
3233 (thunderx2t99_fp_loadpair_basic): Delete.
3234 (thunderx2t99_fp_storepair_basic): Delete.
3235 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
3236 (thunderx2t99_asimd_polynomial): Delete.
3237 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
3238 and neon_fp_mul_d_scalar_q.
3239 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
3240 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
3241 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
3242 (thunderx2t99_asimd_lut): Add missing tbl types.
3243 (thunderx2t99_asimd_ext): Delete.
3244 (thunderx2t99_asimd_load1_1_mult): Delete.
3245 (thunderx2t99_asimd_load1_2_mult): Delete.
3246 (thunderx2t99_asimd_load1_ldp): New.
3247 (thunderx2t99_asimd_load1): New.
3248 (thunderx2t99_asimd_load2): Add missing *load2* types.
3249 (thunderx2t99_asimd_load3): New.
3250 (thunderx2t99_asimd_load4): New.
3251 (thunderx2t99_asimd_store1_1_mult): Delete.
3252 (thunderx2t99_asimd_store1_2_mult): Delete.
3253 (thunderx2t99_asimd_store2_mult): Delete.
3254 (thunderx2t99_asimd_store2_onelane): Delete.
3255 (thunderx2t99_asimd_store_stp): New.
3256 (thunderx2t99_asimd_store1): New.
3257 (thunderx2t99_asimd_store2): New.
3258 (thunderx2t99_asimd_store3): New.
3259 (thunderx2t99_asimd_store4): New.
3261 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
3263 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
3264 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
3266 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
3267 Segher Boessenkool <segher@kernel.crashing.org>
3270 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
3273 2018-05-17 Richard Biener <rguenther@suse.de>
3275 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
3276 for pruning loop and prune defs feeding only already visited PHIs.
3278 2018-05-17 Richard Biener <rguenther@suse.de>
3280 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
3282 2018-05-17 Bin Cheng <bin.cheng@arm.com>
3283 Richard Biener <rguenther@suse.de>
3285 PR tree-optimization/85793
3286 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
3287 for VMAT_ELEMENTWISE.
3289 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3291 * internal-fn.h (lookup_internal_fn): Declare
3292 * internal-fn.c (lookup_internal_fn): New function.
3293 * gimple.c (gimple_build_call_from_tree): Handle calls to
3295 * gimple-pretty-print.c (dump_gimple_call): Print "." before
3296 internal function names.
3297 * tree-pretty-print.c (dump_generic_node): Likewise.
3298 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
3300 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3302 * gimple-fold.h (gimple_build): Make the function forms take
3303 combined_fn rather than built_in_function.
3304 (gimple_simplify): Likewise.
3305 * gimple-match-head.c (gimple_simplify): Likewise.
3306 * gimple-fold.c (gimple_build): Likewise.
3307 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
3308 rather than gimple_build_call_internal.
3309 (get_initial_defs_for_reduction): Likewise.
3310 (vect_create_epilog_for_reduction): Likewise.
3311 (vectorizable_live_operation): Likewise.
3313 2018-05-17 Martin Liska <mliska@suse.cz>
3315 * gimple-ssa-sprintf.c (format_directive): Do not use
3316 space in between 'G_' and '('.
3318 2018-05-17 Jakub Jelinek <jakub@redhat.com>
3321 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
3322 even if the mask is not all ones.
3325 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
3327 (ix86_gimple_fold_builtin): Likewise.
3330 * config/i386/i386.c: Include tree-vector-builder.h.
3331 (ix86_vector_shift_count): New function.
3332 (ix86_fold_builtin): Fold shift builtins by scalar count.
3333 (ix86_gimple_fold_builtin): Likewise.
3335 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
3336 _mm512_setzero): New intrinsics.
3338 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
3339 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3341 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
3342 code generation for cases where splatting a value is not useful.
3343 * simplify-rtx.c (simplify_ternary_operation): Simplify
3344 vec_merge across a vec_duplicate and a paradoxical subreg forming
3345 a vector mode to a vec_concat.
3347 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
3349 * config.gcc: Support "goldmont-plus".
3350 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3352 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3353 PROCESSOR_GOLDMONT_PLUS.
3354 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
3355 (processor_target_table): Add "goldmont-plus".
3356 (PTA_GOLDMONT_PLUS): Define.
3357 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
3358 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
3359 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
3360 (fold_builtin_cpu): Add "goldmont-plus".
3361 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
3362 (ix86_option_override_internal): Add "goldmont-plus".
3363 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
3364 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
3365 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
3366 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
3368 2018-05-17 Richard Biener <rguenther@suse.de>
3370 PR tree-optimization/85757
3371 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
3372 remove defs that only feed that PHI from further processing.
3374 2018-05-16 Jim Wilson <jimw@sifive.com>
3376 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
3378 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
3380 2018-05-16 Mark Wielaard <mark@klomp.org>
3382 * dwarf2out.c (count_index_strings): New function.
3383 (output_indirect_strings): Call count_index_strings and generate
3384 header for dwarf_version >= 5.
3386 2018-05-16 Mark Wielaard <mark@klomp.org>
3388 * dwarf2out.c (dwarf_FORM): New function.
3389 (set_indirect_string): Use dwarf_FORM.
3390 (reset_indirect_string): Likewise.
3391 (size_of_die): Likewise.
3392 (value_format): Likewise.
3393 (output_die): Likewise.
3394 (add_skeleton_AT_string): Likewise.
3395 (output_macinfo_op): Likewise.
3396 (index_string): Likewise.
3397 (output_index_string_offset): Likewise.
3398 (output_index_string): Likewise.
3399 (count_index_strings): Likewise.
3401 2018-05-16 Carl Love <cel@us.ibm.com>
3403 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3404 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
3406 2018-05-16 Martin Jambor <mjambor@suse.cz>
3408 * ipa-prop.c (ipa_free_all_edge_args): Remove.
3409 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
3411 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
3413 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
3414 (fnma<mode>4): Likewise.
3415 (fms<mode>4): Likewise.
3416 (fnms<mode>4): Likewise.
3417 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
3418 (aarch64_fnma<mode>4): Likewise.
3419 (aarch64_fms<mode>4): Likewise.
3420 (aarch64_fnms<mode>4): Likewise.
3421 (aarch64_fnmadd<mode>4): Likewise.
3423 2018-05-16 Jason Merrill <jason@redhat.com>
3425 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
3427 2018-05-16 Richard Biener <rguenther@suse.de>
3429 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
3430 (dump_stmt_cost): Declare.
3431 (add_stmt_cost): Dump cost we add.
3432 (add_stmt_costs): New function.
3433 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
3435 (vect_analyze_stmt): Adjust prototype.
3436 (vectorizable_condition): Likewise.
3437 (vectorizable_live_operation): Likewise.
3438 (vectorizable_reduction): Likewise.
3439 (vectorizable_induction): Likewise.
3440 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
3441 cost vector to pass to vectorizable_ and record afterwards.
3442 (vect_model_reduction_cost): Take cost vector argument and adjust.
3443 (vect_model_induction_cost): Likewise.
3444 (vectorizable_reduction): Likewise.
3445 (vectorizable_induction): Likewise.
3446 (vectorizable_live_operation): Likewise.
3447 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
3448 SLP_TREE_NUMBER_OF_VEC_STMTS.
3449 (vect_analyze_slp_cost_1): Remove.
3450 (vect_analyze_slp_cost): Likewise.
3451 (vect_slp_analyze_node_operations): Take visited args and
3452 a target cost vector. Avoid processing already visited stmt sets.
3453 (vect_slp_analyze_operations): Use a local cost vector to gather
3454 costs and register those of non-discarded instances.
3455 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
3456 (vect_schedule_slp_instance): Remove copying of
3457 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
3459 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
3460 adding cost. Record cost entry location.
3461 (vect_prologue_cost_for_slp_op): Function to compute cost of
3462 a constant or invariant generated for SLP vect in the prologue,
3463 split out from vect_analyze_slp_cost_1.
3464 (vect_model_simple_cost): Make static. Adjust for SLP costing.
3465 (vect_model_promotion_demotion_cost): Likewise.
3466 (vect_model_store_cost): Likewise, make static.
3467 (vect_model_load_cost): Likewise.
3468 (vectorizable_bswap): Add cost vector arg and adjust.
3469 (vectorizable_call): Likewise.
3470 (vectorizable_simd_clone_call): Likewise.
3471 (vectorizable_conversion): Likewise.
3472 (vectorizable_assignment): Likewise.
3473 (vectorizable_shift): Likewise.
3474 (vectorizable_operation): Likewise.
3475 (vectorizable_store): Likewise.
3476 (vectorizable_load): Likewise.
3477 (vectorizable_condition): Likewise.
3478 (vectorizable_comparison): Likewise.
3479 (can_vectorize_live_stmts): Likewise.
3480 (vect_analyze_stmt): Likewise.
3481 (vect_transform_stmt): Adjust calls to vectorizable_*.
3482 * tree-vectorizer.c: Include gimple-pretty-print.h.
3483 (dump_stmt_cost): New function.
3485 2018-05-16 Richard Biener <rguenther@suse.de>
3487 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
3488 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
3489 * tree-ssa-dse.c: Include tree-ssa-loop.h.
3490 (check_name): New callback.
3491 (dse_classify_store): Track cycles via a visited bitmap of PHI
3492 defs and simplify handling of in-loop and across loop dead stores
3493 and properly fail for loop-variant refs. Handle byte-tracking with
3494 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
3497 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
3499 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
3500 (vect_get_mask_type_for_stmt): Likewise.
3501 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
3503 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
3504 to determine the statement's vector type and the vector type that
3505 should be used for calculating nunits. Deal with cases in which
3506 the type has to be deferred.
3507 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
3508 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
3509 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
3510 (vect_determine_vf_for_stmt): New functions, split out from...
3511 (vect_determine_vectorization_factor): ...here.
3512 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
3513 (vect_get_mask_type_for_stmt): New functions, split out from
3514 vect_determine_vectorization_factor.
3516 2018-05-16 Richard Biener <rguenther@suse.de>
3518 * tree-cfg.c (verify_gimple_assign_ternary): Properly
3519 verify the [VEC_]COND_EXPR embedded comparison.
3521 2018-05-15 Martin Sebor <msebor@redhat.com>
3523 PR tree-optimization/85753
3524 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
3525 RECORD_TYPE in addition to ARRAY_TYPE.
3527 2018-05-15 Martin Sebor <msebor@redhat.com>
3530 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
3532 2018-05-15 Richard Biener <rguenther@suse.de>
3534 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
3535 add by_clobber_p one. Change algorithm to collect all defs
3536 representing uses we need to walk and try reducing them to
3537 a single one before failing.
3538 (dse_dom_walker::dse_optimize_stmt): Adjust.
3540 2018-05-13 Mark Wielaard <mark@klomp.org>
3542 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
3543 (size_of_loc_descr): Likewise.
3544 (output_loc_operands): Likewise.
3545 (output_loc_operands_raw): Likewise.
3546 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
3547 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
3548 (hash_loc_operands): Likewise.
3549 (compare_loc_operands): Likewise.
3551 2018-05-14 Mark Wielaard <mark@klomp.org>
3553 * dwarf2out.c (count_index_addrs): New function.
3554 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
3556 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3558 PR tree-optimization/83648
3559 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
3560 return value as malloc candidate.
3562 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3565 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
3566 param as true in call to suggest_attribute.
3568 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
3570 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
3571 -mreadonly-in-sdata.
3573 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3575 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
3577 (aarch64_crypto_aesd_fused): Likewise.
3579 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
3581 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
3582 (movsi_aarch64): Likewise.
3583 (load_pairsi): Likewise.
3584 (load_pairdi): Likewise.
3585 (store_pairsi): Likewise.
3586 (store_pairdi): Likewise.
3587 (load_pairsf): Likewise.
3588 (load_pairdf): Likewise.
3589 (store_pairsf): Likewise.
3590 (store_pairdf): Likewise.
3591 (zero_extend): Likewise.
3592 (trunc): Swap alternatives.
3593 (fcvt_target): Add '?' to prefer w over r.
3595 2018-05-14 Jakub Jelinek <jakub@redhat.com>
3598 * config/i386/i386.md: Disallow non-commutative arithmetics in
3599 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
3600 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
3601 in the peephole2 before it.
3603 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
3605 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
3606 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
3607 (ix86_handle_option): Handle -mcldemote.
3608 * config.gcc: New header.
3609 * config/i386/cldemoteintrin.h: New file.
3610 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
3611 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3613 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3614 OPTION_MASK_ISA_CLDEMOTE.
3615 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
3616 (ix86_valid_target_attribute_inner_p): Ditto.
3617 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
3618 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
3619 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
3620 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
3621 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
3623 * config/i386/i386.opt: Add -mcldemote.
3624 * config/i386/x86intrin.h: New header.
3625 * doc/invoke.texi: Add -mcldemote.
3627 2018-05-14 Richard Biener <rguenther@suse.de>
3629 * doc/match-and-simplify.texi: Adjust :s documentation.
3631 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
3633 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
3634 intended memcpy size.
3635 (REORDER_45): Likewise.
3637 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
3639 * sort.cc: New file.
3640 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
3641 * vec.c (qsort_chk): Use gcc_qsort.
3642 * Makefile.in (OBJS-libcommon): Add sort.o.
3643 (build/sort.o): New target. Use it...
3644 (BUILD_RTL): ... here, and...
3645 (build/gencfn-macros): ... here, and...
3646 (build/genmatch): ... here.
3648 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3649 Chung-Ju Wu <jasonwucj@gmail.com>
3651 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
3652 * config/nds32/nds32-graywolf.md: New file.
3653 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
3654 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
3656 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
3657 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
3658 * config/nds32/nds32.md (pipeline_model): Add graywolf.
3659 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
3660 * config/nds32/pipelines.md: Include n15 settings.
3662 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3663 Chung-Ju Wu <jasonwucj@gmail.com>
3665 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
3666 * config/nds32/nds32-n13.md: New file.
3667 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
3668 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
3670 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
3671 * config/nds32/nds32.md (pipeline_model): Add n13.
3672 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
3673 * config/nds32/pipelines.md: Include n13 settings.
3675 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3676 Chung-Ju Wu <jasonwucj@gmail.com>
3678 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
3679 * config/nds32/nds32-n10.md: New file.
3680 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
3681 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
3683 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
3684 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
3685 * config/nds32/nds32.md (pipeline_model): Add n10.
3686 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
3687 * config/nds32/pipelines.md: Include n10 settings.
3689 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
3690 Kito Cheng <kito.cheng@gmail.com>
3691 Chung-Ju Wu <jasonwucj@gmail.com>
3693 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
3694 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3695 Add enum values for DSP extension instructions.
3696 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
3698 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
3699 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
3701 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
3702 * config/nds32/nds32-dspext.md: New file for DSP implementation.
3703 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
3704 * config/nds32/nds32-intrinsic.md: Likewise.
3705 * config/nds32/nds32_intrinsic.h: Likewise.
3706 * config/nds32/nds32-md-auxiliary.c: Likewise.
3707 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
3708 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
3709 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
3710 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
3711 * config/nds32/nds32-protos.h: New declarations for DSP extension.
3712 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
3713 TYPE_DMAC in switch statement.
3714 * config/nds32/nds32.c: New checking and implementation for DSP
3715 extension instructions.
3716 * config/nds32/nds32.h: Likewise.
3717 * config/nds32/nds32.md: Likewise.
3718 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
3719 * config/nds32/predicates.md: Implement new predicates for DSP
3722 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
3724 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
3725 Reformat alternatives and attributes so it is easier to identify
3726 which constraints/attributes go with which instruction.
3727 (mov<mode>_hardfloat32, FMOVE64): Likewise.
3728 (mov<mode>_softfloat32, FMOVE64): Likewise.
3729 (mov<mode>_hardfloat64, FMOVE64): Likewise.
3730 (mov<mode>_softfloat64, FMOVE64): Likewise.
3732 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
3734 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3736 (Basic PowerPC Built-in Functions): The new name of the
3737 subsection previously known as "PowerPC Built-in Functions".
3738 (Basic PowerPC Built-in Functions Available on all Configurations):
3740 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
3741 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
3742 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
3743 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
3745 2018-05-11 Martin Jambor <mjambor@suse.cz>
3748 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
3751 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
3754 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
3756 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
3758 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
3759 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
3760 (ix86_handle_option): Handle -mwaitpkg.
3761 * config.gcc: New header.
3762 * config/i386/cpuid.h (bit_WAITPKG): New bit.
3763 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
3764 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
3766 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3767 OPTION_MASK_ISA_WAITPKG.
3768 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
3769 (ix86_option_override_internal): Add PTA_WAITPKG.
3770 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
3771 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
3772 IX86_BUILTIN_TPAUSE.
3773 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
3774 __builtin_ia32_umwait and __builtin_ia32_tpause.
3775 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
3776 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
3777 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
3778 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
3779 UNSPECV_TPAUSE): New.
3780 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
3781 * config/i386/i386.opt: Add -mwaitpkg.
3782 * config/i386/waitpkgintrin.h: New file.
3783 * config/i386/x86intrin.h: New header.
3784 * doc/invoke.texi: Add -mwaitpkg.
3786 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
3789 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
3791 (cortex-m0): Use armv6s-m isa.
3792 (cortex-m0plus): Likewise.
3793 (cortex-m1): Likewise.
3794 (cortex-m0.small-multiply): Likewise.
3795 (cortex-m0plus.small-multiply): Likewise.
3796 (cortex-m1.small-multiply): Likewise.
3798 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
3799 Jakub Jelinek <jakub@redhat.com>
3801 PR tree-optimization/85692
3802 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
3803 source permute as well.
3805 2018-05-11 Martin Liska <mliska@suse.cz>
3808 * doc/extend.texi: Document LLVM style format for no_sanitize
3811 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
3813 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
3814 mode_supports_vsx_dform_quad to mode_supports_dq_form.
3815 (mode_supports_vsx_dform_quad): Likewise.
3816 (mode_supports_vmx_dform): Move these functions to be next to the
3817 other mode_supports functions.
3818 (mode_supports_dq_form): Likewise.
3819 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
3820 mode_supports_dq_form.
3821 (reg_offset_addressing_ok_p): Likewise.
3822 (offsettable_ok_by_alignment): Likewise.
3823 (rs6000_legitimate_offset_address_p): Likewise.
3824 (legitimate_lo_sum_address_p): Likewise.
3825 (rs6000_legitimize_address): Likewise.
3826 (rs6000_legitimize_reload_address): Likewise.
3827 (rs6000_secondary_reload_inner): Likewise.
3828 (rs6000_preferred_reload_class): Likewise.
3829 (rs6000_output_move_128bit): Likewise.
3831 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
3833 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
3834 Generate SImode target register for null target.
3835 <case IX86_BUILTIN_XGETBV>: Ditto.
3836 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
3837 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
3839 2018-05-10 Carl Love <cel@us.ibm.com>
3841 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3842 dcbtt and dcbtstt if operands[2] is 0.
3844 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
3847 * config/i386/sse.md (usadv64qi): New expander.
3849 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
3851 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
3852 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
3853 -maltivec=be support.
3854 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
3855 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
3856 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
3857 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
3858 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
3859 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
3860 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
3861 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
3862 altivec_vsumsws): Adjust.
3863 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
3864 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
3865 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
3867 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
3868 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
3869 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
3870 (altivec_lve<VI_char>x): Delete expand.
3871 (*altivec_lve<VI_char>x_internal): Rename to...
3872 (altivec_lve<VI_char>x): ... this.
3873 (altivec_lvxl_<mode>): Delete expand.
3874 (*altivec_lvxl_<mode>_internal): Rename to ...
3875 (altivec_lvxl_<mode>): ... this.
3876 (altivec_stvxl_<mode>): Delete expand.
3877 (*altivec_stvxl_<mode>_internal): Rename to ...
3878 (altivec_stvxl_<mode>): ... this.
3879 (altivec_stve<VI_char>x): Delete expand.
3880 (*altivec_stve<VI_char>x_internal): Rename to ...
3881 (altivec_stve<VI_char>x): ... this.
3882 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
3883 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
3884 reduc_plus_scal_<mode>): Adjust.
3885 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
3887 (rs6000_cpu_cpp_builtins): Adjust.
3888 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
3889 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
3890 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
3891 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
3892 -maltivec=be support.
3893 (rs6000_split_vec_extract_var): Adjust.
3894 (rs6000_split_v4si_init): Adjust.
3895 (swap_selector_for_mode): Delete.
3896 (altivec_expand_lvx_be, altivec_expand_stvx_be,
3897 altivec_expand_stvex_be): Delete.
3898 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
3899 -maltivec=be support.
3900 (rs6000_gimple_fold_builtin): Ditto.
3901 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
3903 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
3904 (TARGET_DIRECT_MOVE_64BIT): Adjust.
3905 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
3906 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
3907 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
3908 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
3909 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
3910 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
3911 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
3912 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
3913 anonymous split): Adjust.
3914 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
3915 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
3917 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
3919 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
3920 when --with-gxx-include-dir is also specified.
3921 * configure: Regenerate.
3923 2018-05-09 Jim Wilson <jimw@sifive.com>
3926 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
3927 * config/riscv/t-withmultilib: New.
3928 * config/riscv/withmultilib.h: New.
3929 * doc/install.texi: Document RISC-V --with-multilib-list support.
3931 2018-05-09 Richard Biener <rguenther@suse.de>
3933 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
3935 (vect_bb_vectorization_profitable_p): Adjust. Compute
3936 actual scalar cost using the cost vector and the add_stmt_cost
3939 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3941 PR rtl-optimization/85645
3942 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
3943 in the REG_CFA_REGISTER note for LR, don't leave it empty.
3945 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3947 PR rtl-optimization/85645
3948 * shrink-wrap.c (spread_components): Return a boolean saying if
3949 anything was changed.
3950 (try_shrink_wrapping_separate): Iterate spread_components until
3951 nothing changes anymore.
3953 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3955 PR rtl-optimization/85645
3956 * regrename.c (build_def_use): Also kill the chains that include the
3957 destination of a REG_CFA_REGISTER note.
3959 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3961 PR rtl-optimization/85645
3962 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
3963 insn that has a REG_CFA_REGISTER note.
3965 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
3967 * cfgexpand.c (expand_clobber): New function.
3968 (expand_gimple_stmt_1): Use it.
3969 * tree-vect-stmts.c (vect_clobber_variable): New function,
3971 (vectorizable_simd_clone_call): ...here.
3972 (vectorizable_store): Emit a clobber either side of an
3973 IFN_STORE_LANES sequence.
3974 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
3976 2018-05-09 Tom de Vries <tom@codesourcery.com>
3979 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
3980 (define_insn "trap_if_false"): Add exit after trap.
3982 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
3984 PR rtl-optimization/85638
3985 * bb-reorder.c: Include common/common-target.h.
3986 (create_forwarder_block): New function extracted from...
3987 (fix_up_crossing_landing_pad): ...here. Rename into...
3988 (dw2_fix_up_crossing_landing_pad): ...this.
3989 (sjlj_fix_up_crossing_landing_pad): New function.
3990 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
3991 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
3992 from both partitions and exit the loop after one iteration.
3994 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
3997 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3999 (Basic PowerPC Built-in Functions): The new name of the
4000 subsection previously known as "PowerPC Built-in Functions".
4001 (Basic PowerPC Built-in Functions Available on all Configurations):
4003 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4005 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4007 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4009 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4012 2018-05-08 Jim Wilson <jimw@sifive.com>
4014 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
4015 (LD_EMUL_SUFFIX): New.
4016 (LINK_SPEC): Use it.
4018 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4020 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4022 (Basic PowerPC Built-in Functions): The new name of the
4023 subsection previously known as "PowerPC Built-in Functions".
4024 (Basic PowerPC Built-in Functions Available on all Configurations):
4026 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4028 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4030 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4032 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4035 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4038 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4039 after cmpelim optimization.
4041 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
4043 * config.gcc: Support "goldmont".
4044 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4045 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4047 * config/i386/i386.c (m_GOLDMONT): Define.
4048 (processor_target_table): Add "goldmont".
4049 (PTA_GOLDMONT): Define.
4050 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4051 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4052 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4053 (fold_builtin_cpu): Add "goldmont".
4054 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4055 (ix86_option_override_internal): Add "goldmont".
4056 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4057 (processor_type): Add PROCESSOR_GOLDMONT.
4058 * config/i386/i386.md: Add CPU "glm".
4059 * config/i386/glm.md: New file.
4060 * config/i386/x86-tune.def: Add m_GOLDMONT.
4061 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4063 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4066 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4068 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4069 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
4070 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
4073 * config/i386/i386.c (ix86_fold_builtin): Handle
4074 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4077 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4078 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4080 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4083 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4084 (check_arch): Likewise.
4085 (check_fpu): Return the result rather than printing it.
4086 (end arch): Fix operator precedence.
4087 (end cpu): Likewise.
4088 (END): Print the result from check_fpu.
4090 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4091 Alan Hayward <alan.hayward@arm.com>
4092 David Sherwood <david.sherwood@arm.com>
4094 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4095 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4096 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4097 (*fcmuo<mode>_and): New patterns.
4099 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4101 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4102 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4103 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4104 (cmp_op, sve_imm_con): New code attributes.
4105 (SVE_COND_INT_CMP, imm_con): Delete.
4106 (cmp_op): Remove above unspecs from int attribute.
4107 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4109 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
4110 comparison-specific unspecs.
4111 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4112 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4113 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4114 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4115 (*vec_fcm<cmp_op><mode>): Rename to...
4116 (*fcm<cmp_op><mode>): ...this and adjust likewise.
4117 (*vec_fcmuo<mode>): Rename to...
4118 (*fcmuo<mode>): ...this and adjust likewise.
4119 (*pred_fcm<cmp_op><mode>): New pattern.
4120 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4121 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4123 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4125 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4126 (aarch64_emit_sve_predicated_cond): New function.
4127 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4128 (aarch64_emit_unspec_cond_or): Replace with...
4129 (aarch64_emit_sve_or_conds): ...this new function. Use
4130 aarch64_emit_sve_ptrue_op for the individual comparisons and
4131 aarch64_emit_binop to OR them together.
4132 (aarch64_emit_inverted_unspec_cond): Replace with...
4133 (aarch64_emit_sve_inverted_cond): ...this new function. Use
4134 aarch64_emit_sve_ptrue_op for the comparison and
4135 aarch64_emit_unop to invert the result.
4136 (aarch64_expand_sve_vec_cmp_float): Update after the above
4137 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
4139 2018-05-07 Nathan Sidwell <nathan@acm.org>
4141 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4142 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4143 (Backwards Compatibility): Likewise.
4145 2018-05-07 Luis Machado <luis.machado@linaro.org>
4149 2018-05-07 Luis Machado <luis.machado@linaro.org>
4151 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4152 <prefetch_dynamic_strides>: New const bool field.
4153 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4154 prefetch_dynamic_strides.
4155 (exynosm1_prefetch_tune): Likewise.
4156 (thunderxt88_prefetch_tune): Likewise.
4157 (thunderx_prefetch_tune): Likewise.
4158 (thunderx2t99_prefetch_tune): Likewise.
4159 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4161 (aarch64_override_options_internal): Update to set
4162 PARAM_PREFETCH_DYNAMIC_STRIDES.
4163 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4164 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4165 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4166 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4167 prefetch-dynamic-strides setting.
4169 2018-05-07 Luis Machado <luis.machado@linaro.org>
4171 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4172 <minimum_stride>: New const int field.
4173 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4174 minimum_stride field.
4175 (exynosm1_prefetch_tune): Likewise.
4176 (thunderxt88_prefetch_tune): Likewise.
4177 (thunderx_prefetch_tune): Likewise.
4178 (thunderx2t99_prefetch_tune): Likewise.
4179 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4180 (aarch64_override_options_internal): Update to set
4181 PARAM_PREFETCH_MINIMUM_STRIDE.
4182 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4183 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4184 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4185 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4186 stride is constant and is below the minimum stride threshold.
4188 2018-05-07 Luis Machado <luis.machado@linaro.org>
4190 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
4193 2018-05-07 Luis Machado <luis.machado@linaro.org>
4195 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4196 <prefetch_dynamic_strides>: New const bool field.
4197 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4198 prefetch_dynamic_strides.
4199 (exynosm1_prefetch_tune): Likewise.
4200 (thunderxt88_prefetch_tune): Likewise.
4201 (thunderx_prefetch_tune): Likewise.
4202 (thunderx2t99_prefetch_tune): Likewise.
4203 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4205 (aarch64_override_options_internal): Update to set
4206 PARAM_PREFETCH_DYNAMIC_STRIDES.
4207 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4208 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4209 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4210 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4211 prefetch-dynamic-strides setting.
4213 2018-05-07 Luis Machado <luis.machado@linaro.org>
4215 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4216 <minimum_stride>: New const int field.
4217 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4218 minimum_stride field.
4219 (exynosm1_prefetch_tune): Likewise.
4220 (thunderxt88_prefetch_tune): Likewise.
4221 (thunderx_prefetch_tune): Likewise.
4222 (thunderx2t99_prefetch_tune): Likewise.
4223 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4224 (aarch64_override_options_internal): Update to set
4225 PARAM_PREFETCH_MINIMUM_STRIDE.
4226 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4227 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4228 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4229 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4230 stride is constant and is below the minimum stride threshold.
4232 2018-05-06 Jakub Jelinek <jakub@redhat.com>
4235 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4236 the type is addressable. Don't force op into register if it has
4239 2018-05-05 Roland McGrath <mcgrathr@google.com>
4242 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
4243 any section for which we don't know a specific type it should have,
4244 regardless of name. Previously this was done only for the exact
4245 names ".init_array", ".fini_array", and ".preinit_array".
4246 (default_elf_asm_named_section): Add comment about
4247 relationship with default_section_type_flags and SECTION_NOTYPE.
4248 (get_section): Don't consider it a type conflict if one side has
4249 SECTION_NOTYPE and the other doesn't, as long as neither has the
4250 SECTION_BSS et al used in the default_section_type_flags logic.
4252 2018-05-05 Tom de Vries <tom@codesourcery.com>
4255 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
4256 (workaround_barsyncs): New function.
4257 (nvptx_reorg): Use workaround_barsyncs.
4258 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4259 (define_expand "nvptx_membar_cta"): New define_expand.
4260 (define_insn "*nvptx_membar_cta"): New insn.
4262 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
4264 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
4265 To improve optimization opportunities.
4266 * builtin-types.def: The new needed builtin types for the above.
4268 2018-05-04 Richard Biener <rguenther@suse.de>
4270 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4271 * gimple-ssa-store-merging.c
4272 (imm_store_chain_info::output_merged_store): Remove redundant create,
4273 release split_store vector contents on failure.
4274 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4275 scalar stmt vector on cache hit.
4277 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
4279 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
4281 * config.gcc (powerpc-xilinx-eabi*): Remove.
4282 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
4284 (fusion_addis_mem_combo_load): Ditto.
4285 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
4287 (rs6000_cpu_cpp_builtins): Ditto.
4288 * config/rs6000/rs6000-linux.c
4289 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
4290 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
4291 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
4293 (rs6000_setup_reg_addr_masks): Ditto.
4294 (rs6000_init_hard_regno_mode_ok): Ditto.
4295 (rs6000_option_override_internal): Ditto.
4296 (legitimate_lo_sum_address_p): Ditto.
4297 (rs6000_legitimize_address): Ditto.
4298 (rs6000_legitimize_reload_address): Ditto.
4299 (rs6000_legitimate_address_p): Ditto.
4300 (abi_v4_pass_in_fpr): Ditto.
4301 (setup_incoming_varargs): Ditto.
4302 (rs6000_gimplify_va_arg): Ditto.
4303 (rs6000_split_multireg_move): Ditto.
4304 (rs6000_savres_strategy): Ditto.
4305 (rs6000_emit_prologue_components): Ditto.
4306 (rs6000_emit_epilogue_components): Ditto.
4307 (rs6000_emit_prologue): Ditto.
4308 (rs6000_emit_epilogue): Ditto.
4309 (rs6000_elf_file_end): Ditto.
4310 (rs6000_function_value): Ditto.
4311 (rs6000_libcall_value): Ditto.
4312 * config/rs6000/rs6000.h: Ditto.
4313 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
4314 (TARGET_MINMAX): ... this. New.
4315 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
4316 * config/rs6000/rs6000.md: Remove Xilinx FP support.
4317 (*movsi_internal1_single): Delete.
4318 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
4319 mfpu=, mxilinx-fpu): Delete.
4320 * config/rs6000/singlefp.h: Delete.
4321 * config/rs6000/sysv4.h: Remove Xilinx FP support.
4322 * config/rs6000/t-rs6000: Ditto.
4323 * config/rs6000/t-xilinx: Delete.
4324 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
4325 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
4326 (VStype_simple): Delete.
4327 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
4328 * config/rs6000/xfpu.h: Delete.
4329 * config/rs6000/xfpu.md: Delete.
4330 * config/rs6000/xilinx.h: Delete.
4331 * config/rs6000/xilinx.opt: Delete.
4332 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
4333 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
4335 2018-05-04 Tom de Vries <tom@codesourcery.com>
4338 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
4341 2018-05-04 Richard Biener <rguenther@suse.de>
4344 * tree-complex.c (update_complex_assignment): We are always in SSA form.
4345 (expand_complex_div_wide): Likewise.
4346 (expand_complex_operations_1): Likewise.
4347 (expand_complex_libcall): Preserve EH info of the original stmt.
4348 (tree_lower_complex): Handle removed blocks.
4349 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
4350 on complex multiplication and division libcall builtins.
4352 2018-05-04 Richard Biener <rguenther@suse.de>
4355 * fold-const.c (negate_expr_p): Restrict negation of operand
4356 zero of a division to when we know that can happen without
4358 (fold_negate_expr_1): Likewise.
4360 2018-05-04 Jakub Jelinek <jakub@redhat.com>
4363 * real.h (real_nextafter): Declare.
4364 * real.c (real_nextafter): New function.
4365 * fold-const-call.c (fold_const_nextafter): New function.
4366 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
4367 CASE_CFN_NEXTTOWARD.
4368 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
4369 even when arg1_mode is different from arg0_mode.
4371 2018-05-03 Nathan Sidwell <nathan@acm.org>
4373 * doc/extend.texi (Deprecated Features): Remove
4375 (Backwards Compatibility): Likewise.
4376 * doc/invoke.texi (C++ Language Options): Likewise.
4377 (C++ Dialect Options): Likewise.
4379 2018-05-03 Jakub Jelinek <jakub@redhat.com>
4382 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
4383 _mm512_mask_mullox_epi64): New intrinsics.
4385 2018-05-03 Tom de Vries <tom@codesourcery.com>
4388 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4389 dump files): Add offload-tree.
4391 2018-05-03 Richard Biener <rguenther@suse.de>
4393 PR tree-optimization/85615
4394 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
4395 to loops not nested in BBs loop father to avoid creating multi-entry
4398 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4400 PR tree-optimization/70291
4401 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
4402 arguments. Change return type to tree. Emit libcall as a new
4403 statement rather than replacing existing one when inplace_p is true.
4404 (expand_complex_multiplication_components): New function.
4405 (expand_complex_multiplication): Expand floating-point complex
4406 multiplication using the above.
4407 (expand_complex_division): Rename inner_type parameter to type.
4408 Update expand_complex_libcall call-site.
4409 (expand_complex_operations_1): Update expand_complex_multiplication
4410 and expand_complex_division call-sites.
4412 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4415 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4416 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4417 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
4418 the highest significant bit of the shift count mask is clear. In
4419 check whether and[sq]i3 is needed verify that all significant bits
4420 of the shift count other than the highest are set.
4422 2018-05-02 Tom de Vries <tom@codesourcery.com>
4425 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
4426 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
4427 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
4428 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
4429 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
4430 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
4431 __builtin_goacc_parlevel_size.
4433 2018-05-02 Richard Biener <rguenther@suse.de>
4435 PR tree-optimization/85597
4436 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
4437 do not use split vect_get_vec_defs call but call vect_get_slp_defs
4440 2018-05-02 Tom de Vries <tom@codesourcery.com>
4443 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4444 dump files): Add ltrans-tree.
4446 2018-05-02 Tom de Vries <tom@codesourcery.com>
4449 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4450 dump files): Add wpa-ipa.
4452 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
4454 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
4455 powerpc*-*-linux*paired* target.
4456 * config/rs6000/750cl.h: Delete.
4457 * config/rs6000/paired.h: Delete.
4458 * config/rs6000/paired.md: Delete.
4459 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
4461 * config/rs6000/rs6000-builtin.def: Remove paired float support.
4462 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
4463 comment. Remove paired float support.
4464 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
4465 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4467 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
4468 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
4470 * config/rs6000/rs6000.c: Remove paired float support.
4471 (paired_expand_vector_init, paired_expand_vector_move,
4472 paired_emit_vector_compare, paired_emit_vector_cond_expr,
4473 (paired_expand_lv_builtin, paired_expand_stv_builtin,
4474 paired_expand_builtin, paired_expand_predicate_builtin,
4475 paired_init_builtins): Delete.
4476 * config/rs6000/rs6000.h: Remove paired float support.
4477 * config/rs6000/rs6000.md: Remove paired float support.
4478 (move_from_CR_ov_bit): Delete.
4479 * config/rs6000/rs6000.opt (mpaired): Delete.
4480 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
4481 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
4483 2018-05-02 Richard Biener <rguenther@suse.de>
4486 * gimplify.c (gimplify_save_expr): When in SSA form allow
4487 SAVE_EXPRs to compute to SSA vars.
4489 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4492 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4493 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4494 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
4495 clobber operands[2], instead use a new pseudo. Formatting fixes.
4497 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
4499 PR tree-optimization/85586
4500 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
4501 exit early for statements in the same group if the accesses are
4504 2018-05-02 Tom de Vries <tom@codesourcery.com>
4507 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
4510 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
4512 PR tree-optimization/85143
4513 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
4515 2018-05-01 Tom de Vries <tom@codesourcery.com>
4518 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
4519 not found" error message.
4521 2018-05-01 Tom de Vries <tom@codesourcery.com>
4524 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
4525 * vec.c (test_ordered_remove_if): New function.
4526 (vec_c_tests): Call test_ordered_remove_if.
4527 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
4528 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
4529 * tree-vect-patterns.c (vect_pattern_recog_1): Use
4530 VEC_ORDERED_REMOVE_IF.
4532 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4534 PR tree-optimization/82665
4535 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
4536 pointer subtraction where arguments come from a memchr call.
4538 2018-05-01 Jakub Jelinek <jakub@redhat.com>
4540 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
4541 --push-state --as-needed and --pop-state instead of --as-needed and
4542 --no-as-needed if ld supports it.
4543 * configure: Regenerated.
4546 * doc/install.texi2html: Replace _002d with - and _002a with * in
4547 generated html files using sed.
4549 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4552 * gcc-rich-location.c (blank_line_before_p): New function.
4553 (use_new_line): New function.
4554 (gcc_rich_location::add_fixit_insert_formatted): New function.
4555 * gcc-rich-location.h
4556 (gcc_rich_location::add_fixit_insert_formatted): New function.
4558 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4560 * selftest.c (assert_streq): Rename "expected" and "actual" to
4561 "val1" and "val2". Extend NULL-handling to cover both inputs
4562 symmetrically, while still requiring both to be non-NULL for a pass.
4563 * selftest.h (assert_streq): Rename "expected" and "actual" to
4565 (ASSERT_EQ): Likewise.
4566 (ASSERT_EQ_AT): Likewise.
4567 (ASSERT_KNOWN_EQ): Likewise.
4568 (ASSERT_KNOWN_EQ_AT): Likewise.
4569 (ASSERT_NE): Likewise.
4570 (ASSERT_MAYBE_NE): Likewise.
4571 (ASSERT_MAYBE_NE_AT): Likewise.
4572 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
4573 the assertion to pass.
4574 (ASSERT_STREQ_AT): Likewise.
4576 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4578 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
4579 interaction with -pie.
4581 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4583 * selftest.h: Fix alphabetization of per-source-file selftest
4586 2018-04-30 Jason Merrill <jason@redhat.com>
4588 PR c++/61982 - dead stores to destroyed objects.
4589 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
4592 2018-04-30 Jason Merrill <jason@redhat.com>
4594 * tree.c (build_clobber): New.
4595 * tree.h: Declare it.
4596 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
4598 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4600 * diagnostic-show-locus.c (layout::layout): Update for
4601 location_get_source_line returning a char_span.
4602 (struct char_span): Move to input.h.
4603 (struct correction): Update for fields in char_span becoming
4605 (struct source_line): Update for location_get_source_line
4606 returning a char_span.
4607 (layout::print_line): Likewise.
4608 * edit-context.c (edited_file::print_content): Likewise.
4609 (edited_file::print_diff_hunk): Likewise.
4610 (edited_file::print_run_of_changed_lines): Likewise.
4611 (edited_file::get_num_lines): Likewise.
4612 (edited_line::edited_line): Likewise.
4613 * final.c (asm_show_source): Likewise.
4614 * input.c (location_get_source_line): Convert return type
4615 from const char * to char_span, losing the final "line_len"
4617 (dump_location_info): Update for the above.
4618 (get_substring_ranges_for_loc): Likewise. Use a char_span
4619 when handling the literal within the line.
4620 (test_reading_source_line): Update for location_get_source_line
4621 returning a char_span.
4622 * input.h (class char_span): Move here from
4623 diagnostic-show-locus.c, converting from a struct to a class.
4624 Make data members private.
4625 (char_span::operator bool): New.
4626 (char_span::length): New.
4627 (char_span::get_buffer): New.
4628 (char_span::operator[]): New.
4629 (char_span::subspan): Make const.
4630 (char_span::xstrdup): New.
4631 (location_get_source_line): Convert return type from const char *
4632 to char_span, losing the final "line_size" param.
4634 2018-04-30 Jan Hubicka <jh@suse.cz>
4636 * lto-wrapper.c (ltrans_priorities): New static var.
4637 (cmp_priority): New.
4638 (run_gcc): Read priorities and if doing parallel build order
4639 the Makefile by them.
4641 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4643 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
4645 2018-04-30 Richard Biener <rguenther@suse.de>
4647 * tree-cfg.c (verify_address): Remove base argument, add
4648 flag whether to check TREE_ADDRESSABLE and do that.
4649 (verify_expr): Remove.
4650 (verify_types_in_gimple_reference): Add pieces from verify_expr.
4651 (verify_gimple_assign_single): Likewise.
4652 (verify_gimple_switch): Likewise.
4653 (verify_expr_location_1): Dereference tp once. Add (disabled)
4654 piece from verify_expr.
4655 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
4657 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4659 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
4661 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4663 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
4664 (small_data_pattern): Likewise.
4665 (arc_rewrite_small_data): Likewise.
4666 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
4667 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
4668 (get_symbol_alignment): New function.
4669 (legitimate_small_data_address_p): Likewise.
4670 (legitimate_scaled_address): Update, call
4671 legitimate_small_data_address_p.
4672 (output_sdata): New static variable.
4673 (arc_print_operand): Update how we handle small data operands.
4674 (arc_print_operand_address): Likewise.
4675 (arc_legitimate_address_p): Update, use
4676 legitimate_small_data_address_p.
4677 (arc_rewrite_small_data_p): Remove.
4678 (arc_rewrite_small_data_1): Likewise.
4679 (arc_rewrite_small_data): Likewise.
4680 (small_data_pattern): Likewise.
4681 (compact_sda_memory_operand): Update to use
4682 legitimate_small_data_address_p and get_symbol_alignment.
4683 (prepare_move_operands): Don't rewite sdata pattern.
4684 (prepare_extend_operands): Remove.
4685 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
4687 (zero_extendqisi2): Likewise.
4688 (zero_extendhisi2): Likewise.
4689 (extendqihi2): Likewise.
4690 (extendqisi2): Likewise.
4691 (extendhisi2): Likewise.
4695 * config/arc/constraints.md (Usd): Change it to memory constraint.
4697 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4699 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
4700 as source of std instructions.
4701 * config/arc/arc.md (movsi_insn): Update pattern predicate to
4702 allow 6-bit constants as source for store instructions.
4703 (movdi_insn): Update instruction pattern to allow 6-bit constants
4704 as source for store instructions.
4706 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4708 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
4710 2018-04-30 Nathan Sidwell <nathan@acm.org>
4711 Sandra Loosemore <sandra@codesourcery.com>
4713 * dumpfile.c (dump_open): Allow '-' for stdout.
4714 * doc/invoke.texi (Developer Options): Document dump filename
4715 determination early. Document stdin/stdout selection.
4717 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
4719 Microblaze Target: PIC data text relative
4721 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
4722 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
4724 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
4725 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
4726 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
4727 New addressing mode for data-text relative position indepenedent code.
4728 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
4729 'ADDRESS_SYMBOLIC_TXT_REL'.
4730 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
4731 (microblaze_legitimate_pic_operand): Exclude function calls from
4732 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
4733 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
4735 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4736 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4737 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
4738 for 'address + offset'.
4739 (microblaze_expand_prologue): Add new function prologue call for
4741 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
4742 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
4743 table in case of TARGET_PIC_DATA_TEXT_REL.
4744 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
4745 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
4746 Add new macros 'UNSPEC_TEXT',
4747 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
4748 + exclude function calls from 'UNSPEC_PLT' in case of data text
4750 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4751 new target hook for generating address diff vector tables in case of
4753 * doc/tm.texi : Regenerate.
4754 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
4755 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
4756 of addr diff vector generation.
4757 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4758 target hook definition.
4759 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
4760 Add default function for generate_pic_addr_diff_vec -> flag_pic.
4761 * doc/invoke.texi (Add new pic option): Add new microblaze pic
4762 option for data text relative.
4764 2018-04-30 Richard Biener <rguenther@suse.de>
4766 * tree-chrec.h (evolution_function_is_constant_p): Remove
4768 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
4770 2018-04-30 Richard Biener <rguenther@suse.de>
4773 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
4775 2018-04-30 Richard Biener <rguenther@suse.de>
4777 PR tree-optimization/28364
4778 PR tree-optimization/85275
4779 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
4780 copying first exit test.
4782 2018-04-28 Mark Wielaard <mark@klomp.org>
4784 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
4786 (dwarf_AT): Handle DW_AT_addr_base.
4787 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
4789 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
4792 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
4793 (*ashl<dwi>3_doubleword_mask_1): Ditto.
4794 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
4795 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
4797 2018-04-28 Richard Biener <rguenther@suse.de>
4799 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
4800 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
4801 to reflect use. Only add interesting stmts.
4803 2018-04-27 Martin Jambor <mjambor@suse.cz>
4806 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
4807 the jump function allows for passing through aggregate values.
4809 2018-04-27 David Malcolm <dmalcolm@redhat.com>
4811 * input.h (in_system_header_at): Convert from macro to inline
4813 (from_macro_expansion_at): Likewise.
4814 (from_macro_definition_at): Likewise.
4816 2018-04-27 Jeff Law <law@redhat.com>
4818 * config.gcc: Mark tile* targets as deprecated/obsolete.
4820 2018-04-27 Richard Biener <rguenther@suse.de>
4822 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
4825 2018-04-27 Richard Biener <rguenther@suse.de>
4827 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
4829 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
4831 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
4832 with Yd constraint. Set "preferred_for_speed" attribute from
4833 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
4835 (*movdi_internal): Ditto.
4836 (movti_interunit splitters): Remove
4837 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
4838 (movdi_interunit splitters): Ditto.
4839 * config/i386/constraints.md (Ye): Remove.
4840 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
4842 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4845 * config/aarch64/constraints.md (Usg): Limit to 31.
4848 2018-04-27 Jakub Jelinek <jakub@redhat.com>
4850 PR tree-optimization/85529
4851 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
4852 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
4853 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
4854 zero extension or masking of the MSB bit.
4855 (optimize_range_tests): Add FIRST_BB argument, pass it through
4856 to optimize_range_tests_var_bound.
4857 (maybe_optimize_range_tests, reassociate_bb): Adjust
4858 optimize_range_tests callers.
4860 2018-04-26 Richard Biener <rguenther@suse.de>
4861 Jakub Jelinek <jakub@redhat.com>
4863 * cgraph.h (symbol_table): Just declare debug method here.
4864 * symtab.c (symbol_table::debug): Define.
4866 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
4868 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
4870 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
4872 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
4873 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
4874 (*movdi_internal): Substitute Yi and Yj constraint with x
4875 and Ym and Yn constraint with y constraint. Update "isa"
4876 attribute and set "preferred_for_speed" attribute from
4877 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
4878 (*movsi_internal): Ditto.
4879 (*movdf_internal): Ditto.
4880 (*movsf_internal): Ditto.
4881 (*zero_extendsidi2): Ditto.
4882 * config/i386/sse.md (vec_set<mode>_0): Ditto.
4883 (sse2_loadld): Ditto.
4884 (*vec_extract<ssevecmodelower>_0): Ditto.
4885 (*vec_extractv4si_0_zext_sse4): Ditto.
4886 (vec_concatv2di): Ditto.
4887 (*vec_dup<mode>): Ditto.
4888 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
4889 * config/i386/constraints.md (Yi): Remove.
4894 2018-04-26 Nathan Sidwell <nathan@acm.org>
4896 * dumpfile.c (dump_open): New.
4897 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
4898 (dump_finish): Detect stdio/stderr by value not name.
4900 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
4902 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
4904 2018-04-26 Tom de Vries <tom@codesourcery.com>
4907 * config/nvptx/nvptx.c (verify_neutering_jumps)
4908 (verify_neutering_labels): New function
4909 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
4911 2018-04-26 Tom de Vries <tom@codesourcery.com>
4914 * config/nvptx/nvptx.c (needs_neutering_p): New function.
4915 (nvptx_single): Use needs_neutering_p to skip over insns that do not
4918 2018-04-26 Richard Biener <rguenther@suse.de>
4919 Tom de Vries <tom@codesourcery.com>
4922 * lto-streamer-out.c (output_function): Fixup loops if required to match
4923 discovery done in the reader.
4925 2018-04-26 Richard Biener <rguenther@suse.de>
4927 PR tree-optimization/85116
4928 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
4929 have a loop exit from the single latch predecessor. Remove
4930 case of header with just condition.
4931 (ch_base::copy_headers): Exclude infinite loops from any
4933 (pass_ch::execute): Record exits.
4935 2018-04-26 Richard Biener <rguenther@suse.de>
4937 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
4938 prologue cost vector and pass it to vect_get_load_cost.
4939 (vect_get_peeling_costs_all_drs): Likewise.
4940 (vect_peeling_hash_get_lowest_cost): Likewise.
4941 (vect_enhance_data_refs_alignment): Likewise.
4943 2018-04-26 Richard Biener <rguenther@suse.de>
4946 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
4947 checking of integer<->pointer conversions.
4948 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
4949 sign-/zero-extending pointer types.
4950 (expand_omp_for_static_chunk): Likewise.
4952 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
4953 Jean Lee <xiaoyur347@gmail.com>
4955 * config/mips/mips.c (mips_asan_shadow_offset): New function.
4956 (TARGET_ASAN_SHADOW_OFFSET): Define.
4957 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
4958 true for -fsanitize=address.
4960 2018-04-25 Mark Wielaard <mark@klomp.org>
4962 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
4965 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4967 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
4968 than "alu", remove explicit "memory" and "imm_disp" attributes.
4969 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
4972 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
4973 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
4974 gen_lowpart_no_emit.
4976 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
4979 * config/i386/i386.c (ix86_expand_builtin): Change memory
4980 operand to XI, extend p0 to Pmode.
4981 * config/i386/i386.md: Change unspec volatile and operand
4982 1 mode to XI, change operand 0 mode to P.
4984 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4986 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
4987 GET_MODE_MASK before any checking.
4988 (nds32_can_use_bset_p): Likewise.
4989 (nds32_can_use_btgl_p): Likewise.
4991 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4993 * config/nds32/nds32-doubleword.md: New define_split pattern for
4994 illegal register number.
4996 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4998 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
5000 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5002 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
5004 2018-04-25 Richard Biener <rguenther@suse.de>
5006 * lto-streamer.h (LTO_major_version): Bump to 8.
5008 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5010 * BASE-VER: Set to 9.0.0.
5012 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
5014 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
5015 in __abskf2 and __powikf2.
5017 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5020 * config/aarch64/constraints.md (Usg, Usj): New constraints.
5021 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5022 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5023 Use the above on operand 2. Reindent.
5024 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5026 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5029 * common/config/i386/i386-common.c (ix86_handle_option): Don't
5031 * config/i386/i386.opt (mcet): Removed.
5032 * doc/install.texi: Remove -mcet documentation.
5033 * doc/invoke.texi: Likewise.
5035 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5038 * doc/install.texi: Remove -mcet from bootstrap-cet.
5040 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5043 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5044 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5048 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5049 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5050 containing a CONST_VECTOR.
5052 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
5054 * doc/install.texi: Update newlib dependency for nvptx.
5056 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5059 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5060 instead of INTVAL when shifting x left.
5062 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
5064 PR tree-optimization/85478
5065 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5066 vect_grouped_store_supported for single element vectors.
5068 2018-04-24 Richard Biener <rguenther@suse.de>
5071 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5072 load cost increase to the case of non-constant step.
5074 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5077 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5078 destination if any_malformed_asm.
5080 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
5083 * expr.c (store_field): In the bitfield case, if the value comes from
5084 a function call and is returned in registers by means of a PARALLEL,
5085 do not change the mode of the temporary unless BLKmode and VOIDmode.
5087 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
5089 PR rtl-optimization/85423
5090 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5091 dependencies to debug insns when the previous insn is non-debug.
5093 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5095 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5096 enums into a single definition.
5097 (fls): Fix predicates and printing.
5100 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5102 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5103 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5104 and short u6 immediate.
5105 (check_if_valid_sleep_operand): Remove.
5106 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5108 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5110 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5111 flag_always_save_lp condition.
5112 * config/nds32/nds32.opt (malways-save-lp): New option.
5114 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5116 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5117 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5118 * config/nds32/nds32.h
5119 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5120 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5122 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5124 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5125 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5127 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5128 Chung-Ju Wu <jasonwucj@gmail.com>
5130 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5132 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5133 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5135 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5137 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5139 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5141 * config/nds32/nds32-protos.h (nds32_data_alignment,
5142 nds32_local_alignment): Declare.
5143 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5144 nds32_local_alignment): New functions.
5145 (TARGET_CONSTANT_ALIGNMENT): Define.
5146 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5148 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5150 * config/nds32/nds32.c
5151 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5152 (TARGET_MODES_TIEABLE_P): Likewise.
5154 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5156 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5159 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
5160 Chung-Ju Wu <jasonwucj@gmail.com>
5162 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5163 for unaligned access.
5164 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5166 * config/nds32/nds32-intrinsic.md: Likewise.
5167 * config/nds32/nds32_intrinsic.h: Likewise.
5168 * config/nds32/nds32.h (nds32_builtins): Likewise.
5169 * config/nds32/nds32.opt (munaligned-access): New option.
5170 * config/nds32/nds32.c (nds32_asm_file_start): Display
5171 flag_unaligned_access status.
5173 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5175 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5176 -mno-relax is present.
5177 * config/riscv/linux.h (LINK_SPEC): Ditto.
5179 2018-04-20 Martin Sebor <msebor@redhat.com>
5182 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
5184 (gimple_fold_builtin_stxcpy_chk): Same.
5185 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
5187 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
5190 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
5191 __powikf2 when long double is IEEE 128-bit.
5193 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5195 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
5196 step to make sure stack always aligned.
5198 2018-04-20 Carl Love <cel@us.ibm.com>
5201 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
5202 size check for arg0.
5204 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
5205 Tom de Vries <tom@codesourcery.com>
5208 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
5209 Emit insns for calls too.
5210 (nvptx_find_par): Always look for worker-level predecessor insn.
5211 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
5213 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
5214 (nvptx_process_pars): Propagate frames for calls.
5216 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
5219 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
5221 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
5222 (ix86_handle_option): Don't handle OPT_mibt.
5223 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
5225 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
5227 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
5228 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
5229 (ix86_target_macros): Define __CET__ with flag_cf_protection
5230 for -fcf-protection.
5231 * config/i386/i386.c (isa2_opts): Remove -mibt.
5232 * config/i386/i386.h (TARGET_IBT): Removed.
5233 (TARGET_IBT_P): Likewise.
5234 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
5235 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
5236 * config/i386/i386.opt (mcet): Update help message.
5239 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
5240 -mcet as an alias for -mshstk.
5242 2018-04-20 Richard Biener <rguenther@suse.de>
5245 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
5246 complexity by forcing a single use of the multiply operand.
5248 2018-04-20 Martin Jambor <mjambor@suse.cz>
5251 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
5252 recursion dependency to only apply to non-clones.
5254 2018-04-20 Martin Jambor <mjambor@suse.cz>
5257 * ipa-cp.c (create_specialized_node): Check that clones of
5258 self-recursive edges exist during IPA-CP.
5260 2018-04-19 Toon Moene <toon@moene.org>
5262 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
5265 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5267 PR tree-optimization/85467
5268 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
5269 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
5270 VECTOR_CST element to type.
5272 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5275 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
5276 * config/i386/i386.md (builtin_setjmp_setup): Removed.
5277 (builtin_longjmp): Likewise.
5278 (save_stack_nonlocal): New pattern.
5279 (restore_stack_nonlocal): Likewise.
5281 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5284 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5285 Replace ASM_OUTPUT_LABEL with fprintf.
5287 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5290 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5291 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
5292 * config/i386/i386-c.c (ix86_target_macros_internal): Also
5293 define __IBT__ and __SHSTK__ for -fcf-protection.
5294 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
5296 (ix86_trampoline_init): Likewise.
5297 (x86_output_mi_thunk): Likewise.
5298 (ix86_notrack_prefixed_insn_p): Likewise.
5299 (ix86_option_override_internal): Don't disallow -fcf-protection.
5300 * config/i386/i386.md (rdssp<mode>): Also enable for
5302 (incssp<mode>): Likewise.
5303 (nop_endbr): Likewise.
5304 * config/i386/i386.opt (mcet): Change help message to built-in
5308 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
5309 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
5310 enable CET built-in functions.
5312 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
5314 * common/config/i386/i386-common.c
5315 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
5316 OPTION_MASK_ISA_MOVDIRI_UNSET,
5317 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
5318 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
5319 * config.gcc (movdirintrin.h): New header.
5320 * config/i386/cpuid.h (bit_MOVDIRI,
5321 bit_MOVDIR64B): New bits.
5322 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
5324 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
5325 (VOID, PVOID, PCVOID)): New function types.
5326 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
5327 __builtin_ia32_directstoreu_u64,
5328 __builtin_ia32_movdir64b): New builtins.
5329 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
5330 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
5332 (ix86_valid_target_attribute_inner_p): Ditto.
5333 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
5334 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
5335 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
5336 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
5337 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
5338 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
5339 (movdiri<mode>, movdir64b_<mode>): New.
5340 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
5341 * config/i386/immintrin.h: Include movdirintrin.h.
5342 * config/i386/movdirintrin.h: New file.
5343 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
5345 2018-04-19 Richard Biener <rguenther@suse.de>
5348 * cfg.c (clear_bb_flags): When loop state says we have
5349 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
5351 2018-04-19 Richard Biener <rguenther@suse.de>
5353 PR tree-optimization/84737
5354 * tree-vect-data-refs.c (vect_copy_ref_info): New function
5355 copying restrict info.
5356 (vect_setup_realignment): Use it.
5357 * tree-vectorizer.h (vect_copy_ref_info): Declare.
5358 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
5359 the first DR to all generated stores.
5360 (vectorizable_load): Likewise for loads.
5362 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5364 PR tree-optimization/85446
5365 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
5366 the integral and pointer types to have the same precision.
5368 * doc/install.texi: Document --disable-cet being the default and
5371 2018-04-18 Martin Liska <mliska@suse.cz>
5373 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
5376 2018-04-18 Martin Liska <mliska@suse.cz>
5379 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5382 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5383 arguments if they are comparable.
5385 2018-04-18 Martin Liska <mliska@suse.cz>
5388 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5391 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5394 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
5397 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
5398 ENDBR after calling __morestack.
5400 2018-04-18 David Malcolm <dmalcolm@redhat.com>
5403 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
5404 by using gcc_base_ver to generate a gcc_driver_version, and use
5405 it when generating GCC_DRIVER_NAME.
5406 * configure: Regenerate.
5408 2018-04-18 Jakub Jelinek <jakub@redhat.com>
5411 * config.gcc: Obsolete powerpc*-*-*spe*.
5413 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5416 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
5417 (stabstr_D): Change type of unum from unsigned int to
5418 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
5421 2018-04-17 Jim Wilson <jimw@sifive.com>
5424 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
5425 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
5426 Set arg_pointer_offset after using pretend_args_size.
5428 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5430 PR rtl-optimization/85431
5431 * dse.c (record_store): Ignore zero width stores.
5434 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
5435 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
5436 __builtin_stack_restore rather than after it.
5437 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
5438 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
5439 argument instead of virtual_dynamic_stack_rtx.
5441 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
5443 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
5445 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5446 Add note to error message to explain internal mapping of overloaded
5447 built-in function name to non-overloaded built-in function name.
5448 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
5451 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5454 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
5455 where the inputs overlap with the output.
5457 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5459 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
5460 (=v, v) alternative and explicit "memory" attribute.
5461 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
5462 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5464 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
5465 "sselog1" type instead of "sselog".
5466 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
5467 "sselog". Remove explicit "memory" attribute.
5468 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
5469 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5471 (vec_extract_hi_v32hi): Merge all alternatives into one, use
5472 "sselog1" type instead of "sselog". Remove explicit "memory"
5474 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
5475 use "sselog1" type instead of "sselog". Remove explicit "memory"
5477 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
5478 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5480 (vec_extract_hi_v64qi): Merge all alternatives into one, use
5481 "sselog1" type instead of "sselog". Remove explicit "memory"
5483 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
5484 use "sselog1" type instead of "sselog". Remove explicit "memory"
5488 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
5491 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
5494 2018-04-17 Martin Jambor <mjambor@suse.cz>
5497 * ipa-cp.c (create_specialized_node): Call
5498 expand_all_artificial_thunks if necessary.
5500 2018-04-17 Martin Liska <mliska@suse.cz>
5503 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
5504 in message, remote space in between '_G' and '('.
5506 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5509 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
5510 avx512f_vmcmp<mode>3<round_saeonly_name>,
5511 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
5512 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
5513 avx512f_rndscale<mode><round_saeonly_name>,
5514 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
5515 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
5516 Use %<iptr>2 instead of %2 for -masm=intel.
5517 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
5518 avx512f_vcvttss2usi<round_saeonly_name>,
5519 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
5521 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
5522 avx512f_vcvttsd2usi<round_saeonly_name>,
5523 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
5524 Use %q1 instead of %1 for -masm=intel.
5525 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
5526 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
5527 of %3 for -masm=intel.
5528 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
5530 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
5532 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
5534 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
5535 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
5537 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
5539 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
5540 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
5541 %g1 and one with %0 and %1.
5542 (avx512er_vmrcp28<mode><round_saeonly_name>,
5543 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
5545 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
5546 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
5547 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
5548 of %0 and %{%4%} for -masm=intel.
5549 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
5550 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
5551 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
5552 order of %0 and %{%5%}%{z%} for -masm=intel.
5554 2018-04-17 Jan Hubicka <jh@suse.cz>
5557 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
5559 2018-04-17 Martin Liska <mliska@suse.cz>
5562 * multiple_target.c (create_dispatcher_calls): Set apostrophes
5563 for target_clone error message. Make default implementation
5564 clone to be a local declaration.
5565 (separate_attrs): Add new argument and check for an empty
5567 (expand_target_clones): Handle it.
5568 (ipa_target_clone): Make redirection just for target_clones
5571 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
5572 Tom de Vries <tom@codesourcery.com>
5575 * omp-expand.c (expand_oacc_for): Add dummy false branch for
5576 tiled basic blocks without omp continue statements.
5578 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
5581 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
5582 vec_extract expression as having side effects to make sure it gets
5585 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
5588 * config/i386/i386.c (get_builtin_code_for_version): Check
5591 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
5594 * gcc/config.gcc: Support "skylake".
5595 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
5597 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
5598 (processor_target_table): Add "skylake".
5599 (ix86_option_override_internal): Add "skylake".
5600 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
5601 PROCESSOR_CANNONLAKE.
5602 (get_builtin_code_for_version): Fix priority for
5603 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
5604 PROCESSOR_SKYLAKE-AVX512.
5605 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
5606 (processor_type): Add PROCESSOR_SKYLAKE.
5608 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
5609 Jason Merrill <jason@redhat.com>
5612 * convert.c (convert_to_integer_1): Use direct recursion for
5613 enumeral types and types with a precision less than the number
5614 of bits in their mode.
5616 2018-04-16 Julia Koval <julia.koval@intel.com>
5619 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
5620 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
5622 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
5625 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
5626 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
5627 and -mno-direct-move.
5629 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
5632 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
5633 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
5634 Ensure negative shifts result in {0}.
5636 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
5638 PR rtl-optimization/79916
5639 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
5640 regs (if any) to define how to gnerate SD moves when LRA is in
5643 2018-04-13 Jakub Jelinek <jakub@redhat.com>
5645 PR rtl-optimization/85393
5646 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
5647 * except.c (expand_dw2_landing_pad_for_region): Make static.
5648 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
5649 a label and unconditional jump to old_bb, rather than
5650 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
5653 PR rtl-optimization/85376
5654 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
5655 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
5656 instead of a specific value.
5658 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5659 Bin Cheng <bin.cheng@arm.com>
5661 PR tree-optimization/82965
5662 PR tree-optimization/83991
5663 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
5664 by_profile_only parameter.
5665 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
5666 information if the loop was predicted to iterate too many times.
5667 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
5669 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5672 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
5675 2018-04-13 Martin Liska <mliska@suse.cz>
5676 Jakub Jelinek <jakub@redhat.com>
5679 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
5680 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
5681 * builtins.c (expand_builtin_memory_copy_args): Use
5682 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
5683 handle dest_addr == pc_rtx.
5685 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
5688 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
5689 asked to not generate direct moves.
5690 (fix_trunc<mode>si2_stfiwx): Similar.
5691 (fix_trunc<mode>si2_internal): Similar.
5693 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5696 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
5697 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
5698 lookup if dest in some wider mode is known to be const0_rtx and
5699 if so, record permanent equivalence for it to be ZERO_EXTEND of
5700 the narrower mode destination.
5702 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5704 * lto-streamer-out.c (output_function): Revert 259346.
5705 * omp-expand.c (expand_oacc_for): Likewise.
5707 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
5709 PR rtl-optimization/85354
5710 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
5711 * sel-sched.c (sel_global_init): ... here.
5713 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
5716 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
5717 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
5718 mode for PE-COFF targets.
5719 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
5720 (i386_pe_asm_lto_end): Likewise.
5721 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
5722 (TARGET_ASM_LTO_END): Likewise.
5723 * config/i386/winnt.c (saved_debug_info_level): New static variable.
5724 (i386_pe_asm_lto_start): New function.
5725 (i386_pe_asm_lto_end): Likewise.
5727 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5728 Richard Biener <rguenther@suse.de>
5731 * lto-streamer-out.c (output_function): Fix CFG loop state before
5733 * omp-expand.c (expand_oacc_for): Handle calls to internal
5734 functions like regular functions.
5736 2018-04-12 Richard Biener <rguenther@suse.de>
5739 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
5740 for the early LTO debug to properly generate references to it
5741 during DIE emission. Do not re-use that for the skeleton for
5743 (dwarf2out_early_finish): Likewise.
5745 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5748 * config/i386/sse.md
5749 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
5750 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
5751 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
5752 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
5753 and output is a reg, avoid creating invalid lowpart subreg, but
5754 instead split into a 512-bit move. Don't split if not AVX512VL,
5755 input is xmm16+ reg and output is a mem.
5756 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
5757 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
5758 xmm16+ reg and output is a mem.
5760 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5762 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
5763 also for flag_dwarf2_cfi_asm.
5765 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5767 PR rtl-optimization/85342
5768 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
5769 a bool scalar var inside of the loop instead. Don't try to update
5770 recog_data.operand after failed apply_change_group.
5772 2018-04-12 Tom de Vries <tom@codesourcery.com>
5775 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
5776 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
5777 array with flexible array member as array without given dimension.
5778 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
5779 argument for undefined param to true.
5781 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
5784 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5785 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
5786 from PowerPC section.
5787 * config/rs6000/sysv4.opt (mcall-): Improve help text.
5788 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
5789 help text that is too long.
5790 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
5791 help text that is too long.
5792 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
5793 help text that is too long.
5795 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
5797 * config/alpha/alpha.md (stack_probe_internal): Rename
5798 from "probe_stack". Update all callers.
5800 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5802 PR rtl-optimization/84566
5803 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
5804 sched_macro_fuse_insns.
5806 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5809 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
5810 (compute_block_dependences): ... from here.
5812 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5814 PR tree-optimization/85331
5815 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
5816 from int to HOST_WIDE_INT.
5818 2018-04-11 Martin Jambor <mjambor@suse.cz>
5821 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
5822 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
5823 not the same as the source val.
5824 (cgraph_edge_brings_value_p): New parameter.
5825 (gather_edges_for_value): Pass destination value to
5826 cgraph_edge_brings_value_p.
5827 (perhaps_add_new_callers): Likewise.
5828 (get_info_about_necessary_edges): Likewise and exclude values brought
5829 only by self-recursive edges.
5830 (create_specialized_node): Redirect only clones of self-calling edges.
5831 (+self_recursive_pass_through_p): New function.
5832 (find_more_scalar_values_for_callers_subset): Use it.
5833 (find_aggregate_values_for_callers_subset): Likewise.
5834 (known_aggs_to_agg_replacement_list): Removed.
5835 (decide_whether_version_node): Re-calculate known constants for all
5836 remaining context clones.
5838 2018-04-11 Richard Biener <rguenther@suse.de>
5841 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
5842 from early DWARF output.
5843 (dwarf2out_early_finish): Output line info unconditionally into
5844 early DWARF and add reference to it.
5846 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5849 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
5850 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
5851 other than V2DFmode using iptr mode attribute.
5852 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
5854 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5856 PR rtl-optimization/84659
5857 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
5859 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5862 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
5864 (output_loc_list): Pass address of a dummy size variable even in the
5865 locview handling loop.
5866 (index_location_lists): Add comment on why skip_loc_list_entry can't
5869 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
5872 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
5875 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
5878 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5879 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
5880 and -mstring-compare-inline-limit.
5882 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5885 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
5886 for stack clash protection in a register whenever we need it to be in
5889 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5891 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
5892 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
5894 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5897 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
5899 (mlong-double-): Ditto.
5900 * config/rs6000/sysv4.opt (msdata=): Ditto.
5901 (mtls-size=): Ditto.
5903 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5905 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5906 erroneous entries for
5907 "vector int vec_ldl (int, long int *)", and
5908 "vector unsigned int vec_ldl (int, unsigned long int *)".
5909 Add comments and entries for
5910 "vector bool char vec_ldl (int, bool char *)",
5911 "vector bool short vec_ldl (int, bool short *)",
5912 "vector bool int vec_ldl (int, bool int *)",
5913 "vector bool long long vec_ldl (int, bool long long *)",
5914 "vector pixel vec_ldl (int, pixel *)",
5915 "vector long long vec_ldl (int, long long *)",
5916 "vector unsigned long long vec_ldl (int, unsigned long long *)".
5917 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
5918 type tree bool_long_long_type_node and correct definition of
5919 bool_V2DI_type_node to make reference to this new type tree.
5920 (rs6000_mangle_type): Replace erroneous reference to
5921 bool_long_type_node with bool_long_long_type_node.
5922 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
5923 comments to emphasize sign distinctions for char and int types and
5924 replace RS6000_BTI_bool_long constant with
5925 RS6000_BTI_bool_long_long constant. Also add comment to restrict
5926 use of RS6000_BTI_pixel.
5927 (bool_long_type_node): Remove this macro definition.
5928 (bool_long_long_type_node): New macro definition
5930 2018-04-10 Jakub Jelinek <jakub@redhat.com>
5932 PR rtl-optimization/85300
5933 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
5934 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
5935 simplify_unary_operation fails.
5937 2018-04-10 Martin Liska <mliska@suse.cz>
5939 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
5940 cgraph_edge and ipa_ref.
5942 2018-04-10 Jakub Jelinek <jakub@redhat.com>
5946 * config/i386/sse.md
5947 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
5948 computation of the VEC_MERGE selector from mask.
5949 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
5950 Fix decoding of the VEC_MERGE selector into mask.
5952 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5954 PR tree-optimization/85286
5955 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
5957 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5959 * final.c (final_1): Set insn_last_address as well as
5960 insn_current_address.
5962 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5965 * explow.c (emit_stack_probe): Call validize_mem on memory location
5966 before passing it to gen_probe_stack. Create address operand and
5967 legitimize it for the probe_stack_address case.
5969 2018-04-09 Jan Hubicka <jh@suse.cz>
5972 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
5973 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
5974 * tree.c (free_lang_data_in_type): Fix handling of binfos;
5976 (free_lang_data): Rebuild type inheritance graph.
5978 2018-04-09 Martin Sebor <msebor@redhat.com>
5980 * invoke.texi (-finline-small-functions): Mention other optimization
5982 (-findirect-inlining, -fpartial-inlining): Same.
5983 (-finline-functions-called-once): Same.
5984 (-freorder-blocks-and-partition): Same.
5986 2018-04-09 Jan Hubicka <jh@suse.cz>
5989 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
5990 jumps; choose last target that matches the criteria (i.e.
5991 no partition changes for non-crossing jumps).
5992 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
5993 support for redirecting crossing jumps to non-crossing.
5995 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
5997 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
5998 also for naked functions.
6000 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
6002 * config/arc/arc.md (add_shift): New pattern.
6003 (add_shift2): Likewise.
6004 (sub_shift): Likewise.
6005 (sub_shift_cmp0_noout): Likewise.
6006 (compare_si_ashiftsi): Likewise.
6007 (xbfu_cmp0_noout): New combine pattern.
6008 (xbfu_cmp0"): Likewise.
6009 (movsi_set_cc_insn): Place the predicable variant first.
6010 (commutative_binary_cmp0_noout): Remove clobber.
6011 (commutative_binary_cmp0): New pattern.
6012 (noncommutative_binary_cmp0): Likewise.
6013 (noncommutative_binary_cmp0_noout): Likewise.
6014 (noncommutative_binary_comparison_result_used): Removed.
6015 (rsub_cmp0): New pattern.
6016 (rsub_cmp0_noout): Likewise.
6017 (extzvsi): Changed, keep only meaningful variants.
6018 (SQH, SEZ): New iterators.
6019 (SQH_postfix): New mode attribute.
6020 (SEZ_prefix): New code attribute.
6021 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6022 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6023 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6025 (noncommutative_operator): Check the availability of barrel
6028 2018-04-09 Richard Biener <rguenther@suse.de>
6030 PR tree-optimization/85284
6031 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6032 Only use the niter constraining form of simple_iv when the exit
6035 2018-04-09 Tom de Vries <tom@codesourcery.com>
6038 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6039 (define_expand "*memory_barrier"): New define_expand.
6040 (define_insn "memory_barrier"): New insn.
6042 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6044 PR rtl-optimization/80463
6045 PR rtl-optimization/83972
6046 PR rtl-optimization/83480
6048 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6049 correct producer for the insn.
6050 (tidy_control_flow): Fixup seqnos in case of debug insns.
6052 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6054 PR rtl-optimization/83913
6056 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6057 different sched-times when merging exprs.
6059 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6061 PR rtl-optimization/83962
6063 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6064 tidy_fallthru_edge and tidy_control_flow.
6066 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6068 PR rtl-optimization/83530
6070 * sel-sched.c (force_next_insn): New global variable.
6071 (remove_insn_for_debug): When force_next_insn is true, also leave only
6072 next insn in the ready list.
6073 (sel_sched_region): When the region wasn't scheduled, make another pass
6074 over it with force_next_insn set to 1.
6076 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6078 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6080 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6081 for interrupt control.
6082 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6083 functions for interrupt control.
6084 * config/nds32/nds32-intrinsic.md: Likewise.
6085 * config/nds32/nds32_intrinsic.h: Likewise.
6086 * config/nds32/nds32.h (nds32_builtins): Likewise.
6088 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6090 * config/nds32/nds32.c (nds32_init_machine_status,
6091 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6092 strict_aligned_p field.
6093 (nds32_expand_to_rtl_hook): New function.
6094 (TARGET_EXPAND_TO_RTL_HOOK): Define.
6095 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6097 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6098 Chung-Ju Wu <jasonwucj@gmail.com>
6100 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6101 * config/nds32/nds32-n7.md: New file.
6102 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6103 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6105 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6106 * config/nds32/nds32.md (pipeline_model): Add n7.
6107 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6108 * config/nds32/pipelines.md: Include n7 settings.
6110 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6111 Chung-Ju Wu <jasonwucj@gmail.com>
6113 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6114 * config/nds32/nds32-e8.md: New file.
6115 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6116 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6118 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6119 * config/nds32/nds32.md (pipeline_model): Add e8.
6120 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6121 * config/nds32/pipelines.md: Include e8 settings.
6123 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6124 Chung-Ju Wu <jasonwucj@gmail.com>
6126 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6127 * config/nds32/nds32-n8.md: New file.
6128 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6129 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6131 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6132 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6133 * config/nds32/nds32.md (pipeline_model): Add n8.
6134 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6135 * config/nds32/pipelines.md: Include n8 settings.
6137 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6138 Chung-Ju Wu <jasonwucj@gmail.com>
6140 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6141 * config/nds32/nds32-n9-2r1w.md: New file.
6142 * config/nds32/nds32-n9-3r2w.md: New file.
6143 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6144 nds32_register_ports): New or modify for cpu n9.
6145 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6147 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6148 * config/nds32/nds32-utils.c: New file.
6149 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6150 TARGET_MUL_SLOW): Define.
6151 * config/nds32/nds32.md (pipeline_model): New attribute.
6152 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6153 New options that support cpu n9.
6154 * config/nds32/pipelines.md: Include n9 settings.
6155 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6157 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6159 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6160 information if necessary.
6161 (output_cond_branch_compare_zero): Likewise.
6162 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6163 (nds32_target_alignment): Refine for alignment.
6164 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6165 (FUNCTION_BOUNDARY): Modify.
6166 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6168 * config/nds32/nds32.opt (malways-align, malign-functions): New.
6170 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6172 * config/nds32/constants.md (unspec_volatile_element): Add values for
6173 TLB operation and data prefetch.
6174 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6175 functions for TLB operation and data prefetch.
6176 * config/nds32/nds32-intrinsic.md: Likewise.
6177 * config/nds32/nds32_intrinsic.h: Likewise.
6178 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
6179 (nds32_print_operand): Likewise.
6180 * config/nds32/nds32.h (nds32_builtins): Likewise.
6182 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
6183 Andrew Pinski <pinsika@gcc.gnu.org>
6186 * match.pd: Use constant_boolean_node of correct type instead of
6187 boolean_true_node or boolean_false_node for simplifying
6188 pointer comparisons to zero.
6190 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6192 PR tree-optimization/80021
6193 * tree.c (verify_type_variant): Make error call in verify_variant_match
6194 translatable and remove final full stop.
6196 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6198 * config/nds32/constants.md (unspec_volatile_element): Add
6199 UNSPEC_VOLATILE_EH_RETURN.
6200 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
6201 nds32_output_stack_pop): Support dwarf exception handling process.
6202 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
6203 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
6204 exception handling process.
6205 (nds32_compute_stack_frame): Likewise.
6206 (nds32_return_addr_rtx): Likewise.
6207 (nds32_initial_elimination_offset): Likewise.
6208 (nds32_expand_prologue): Likewise.
6209 (nds32_expand_epilogue): Likewise.
6210 (nds32_dynamic_chain_address): New function.
6211 * config/nds32/nds32.h (machine_function): Add fields for dwarf
6213 (DYNAMIC_CHAIN_ADDRESS): Define.
6214 (EH_RETURN_DATA_REGNO): Define.
6215 (EH_RETURN_STACKADJ_RTX): Define.
6216 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
6217 patterns for dwarf exception handling.
6219 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6221 * config/nds32/nds32.h: Clean up obsolete macros.
6223 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6225 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6226 Add enum values for particular instructions.
6227 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6228 particular intrinsic functions.
6229 * config/nds32/nds32-intrinsic.md: Likewise.
6230 * config/nds32/nds32_intrinsic.h: Likewise.
6231 * config/nds32/nds32.h (nds32_builtins): Likewise.
6232 * config/nds32/nds32.md (type): Add pbsad and pbsada.
6233 (btst, ave): New patterns for particular instructions.
6235 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6237 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6238 Add enum values for atomic load/store and memory sync.
6239 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
6241 * config/nds32/nds32-intrinsic.md: Likewise.
6242 * config/nds32/nds32_intrinsic.h: Likewise.
6243 * config/nds32/nds32.h (nds32_builtins): Likewise.
6245 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6247 PR tree-optimization/85257
6248 * fold-const.c (native_encode_vector): If not all elts could fit
6249 and off is -1, return 0 rather than offset.
6250 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
6251 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
6252 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
6253 adjust buffer in native_interpret_expr call.
6255 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6257 * config/nds32/constants.md (unspec_volatile_element): Add cache
6258 control enum values.
6259 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
6260 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
6261 * config/nds32/nds32.c (nds32_cctl_names): New.
6262 (nds32_print_operand): Handle cache control register names.
6263 * config/nds32/nds32.h (nds32_builtins): New enum values.
6264 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
6266 * config/nds32/nds32.md (type): Add mmu.
6267 * config/nds32/pipelines.md (simple_insn): Add mmu.
6269 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6271 * config/nds32/nds32.md (type): Remove call.
6272 * config/nds32/pipelines.md (simple_insn): Likewise.
6274 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6276 * config/nds32/constants.md (unspec_volatile_element): Add
6277 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
6278 UNSPEC_VOLATILE_FMFCFG.
6279 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
6280 description for fmfcfg and fmfcsr.
6281 (bdesc_1arg): Add fmtcsr.
6282 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
6283 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
6284 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
6285 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
6286 unspec_fmfcfg): New patterns.
6287 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
6288 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
6289 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
6290 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
6291 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
6292 __nds32__fmfcfg): Define.
6294 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6296 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
6297 intrinsic register names.
6298 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
6299 intrinsic register enum values and macros.
6301 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6303 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
6304 for load/store addressing form.
6305 (nds32_print_operand_address): Likewise.
6307 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
6310 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
6311 based on LABEL_REF. Remove useless assertion.
6312 (pic_address_needs_scratch): Fix formatting.
6313 (sparc_legitimize_pic_address): Minor tweaks.
6314 (sparc_delegitimize_address): Adjust assertion accordingly.
6315 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
6316 into symbolic_operand.
6317 (movsi_high_pic_label_ref): Likewise.
6318 (movsi_lo_sum_pic_label_ref): Likewise.
6319 (movdi_pic_label_ref): Likewise.
6320 (movdi_high_pic_label_ref): Likewise.
6321 (movdi_lo_sum_pic_label_ref): Likewise.
6323 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
6325 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
6326 custom LIB_SPEC setup.
6328 2018-04-06 Ruslan Bukin <br@bsdpad.com>
6329 Kito Cheng <kito.cheng@gmail.com>
6331 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
6332 * config/riscv/freebsd.h: New.
6334 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6336 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
6337 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
6340 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6341 Kito Cheng <kito.cheng@gmail.com>
6343 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
6344 nds32_output_call, nds32_symbol_binds_local_p): New functions.
6345 * config/nds32/nds32-protos.h (nds32_output_call,
6346 nds32_output_return): Declare.
6347 * config/nds32/nds32.md: Refine all the call and return patterns.
6349 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6352 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
6353 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
6355 PR rtl-optimization/84872
6356 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
6357 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
6360 2018-04-06 Tamar Christina <tamar.christina@arm.com>
6362 * expr.c (copy_blkmode_to_reg): Revert 254862.
6363 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
6365 2018-04-06 Richard Biener <rguenther@suse.de>
6368 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
6369 after seeing a component reference with an adjacent field. Treat
6370 refs to arrays at struct end of external decls similar to
6371 refs to unconstrained commons.
6373 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6376 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
6377 look through SAVE_EXPRs with non-side-effects argument. Adjust
6379 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
6382 2018-04-06 Richard Biener <rguenther@suse.de>
6385 * alias.c (find_base_term): New wrapper around find_base_term
6386 unwinding CSELIB_VAL_PTR changes.
6387 (find_base_term): Do not restore CSELIB_VAL_PTR during the
6390 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6392 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
6394 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
6395 constant definitions.
6396 ("nop"): lr 0,0 -> nopr r0
6397 ("nop_lr0", "nop_lr1"): New insn definitions.
6399 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6401 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
6402 NDS32_V3PUSH_AVAILABLE_P macro.
6404 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
6405 Chung-Ju Wu <jasonwucj@gmail.com>
6407 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
6408 (nds32*-*-*): Add float and fpu_config into supported_defaults.
6409 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
6410 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
6411 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
6412 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
6413 * config/nds32/constraints.md: New constraints and checking for hard
6414 float configuration.
6415 * config/nds32/iterators.md: New mode iterator and attribute for hard
6416 float configuration.
6417 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
6419 * config/nds32/nds32-fpu.md: New file.
6420 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
6421 deal with hard float code generation.
6422 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
6424 (abi_type, float_reg_number): New enum type.
6425 * config/nds32/nds32-predicates.c: New predicates for hard float.
6426 * config/nds32/nds32-protos.h: Declare functions for hard float.
6427 * config/nds32/nds32.c: Implementation for hard float configuration.
6428 * config/nds32/nds32.h: Definitions for hard float configuration.
6429 * config/nds32/nds32.md: Include hard float machine description and
6430 modify patterns for hard float configuration.
6431 * config/nds32/nds32.opt: New options for hard float configuration.
6432 * config/nds32/predicates.md: New predicates for hard float
6435 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6437 * common/config/nds32/nds32-common.c
6438 (nds32_option_optimization_table): Enable -mreleax-hint by default.
6440 2018-04-05 Jakub Jelinek <jakub@redhat.com>
6443 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
6444 CONSTRUCTOR_ELT (ctor, ...)->value.
6446 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
6449 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
6451 2018-04-05 Tom de Vries <tom@codesourcery.com>
6454 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
6457 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6458 Kito Cheng <kito.cheng@gmail.com>
6460 * config/nds32/constraints.md (U33): Fine-tune checking condition.
6461 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
6462 * config/nds32/nds32.h (nds32_16bit_address_type): Add
6463 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
6465 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6466 Kito Cheng <kito.cheng@gmail.com>
6468 * config/nds32/constraints.md (Ufe): New memory constraint.
6469 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
6470 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
6471 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
6473 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
6474 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
6476 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6478 * config/nds32/nds32.md: Use optimize_size in the condition for
6479 alu-shift instructions.
6481 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6483 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
6485 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6487 * config/nds32/nds32.md (negsi2): Refine pattern.
6489 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6490 Chung-Ju Wu <jasonwucj@gmail.com>
6492 * config/nds32/iterators.md (shift_rotate): New code iterator.
6493 (shift): New code attribute.
6494 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
6495 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
6496 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
6497 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
6498 bit-wise operations.
6499 (andsi3, *andsi3): Ditto.
6500 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
6501 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
6502 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
6503 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
6504 nds32_ior_operand, nds32_xor_operand): New predicates.
6506 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6508 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
6509 (addsi3, subsi3): ... this.
6511 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6513 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
6515 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6517 * config/nds32/nds32.md: Adjust indention.
6519 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6521 * config/nds32/nds32.md (feature): New attribute.
6523 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6525 * config/nds32/nds32.md (subtype): New attribute.
6527 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
6530 * config/arm/arm-builtins.c (arm_expand_builtin): Change
6531 expansion to perform a bitwise AND of the argument followed by a
6532 boolean negation of the result.
6534 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
6536 PR rtl-optimization/84878
6537 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
6538 the basic block. Assert the use reference is not artificial and that
6539 it has an associated insn.
6541 2018-04-04 Michael Matz <matz@suse.de>
6543 * builtins.c (compute_objsize): Pass correct operand
6544 to array_at_struct_end_p.
6546 2018-04-04 Richard Biener <rguenther@suse.de>
6549 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
6550 from contexts for DINFO_LEVEL_TERSE and below.
6552 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6554 * config/nds32/nds32-doubleword.md (move_<mode>): Require
6555 resiter_operand condition.
6556 * config/nds32/nds32.md (*move<mode>): Ditto.
6558 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6559 Monk Chiang <sh.chiang04@gmail.com>
6561 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
6563 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6565 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
6567 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6568 Kito Cheng <kito.cheng@gmail.com>
6570 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
6571 nds32_cond_code_str, output_cond_branch,
6572 output_cond_branch_compare_zero, nds32_expand_cbranch,
6573 nds32_expand_cstore, nds32_expand_movcc,
6574 nds32_output_cbranchsi4_equality_zero,
6575 nds32_output_cbranchsi4_equality_reg,
6576 nds32_output_cbranchsi4_equality_reg_or_const_int,
6577 nds32_output_cbranchsi4_greater_less_zero: New functions.
6578 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
6579 nds32_expand_cstore, nds32_expand_movcc,
6580 nds32_output_cbranchsi4_equality_zero,
6581 nds32_output_cbranchsi4_equality_reg,
6582 nds32_output_cbranchsi4_equality_reg_or_const_int,
6583 nds32_output_cbranchsi4_greater_less_zero): Declare.
6584 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
6585 nds32_rimm11s_operand): New predicates.
6586 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
6587 * config/nds32/nds32.md: Rewrite all the branch and conditional move
6590 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6592 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
6593 * config/nds32/nds32.md: Ditto.
6594 * config/nds32/pipelines.md: Ditto.
6596 2018-04-04 Richard Biener <rguenther@suse.de>
6598 PR tree-optimization/85168
6599 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
6600 propagating abnormals.
6602 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6604 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
6606 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6607 Kito Cheng <kito.cheng@gmail.com>
6609 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
6610 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
6611 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
6612 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
6613 * config/nds32/nds32.md (sibcall_internal): New.
6614 (sibcall_register): Remove.
6615 (sibcall_immediate): Remove.
6616 (sibcall_value_internal): New.
6617 (sibcall_value_register): Remove.
6618 (sibcall_value_immediate): Remove.
6619 * config/nds32/predicates.md (nds32_general_register_operand): New.
6620 (nds32_call_address_operand): New.
6622 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6624 PR rtl-optimization/85167
6625 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
6626 bb_defs if *split_p, instead preinitialize it to NULL.
6628 PR tree-optimization/85156
6629 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
6630 evaluating the argument multiple times.
6632 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
6634 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
6636 (_mm_cvtpd_ps): Likewise.
6637 (_mm_cvttpd_epi32): Likewise.
6638 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
6639 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
6640 vector, pixel, and bool following altivec.h include.
6642 2018-04-03 Martin Sebor <msebor@redhat.com>
6644 * doc/extend.texi (Common Function Attributes): Clarify.
6645 (const attribute): Likewise.
6646 (pure attribute): Likewise.
6648 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6651 * config/i386/i386.c (ix86_expand_vector_set): Use
6652 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
6654 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
6656 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
6657 instructions when changing rounding bits to preserve precision bits
6658 in the x87 control word.
6660 2018-04-03 Martin Liska <mliska@suse.cz>
6662 PR tree-optimization/82491
6663 * rtl.h (strip_offset_and_add): Replace += suboffset with
6664 poly_uint64 () + suboffset.
6666 2018-03-29 Martin Liska <mliska@suse.cz>
6667 Martin Jambor <mjambor@suse.cz>
6670 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
6671 param_type is not an integral or pointer type.
6673 2018-04-03 Richard Biener <rguenther@suse.de>
6675 * sese.h (recompute_all_dominators): Remove.
6677 2018-04-02 Martin Sebor <msebor@redhat.com>
6679 * doc/invoke.texi (-Wrestrict): Fix typos.
6681 2018-04-02 Jim Wilson <jimw@sifive.com>
6683 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
6684 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
6685 (<optab>di3, <optab>si3_extend): Likewise.
6686 (<optab>si3_mask, <optab>si3_mask_1): New.
6687 (<optab>di3_mask, <optab>di3_mask_1): New.
6688 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
6689 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
6690 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
6692 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
6694 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
6697 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
6699 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
6700 (nds32_canonicalize_comparison): New function.
6702 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6703 Kito Cheng <kito.cheng@gmail.com>
6704 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6706 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
6707 * config/nds32/constants.md (unspec_volatile_element): Add
6708 UNSPEC_VOLATILE_RELAX_GROUP.
6709 * config/nds32/nds32-relax-opt.c: New file.
6710 * config/nds32/nds32-predicates.c
6711 (nds32_symbol_load_store_p): New function.
6712 * config/nds32/nds32-protos.h
6713 (nds32_symbol_load_store_p): Declare function.
6714 (make_pass_nds32_relax_opt): Declare new rtl pass function.
6715 * config/nds32/nds32.c
6716 (nds32_register_pass): New function to register pass.
6717 (nds32_register_passes): New function to register passes.
6718 * config/nds32/nds32.md (relax_group): New pattern.
6719 * config/nds32/nds32.opt (mrelax-hint): New option.
6720 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
6722 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
6724 * config/nds32/t-nds32: Modify files dependency.
6726 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6728 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
6729 (PROFILE_HOOK): Define its implementation.
6731 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6733 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
6734 type and 32-bit size.
6736 2018-04-01 Jakub Jelinek <jakub@redhat.com>
6739 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
6740 (V_128_256): New mode iterator.
6741 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
6742 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
6743 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
6745 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
6746 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
6748 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
6751 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
6752 NaN inputs correctly.
6754 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
6757 * config/rs6000/vsx.md (??r): New mode attribute.
6758 (*vsx_mov<mode>_64bit): Use it.
6759 (*vsx_mov<mode>_32bit): Likewise.
6761 2018-03-30 Martin Sebor <msebor@redhat.com>
6763 PR tree-optimization/84818
6764 * builtins.c (check_access): Use warning_n.
6766 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6769 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
6771 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
6774 2018-03-30 Julia Koval <julia.koval@intel.com>
6777 * x86-tune.def (movx, partial_reg_dependency): Enable for
6780 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
6783 * lra-constraints.c (process_alt_operands): Move setting
6784 this_alternative_matches below.
6786 2018-03-29 Martin Liska <mliska@suse.cz>
6789 * doc/invoke.texi: Document how LTO works with debug info.
6790 Describe auto-load support of binutils. Mention 'x86-64'
6791 as valid option value of -march option.
6793 2018-03-29 Jakub Jelinek <jakub@redhat.com>
6795 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
6798 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
6799 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
6800 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
6803 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
6806 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
6807 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
6808 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
6809 for RS6000_BTM_POWERPC64.
6810 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
6811 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
6812 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
6817 2018-03-28 Carl Love <cel@us.ibm.com>
6820 2017-09-27 Carl Love <cel@us.ibm.com>
6822 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
6823 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
6824 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
6827 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6829 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
6830 instead of __vector bool.
6831 (_mm_max_pu8): Likewise.
6832 (_mm_min_pi16): Likewise.
6834 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
6837 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
6838 (DIVWEUO): Likewise.
6840 (DIVDEUO): Likewise.
6841 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
6842 DIVWEUO and DIVDEUO.
6843 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
6844 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
6845 (div_extend): Likewise.
6846 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
6848 (__builtin_divweuo): Likewise.
6849 (__builtin_divdeo): Likewise.
6850 (__builtin_divdeuo): Likewise.
6852 2018-03-28 Jakub Jelinek <jakub@redhat.com>
6855 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
6856 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
6858 PR tree-optimization/82004
6859 * gimple-match-head.c (optimize_pow_to_exp): New function.
6860 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
6861 Don't fold to exp if optimize_pow_to_exp is false.
6863 2018-03-28 Martin Liska <mliska@suse.cz>
6866 * calls.c (initialize_argument_information): Fix trailing space.
6867 * common.opt: Fix typo and provide better explanation for
6868 -fsanitize-coverage option.
6869 * config/i386/i386.opt: Fix typo.
6871 2018-03-28 Jakub Jelinek <jakub@redhat.com>
6872 Martin Liska <mliska@suse.cz>
6875 * gimplify.c (asan_poison_variable): Don't do the check for
6876 gimplify_omp_ctxp here.
6877 (gimplify_decl_expr): Do it here.
6878 (gimplify_target_expr): Likewise.
6880 2018-03-28 Martin Liska <mliska@suse.cz>
6883 * config/i386/i386.c (ix86_function_arg_advance): Do not call
6884 chkp_type_bounds_count if MPX is not enabled.
6886 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
6888 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
6890 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
6893 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
6894 function to create the function decl for complex long double
6895 multiply and divide for -mabi=ieeelongdouble.
6896 (init_float128_ieee): Call it.
6898 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
6901 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
6902 -fcf-protection=branch -mibt.
6903 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
6905 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6908 * config/arm/arm.c (arm_valid_symbolic_address): Handle
6909 arm_word_relocations.
6911 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
6914 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
6915 extern array declarations.
6917 2018-03-27 Richard Biener <rguenther@suse.de>
6920 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
6921 explicit single_use checks.
6923 2018-03-27 Richard Biener <rguenther@suse.de>
6925 PR tree-optimization/85082
6926 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
6929 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6931 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
6932 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
6933 Turn on fasynchronous-unwind-tables and funwind-tables.
6935 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
6938 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
6939 (*bmi_blsr_<mode>_ccz): Ditto.
6941 2018-03-26 Tom de Vries <tom@codesourcery.com>
6943 PR tree-optimization/85063
6944 * omp-general.c (offloading_function_p): New function. Factor out
6946 * omp-offload.c (pass_omp_target_link::gate): ... here.
6947 * omp-general.h (offloading_function_p): Declare.
6948 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
6949 with attribute omp declare target for offloading functions.
6951 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
6953 PR tree-optimization/84005
6954 * tree-data-ref.h (get_base_for_alignment): Declare.
6955 * tree-data-ref.c (get_base_for_alignment_1): New function.
6956 (get_base_for_alignment): Likewise.
6957 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
6958 get_base_for_alignment to find a suitable base object, instead
6959 of always using drb->base_address.
6961 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6964 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
6965 known size by default.
6967 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
6970 * lra-constraints.c (process_alt_operands): Don't match BLKmode
6971 and non BLKmode operands.
6973 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6976 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
6977 Clean up attributes.
6979 2018-03-23 Richard Biener <rguenther@suse.de>
6982 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
6983 we are going to emit early debug for LTO.
6985 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6988 * function.c (match_asm_constraints_1): Don't optimize if input
6989 doesn't satisfy general_operand predicate for output's mode.
6992 * alias.c (write_dependence_p): Don't require for x_canonicalized
6993 non-VOIDmode if x has VOIDmode.
6996 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
6997 just don't try to optimize it rather than assert it never happens.
6999 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7001 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
7002 macro expansions for definition of ST_INTERNAL_<mode> and
7003 LD_INTERNAL_<mode> builtins.
7004 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
7006 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
7008 (altivec_expand_st_builtin): Likewise.
7009 (altivec_expand_builtin): Remove calls to deleted functions.
7010 (rs6000_address_for_altivec): Delete this function.
7011 * config/rs6000/vector.md: Remove expands for
7012 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
7014 2018-03-22 Sudakshina Das <sudi.das@arm.com>
7017 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7018 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7019 re-computing once computed.
7020 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7021 (arm_init_machine_status): Initialize
7022 machine->static_chain_stack_bytes.
7024 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7027 * doc/extend.texi: Add four new prototypes for vec_ld.
7028 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7029 definitions for more logical presentation.
7030 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7031 entries for V1TI variants of __builtin_altivec_ld builtin.
7032 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7033 handling of V1TI variant of LVX icode pattern.
7034 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7035 (rs6000_gimple_fold_builtin): Likewise.
7036 (altivec_init_builtins): Add code to define
7037 __builtin_altivec_lvx_v1ti function.
7039 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7042 * function.c (match_asm_constraints_1): Don't do the optimization
7043 if input isn't a REG, SUBREG, MEM or constant.
7045 2018-03-22 Tom de Vries <tom@codesourcery.com>
7047 PR tree-optimization/84956
7048 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7049 bb_has_abnormal_pred.
7051 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7054 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7055 DECL_INITIAL (decl) to decl at the end.
7056 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7059 2018-03-21 Joseph Myers <joseph@codesourcery.com>
7061 * doc/extend.texi (__builtin_tgmath): Document when complex
7062 integer types are treated as _Complex _Float64.
7064 2018-03-21 Tom de Vries <tom@codesourcery.com>
7066 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7068 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7070 PR tree-optimization/84960
7071 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7072 if it is ENTRY block, move them into single succ of ENTRY in that case.
7074 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
7076 PR tree-optimization/84811
7077 * poly-int.h (poly_span_traits): Remove the T3 parameter and
7078 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7079 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7080 (known_subrange_p): Update accordingly. Cast each value involved
7081 in the size comparison, rather than casting the result of the
7084 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7086 PR tree-optimization/84982
7087 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7088 by flipping the least significant bit rather than all bits from
7089 bitpos to bitpos + bitsize - 1.
7091 2018-03-21 Nathan Sidwell <nathan@acm.org>
7093 * doc/extend.texi (Deprecated Features): Remove mention of
7094 long-deleted deprecations.
7096 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7099 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7100 * configure: Regenerate.
7102 2018-03-21 Tom de Vries <tom@codesourcery.com>
7104 PR tree-optimization/83126
7105 * tree-parloops.c (num_phis): New function.
7106 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7108 2018-03-21 Nathan Sidwell <nathan@acm.org>
7110 * doc/extend.texi (Deprecated Features): Update deprecated flags,
7111 mention anon-struct/union members and trailing attributes.
7113 2018-03-21 Bin Cheng <bin.cheng@arm.com>
7115 PR tree-optimization/84969
7116 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7117 builtin memset partitions if they set different rhs values.
7119 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7121 PR rtl-optimization/84989
7122 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7123 VEC_DUPLICATE with scalar result mode.
7125 2018-03-21 Martin Liska <mliska@suse.cz>
7128 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7129 not intended return statement.
7131 2018-03-21 Martin Liska <mliska@suse.cz>
7134 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7135 (chkp_find_bound_slots_1): Limit number of iterations.
7137 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
7140 * Minor grammar fixes for x86 options.
7142 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7145 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7146 holding REG_CFA_RESTORE notes, instead turn them into a USE.
7148 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
7151 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7152 (altivec_lvx_<mode>_1op): Likewise.
7153 (altivec_stvx_<mode>_2op): Likewise.
7154 (altivec_stvx_<mode>_1op): Likewise.
7155 (altivec_lvx_<VM2:mode>): New define_expand.
7156 (altivec_stvx_<VM2:mode>): Likewise.
7157 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7158 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7159 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7160 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7161 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7162 (rs6000_gen_lvx): Likewise.
7163 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7164 (altivec_expand_stv_builtin): Likewise.
7165 (altivec_expand_builtin): Likewise.
7166 * config/rs6000/vector.md: Likewise.
7168 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7171 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7174 2018-03-20 Richard Biener <rguenther@suse.de>
7177 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
7178 sign-conversions as zero, fall back to standard scalar_stmt
7181 2018-03-20 Martin Liska <mliska@suse.cz>
7184 * predict.c (rebuild_frequencies): Handle case when we have
7185 PROFILE_ABSENT, but flag_guess_branch_prob is false.
7187 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7190 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
7191 flag_section_anchors.
7192 * varasm.c (use_blocks_for_decl_p): Remove hack for
7193 dw2_force_const_mem.
7196 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
7198 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
7199 be created, use lowpart_subreg of operands[0] rather than operands[0]
7201 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
7202 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
7203 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
7204 and n constraint instead of aarch64_shift_imm_di and Usd.
7205 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
7206 (*aarch64_<optab>_reg_minus<mode>3): ... this.
7208 2018-03-20 Sudakshina Das <sudi.das@arm.com>
7211 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
7212 to favor GPR over NEON registers.
7213 (<shift>di3_neon): Likewise.
7215 2018-03-20 Tom de Vries <tom@codesourcery.com>
7218 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
7219 (nvptx_process_pars): Emit bar.sync asap and alap.
7221 2018-03-20 Tom de Vries <tom@codesourcery.com>
7224 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
7225 seen_label if seen_label is already set.
7227 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7230 * config/i386/i386.c (fold_builtin_cpu): For features above 31
7231 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
7232 Use 1U instead of 1. Formatting fixes.
7235 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
7236 instead of TREE_TYPE (s1) for the return value.
7238 2018-03-19 Jakub Jelinek <jakub@redhat.com>
7240 PR tree-optimization/84946
7241 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
7242 bitsize + bitsize in poly_uint64 rather than poly_int64.
7245 * dwarf2asm.c: Include fold-const.c.
7246 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
7247 of decl rather than decl itself.
7249 PR rtl-optimization/84643
7250 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
7252 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
7255 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
7256 calling assemble_variable.
7258 2018-03-19 Sudakshina Das <sudi.das@arm.com>
7261 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
7262 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
7264 2018-03-19 Jim Wilson <jimw@sifive.com>
7267 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
7268 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
7269 (riscv_first_stack_step): Likewise.
7270 (riscv_option_override): Use STACK_BOUNDARY instead of
7272 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
7274 (BIGGEST_ALIGNMENT): Set to 128.
7275 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
7276 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
7279 2018-03-19 Richard Biener <rguenther@suse.de>
7281 PR tree-optimization/84933
7282 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
7283 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
7285 2018-03-19 Richard Biener <rguenther@suse.de>
7287 PR tree-optimization/84859
7288 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
7289 (cond_if_else_store_replacement): Perform sinking operation on
7290 single-store BBs regardless of MAX_STORES_TO_SINK setting.
7291 Generalize what a BB with a single eligible store is.
7293 2018-03-19 Richard Biener <rguenther@suse.de>
7295 PR tree-optimization/84929
7296 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
7297 chrec_is_positive against non-chrec arg.
7299 2018-03-19 Tamar Christina <tamar.christina@arm.com>
7302 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
7304 2018-03-18 Martin Liska <mliska@suse.cz>
7306 PR rtl-optimization/84635
7307 * regrename.c (build_def_use): Use matches_mode only when
7310 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
7312 PR tree-optimization/84913
7313 * tree-vect-loop.c (vectorizable_reduction): Don't try to
7314 vectorize chains of COND_EXPRs.
7316 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7318 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
7320 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7322 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
7324 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7326 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
7328 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7329 Kito Cheng <kito.cheng@gmail.com>
7331 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
7332 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
7333 (nds32_adjust_reg_alloc_order): New function.
7334 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
7336 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
7338 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
7339 nds32_print_operand, nds32_print_operand_address): Use
7340 HOST_WIDE_INT_PRINT_DEC instead.
7342 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7344 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
7346 2018-03-17 Jakub Jelinek <jakub@redhat.com>
7349 * config/i386/i386.c (initial_ix86_tune_features,
7350 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
7352 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
7353 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
7354 rather than 1u << ix86_tune. Formatting fix.
7355 (ix86_option_override_internal): Change ix86_arch_mask from
7356 unsigned int to unsigned HOST_WIDE_INT, initialize to
7357 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
7358 (ix86_function_specific_restore): Likewise.
7360 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7363 * postreload.c (reload_combine_recognize_pattern): Perform
7364 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
7365 truncate_int_for_mode the result for the destination's mode.
7368 * hsa-gen.c (mem_type_for_type): Fix comment typo.
7369 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
7371 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7374 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
7377 * lra-assigns.c (lra_split_hard_reg_for): Don't use
7378 regno_allocno_class_array and sorted_pseudos.
7379 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
7380 insns where regno is used.
7382 2018-03-16 Martin Liska <mliska@suse.cz>
7385 * multiple_target.c (create_dispatcher_calls): Redirect
7386 reference in the symbol table.
7388 2018-03-16 Martin Liska <mliska@suse.cz>
7391 * multiple_target.c (create_dispatcher_calls): Redirect also
7394 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7398 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
7399 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
7400 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
7402 2018-03-16 Julia Koval <julia.koval@intel.com>
7404 * doc/invoke.texi (Skylake Server): Add CLWB.
7405 Cannonlake): Remove CLWB.
7407 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7409 PR tree-optimization/84841
7410 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
7412 (FLOAT_ONE_CONST_TYPE): Define.
7413 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
7414 (sort_by_operand_rank): Put entries with higher constant_type last
7415 rather than first to match comments.
7417 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
7419 * config/nios2/nios2.md (movsi_internal): Fix thinko in
7422 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7425 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
7426 check and use address of target always.
7428 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
7431 * config/i386/i386.c (indirect_thunk_needed): Update comments.
7432 (indirect_thunk_bnd_needed): Likewise.
7433 (indirect_thunks_used): Likewise.
7434 (indirect_thunks_bnd_used): Likewise.
7435 (indirect_return_needed): New.
7436 (indirect_return_bnd_needed): Likewise.
7437 (output_indirect_thunk_function): Add a bool argument for
7439 (output_indirect_thunk_function): Don't generate alias for
7440 function return thunk.
7441 (ix86_code_end): Call output_indirect_thunk_function to generate
7442 function return thunks.
7443 (ix86_output_function_return): Set indirect_return_bnd_needed
7444 and indirect_return_needed instead of indirect_thunk_bnd_needed
7445 and indirect_thunk_needed.
7447 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
7449 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
7450 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
7451 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
7453 2018-03-15 David Malcolm <dmalcolm@redhat.com>
7454 Paul Hua <paul.hua.gm@gmail.com>
7457 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
7459 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
7461 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
7462 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
7465 2018-03-15 Tamar Christina <tamar.christina@arm.com>
7468 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7469 instead of GET_MODE_SIZE when comparing Units.
7471 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7474 * varasm.c (hash_section): Return an unchangeble hash value
7475 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
7476 Return !aarch64_can_use_per_function_literal_pools_p ().
7478 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7481 * optabs.c (emit_conditional_move): Pass address of cmode's copy
7482 rather than address of cmode as last argument to prepare_cmp_insn.
7484 2018-03-15 Julia Koval <julia.koval@intel.com>
7486 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
7487 F_AVX512VNNI, F_AVX512BITALG): New.
7489 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
7492 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
7493 insn for floating-point loads and stores.
7495 2018-03-14 Carl Love <cel@us.ibm.com>
7497 * config/rs6000/rs6000-c.c: Add macro definitions for
7498 ALTIVEC_BUILTIN_VEC_PERMXOR.
7499 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
7500 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
7501 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
7502 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
7504 * config/doc/extend.texi: Add prototypes for vec_permxor.
7506 2018-03-14 David Malcolm <dmalcolm@redhat.com>
7509 * diagnostic-show-locus.c (class layout_point): Convert m_line
7510 from int to linenum_type.
7511 (line_span::comparator): Use linenum "compare" function when
7512 comparing line numbers.
7513 (test_line_span): New function.
7514 (layout_range::contains_point): Convert param "row" from int to
7516 (layout_range::intersects_line_p): Likewise.
7517 (layout::will_show_line_p): Likewise.
7518 (layout::print_source_line): Likewise.
7519 (layout::should_print_annotation_line_p): Likewise.
7520 (layout::print_annotation_line): Likewise.
7521 (layout::print_leading_fixits): Likewise.
7522 (layout::annotation_line_showed_range_p): Likewise.
7523 (struct line_corrections): Likewise for field m_row.
7524 (line_corrections::line_corrections): Likewise for param "row".
7525 (layout::print_trailing_fixits): Likewise.
7526 (layout::get_state_at_point): Likewise.
7527 (layout::get_x_bound_for_row): Likewise.
7528 (layout::print_line): Likewise.
7529 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
7530 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
7531 * input.c (selftest::test_linenum_comparisons): New function.
7532 (selftest::input_c_tests): Call it.
7533 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
7534 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
7535 * selftest.h (ASSERT_GT): New macro.
7536 (ASSERT_GT_AT): New macro.
7537 (ASSERT_LT): New macro.
7538 (ASSERT_LT_AT): New macro.
7540 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
7542 PR rtl-optimization/84780
7543 * combine.c (distribute_links): Don't make a link based on pc_rtx.
7545 2018-03-14 Martin Liska <mliska@suse.cz>
7547 * tree.c (record_node_allocation_statistics): Use
7548 get_stats_node_kind.
7549 (get_stats_node_kind): New function extracted from
7550 record_node_allocation_statistics.
7551 (free_node): Use get_stats_node_kind.
7553 2018-03-14 Richard Biener <rguenther@suse.de>
7555 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
7556 that the value-set of ANTIC_IN doesn't grow.
7559 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7561 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7562 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7564 2018-03-14 Julia Koval <julia.koval@intel.com>
7566 * config.gcc (icelake-client, icelake-server): New.
7568 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
7569 (initial_ix86_arch_features): Ditto.
7570 (PTA_SKYLAKE): Add SGX.
7571 (PTA_ICELAKE): Remove.
7572 (PTA_ICELAKE_CLIENT): New.
7573 (PTA_ICELAKE_SERVER): New.
7574 (ix86_option_override_internal): Split up icelake on icelake client and
7576 (get_builtin_code_for_version): Ditto.
7577 (fold_builtin_cpu): Ditto.
7578 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
7579 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
7580 * config/i386/i386.h (processor_type): Ditto.
7581 * doc/invoke.texi: Ditto.
7583 2018-03-14 Jakub Jelinek <jakub@redhat.com>
7586 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
7587 INTEGER_CST offset, add it together with bitpos / 8 and
7588 sign extend based on POINTER_SIZE.
7592 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
7595 * config/i386/constraints.md (Yc): New register constraint.
7596 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
7597 Use Yc constraint for alternative 2 of operand 0. Remove
7598 preferred_for_speed attribute.
7600 2018-03-14 Richard Biener <rguenther@suse.de>
7602 PR tree-optimization/84830
7603 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
7604 with the old one to avoid oscillations.
7606 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
7609 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
7611 (assign_by_spills): Return a flag of reload assignment failure.
7612 Do not process the reload assignment failures. Do not spill other
7613 reload pseudos if they has the same reg class. Update n if
7615 (lra_assign): Add a return arg. Set up from the result of
7616 assign_by_spills call.
7617 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7618 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7619 usage_insns if it is not NULL.
7620 (spill_hard_reg_in_range): New function.
7621 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7622 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7623 function prototypes.
7624 (lra_assign): Change prototype.
7625 * lra.c (lra): Add code to deal with fails by splitting hard reg
7628 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
7630 * config/riscv/riscv.opt (mrelax): New option.
7631 * config/riscv/riscv.c (riscv_file_start): Emit ".option
7632 "norelax" when riscv_mrelax is disabled.
7633 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
7635 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7638 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
7639 reassociation for int modes.
7641 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7643 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
7644 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
7646 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
7647 * config/aarch64/aarch64-sve.md
7648 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
7649 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
7650 (*extend<mode><Vwide>2): Rename to...
7651 (aarch64_sve_extend<mode><Vwide>2): ...this.
7652 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
7653 renaming the old pattern to...
7654 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
7656 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
7657 define_expand, renaming the old pattern to...
7658 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
7659 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
7660 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
7661 account when deciding which SVE instruction the optab should use.
7662 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
7664 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7666 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
7667 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
7668 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
7669 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
7670 (tlsdesc_small_<mode>): Turn a define_expand and use
7671 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
7672 (tlsdesc_small_advsimd_<mode>): ...this.
7673 (tlsdesc_small_sve_<mode>): New pattern.
7675 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7677 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
7678 (UNSPEC_UMUL_HIGHPART): New constants.
7679 (MUL_HIGHPART): New int iteraor.
7680 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
7681 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
7683 (*<su>mul<mode>3_highpart): New define_insn.
7685 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7688 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7691 2018-03-13 Martin Liska <mliska@suse.cz>
7694 * (sem_item_optimizer::sem_item_optimizer): Initialize new
7696 (sem_item_optimizer::~sem_item_optimizer): Release it.
7697 (sem_item_optimizer::merge_classes): Register variable aliases.
7698 (sem_item_optimizer::fixup_pt_set): New function.
7699 (sem_item_optimizer::fixup_points_to_sets): Likewise.
7700 * ipa-icf.h: Declare new variables and functions.
7702 2018-03-13 Jakub Jelinek <jakub@redhat.com>
7705 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
7706 integer_pow2p@2 and test integer_pow2p in condition.
7707 (A < 0 ? C : 0): Similarly for @1.
7710 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
7711 characters starting at p contain '\0' character, don't look beyond
7715 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
7716 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
7719 * reg-stack.c (change_stack): Change update_end var from int to
7720 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
7721 also call set_block_for_insn on the newly added insns and rescan.
7724 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
7725 on the last operand.
7728 * tree.c (stabilize_reference_1): Return save_expr (e) for
7729 STATEMENT_LIST even if it doesn't have side-effects.
7731 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
7733 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
7735 2018-03-12 Renlin Li <renlin.li@arm.com>
7737 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
7738 aarch64_output_scalar_simd_mov_immediate.
7740 2018-03-12 Martin Sebor <msebor@redhat.com>
7742 PR tree-optimization/83456
7743 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
7744 for perfectly overlapping calls to memcpy.
7745 (gimple_fold_builtin_memory_chk): Same.
7746 (gimple_fold_builtin_strcpy): Handle no-warning.
7747 (gimple_fold_builtin_stxcpy_chk): Same.
7748 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
7750 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7752 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
7753 parameter. Use it for SFmode.
7754 (rs6000_function_arg_advance_1): Adjust.
7755 (rs6000_function_arg): Adjust.
7756 (rs6000_gimplify_va_arg): Pass false for that new parameter.
7758 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7760 PR rtl-optimization/84169
7761 PR rtl-optimization/84780
7762 * combine.c (can_combine_p): Check for a 2-insn combination whether
7763 the destination register is used between the two insns, too.
7765 2018-03-12 Richard Biener <rguenther@suse.de>
7767 PR tree-optimization/84803
7768 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
7769 for refs DR analysis didn't process.
7771 2018-03-12 Richard Biener <rguenther@suse.de>
7773 PR tree-optimization/84777
7774 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
7775 force-vectorize loops ignore whether we are optimizing for size.
7777 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
7779 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
7780 (TARGET_MD_ASM_ADJUST): Define.
7782 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
7783 Kito Cheng <kito.cheng@gmail.com>
7784 Chung-Ju Wu <jasonwucj@gmail.com>
7786 * config/nds32/nds32.c (nds32_compute_stack_frame,
7787 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
7788 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
7789 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
7790 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
7791 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
7792 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
7793 * config/nds32/nds32.md (prologue, epilogue): Use macro
7794 NDS32_V3PUSH_AVAILABLE_P to do checking.
7796 2018-03-11 Jakub Jelinek <jakub@redhat.com>
7799 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
7800 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
7801 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
7802 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
7803 addition of most attributes on !orig_type_die or the attribute not
7804 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
7806 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7807 Chung-Ju Wu <jasonwucj@gmail.com>
7809 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
7811 * config/nds32/nds32.opt (mvh): New option.
7813 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7814 Chung-Ju Wu <jasonwucj@gmail.com>
7816 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
7818 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
7819 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
7822 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7823 Chung-Ju Wu <jasonwucj@gmail.com>
7825 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
7827 * config/nds32/nds32-multiple.md (strlensi): New pattern.
7828 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
7830 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
7831 Kito Cheng <kito.cheng@gmail.com>
7832 Chung-Ju Wu <jasonwucj@gmail.com>
7834 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
7835 UNSPEC_FFMISM and UNSPEC_FLMISM.
7836 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
7837 for ffb, ffmism and flmism.
7838 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
7839 (unspec_ffmism): Ditto.
7840 (unspec_flmism): Ditto.
7841 (nds32_expand_builtin_impl): Check if string extension is available.
7842 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
7843 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
7845 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
7848 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
7851 * lra-assigns.c (assign_by_spills): Return a flag of reload
7852 assignment failure. Do not process the reload assignment
7853 failures. Do not spill other reload pseudos if they has the same
7855 (lra_assign): Add a return arg. Set up from the result of
7856 assign_by_spills call.
7857 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7858 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7859 usage_insns if it is not NULL.
7860 (spill_hard_reg_in_range): New function.
7861 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7862 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7863 function prototypes.
7864 (lra_assign): Change prototype.
7865 * lra.c (lra): Add code to deal with fails by splitting hard reg
7868 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
7871 * config/i386/i386.opt: Replace Enforcment with Enforcement.
7873 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
7876 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
7877 (dw_val_node): Add val_symbolic_view.
7878 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
7879 (symview_upper_bound): New.
7880 (new_line_info_table): Initialize symviews_since_reset.
7881 (dwarf2out_source_line): Count symviews_since_reset and set
7882 symview_upper_bound.
7883 (dw_val_equal_p): Handle symview.
7884 (add_AT_symview): New.
7885 (print_dw_val): Handle symview.
7886 (attr_checksum, attr_checksum_ordered): Likewise.
7887 (same_dw_val_p, size_of_die): Likewise.
7888 (value_format, output_die): Likewise.
7889 (add_high_low_attributes): Use add_AT_symview for entry_view.
7890 (dwarf2out_finish): Reset symview_upper_bound, clear
7893 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
7896 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
7897 Add strict argument and use it.
7898 (rs6000_split_multireg_move): Update for new strict argument.
7899 (mem_operand_gpr): Disallow all non-offsettable addresses.
7900 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
7902 2018-03-09 Jakub Jelinek <jakub@redhat.com>
7905 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
7906 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
7907 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
7910 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
7911 decl, use remap_type if we want to use the type.
7913 2018-03-09 Martin Sebor <msebor@redhat.com>
7915 PR tree-optimization/84526
7916 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7918 (builtin_access::generic_overlap): Be prepared to handle non-array
7921 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
7923 PR rtl-optimization/84682
7924 * lra-constraints.c (process_address_1): Check is_address flag
7925 for address constraints.
7926 (process_alt_operands): Likewise.
7927 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
7928 preprocess_constraints.
7929 * recog.h (preprocess_constraints): Add oploc parameter.
7931 * recog.c (preprocess_constraints): Test address_operand for
7932 CT_ADDRESS constraints.
7934 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
7937 * lra-assigns.c (assign_by_spills): Return a flag of reload
7938 assignment failure. Do not process the reload assignment
7939 failures. Do not spill other reload pseudos if they has the same
7941 (lra_assign): Add a return arg. Set up from the result of
7942 assign_by_spills call.
7943 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7944 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7945 usage_insns if it is not NULL.
7946 (spill_hard_reg_in_range): New function.
7947 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7948 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7949 function prototypes.
7950 (lra_assign): Change prototype.
7951 * lra.c (lra): Add code to deal with fails by splitting hard reg
7954 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7957 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
7958 Accept complain bool parameter. Only emit errors if it is true.
7959 (arm_parse_cpu_option_name): Likewise.
7960 (arm_target_thumb_only): Adjust callers of the above.
7961 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
7962 prototype to take a default true bool parameter.
7963 (arm_parse_arch_option_name): Likewise.
7965 2018-03-09 David Malcolm <dmalcolm@redhat.com>
7966 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
7970 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
7971 * configure: Regenerate.
7972 * configure.ac ("linker --version-script option"): New.
7973 ("linker soname option"): New.
7975 2018-03-09 Richard Biener <rguenther@suse.de>
7977 PR tree-optimization/84775
7978 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
7979 immediate uses of predicate stmts and mark them modified.
7982 PR tree-optimization/84178
7983 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7985 (version_loop_for_if_conversion): Delay update_ssa call.
7986 (tree_if_conversion): Delay update_ssa until after predicate
7989 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
7992 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
7993 when the function accesses prior frames.
7995 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7998 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
7999 gen_llsym, otherwise call maybe_gen_llsym.
8002 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
8003 has ',' character inside of it.
8005 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8008 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
8009 as clobbering CC_REGNUM.
8011 2018-03-08 Richard Biener <rguenther@suse.de>
8014 * tree-scalar-evolution.c: Include tree-into-ssa.h.
8015 (follow_copies_to_constant): Do not follow SSA names registered
8018 2018-03-08 Richard Biener <rguenther@suse.de>
8020 PR tree-optimization/84178
8021 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8023 (version_loop_for_if_conversion): Delay update_ssa call.
8024 (tree_if_conversion): Delay update_ssa until after predicate
8027 2018-03-08 David Malcolm <dmalcolm@redhat.com>
8029 PR tree-optimization/84178
8030 * tree-if-conv.c (release_bb_predicate): Remove the
8031 the assertion that the stmts have NULL use_ops.
8032 Discard the statements, asserting that they haven't
8033 yet been added to a BB.
8035 2018-03-08 Richard Biener <rguenther@suse.de>
8037 PR tree-optimization/84746
8038 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8039 (phi_translate): Pass in destination ANTIC_OUT set.
8040 (phi_translate_1): Likewise. For a simplified result lookup
8041 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8042 (phi_translate_set): Adjust.
8043 (do_pre_regular_insertion): Likewise.
8044 (do_pre_partial_partial_insertion): Likewise.
8046 2018-03-08 Martin Liska <mliska@suse.cz>
8048 PR gcov-profile/84735
8049 * doc/gcov.texi: Document usage of profile files.
8050 * gcov-io.h: Document changes in the format.
8052 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
8056 * dwarf2out.c (struct dw_line_info_table): Update comments for
8058 (FORCE_RESET_NEXT_VIEW): New.
8059 (FORCE_RESETTING_VIEW_P): New.
8060 (RESETTING_VIEW_P): Check for -1 too.
8061 (ZERO_VIEW_P): Likewise.
8062 (new_line_info_table): Force-reset next view.
8063 (dwarf2out_begin_function): Likewise.
8064 (dwarf2out_source_line): Simplify zero_view_p initialization.
8065 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8066 view directly. Omit view when omitting .loc at line 0.
8068 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8070 PR tree-optimization/84740
8071 * tree-switch-conversion.c (process_switch): Call build_constructors
8072 only if info.phi_count is non-zero.
8074 PR tree-optimization/84739
8075 * tree-tailcall.c (find_tail_calls): Check call arguments against
8076 DECL_ARGUMENTS (current_function_decl) rather than
8077 DECL_ARGUMENTS (func) when checking for tail recursion.
8079 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8081 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8082 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8083 Volker Reichelt's entry and add entries for people that perform
8084 GCC fuzzy testing and report numerous bugs.
8086 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
8089 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8090 readonly data in sdata, if that is disabled.
8091 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8092 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8093 -mreadonly-in-sdata option.
8095 2018-03-07 Martin Sebor <msebor@redhat.com>
8097 PR tree-optimization/84468
8098 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8099 basic block when looking for nul assignment.
8101 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
8104 * except.h (output_function_exception_table): Adjust prototype.
8105 * except.c (output_function_exception_table): Remove FNNAME parameter
8106 and add SECTION parameter. Ouput one part of the table at a time.
8107 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8108 the first part of the exception table and emit unwind directives.
8109 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8110 (i386_pe_seh_cold_init): Likewise.
8111 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8112 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8113 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8114 (ix86_output_call_insn): Emit a nop in one more case for SEH.
8115 * config/i386/winnt.c: Include except.h.
8116 (struct seh_frame_state): Add reg_offset, after_prologue and
8117 in_cold_section fields.
8118 (i386_pe_seh_end_prologue): Set seh->after_prologue.
8119 (i386_pe_seh_cold_init): New function.
8120 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8121 to seh->in_cold_section.
8122 (seh_emit_push): Record the offset of the push.
8123 (seh_emit_save): Record the offet of the save.
8124 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8125 Test seh->after_prologue to disregard the epilogue.
8126 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8127 (i386_pe_end_cold_function): New function.
8129 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8132 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8133 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8136 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8138 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8139 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8142 * multiple_target.c: Include tree-inline.h and intl.h.
8143 (expand_target_clones): Diagnose and fail if node->definition and
8144 !tree_versionable_function_p (node->decl).
8146 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
8148 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8150 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8151 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8152 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8154 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8157 * combine.c (try_combine): Use reg_or_subregno instead of handling
8158 just paradoxical SUBREGs and REGs.
8160 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
8162 * config/arc/arc.c (arc_finalize_pic): Remove function.
8163 (arc_must_save_register): We use single base PIC register, remove
8164 checks to save/restore the PIC register.
8165 (arc_expand_prologue): Likewise.
8166 * config/arc/arc-protos.h (arc_set_default_type_attributes):
8168 (arc_verify_short): Likewise.
8169 (arc_attr_type): Likewise.
8170 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8171 (walk_stores): Likewise.
8172 (arc_address_cost): Make it static.
8173 (arc_verify_short): Likewise.
8174 (branch_dest): Likewise.
8175 (arc_attr_type): Likewise.
8176 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
8177 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
8178 (arc_final_prescan_insn): Remove inserting the nops due to
8179 hardware hazards. It is done in reorg step.
8180 (insn_length_variant_t): Remove.
8181 (insn_length_parameters_t): Likewise.
8182 (arc_insn_length_parameters): Likewise.
8183 (arc_get_insn_variants): Likewise.
8184 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
8186 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8189 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
8192 PR tree-optimization/84687
8193 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
8195 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8197 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8199 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
8200 Rename to ppc_speculation_barrier.
8201 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
8202 __builtin_ppc_speculation_barrier.
8204 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8207 * combine.c (combine_simplify_rtx): Don't try to simplify if
8208 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
8211 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
8213 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
8214 to 32 bytes when compiling for POWER9.
8216 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8219 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
8220 regparm >= 3 with no arg reg available also for calls with
8221 flag_force_indirect_call. Pass decl to ix86_function_regparm.
8224 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
8226 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
8228 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
8231 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
8233 2018-03-05 Richard Biener <rguenther@suse.de>
8235 PR tree-optimization/84486
8236 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
8237 When inserting a __builtin_assume_aligned call set the LHS
8238 SSA name alignment info accordingly.
8240 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
8242 PR tree-optimization/84114
8243 * config/aarch64/aarch64.c (aarch64_reassociation_width)
8244 Avoid reassociation of FLOAT_MODE addition.
8246 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
8248 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
8249 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
8250 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
8251 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
8252 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
8253 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
8254 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
8256 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
8257 __builtin_ia32_wbinvd): New builtins.
8258 (SPECIAL_ARGS2): New.
8259 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
8260 (SPECIAL_ARGS2): New.
8261 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
8262 (ix86_valid_target_attribute_inner_p): Ditto.
8263 (ix86_init_mmx_sse_builtins): Add special_args2.
8264 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
8265 TARGET_WBNOINVD_P): New.
8266 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
8267 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
8268 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
8269 * config/i386/immintrin.h (_wbinvd): New intrinsic.
8270 * config/i386/pconfigintrin.h: New file.
8271 * config/i386/wbnoinvdintrin.h: Ditto.
8272 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
8274 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
8276 2018-03-05 Richard Biener <rguenther@suse.de>
8278 PR tree-optimization/84670
8279 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8281 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8282 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8283 (compute_antic_aux): Only assert the number of values in ANTIC_IN
8284 doesn't grow if all successors (recursively) were visited at least
8287 2018-03-05 Richard Biener <rguenther@suse.de>
8289 PR tree-optimization/84650
8290 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
8291 if executed in the loop pipeline.
8293 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
8295 * doc/configfiles.texi (Configuration Files): Move info about
8296 conditionalizing $target-protos.h to...
8297 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
8298 differs from $target-protos.h.
8300 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
8301 Chung-Ju Wu <jasonwucj@gmail.com>
8303 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
8304 * config/nds32/nds32-multiple.md (setmemsi): Define.
8305 * config/nds32/nds32-memory-manipulation.c
8306 (nds32_gen_dup_4_byte_to_word_value): New.
8307 (emit_setmem_word_loop): New.
8308 (emit_setmem_byte_loop): New.
8309 (nds32_expand_setmem_loop): New.
8310 (nds32_expand_setmem_loop_v3m): New.
8311 (nds32_expand_setmem_unroll): New.
8312 (nds32_expand_setmem): New.
8314 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8315 Chung-Ju Wu <jasonwucj@gmail.com>
8317 * config/nds32/nds32-memory-manipulation.c
8318 (nds32_emit_load_store): New.
8319 (nds32_emit_post_inc_load_store): New.
8320 (nds32_emit_mem_move): New.
8321 (nds32_emit_mem_move_block): New.
8322 (nds32_expand_movmemsi_loop_unknown_size): New.
8323 (nds32_expand_movmemsi_loop_known_size): New.
8324 (nds32_expand_movmemsi_loop): New.
8325 (nds32_expand_movmemsi_unroll): New.
8326 (nds32_expand_movmemqi): Rename ...
8327 (nds32_expand_movmemsi): ... to this.
8328 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
8329 (movmemsi): ... to this.
8330 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
8331 (nds32_expand_movmemsi): ... to this.
8333 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8334 Monk Chiang <sh.chiang04@gmail.com>
8335 Chung-Ju Wu <jasonwucj@gmail.com>
8337 * config/nds32/nds32-protos.h
8338 (nds32_expand_load_multiple): New arguments.
8339 (nds32_expand_store_multiple): Ditto.
8340 (nds32_valid_multiple_load_store): Rename ...
8341 (nds32_valid_multiple_load_store_p): ... to this.
8342 * config/nds32/nds32-memory-manipulation.c
8343 (nds32_expand_load_multiple): Refine implementation.
8344 (nds32_expand_store_multiple): Ditto.
8345 * config/nds32/nds32-multiple.md
8346 (load_multiple): Update nds32_expand_load_multiple interface.
8347 (store_multiple): Update nds32_expand_store_multiple interface.
8348 * config/nds32/nds32-predicates.c
8349 (nds32_valid_multiple_load_store): Rename ...
8350 (nds32_valid_multiple_load_store_p): ... to this and refine
8352 * config/nds32/predicates.md
8353 (nds32_load_multiple_and_update_address_operation): New predicate.
8354 (nds32_store_multiple_and_update_address_operation): New predicate.
8356 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8357 Chung-Ju Wu <jasonwucj@gmail.com>
8359 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
8360 (combo): New attribute.
8361 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
8363 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
8365 * config/nds32/nds32.opt: Change -mcmodel= default value.
8367 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
8368 Monk Chiang <sh.chiang04@gmail.com>
8369 Chung-Ju Wu <jasonwucj@gmail.com>
8371 * config/nds32/constants.md (unspec_element): New enum.
8372 * config/nds32/constraints.md (Umw): New constraint.
8373 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
8374 * config/nds32/nds32-intrinsic.md: Likewise.
8375 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
8376 (nds32_valid_smw_lwm_base_p): New.
8377 (nds32_output_smw_single_word): New.
8378 (nds32_output_lmw_single_word): New.
8379 (nds32_expand_unaligned_load): New.
8380 (nds32_expand_unaligned_store): New.
8381 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
8382 (nds32_output_smw_single_word): Declare.
8383 (nds32_output_lmw_single_word): Declare.
8384 (nds32_expand_unaligned_load): Declare.
8385 (nds32_expand_unaligned_store): Declare.
8386 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
8387 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
8388 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
8389 NDS32_BUILTIN_UASTORE_DW.
8390 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
8393 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
8394 Kito Cheng <kito.cheng@gmail.com>
8395 Chung-Ju Wu <jasonwucj@gmail.com>
8397 * config/nds32/nds32-intrinsic.c
8398 (nds32_expand_builtin_null_ftype_reg): Delete.
8399 (nds32_expand_builtin_reg_ftype_imm): Ditto.
8400 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
8401 (nds32_read_argument): New.
8402 (nds32_legitimize_target): Ditto.
8403 (nds32_legitimize_argument): Ditto.
8404 (nds32_check_constant_argument): Ditto.
8405 (nds32_expand_unop_builtin): Ditto.
8406 (nds32_expand_unopimm_builtin): Ditto.
8407 (nds32_expand_binop_builtin): Ditto.
8408 (nds32_builtin_decl_impl): Ditto.
8409 (builtin_description): Ditto.
8410 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
8411 (nds32_init_builtins_impl): Ditto.
8412 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
8413 (nds32_builtin_decl): New.
8414 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
8415 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
8417 2018-03-02 Jeff Law <law@redhat.com>
8419 * reorg.c (stop_search_p): Handle DEBUG_INSN.
8420 (redundant_insn, fill_simple_delay_slots): Likewise.
8421 (fill_slots_from_thread): Likewise.
8422 * resource.c (mark_referenced_resources): Likewise.
8423 (mark_set_resources, find_dead_or_set_registers): Likewise.
8425 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8427 * substring-locations.h (format_warning_va): Formatting fix for
8429 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
8431 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
8432 * substring-locations.c: Include intl.h.
8433 (format_warning_va): Turned into small wrapper around
8434 format_warning_n_va, renamed to ...
8435 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
8436 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
8438 (format_warning_at_substring_n): New function.
8439 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
8440 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
8441 format_warning_at_substring with just a shorter name instead of
8442 const function pointer.
8443 (fmtwarn_n): New function.
8444 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
8445 appropriate, get rid of all the fmtstr temporaries, move conditionals
8446 with G_() wrapped string literals directly into fmtwarn arguments,
8447 cast dir.len to (int), formatting fixes.
8449 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
8451 * doc/invoke.texi: Remove "Cilk Plus" references.
8453 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8454 Richard Biener <rguenther@suse.de>
8457 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
8458 for error or warning attributes if CALL_FROM_THUNK_P is set.
8461 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8464 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
8465 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
8468 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
8469 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
8471 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
8472 instead of -1U in last predictors element's probability member.
8474 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8477 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8478 arguments if they are comparable.
8480 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8482 PR tree-optimization/84634
8483 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
8484 masks and masked_loop_p with a single loop_masks, making sure it's
8485 null for bb vectorization.
8487 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8489 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8490 (vect_analyze_data_ref_access): Use loop->safe_len rather than
8491 loop->force_vectorize to check whether there is no alias.
8493 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8496 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
8498 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
8500 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
8501 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
8502 instead of a loop around prev_real_insn.
8503 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
8507 * config/i386/i386.c (ix86_print_operand): Use conditional
8508 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
8511 2018-03-02 Richard Biener <rguenther@suse.de>
8513 PR tree-optimization/84427
8514 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
8515 (bitmap_set_subtract_values): Rewrite to handle multiple
8518 (prune_clobbered_mems): Likewise.
8519 (phi_translate): Take edge instead of pred/phiblock.
8520 (phi_translate_1): Likewise.
8521 (phi_translate_set): Likewise. Insert all translated
8522 exprs for a value into the set, keeping possibly multiple
8523 expressions per value.
8524 (compute_antic_aux): Adjust for phi_translate changes.
8525 When intersecting union the expressions and prune those
8526 not in the final value set, keeping possibly multiple
8527 expressions per value. Do not use value-insertion
8528 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
8529 all expressions. Add verification that the value-sets
8530 only shrink during iteration.
8531 (compute_partial_antic_aux): Adjust for the phi_translate changes.
8532 (do_pre_regular_insertion): Likewise.
8533 (do_pre_partial_partial_insertion): Likewise.
8535 2018-03-02 Richard Biener <rguenther@suse.de>
8538 * config/darwin.c (saved_debug_info_level): New static global.
8539 (darwin_asm_lto_start): Disable debug info generation for LTO out.
8540 (darwin_asm_lto_end): Restore debug info generation settings.
8542 2018-03-01 Martin Liska <mliska@suse.cz>
8545 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
8548 2018-03-01 Richard Biener <rguenther@suse.de>
8551 * dwarf2out.c (gen_variable_die): Properly handle late VLA
8552 type annotation with LTO when debug was disabled at compile-time.
8554 2018-03-01 Matthew Fortune <mfortune@gmail.com>
8556 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
8558 (mips_final_postscan_insn): Likewise.
8560 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
8562 PR rtl-optimization/84528
8563 * alias.c (init_alias_target): Add commentary.
8564 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
8565 a unique base value if the frame pointer is not eliminated
8566 to the stack pointer.
8568 2018-03-01 Tom de Vries <tom@codesourcery.com>
8570 PR rtl-optimization/83327
8571 * lra-int.h (hard_regs_spilled_into): Declare.
8572 * lra.c (hard_regs_spilled_into): Define.
8573 (init_reg_info): Init hard_regs_spilled_into.
8574 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
8575 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
8576 (process_bb_lives): Handle hard_regs_spilled_into.
8577 (lra_create_live_ranges_1): Before doing liveness propagation, clear
8578 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
8580 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
8582 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
8583 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
8584 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
8585 * config/rs6000/aix72.h: New file.
8587 2018-02-28 Jakub Jelinek <jakub@redhat.com>
8589 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
8590 instead of warning_at with conditional singular and plural messages
8594 * stor-layout.c (update_alignment_for_field): For
8595 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
8596 && !DECL_PACKED (field), do the alignment update, just use
8597 only desired_align instead of MAX (type_align, desired_align)
8599 (place_field): Don't do known_align < desired_align handling
8600 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
8601 is non-NULL, instead do it after rli->prev_field handling and
8602 only if not within a bitfield word. For DECL_PACKED (field)
8603 use type_align of BITS_PER_UNIT.
8605 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
8607 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
8608 superfluous parentheses and trailing spaces.
8610 2018-02-28 Richard Biener <rguenther@suse.de>
8612 PR tree-optimization/84584
8613 * graphite-scop-detection.c (scop_detection::add_scop): Discard
8614 SCoPs with fake exit edge.
8616 2018-02-28 Martin Liska <mliska@suse.cz>
8619 * timevar.c (timer::print): Fix format to properly print 100%
8622 2018-02-28 Richard Biener <rguenther@suse.de>
8625 * genmatch.c (capture_info::walk_match): Do not mark
8626 captured expressions without operands as expr_p given
8627 they act more like predicates and should be subject to
8628 "lost tail" side-effect preserving.
8630 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
8632 PR rtl-optimization/81611
8633 * auto-inc-dec.c (attempt_change): Move dead note from
8634 mem_insn if it's the next use of regno
8635 (find_address): Take address use of reg holding
8636 non-incremented value. Add parm to limit search to the named
8638 (merge_in_block): Attempt to use a mem insn that is the next
8639 use of the original regno.
8641 2018-02-27 Martin Sebor <msebor@redhat.com>
8644 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
8645 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
8647 2018-02-27 Martin Sebor <msebor@redhat.com>
8649 PR translation/84207
8650 * diagnostic-core.h (warning_n, error_n, inform_n): Change
8651 n argument to unsigned HOST_WIDE_INT.
8652 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
8653 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
8654 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
8655 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
8657 2018-02-27 Richard Biener <rguenther@suse.de>
8659 PR tree-optimization/84512
8660 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
8661 Do not use the estimate returned from record_stmt_cost for
8662 the scalar iteration cost but sum properly using add_stmt_cost.
8664 2018-02-27 Richard Biener <rguenther@suse.de>
8666 PR tree-optimization/84466
8667 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8668 Adjust last change to less strictly validate use operands.
8670 2018-02-27 Martin Liska <mliska@suse.cz>
8672 PR gcov-profile/84548
8673 * gcov.c (process_file): Allow partial overlap and consider it
8674 also as group functions.
8675 (output_lines): Properly calculate range of lines for a group.
8677 2018-02-27 Martin Liska <mliska@suse.cz>
8679 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
8680 'ggc' suffixes. Change first column width.
8681 (timer::print): Fix formatting of the column.
8683 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
8685 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
8686 preserve inline entry blocks for the sake of debug inline
8687 entry point markers alone.
8688 (remove_unused_locals): Suggest in comments a better place to
8689 force the preservation of inline entry blocks that are
8690 otherwise unused, but do not preserve them.
8692 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8694 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
8696 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8699 * config/i386/constraints.md (Bs): Replace
8700 ix86_indirect_branch_register with
8701 TARGET_INDIRECT_BRANCH_REGISTER.
8703 * config/i386/i386.md (indirect_jump): Likewise.
8704 (tablejump): Likewise.
8705 (*sibcall_memory): Likewise.
8706 (*sibcall_value_memory): Likewise.
8707 Peepholes of indirect call and jump via memory: Likewise.
8708 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
8709 (*sibcall_value_GOT_32): Likewise.
8710 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8711 (GOT_memory_operand): Likewise.
8712 (call_insn_operand): Likewise.
8713 (sibcall_insn_operand): Likewise.
8714 (GOT32_symbol_operand): Likewise.
8715 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
8717 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8719 PR rtl-optimization/83496
8720 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
8721 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
8722 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
8723 redundant insn, if any.
8724 (relax_delay_slots): Likewise.
8725 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
8727 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
8729 PR tree-optimization/83965
8730 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
8731 that grouped statements are part of a reduction chain. Return
8732 true if the statement is not marked as a reduction itself but
8734 (vect_recog_dot_prod_pattern): Don't check whether the statement
8735 is part of a group here.
8736 (vect_recog_sad_pattern): Likewise.
8737 (vect_recog_widen_sum_pattern): Likewise.
8739 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8742 * final.c (rest_of_clean_state): Also look for calls inside sequences.
8744 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8747 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
8749 (ix86_output_indirect_function_return): New prototype.
8750 (ix86_split_simple_return_pop_internal): Likewise.
8751 * config/i386/i386.c (indirect_return_via_cx): New.
8752 (indirect_return_via_cx_bnd): Likewise.
8753 (indirect_thunk_name): Handle return va CX_REG.
8754 (output_indirect_thunk_function): Create alias for
8755 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
8756 (ix86_output_indirect_jmp): Remove the bool argument.
8757 (ix86_output_indirect_function_return): New function.
8758 (ix86_split_simple_return_pop_internal): Likewise.
8759 * config/i386/i386.md (*indirect_jump): Don't pass false
8760 to ix86_output_indirect_jmp.
8761 (*tablejump_1): Likewise.
8762 (simple_return_pop_internal): Change it to define_insn_and_split.
8763 Call ix86_split_simple_return_pop_internal to split it for
8765 (simple_return_indirect_internal): Call
8766 ix86_output_indirect_function_return instead of
8767 ix86_output_indirect_jmp.
8769 2018-02-26 Jakub Jelinek <jakub@redhat.com>
8772 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
8773 memset and value initialization afterwards.
8775 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
8777 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
8779 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8782 * common/config/aarch64/aarch64-common.c
8783 (aarch_option_optimization_table[]): Switch
8784 off fomit-frame-pointer
8786 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
8787 Chung-Ju Wu <jasonwucj@gmail.com>
8789 * config/nds32/nds32-multiple.md (load_multiple): Disallow
8791 (store_multiple): Ditto.
8793 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
8795 * config.gcc: Add --with-cpu support for nds32 target.
8796 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
8797 * config/nds32/nds32.opt: Add -mcpu= option.
8799 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
8801 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
8802 isel=yes): Warn for these deprecated options.
8804 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
8806 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
8807 ISA_2_5_MASKS_EMBEDDED.
8809 2018-02-23 Jakub Jelinek <jakub@redhat.com>
8811 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
8812 p->max as pointers rather than using iterative_hash_expr.
8814 2018-02-23 Carl Love <cel@us.ibm.com>
8816 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
8817 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
8819 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
8820 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
8821 P8V_BUILTIN_VEC_VUNSIGNED2.
8823 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
8826 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
8827 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
8829 * lra-constraints.c (curr_insn_transform): Set up
8830 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
8832 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
8833 * lra-eliminations.c (spill_pseudos): Ditto.
8834 (process_insn_for_elimination): Ditto.
8835 * lra-lives.c (reg_early_clobber_p): Use the new macros.
8836 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
8837 LRA_NON_CLOBBERED_ALT.
8839 2018-02-22 Martin Sebor <msebor@redhat.com>
8841 PR tree-optimization/84480
8842 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
8843 to maybe_diag_stxncpy_trunc. Call it.
8844 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
8845 from gimple_fold_builtin_strcpy. Print inlining stack.
8846 (handle_builtin_stxncpy): Print inlining stack.
8847 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
8849 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
8852 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
8853 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
8854 and -fcheck-pointer-bounds are used together.
8855 (indirect_thunk_prefix): New enum.
8856 (indirect_thunk_need_prefix): New function.
8857 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
8858 "_nt" instead of "_bnd" for NOTRACK prefix.
8859 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
8860 (output_indirect_thunk_function): Likewise.
8862 (ix86_code_end): Update output_indirect_thunk_function calls.
8863 (ix86_output_indirect_branch_via_reg): Replace
8864 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
8865 (ix86_output_indirect_branch_via_push): Likewise.
8866 (ix86_output_function_return): Likewise.
8867 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
8868 incompatible with -fcf-protection=branch and
8869 -fcheck-pointer-bounds.
8871 2018-02-22 Steve Ellcey <sellcey@cavium.com>
8874 * config/aarch64/aarch64.c (aarch64_print_address_internal):
8875 Change gcc_assert call to output_operand_lossage.
8877 2018-02-22 Steve Ellcey <sellcey@cavium.com>
8879 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
8881 2018-02-22 DJ Delorie <dj@redhat.com>
8882 Sebastian Perta <sebastian.perta@renesas.com>
8883 Oleg Endo <olegendo@gcc.gnu.org>
8885 * config/rx/rx.c (rx_rtx_costs): New function.
8886 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
8888 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
8890 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
8892 2018-02-22 Martin Liska <mliska@suse.cz>
8895 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
8896 Add "native" as a possible value.
8898 2018-02-22 Martin Liska <mliska@suse.cz>
8901 * config/i386/i386.c (ix86_option_override_internal):
8902 Add "native" as a possible value for -march and -mtune.
8904 2018-02-22 Jakub Jelinek <jakub@redhat.com>
8907 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
8908 to all type variants.
8910 PR tree-optimization/84503
8911 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
8912 width as info->bitpos + info->bitsize - start.
8913 (merged_store_group::merge_overlapping): Simplify width computation.
8914 (check_no_overlap): New function.
8915 (imm_store_chain_info::try_coalesce_bswap): Compute expected
8916 start + width and last_order of the group, fail if check_no_overlap
8918 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
8919 to group if check_no_overlap fails.
8921 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8923 * config/rs6000/altivec.md: Delete contraint arguments to
8924 define_expand, define_split, and define_peephole2, and in
8925 define_insn_and_split if always unused.
8926 * config/rs6000/darwin.md: Ditto.
8927 * config/rs6000/dfp.md: Ditto.
8928 * config/rs6000/rs6000.md: Ditto.
8929 * config/rs6000/sync.md: Ditto.
8930 * config/rs6000/vector.md: Ditto.
8931 * config/rs6000/vsx.md: Ditto.
8933 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8935 * config/rs6000/altivec.md: Write output control strings as braced
8936 blocks instead of double-quoted strings.
8937 * config/rs6000/darwin.md: Ditto.
8938 * config/rs6000/rs6000.md: Ditto.
8939 * config/rs6000/vector.md: Ditto.
8940 * config/rs6000/vsx.md: Ditto.
8942 2018-02-21 Jason Merrill <jason@redhat.com>
8944 PR c++/84314 - ICE with templates and fastcall attribute.
8945 * attribs.c (build_type_attribute_qual_variant): Remove assert.
8947 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
8949 * ipa-cp.c (determine_versionability): Fix comment typos.
8951 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
8954 * ipa-cp.c (determine_versionability): Do not version functions caling
8957 2018-02-21 Martin Liska <mliska@suse.cz>
8960 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
8961 Add "native" as a possible value.
8962 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
8963 the macro when native cpu detection is available.
8965 2018-02-21 Martin Liska <mliska@suse.cz>
8968 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
8969 Add "native" as a possible value.
8970 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
8971 when native cpu detection is available.
8973 2018-02-21 Jakub Jelinek <jakub@redhat.com>
8974 Martin Sebor <msebor@redhat.com>
8976 PR tree-optimization/84478
8977 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
8979 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
8980 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
8981 support which is conservatively correct, for 2 only stay conservative
8982 for maxlen. Formatting and comment capitalization fixes. Add STRICT
8983 argument to the 2 argument get_range_strlen, adjust 6 arg
8984 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
8986 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
8987 (gimple_fold_builtin_strlen): Pass true as last argument to
8990 2018-02-20 Martin Sebor <msebor@redhat.com>
8993 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
8994 (builtin_memref::set_base_and_offset): Same. Handle inner references.
8995 (builtin_memref::builtin_memref): Factor out parts into
8996 set_base_and_offset and call it.
8998 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9001 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
9002 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
9003 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
9004 search at the associated MODE_INT.
9006 2018-02-20 Jeff Law <law@redhat.com>
9009 PR tree-optimization/81592
9011 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
9012 for range data rather than using global data.
9013 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
9014 range data rather than using global data.
9015 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9016 pass it to children as needed.
9017 (struct directive::fmtresult): Similarly.
9018 (struct directive::set_width): Similarly.
9019 (struct directive::set_precision): Similarly.
9020 (format_integer, format_directive, parse_directive): Similarly.
9021 (format_none): Accept unnamed vr_values parameter.
9022 (format_percent, format_floating, format_character): Similarly.
9023 (format_string, format_plain): Similarly.
9024 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9025 the EVRP range analyzer for range data rather than using global data.
9026 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9027 gimple-ssa-evrp-analyze.h
9028 (class sprintf_dom_walker): Add after_dom_children member function.
9029 Add evrp_range_analyzer member.
9030 (sprintf_dom_walker::before_dom_children): Call into the EVRP
9031 range analyzer as needed.
9032 (sprintf_dom_walker::after_dom_children): New member function.
9033 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9035 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9036 (evrp_range_analyzer::pop_to_marker): Likewise.
9038 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9040 PR tree-optimization/84419
9041 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9042 with the required type if its current type is compatible but
9045 2018-02-20 Jakub Jelinek <jakub@redhat.com>
9048 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9049 after vectorization.
9051 2018-02-20 Martin Liska <mliska@suse.cz>
9054 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9055 possible values if we don't have a hint.
9057 2018-02-20 Martin Liska <mliska@suse.cz>
9061 * final.c (shorten_branches): Build align_tab array with one
9063 * opts.c (finish_options): Add alignment option limit check.
9064 (MAX_CODE_ALIGN): Likewise.
9065 (MAX_CODE_ALIGN_VALUE): Likewise.
9066 * doc/invoke.texi: Document maximum allowed option value for
9067 all -falign-* options.
9069 2018-02-19 Jakub Jelinek <jakub@redhat.com>
9072 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9073 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9074 * var-tracking.c (emit_note_insn_var_location): Remove all references
9075 to NOTE_INSN_CALL_ARG_LOCATION.
9076 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9077 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9078 Use copy_rtx_if_shared.
9079 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9080 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9081 (dwarf2out_var_location): Remove handling of
9082 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9084 * final.c (final_scan_insn): Remove all references to
9085 NOTE_INSN_CALL_ARG_LOCATION.
9086 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
9087 before dumping final insns.
9088 * except.c (emit_note_eh_region_end): Remove all references to
9089 NOTE_INSN_CALL_ARG_LOCATION.
9090 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9091 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9092 * config/arc/arc.c (hwloop_optimize): Likewise.
9093 * config/arm/arm.c (create_fix_barrier): Likewise.
9094 * config/s390/s390.c (s390_chunkify_start): Likewise.
9095 * config/sh/sh.c (find_barrier): Likewise.
9096 * config/i386/i386.c (rest_of_insert_endbranch,
9097 ix86_seh_fixup_eh_fallthru): Likewise.
9098 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9099 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9100 * config/frv/frv.c (frv_function_prologue): Likewise.
9101 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
9103 (note_outside_basic_block_p): Remove all references to
9104 NOTE_INSN_CALL_ARG_LOCATION.
9105 * gengtype.c (adjust_field_rtx_def): Likewise.
9106 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9108 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9109 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9112 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9115 PR tree-optimization/84452
9116 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9117 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9120 2018-02-19 Martin Liska <mliska@suse.cz>
9123 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9125 2018-02-19 Martin Liska <mliska@suse.cz>
9126 Richard Sandiford <richard.sandiford@linaro.org>
9128 PR tree-optimization/82491
9129 * gimple-fold.c (get_base_constructor): Make earlier bail out
9132 2018-02-19 Carl Love <cel@us.ibm.com>
9134 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9135 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9137 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9138 P8V_BUILTIN_VEC_NEG.
9140 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
9142 * config/rl78/rl78.md (movdf): New define expand.
9144 2018-02-19 Martin Liska <mliska@suse.cz>
9147 * doc/invoke.texi: Fix typo.
9148 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9150 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
9152 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9153 handle rs6000_single_float and rs6000_double_float specially for
9156 2018-02-16 Jeff Law <law@redhat.com>
9158 * config/rx/rx.c (add_pop_cfi_notes): New function.;
9161 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9164 * ipa-inline.c (inline_small_functions): Fix a typo.
9166 2018-02-16 Nathan Sidwell <nathan@acm.org>
9168 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9170 2018-02-16 Carl Love <cel@us.ibm.com>
9172 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9173 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9174 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9175 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9176 expansion to P8V_BUILTIN_VEC_FLOAT2.
9178 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
9180 PR rtl-optimization/70023
9181 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
9182 src_regno into account.
9184 2018-02-16 Carl Love <cel@us.ibm.com>
9186 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
9187 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
9188 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
9189 * config/rs6000/rs6000.c: Remove case statements for
9190 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
9191 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
9192 and P9V_BUILTIN_VEC_VINSERT4B.
9193 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
9194 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
9195 * config/rs6000/vsx.md:
9196 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
9199 2018-02-16 Carl Love <cel@us.ibm.com>
9201 * config/rs6000/altivec.h: Add builtin names vec_extract4b
9203 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
9205 * config/rs6000/rs6000-c.c: Add the definitions for
9206 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
9207 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
9208 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
9209 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
9210 definition for insert4b and define insn *insert3b_internal.
9211 * doc/extend.texi: Add documentation for vec_extract4b.
9213 2018-02-16 Nathan Sidwell <nathan@acm.org>
9215 * doc/extend.texi (Backwards Compatibility): Mention friend
9216 injection. Note for-scope is deprecated.
9217 * doc/invoke.texi (-ffriend-injection): Deprecate.
9219 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
9221 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
9222 that moved to I2, also allow destinations that are a paradoxical
9223 subreg (instead of a normal reg).
9225 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
9228 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
9231 2018-02-16 Richard Biener <rguenther@suse.de>
9233 PR tree-optimization/84037
9234 PR tree-optimization/84016
9236 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9237 Adjust vec_construct for the fact we need additional higher latency
9238 128bit inserts for AVX256 and AVX512 vector builds.
9239 (ix86_add_stmt_cost): Scale vector construction cost for
9242 2018-02-16 Richard Biener <rguenther@suse.de>
9244 PR tree-optimization/84417
9245 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
9246 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
9247 (non_rewritable_lvalue_p): Likewise, use poly-ints.
9249 2018-02-16 Martin Liska <mliska@suse.cz>
9252 * internal-fn.def (ASAN_CHECK): Set proper flags.
9253 (ASAN_MARK): Likewise.
9255 2018-02-16 Julia Koval <julia.koval@intel.com>
9257 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
9258 from PTA_CANNONLAKE.
9260 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9263 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
9264 Use ++iter rather than iter++ for std::list iterators.
9265 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
9266 defer deleting them until all nodes in the forest are processed. Do
9267 free even leaf nodes. Change to_process into auto_vec.
9270 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
9271 * vec.h (vec_default_construct): Use memset instead of placement new
9272 if BROKEN_VALUE_INITIALIZATION is defined.
9273 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
9274 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
9277 PR rtl-optimization/83723
9278 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
9279 * lra.c (lra_substitute_pseudo): Likewise. If true, use
9280 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
9282 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
9284 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
9286 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
9288 PR rtl-optimization/81443
9289 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
9290 from inner REGs to paradoxical SUBREGs.
9292 2018-02-16 Richard Biener <rguenther@suse.de>
9294 PR tree-optimization/84399
9295 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9296 For operands we can analyze at their definition make sure we can
9297 analyze them at each use as well.
9299 2018-02-16 Richard Biener <rguenther@suse.de>
9301 PR tree-optimization/84190
9302 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
9303 volatile accesses if the decl isn't volatile.
9305 2018-02-15 Jason Merrill <jason@redhat.com>
9307 PR c++/84314 - ICE with templates and fastcall attribute.
9308 * attribs.c (build_type_attribute_qual_variant): Don't clobber
9309 TYPE_CANONICAL on an existing type.
9311 2018-02-15 Jakub Jelinek <jakub@redhat.com>
9313 PR tree-optimization/84383
9314 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
9315 dstoff nor call operand_equal_p if dstbase is NULL.
9317 PR tree-optimization/84334
9318 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
9319 also a CONSTANT_CLASS_P, punt.
9321 2018-02-14 Jim Wilson <jimw@sifive.com>
9323 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
9324 first SMALL_OPERAND check. New local min_second_step. Move assert
9325 to where locals are set. Add TARGET_RVC support.
9326 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
9328 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
9330 * doc/invoke.texi: Correct -Wformat-overflow code sample.
9332 2018-02-14 Martin Sebor <msebor@redhat.com>
9334 PR tree-optimization/83698
9335 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
9336 arrays constrain the offset range to their bounds.
9337 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
9338 (builtin_access::overlap): Avoid setting the size of overlap if it's
9340 (maybe_diag_overlap): Also consider arrays when deciding what values
9341 of offsets to include in diagnostics.
9343 2018-02-14 Martin Sebor <msebor@redhat.com>
9346 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
9347 that correspond to the kind of a declaration.
9349 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
9352 * config/pa/pa.md: Load address of PIC label using the linkage table
9353 if the label is nonlocal.
9355 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
9357 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
9358 warning message if user requests -maltivec=be.
9359 * doc/invoke.texi: Document deprecation of -maltivec=be.
9361 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
9364 * config/rs6000/rs6000-c.c: Update definitions for
9365 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
9366 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
9368 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9371 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
9372 add _get_ssp intrinsics. Remove argument from
9373 __builtin_ia32_rdssp[d|q].
9374 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
9375 * config/i386/i386-builtin.def: Remove argument from
9376 __builtin_ia32_rdssp[d|q].
9377 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
9378 ix86_expand_special_args_builtin for _rdssp[d|q].
9379 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
9380 Clear register before usage.
9381 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
9382 Add documentation for new _get_ssp and _inc_ssp intrinsics.
9384 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
9386 PR tree-optimization/84357
9387 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
9388 operand 1 of an ARRAY_REF too.
9390 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
9393 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
9394 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9396 (set_of_reg): New struct.
9397 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
9398 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
9399 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9401 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
9402 Split into bitclr, bitset, bitinvert patterns if appropriate.
9403 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
9404 use rx_fuse_in_memory_bitop.
9405 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
9406 to named insn, correct maximum insn length.
9408 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
9411 * machmode.def: Define a complex mode for PARTIAL_INT.
9412 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
9414 * doc/rtl.texi: Document CSPImode.
9415 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
9417 (msp430_hard_regno_nregs_with_padding): Likewise.
9419 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
9422 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
9424 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
9426 PR rtl-optimization/84169
9427 * combine.c (try_combine): New variable split_i2i3. Set it to true if
9428 we generated a parallel as new i3 and we split that to new i2 and i3
9429 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
9430 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
9431 those to i2, not i1. Partially rewrite this scan code.
9433 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9436 * stor-layout.c (place_field): For variable length fields, adjust
9437 offset_align afterwards not just based on the field's alignment,
9438 but also on the size.
9441 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
9442 of exps and logs in the use_exp2 case.
9444 2018-02-13 Jeff Law <law@redhat.com>
9446 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
9449 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
9452 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
9456 * common.opt (gas-loc-support, gas-locview-support): New.
9457 (ginline-points, ginternal-reset-location-views): New.
9458 * doc/invoke.texi: Document them. Use @itemx where intended.
9459 (gvariable-location-views): Adjust.
9460 * target.def (reset_location_view): New.
9461 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
9462 (TARGET_RESET_LOCATION_VIEW): New.
9463 * doc/tm.texi: Rebuilt.
9464 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
9465 (dwarf2out_default_as_locview_support): New.
9466 (output_asm_line_debug_info): Use option variables.
9467 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
9468 (output_loc_list): Likewise.
9469 (add_high_low_attributes): Check option variables.
9470 Don't output entry view attribute in strict mode.
9471 (gen_inlined_subroutine_die): Check option variables.
9472 (dwarf2out_inline_entry): Likewise.
9473 (init_sections_and_labels): Likewise.
9474 (dwarf2out_early_finish): Likewise.
9475 (maybe_reset_location_view): New, from...
9476 (dwarf2out_var_location): ... here. Call it.
9477 * debug.h (dwarf2out_default_as_loc_support): Declare.
9478 (dwarf2out_default_as_locview_support): Declare.
9479 * hooks.c (hook_int_rtx_insn_0): New.
9480 * hooks.h (hook_int_rtx_insn_0): Declare.
9481 * toplev.c (process_options): Take -gas-loc-support and
9482 -gas-locview-support from dwarf2out. Enable
9483 -gvariable-location-views by default only with locview
9484 assembler support. Enable -ginternal-reset-location-views by
9485 default only if the target defines the corresponding hook.
9486 Enable -ginline-points by default if location views are
9487 enabled; force it disabled if statement frontiers are
9489 * tree-inline.c (expand_call_inline): Check option variables.
9490 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9492 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
9494 PR tree-optimization/84321
9495 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
9496 handling. Also check whether the anti-range contains any values
9497 that satisfy the mask; switch to a VR_RANGE if not.
9499 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
9502 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
9504 2018-02-13 Martin Jambor <mjambor@suse.cz>
9507 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
9508 of call statements, also set location of a load to a temporary.
9510 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
9512 * config/rl78/rl78.c (add_vector_labels): New function.
9513 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
9514 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
9515 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
9516 which checks that no arguments are passed.
9517 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
9518 * doc/extend.texi: Documentation for the new attribute.
9520 2018-02-13 Andreas Schwab <schwab@suse.de>
9522 * config/riscv/linux.h (CPP_SPEC): Define.
9524 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9527 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
9528 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
9529 OPTION_MASK_ISA_AES as first argument to def_builtin_const
9530 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
9531 instead of OPTION_MASK_ISA_PCLMUL as first argument to
9532 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
9533 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
9534 temporarily for AES and PCLMUL builtins.
9536 PR tree-optimization/84339
9537 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
9538 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
9542 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
9543 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
9544 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
9546 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
9548 * omp-simd-clone.h: New file.
9549 * omp-simd-clone.c: Include omp-simd-clone.h.
9550 (expand_simd_clones): No longer static.
9551 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
9552 cgraph.h and omp-simd-clone.h.
9553 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
9554 (vect_recog_widen_shift_pattern): Formatting fix.
9555 (vect_pattern_recog_1): Don't check optab for calls.
9558 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
9559 operands[2] into a REG before using gen_lowpart on it.
9561 2018-02-12 Jeff Law <law@redhat.com>
9564 * config/sh/sh.c (find_barrier): Consider a sibling call
9567 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
9568 successfully back substituting a reg.
9570 2018-02-12 Richard Biener <rguenther@suse.de>
9572 PR tree-optimization/84037
9573 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
9574 parameter, move visited init to caller.
9575 (vect_slp_analyze_operations): Separate cost from validity
9576 check, initialize visited once for all instances.
9577 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
9579 * tree-vect-stmts.c (vect_model_simple_cost): Make early
9581 (vect_model_promotion_demotion_cost): Likewise.
9582 (vectorizable_bswap): Guard cost modeling with !slp_node
9583 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
9585 (vectorizable_call): Likewise.
9586 (vectorizable_conversion): Likewise.
9587 (vectorizable_assignment): Likewise.
9588 (vectorizable_shift): Likewise.
9589 (vectorizable_operation): Likewise.
9590 (vectorizable_store): Likewise.
9591 (vectorizable_load): Likewise.
9592 (vectorizable_condition): Likewise.
9593 (vectorizable_comparison): Likewise.
9595 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
9598 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
9599 (ASAN_MARK): Fix fnspec to account for return value, change pointer
9600 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
9602 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9605 * params.def (inline-min-speedup): Increase from 8 to 15.
9606 (max-inline-insns-auto): Decrease from 40 to 30.
9607 * ipa-split.c (consider_split): Add some buffer for function to
9608 be considered inlining candidate.
9609 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
9612 2018-02-12 Richard Biener <rguenther@suse.de>
9614 PR tree-optimization/84037
9615 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
9616 matched stmts if we cannot swap the non-matched ones.
9618 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9620 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
9621 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
9622 _mm_maskz_scalef_round_ss): New intrinsics.
9623 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
9624 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
9625 __builtin_ia32_scalefss_round): Remove.
9626 (__builtin_ia32_scalefsd_mask_round,
9627 __builtin_ia32_scalefss_mask_round): New intrinsics.
9628 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
9629 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
9630 ((match_operand:VF_128 2 "<round_nimm_predicate>"
9631 "<round_constraint>")): Changed to ...
9632 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
9633 "<round_scalar_constraint>")): ... this.
9634 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
9635 %0, %1, %2<round_op3>}"): Changed to ...
9636 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
9637 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
9638 %2<round_scalar_mask_op3>}"): ... this.
9639 * config/i386/subst.md (round_scalar_nimm_predicate): New.
9641 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9643 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
9644 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
9645 (_mm_maskz_sqrt_round_ss): New intrinsics.
9646 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
9647 (__builtin_ia32_sqrtsd_mask_round)
9648 (__builtin_ia32_sqrtss_mask_round): New builtins.
9649 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
9650 (__builtin_ia32_sqrtss_round): Remove.
9651 (__builtin_ia32_sqrtsd_mask_round)
9652 (__builtin_ia32_sqrtss_mask_round): New builtins.
9653 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
9654 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
9655 ((match_operand:VF_128 1 "vector_operand"
9656 "xBm,<round_constraint>")): Changed to ...
9657 ((match_operand:VF_128 1 "vector_operand"
9658 "xBm,<round_scalar_constraint>")): ... this.
9659 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
9660 %0, %2, %<iptr>1<round_op3>}): Changed to ...
9661 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
9662 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
9663 %<iptr>1<round_scalar_mask_op3>}): ... this.
9664 ((set_attr "prefix" "<round_prefix>")): Changed to ...
9665 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
9667 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
9670 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
9671 Cast vec_cmpeq result to correct type.
9672 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
9673 Cast vec_cmpgt result to correct type.
9675 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
9677 * final.c (final_scan_insn_1): Renamed from...
9678 (final_scan_insn): ... this. New wrapper, to recover
9679 seen from the outermost call in recursive ones.
9680 * config/sparc/sparc.c (output_return): Drop seen from call.
9681 (output_sibcall): Likewise.
9682 * config/visium/visium.c (output_branch): Likewise.
9684 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
9686 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
9689 2018-02-10 Alan Modra <amodra@gmail.com>
9692 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
9693 Specify LR as an input.
9695 2018-02-10 Jakub Jelinek <jakub@redhat.com>
9698 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
9699 remove_member_access_dummy_vars): New functions.
9700 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
9701 lower_omp_1, execute_lower_omp): Use them.
9703 PR rtl-optimization/84308
9704 * shrink-wrap.c (spread_components): Release todo vector.
9706 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
9708 PR rtl-optimization/57193
9709 * ira-color.c (struct allocno_color_data): Add member
9710 conflict_allocno_hard_prefs.
9711 (update_conflict_allocno_hard_prefs): New.
9712 (bucket_allocno_compare_func): Add a preference based on
9713 conflict_allocno_hard_prefs.
9714 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
9715 (color_allocnos): Remove a dead code. Initiate
9716 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
9718 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9721 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
9722 constraint from =wa to wa. Avoid a subreg on the output operand,
9723 instead use a pseudo and subreg it in a move.
9724 (p9_xxbrd_<mode>): Changed to ...
9725 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
9726 (p9_xxbrd_v2df): New expander.
9727 (p9_xxbrw_<mode>): Changed to ...
9728 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
9729 (p9_xxbrw_v4sf): New expander.
9731 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9733 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
9735 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
9738 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
9739 multiply in 32-bit mode.
9740 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
9741 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
9744 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9746 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
9747 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
9748 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
9749 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
9751 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
9754 * dwarf2out.c (is_trivial_indirect_ref): New function.
9755 (dwarf2out_late_global_decl): Do not generate a location
9756 attribute for variables that have a non-trivial DECL_VALUE_EXPR
9757 and that are not defined in the current unit.
9759 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
9761 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
9762 instead of a libcall for UNORDERED.
9764 2018-02-09 Tamar Christina <tamar.christina@arm.com>
9767 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
9768 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
9770 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9773 * config/s390/s390.c (s390_set_current_function): Invoke
9774 s390_indirect_branch_settings also if fndecl didn't change.
9776 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
9778 * config/rs6000/rs6000.md (blockage): Set length to zero.
9780 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
9782 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
9784 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9787 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
9788 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
9792 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
9793 PARALLEL incoming that failed vt_get_decl_and_offset check.
9796 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
9797 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
9799 (get_variable_section): For decls in named .bss* sections pass true as
9800 second argument to bss_initializer_p.
9802 2018-02-09 Marek Polacek <polacek@redhat.com>
9803 Jakub Jelinek <jakub@redhat.com>
9806 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
9807 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
9808 Sync some changes from cxx_fold_indirect_ref.
9810 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
9812 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
9814 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
9815 (BLOCK_INLINE_ENTRY_LABEL): New.
9816 (dwarf2out_var_location): Disregard inline entry markers.
9817 (inline_entry_data): New struct.
9818 (inline_entry_data_hasher): New hashtable type.
9819 (inline_entry_data_hasher::hash): New.
9820 (inline_entry_data_hasher::equal): New.
9821 (inline_entry_data_table): New variable.
9822 (add_high_low_attributes): Add DW_AT_entry_pc and
9823 DW_AT_GNU_entry_view attributes if a pending entry is found
9824 in inline_entry_data_table. Add old entry_pc attribute only
9825 if debug nonbinding markers are disabled.
9826 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
9827 markers are enabled.
9828 (block_within_block_p, dwarf2out_inline_entry): New.
9829 (dwarf2out_finish): Check that no entries remained in
9830 inline_entry_data_table.
9831 * final.c (reemit_insn_block_notes): Handle inline entry notes.
9832 (final_scan_insn, notice_source_line): Likewise.
9833 (rest_of_clean_state): Skip inline entry markers.
9834 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
9836 * gimple.c (gimple_build_debug_inline_entry): New.
9837 * gimple.h (enum gimple_debug_subcode): Add
9838 GIMPLE_DEBUG_INLINE_ENTRY.
9839 (gimple_build_debug_inline_entry): Declare.
9840 (gimple_debug_inline_entry_p): New.
9841 (gimple_debug_nonbind_marker_p): Adjust.
9842 * insn-notes.def (INLINE_ENTRY): New.
9843 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
9844 inline entry marker notes.
9845 (print_insn): Likewise.
9846 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
9847 (INSN_DEBUG_MARKER_KIND): Likewise.
9848 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
9849 * tree-inline.c (expand_call_inline): Build and insert
9850 debug_inline_entry stmt.
9851 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
9852 inline entry blocks early, if nonbind markers are enabled.
9853 (dump_scope_block): Dump fragment info.
9854 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
9855 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
9856 (gimple_build_debug_inline_entry): New.
9857 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
9858 Enable/disable inline entry points too.
9859 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
9860 (DEBUG_INSN): Describe inline entry markers.
9862 * common.opt (gvariable-location-views): New.
9863 (gvariable-location-views=incompat5): New.
9864 * config.in: Rebuilt.
9865 * configure: Rebuilt.
9866 * configure.ac: Test assembler for view support.
9867 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
9868 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
9869 * dwarf2out.c (var_loc_view): New typedef.
9870 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
9871 (dwarf2out_locviews_in_attribute): New.
9872 (dwarf2out_locviews_in_loclist): New.
9873 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
9874 (enum dw_line_info_opcode): Add LI_adv_address.
9875 (struct dw_line_info_table): Add view.
9876 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
9877 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
9878 (zero_view_p): New variable.
9879 (ZERO_VIEW_P): New macro.
9880 (output_asm_line_debug_info): New.
9881 (struct var_loc_node): Add view.
9882 (add_AT_view_list, AT_loc_list): New.
9883 (add_var_loc_to_decl): Add view param. Test it against last.
9884 (new_loc_list): Add view params. Record them.
9885 (AT_loc_list_ptr): Handle loc and view lists.
9886 (view_list_to_loc_list_val_node): New.
9887 (print_dw_val): Handle dw_val_class_view_list.
9888 (size_of_die): Likewise.
9889 (value_format): Likewise.
9890 (loc_list_has_views): New.
9891 (gen_llsym): Set vl_symbol too.
9892 (maybe_gen_llsym, skip_loc_list_entry): New.
9893 (dwarf2out_maybe_output_loclist_view_pair): New.
9894 (output_loc_list): Output view list or entries too.
9895 (output_view_list_offset): New.
9896 (output_die): Handle dw_val_class_view_list.
9897 (output_dwarf_version): New.
9898 (output_compilation_unit_header): Use it.
9899 (output_skeleton_debug_sections): Likewise.
9900 (output_rnglists, output_line_info): Likewise.
9901 (output_pubnames, output_aranges): Update version comments.
9902 (output_one_line_info_table): Output view numbers in asm comments.
9903 (dw_loc_list): Determine current endview, pass it to new_loc_list.
9904 Call maybe_gen_llsym.
9905 (loc_list_from_tree_1): Adjust.
9906 (add_AT_location_description): Create view list attribute if
9907 needed, check it's absent otherwise.
9908 (convert_cfa_to_fb_loc_list): Adjust.
9909 (maybe_emit_file): Call output_asm_line_debug_info for test.
9910 (dwarf2out_var_location): Reset views as needed. Precompute
9911 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
9912 attribute. Set view.
9913 (new_line_info_table): Reset next view.
9914 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
9915 (dwarf2out_source_line): Likewise. Output view resets and labels to
9916 the assembler, or select appropriate line info opcodes.
9917 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
9918 (optimize_string_length): Catch it. Adjust.
9919 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
9920 dw_val_class_view_list, and remove it if no longer needed.
9921 (hash_loc_list): Hash view numbers.
9922 (loc_list_hasher::equal): Compare them.
9923 (optimize_location_lists): Check whether a view list symbol is
9924 needed, and whether the locview attribute is present, and
9925 whether they match. Remove the locview attribute if no longer
9927 (index_location_lists): Call skip_loc_list_entry for test.
9928 (dwarf2out_finish): Call output_asm_line_debug_info for test.
9929 Use output_dwarf_version.
9930 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
9931 (struct dw_val_node): Add val_view_list.
9932 * final.c (SEEN_NEXT_VIEW): New.
9933 (set_next_view_needed): New.
9934 (clear_next_view_needed): New.
9935 (maybe_output_next_view): New.
9936 (final_start_function): Rename to...
9937 (final_start_function_1): ... this. Take pointer to FIRST,
9938 add SEEN parameter. Emit param bindings in the initial view.
9939 (final_start_function): Reintroduce SEEN-less interface.
9940 (final): Rename to...
9941 (final_1): ... this. Take SEEN parameter. Output final pending
9942 next view at the end.
9943 (final): Reintroduce seen-less interface.
9944 (final_scan_insn): Output pending next view before switching
9945 sections or ending a block. Mark the next view as needed when
9946 outputting variable locations. Notify debug backend of section
9947 changes, and of location view changes.
9948 (rest_of_handle_final): Adjust.
9949 * toplev.c (process_options): Autodetect value for debug variable
9950 location views option. Warn on incompat5 without -gdwarf-5.
9951 * doc/invoke.texi (gvariable-location-views): New.
9952 (gvariable-location-views=incompat5): New.
9953 (gno-variable-location-views): New.
9955 2018-02-08 David Malcolm <dmalcolm@redhat.com>
9957 PR tree-optimization/84136
9958 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
9959 that the result of find_edge is non-NULL.
9961 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9964 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
9965 storing integer register in SImode. Fix cost of 256 and 512
9966 byte aligned SSE register store.
9968 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9970 * config/i386/i386.c (ix86_multiplication_cost): Fix
9971 multiplication cost for TARGET_AVX512DQ.
9973 2018-02-08 Marek Polacek <polacek@redhat.com>
9975 PR tree-optimization/84238
9976 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
9979 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9981 PR tree-optimization/84265
9982 * tree-vect-stmts.c (vectorizable_store): Don't treat
9983 VMAT_CONTIGUOUS accesses as grouped.
9984 (vectorizable_load): Likewise.
9986 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9988 PR tree-optimization/81635
9989 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
9990 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
9991 (test_round_for_mask): New functions.
9992 (wide_int_cc_tests): Call test_round_for_mask.
9993 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
9994 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
9995 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
9996 range returned by get_range_info.
9998 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10001 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
10002 * symtab.c: Include builtins.h
10003 (symtab_node::output_to_lto_symbol_table_p): Move here
10004 from lto-streamer-out.c:output_symbol_p.
10005 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
10006 (output_symbol_p): Move all logic to symtab.c
10007 (produce_symtab): Update.
10009 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10011 * config/s390/s390-opts.h (enum indirect_branch): Define.
10012 * config/s390/s390-protos.h (s390_return_addr_from_memory)
10013 (s390_indirect_branch_via_thunk)
10014 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
10015 (enum s390_indirect_branch_type): Define.
10016 * config/s390/s390.c (struct s390_frame_layout, struct
10017 machine_function): Remove.
10018 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10019 (indirect_branch_table_label_no, indirect_branch_table_name):
10021 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10022 (enum s390_indirect_branch_option): Define.
10023 (s390_return_addr_from_memory): New function.
10024 (s390_handle_string_attribute): New function.
10025 (s390_attribute_table): Add new attribute handler.
10026 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10027 (s390_indirect_branch_via_thunk): New function.
10028 (s390_indirect_branch_via_inline_thunk): New function.
10029 (s390_function_ok_for_sibcall): When jumping via thunk disallow
10030 sibling call optimization for non z10 compiles.
10031 (s390_emit_call): Force indirect branch target to be a single
10032 register. Add r1 clobber for non-z10 compiles.
10033 (s390_emit_epilogue): Emit return jump via return_use expander.
10034 (s390_reorg): Handle JUMP_INSNs as execute targets.
10035 (s390_option_override_internal): Perform validity checks for the
10036 new command line options.
10037 (s390_indirect_branch_attrvalue): New function.
10038 (s390_indirect_branch_settings): New function.
10039 (s390_set_current_function): Invoke s390_indirect_branch_settings.
10040 (s390_output_indirect_thunk_function): New function.
10041 (s390_code_end): Implement target hook.
10042 (s390_case_values_threshold): Implement target hook.
10043 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10045 * config/s390/s390.h (struct s390_frame_layout)
10046 (struct machine_function): Move here from s390.c.
10047 (TARGET_INDIRECT_BRANCH_NOBP_RET)
10048 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10049 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10050 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10051 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10052 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10053 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10054 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10055 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10056 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10057 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10058 (mnemonic attribute): Add values which aren't recognized
10060 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10061 pattern for branch conversion. Fix mnemonic attribute.
10062 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10063 indirect branch via thunk if requested.
10064 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10065 ("*indirect_jump"): Disable for branch conversion using out of
10067 ("indirect_jump_via_thunk<mode>_z10")
10068 ("indirect_jump_via_thunk<mode>")
10069 ("indirect_jump_via_inlinethunk<mode>_z10")
10070 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10071 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10072 ("casesi_jump_via_inlinethunk<mode>_z10")
10073 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10074 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10075 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10076 ("*indirect2_jump"): Disable for branch conversion.
10077 ("casesi_jump"): Turn into expander and expand patterns for branch
10079 ("return_use"): New expander.
10080 ("*return"): Emit return via thunk and rename it to ...
10081 ("*return<mode>"): ... this one.
10082 * config/s390/s390.opt: Add new options and and enum for the
10085 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10087 * lra-constraints.c (match_reload): Unconditionally use
10088 gen_lowpart_SUBREG, rather than selecting between that
10089 and equivalent gen_rtx_SUBREG code.
10091 2018-02-08 Richard Biener <rguenther@suse.de>
10093 PR tree-optimization/84233
10094 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10095 changed flag instead of boguously re-using phi_inserted.
10097 2018-02-08 Martin Jambor <mjambor@suse.cz>
10099 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10100 static local variables.
10102 2018-02-08 Richard Biener <rguenther@suse.de>
10104 PR tree-optimization/84278
10105 * tree-vect-stmts.c (vectorizable_store): When looking for
10106 smaller vector types to perform grouped strided loads/stores
10107 make sure the mode is supported by the target.
10108 (vectorizable_load): Likewise.
10110 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10112 * config/aarch64/aarch64.c (aarch64_components_for_bb):
10113 Increase LDP/STP opportunities by adding adjacent callee-saves.
10115 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10117 PR rtl-optimization/84068
10118 PR rtl-optimization/83459
10119 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10121 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
10123 PR tree-optimization/84224
10124 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10125 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10126 non-zero arguments.
10128 2018-02-07 Iain Sandoe <iain@codesourcery.com>
10131 * config/rs6000/altivec.md (*restore_world): Remove LR use.
10132 * config/rs6000/predicates.md (restore_world_operation): Adjust op
10133 count, remove one USE.
10135 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
10137 * doc/install.texi (Configuration): Document the
10138 --with-long-double-format={ibm,ieee} PowerPC configuration
10142 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10143 Convert from define_expand to be define_insn_and_split. Rework
10144 float/double/_Float128 conversions to QI/HI/SImode to work with
10145 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
10146 conversions to QI/HImode types did a store and then a load to
10147 truncate the value. For conversions to VSX registers, don't split
10148 the insn, instead emit the code directly. Use the code iterator
10149 any_fix to combine signed and unsigned conversions.
10150 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10151 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10152 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10153 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10154 (fix_<mode>di2_hw): Likewise.
10155 (fixuns_<mode>di2_hw): Likewise.
10156 (fix_<mode>si2_hw): Likewise.
10157 (fixuns_<mode>si2_hw): Likewise.
10158 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10159 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10160 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10161 fix<uns>_trunc<SFDF:mode>si2_p8.
10162 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10163 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10164 (fix<uns>_<mode>_mem): Likewise.
10165 (fctiw<u>z_<mode>_mem): Likewise.
10166 (fix<uns>_<mode>_mem): Likewise.
10167 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10168 the register allocator from doing a direct move to the GPRs to do
10169 a store, and instead use the ISA 3.0 store byte/half-word from
10170 vector register instruction. For IEEE 128-bit floating point,
10171 also optimize stores of 32-bit ints.
10172 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10174 2018-02-07 Alan Hayward <alan.hayward@arm.com>
10176 * genextract.c (push_pathstr_operand): New function to support
10178 (walk_rtx): Call push_pathstr_operand.
10179 (print_path): Support [a-zA-Z].
10181 2018-02-07 Richard Biener <rguenther@suse.de>
10183 PR tree-optimization/84037
10184 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
10185 (cse_and_gimplify_to_preheader): Declare.
10186 (vect_get_place_in_interleaving_chain): Likewise.
10187 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10189 (_loop_vec_info::~_loop_vec_info): Delete it.
10190 (cse_and_gimplify_to_preheader): New function.
10191 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
10192 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
10193 (vectorizable_load): Likewise. For grouped stores always base
10194 the IV on the first element.
10195 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
10196 condition before gimplifying.
10198 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10200 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
10201 *DIV_EXPR and *MOD_EXPR.
10203 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
10206 * config/i386/i386.c (ix86_option_override_internal): Mask out
10207 the CF_SET bit when checking -fcf-protection.
10209 2018-02-07 Tom de Vries <tom@codesourcery.com>
10212 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
10215 2018-02-07 Richard Biener <rguenther@suse.de>
10217 PR tree-optimization/84204
10218 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
10221 PR tree-optimization/84205
10222 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
10223 special-case isl_ast_op_zdiv_r.
10225 PR tree-optimization/84223
10226 * graphite-scop-detection.c (gather_bbs::before_dom_children):
10227 Only add conditions from within the region.
10228 (gather_bbs::after_dom_children): Adjust.
10230 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
10233 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
10234 * config/avr/avr.md: Only post-reload split REG-REG moves if
10235 either register is GENERAL_REG_P.
10237 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10239 PR tree-optimization/84235
10240 * tree-ssa-scopedtables.c
10241 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
10242 if the subtraction is performed in floating point type where NaNs are
10243 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
10244 build 1. Formatting fix.
10246 2018-02-06 Jakub Jelinek <jakub@redhat.com>
10249 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
10250 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
10251 and skip it regardless of bb boundaries. Use CALL_P macro,
10252 don't test INSN_P (insn) together with CALL_P or JUMP_P check
10253 unnecessarily, formatting fix.
10255 2018-02-06 Michael Collison <michael.collison@arm.com>
10257 * config/arm/thumb2.md:
10258 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
10259 (*thumb_mov_notscc): Ditto.
10261 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
10264 * config/rs6000/rs6000.md (su code attribute): Use "u" for
10265 unsigned_fix, not "s".
10267 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10269 * configure.ac (gcc_fn_eh_frame_ro): New function.
10270 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
10271 correct .eh_frame permissions.
10272 * configure: Regenerate.
10274 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
10276 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
10277 irrelevant options.
10279 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10281 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10282 Display warning message for -mno-speculate-indirect-jumps.
10284 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
10286 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
10288 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
10290 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
10292 PR tree-optimization/84225
10293 * tree-eh.c (find_trapping_overflow): Only call
10294 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
10296 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10299 * config/i386/i386.c: Reimplement the check of possible options
10300 -mibt/-mshstk conbination. Change error messages.
10301 * doc/invoke.texi: Fix a typo: remove extra '='.
10303 2018-02-06 Marek Polacek <polacek@redhat.com>
10305 PR tree-optimization/84228
10306 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
10308 2018-02-06 Tamar Christina <tamar.christina@arm.com>
10311 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
10312 emitted arch directives.
10313 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
10314 __ARM_FEATURE_COPROC before changing architectures.
10316 2018-02-06 Richard Biener <rguenther@suse.de>
10318 * config/i386/i386.c (print_reg): Fix typo.
10319 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
10321 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
10323 * configure: Regenerate.
10325 2018-02-05 Martin Sebor <msebor@redhat.com>
10327 PR tree-optimization/83369
10328 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
10331 2018-02-05 Martin Liska <mliska@suse.cz>
10333 * doc/invoke.texi: Cherry-pick upstream r323995.
10335 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
10337 * ira.c (ira_init_register_move_cost): Adjust comment.
10339 2018-02-05 Martin Liska <mliska@suse.cz>
10341 PR gcov-profile/84137
10342 * doc/gcov.texi: Fix typo in documentation.
10344 2018-02-05 Martin Liska <mliska@suse.cz>
10346 PR gcov-profile/83879
10347 * doc/gcov.texi: Document necessity of --dynamic-list-data when
10348 using dlopen functionality.
10350 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
10352 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
10353 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
10354 _mm_maskz_range_ss, _mm_mask_range_round_ss,
10355 _mm_maskz_range_round_ss): New intrinsics.
10356 (__builtin_ia32_rangesd128_round)
10357 (__builtin_ia32_rangess128_round): Remove.
10358 (__builtin_ia32_rangesd128_mask_round,
10359 __builtin_ia32_rangess128_mask_round): New builtins.
10360 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
10361 __builtin_ia32_rangess128_round): Remove.
10362 (__builtin_ia32_rangesd128_mask_round,
10363 __builtin_ia32_rangess128_mask_round): New builtins.
10364 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
10365 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
10366 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
10367 "<round_saeonly_constraint>")): Changed to ...
10368 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
10369 "<round_saeonly_scalar_constraint>")): ... this.
10370 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
10371 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
10372 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
10373 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
10374 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
10376 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
10378 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
10380 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
10381 Remove all values except native, 8540 and 8548.
10383 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
10385 * config/i386/i386.c (ix86_output_function_return): Pass
10386 INVALID_REGNUM, instead of -1, as invalid register number to
10387 indirect_thunk_name and output_indirect_thunk.
10389 2018-02-02 Julia Koval <julia.koval@intel.com>
10391 * config.gcc: Add -march=icelake.
10392 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
10393 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
10394 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
10395 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
10396 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
10397 (processor_target_table): Add icelake.
10398 (ix86_option_override_internal): Handle new PTAs.
10399 (get_builtin_code_for_version): Handle icelake.
10400 (M_INTEL_COREI7_ICELAKE): New.
10401 (fold_builtin_cpu): Handle icelake.
10402 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
10403 * doc/invoke.texi: Add -march=icelake.
10405 2018-02-02 Julia Koval <julia.koval@intel.com>
10407 * config/i386/i386.c (ix86_option_override_internal): Change flags type
10408 to wide_int_bitmask.
10409 * wide-int-bitmask.h: New.
10411 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10414 * config/i386/i386.md: Replace Pmode with word_mode in
10415 builtin_setjmp_setup and builtin_longjmp to support x32.
10417 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
10421 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
10423 (rs6000_supported_cpu_names): New static variable.
10424 (linux_cpu_translation_table): Likewise.
10425 (elf_platform) <cpu>: Define new static variable and use it.
10426 Translate kernel AT_PLATFORM name to canonical name if needed.
10427 Error if platform name is unknown.
10429 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
10432 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
10434 2018-02-01 Jeff Law <law@redhat.com>
10437 * config/i386/i386.c (release_scratch_register_on_entry): Add new
10438 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
10439 the scratch if RELEASE_VIA_POP is false.
10440 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
10441 If we have to save a temporary register, decrement SIZE appropriately.
10442 Pass new arguments to release_scratch_register_on_entry.
10443 (ix86_adjust_stack_and_probe): Likewise.
10444 (ix86_emit_probe_stack_range): Pass new arguments to
10445 release_scratch_register_on_entry.
10447 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
10449 PR rtl-optimization/84157
10450 * combine.c (change_zero_ext): Use REG_P predicate in
10451 front of HARD_REGISTER_P predicate.
10453 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
10455 * config/avr/avr.c (avr_option_override): Move disabling of
10456 -fdelete-null-pointer-checks to...
10457 * common/config/avr/avr-common.c (avr_option_optimization_table):
10460 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10462 PR tree-optimization/81635
10463 * tree-data-ref.c (split_constant_offset_1): For types that
10464 wrap on overflow, try to use range info to prove that wrapping
10467 2018-02-01 Renlin Li <renlin.li@arm.com>
10470 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
10471 TAILCALL_ADDR_REGS.
10472 (aarch64_register_move_cost): Likewise.
10473 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
10474 TAILCALL_ADDR_REGS.
10475 (REG_CLASS_NAMES): Likewise.
10476 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
10477 TAILCALL_ADDR_REGS. Remove IP registers.
10478 * config/aarch64/aarch64.md (Ucs): Update register constraint.
10480 2018-02-01 Richard Biener <rguenther@suse.de>
10482 * domwalk.h (dom_walker::dom_walker): Add additional constructor
10483 for specifying RPO order and allow NULL for that.
10484 * domwalk.c (dom_walker::dom_walker): Likewise.
10485 (dom_walker::walk): Handle NULL RPO order.
10486 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
10488 (rewrite_update_dom_walker): Likewise.
10489 (mark_def_dom_walker): Likewise.
10491 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10493 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
10494 (aarch64_maybe_expand_sve_subreg_move): Declare.
10495 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
10496 * config/aarch64/predicates.md (aarch64_any_register_operand): New
10498 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
10499 that are semantically a reverse operation.
10500 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
10501 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
10502 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
10504 (aarch64_can_change_mode_class): For big-endian, forbid changes
10505 between two SVE modes if they have different element sizes.
10507 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10509 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
10510 the TImode handling for big-endian targets.
10512 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10514 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
10515 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
10517 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
10518 Remove BSWAP handing for big-endian targets and use the form of
10519 LD1RQ appropariate for the mode.
10521 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10523 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
10524 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
10525 duplicated element.
10527 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10530 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
10531 check for operands that need to go through aarch64_sve_reload_be.
10533 2018-02-01 Jakub Jelinek <jakub@redhat.com>
10535 PR tree-optimization/81661
10536 PR tree-optimization/84117
10537 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
10538 * tree-eh.c: Include gimplify.h.
10539 (find_trapping_overflow, replace_trapping_overflow,
10540 rewrite_to_non_trapping_overflow): New functions.
10541 * tree-vect-loop.c: Include tree-eh.h.
10542 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
10543 * tree-data-ref.c: Include tree-eh.h.
10544 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
10546 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
10548 PR rtl-optimization/84123
10549 * combine.c (change_zero_ext): Check if hard register satisfies
10550 can_change_dest_mode before calling gen_lowpart_SUBREG.
10552 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
10555 * ira.c (ira_init_register_move_cost): Remove assert.
10557 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10559 PR rtl-optimization/84071
10560 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
10561 * doc/tm.texi: Regenerate.
10563 2018-01-31 Richard Biener <rguenther@suse.de>
10565 PR tree-optimization/84132
10566 * tree-data-ref.c (analyze_miv_subscript): Properly
10567 check whether evolution_function_is_affine_multivariate_p
10568 before calling gcd_of_steps_may_divide_p.
10570 2018-01-31 Julia Koval <julia.koval@intel.com>
10573 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
10574 * config/i386/i386.md (rdpid_rex64) New.
10575 (rdpid): Make 32bit only.
10577 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
10580 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
10581 an IDENTIFIER_NODE for FUNCTION_TYPE's.
10583 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10586 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10588 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10590 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10592 PR rtl-optimization/84071
10593 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
10594 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
10596 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10598 * config/arc/arc.c (arc_handle_aux_attribute): New function.
10599 (arc_attribute_table): Add 'aux' attribute.
10600 (arc_in_small_data_p): Consider aux like variables.
10601 (arc_is_aux_reg_p): New function.
10602 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
10603 (arc_get_aux_arg): New function.
10604 (prepare_move_operands): Handle aux-register access.
10605 (arc_handle_aux_attribute): New function.
10606 * doc/extend.texi (ARC Variable attributes): Add subsection.
10608 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10610 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
10611 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
10612 (arc_attribute_table): Add 'uncached' attribute.
10613 (arc_print_operand): Print '.di' flag for uncached memory
10615 (arc_in_small_data_p): Do not consider for small data the uncached
10617 (arc_is_uncached_mem_p): New function.
10618 * config/arc/predicates.md (compact_store_memory_operand): Check
10619 for uncached memory accesses.
10620 (nonvol_nonimm_operand): Likewise.
10621 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
10623 2018-01-31 Jakub Jelinek <jakub@redhat.com>
10626 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
10627 falign-loops=): Add Optimization flag.
10629 2018-01-30 Jeff Law <law@redhat.com>
10632 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
10633 INT_REGISTERS_SAVED. Check it prior to calling
10634 get_scratch_register_on_entry.
10635 (ix86_adjust_stack_and_probe): Similarly.
10636 (ix86_emit_probe_stack_range): Similarly.
10637 (ix86_expand_prologue): Corresponding changes.
10639 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10642 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
10643 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
10645 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
10648 * lra-constraints.c (curr_insn_transform): Process AND in the
10651 2018-01-30 Jakub Jelinek <jakub@redhat.com>
10653 PR rtl-optimization/83986
10654 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
10655 dependence against last_pending_memory_flush in addition to
10656 pending_jump_insns.
10658 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
10660 PR tree-optimization/81611
10661 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
10664 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10667 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
10670 2018-01-30 Richard Biener <rguenther@suse.de>
10671 Jakub Jelinek <jakub@redhat.com>
10673 PR tree-optimization/84111
10674 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
10675 inner loops added during recursion, as they don't have up-to-date
10678 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10681 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
10682 (can_inline_edge_by_limits_p): ... here.
10683 (can_early_inline_edge_p, check_callers,
10684 update_caller_keys, update_callee_keys, recursive_inlining,
10685 add_new_edges_to_heap, speculation_useful_p,
10686 inline_small_functions,
10687 inline_small_functions, flatten_function,
10688 inline_to_all_callers_1): Update.
10690 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10692 * profile-count.c (profile_count::combine_with_ipa_count): Handle
10695 2018-01-30 Richard Biener <rguenther@suse.de>
10697 PR tree-optimization/83008
10698 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
10699 invariant and constant vector uses in stmts when they need
10700 more than one stmt.
10702 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10705 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
10706 * configure: Regenerate.
10708 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10710 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
10712 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
10713 Use gen_rtx_REG rather than gen_lowpart.
10715 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10717 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
10718 rather than 0 when creating partial subregs.
10720 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10722 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
10725 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
10728 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
10729 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
10730 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
10731 flags. This restores the settings used before the 2017-07-24.
10732 Turning off pre increment/decrement/modify allows IVOPTS to
10733 optimize DF/SF loops where the index is an int.
10735 2018-01-29 Richard Biener <rguenther@suse.de>
10736 Kelvin Nilsen <kelvin@gcc.gnu.org>
10739 * tree-vect-stmts.c (vectorizable_call): Don't call
10740 targetm.vectorize_builtin_md_vectorized_function if callee is
10743 2018-01-22 Carl Love <cel@us.ibm.com>
10745 * doc/extend.tex: Fix typo in second arg in
10746 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
10748 2018-01-29 Richard Biener <rguenther@suse.de>
10750 PR tree-optimization/84086
10751 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
10752 (flush_ssaname_freelist): When SSA names were released reset
10753 the SCEV hash table.
10755 2018-01-29 Richard Biener <rguenther@suse.de>
10757 PR tree-optimization/84057
10758 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
10759 removed paths when removing edges.
10761 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
10763 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
10764 -mfunction-return=@var{choice}.
10766 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
10768 PR diagnostic/84034
10769 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
10770 Handle CR like TAB.
10771 (layout::print_source_line): Likewise.
10772 (test_get_line_width_without_trailing_whitespace): Add test cases.
10774 2018-01-27 Jakub Jelinek <jakub@redhat.com>
10776 PR middle-end/84040
10777 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
10780 2018-01-26 Jim Wilson <jimw@sifive.com>
10782 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
10784 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
10787 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10789 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
10790 and CMP + SUB-immediate -> SUBS.
10792 2018-01-26 Martin Sebor <msebor@redhat.com>
10794 PR tree-optimization/83896
10795 * tree-ssa-strlen.c (get_string_len): Rename...
10796 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
10797 Avoid assuming length is constant.
10798 (handle_char_store): Use HOST_WIDE_INT for string length.
10800 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
10803 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
10804 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
10806 2018-01-26 Richard Biener <rguenther@suse.de>
10808 PR rtl-optimization/84003
10809 * dse.c (record_store): Only record redundant stores when
10810 the earlier store aliases at least all accesses the later one does.
10812 2018-01-26 Jakub Jelinek <jakub@redhat.com>
10814 PR rtl-optimization/83985
10815 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
10816 REG_CFA_RESTORE insns.
10817 (delete_unmarked_insns): Don't ignore separate shrink wrapping
10818 REG_CFA_RESTORE insns here.
10821 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
10822 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
10824 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10826 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
10827 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
10828 (arc_init): Likewise.
10829 (arc_override_options): Likewise.
10830 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
10832 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
10834 * config/arc/arc.h (TARGET_DBNZ): Define.
10835 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
10836 properly set the tune attribute.
10837 (dbnz): Use TARGET_DBNZ guard.
10838 * config/arc/arc.opt (mtune): Add core3 option.
10840 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10842 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
10843 recognize new pic like addresses.
10844 (arc_delegitimize_address): Clean up.
10846 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10848 * config/arc/arc-arches.def: Option mrf16 valid for all
10850 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
10851 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
10852 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
10853 * config/arc/arc-tables.opt: Regenerate.
10854 * config/arc/arc.c (arc_conditional_register_usage): Handle
10855 reduced register file case.
10856 (arc_file_start): Set must have build attributes.
10857 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
10858 mrf16 option value.
10859 * config/arc/arc.opt (mrf16): Add new option.
10860 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
10861 * config/arc/genmultilib.awk: Handle new mrf16 option.
10862 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
10863 * config/arc/t-multilib: Regenerate.
10864 * doc/invoke.texi (ARC Options): Document mrf16 option.
10866 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10868 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
10869 * config/arc/arc.c (arc_handle_secure_attribute): New function.
10870 (arc_attribute_table): Add 'secure_call' attribute.
10871 (arc_print_operand): Print secure call operand.
10872 (arc_function_ok_for_sibcall): Don't optimize tail calls when
10874 (arc_is_secure_call_p): New function. * config/arc/arc.md
10875 (call_i): Add support for sjli instruction.
10876 (call_value_i): Likewise.
10877 * config/arc/constraints.md (Csc): New constraint.
10879 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10880 John Eric Martin <John.Martin@emmicro-us.com>
10882 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
10883 * config/arc/arc.c (_arc_jli_section): New struct.
10884 (arc_jli_section): New type.
10885 (rc_jli_sections): New static variable.
10886 (arc_handle_jli_attribute): New function.
10887 (arc_attribute_table): Add jli_always and jli_fixed attribute.
10888 (arc_file_end): New function.
10889 (TARGET_ASM_FILE_END): Define.
10890 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
10891 (arc_add_jli_section): New function.
10892 (jli_call_scan): Likewise.
10893 (arc_reorg): Call jli_call_scan.
10894 (arc_output_addsi): Remove 'S' from printing asm operand.
10895 (arc_is_jli_call_p): New function.
10896 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
10898 (movhi_insn): Likewise.
10899 (movsi_insn): Likewise.
10900 (movsi_set_cc_insn): Likewise.
10901 (loadqi_update): Likewise.
10902 (load_zeroextendqisi_update): Likewise.
10903 (load_signextendqisi_update): Likewise.
10904 (loadhi_update): Likewise.
10905 (load_zeroextendhisi_update): Likewise.
10906 (load_signextendhisi_update): Likewise.
10907 (loadsi_update): Likewise.
10908 (loadsf_update): Likewise.
10909 (movsicc_insn): Likewise.
10910 (bset_insn): Likewise.
10911 (bxor_insn): Likewise.
10912 (bclr_insn): Likewise.
10913 (bmsk_insn): Likewise.
10914 (bicsi3_insn): Likewise.
10915 (cmpsi_cc_c_insn): Likewise.
10916 (movsi_ne): Likewise.
10917 (movsi_cond_exec): Likewise.
10918 (clrsbsi2): Likewise.
10919 (norm_f): Likewise.
10929 (call_i): Remove 'S' asm letter, add jli instruction.
10930 (call_value_i): Likewise.
10931 * config/arc/arc.op (mjli-always): New option.
10932 * config/arc/constraints.md (Cji): New constraint.
10933 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
10935 (subsf3_fpx): Likewise.
10936 (mulsf3_fpx): Likewise.
10937 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
10939 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
10940 function attrbutes.
10941 * doc/invoke.texi (ARC): Document mjli-always option.
10943 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
10945 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
10946 avoid addition with 0 and use incw and decw where possible.
10948 2018-01-26 Richard Biener <rguenther@suse.de>
10950 PR tree-optimization/81082
10951 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
10952 association if it requires casting to unsigned.
10953 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
10954 from fold_plusminus_mult_expr to catch important cases late when
10955 range info is available.
10957 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10959 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
10960 * configure.ac (hidden_linkonce): New test.
10961 * configure: Regenerate.
10962 * config.in: Regenerate.
10964 2018-01-26 Julia Koval <julia.koval@intel.com>
10966 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
10967 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
10968 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
10969 _mm_mask_bitshuffle_epi64_mask): Fix type.
10970 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
10971 USI_FTYPE_V4DI_V4DI_USI): Remove.
10972 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
10973 __builtin_ia32_vpshufbitqmb256_mask,
10974 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
10975 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
10976 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
10978 2018-01-26 Alan Modra <amodra@gmail.com>
10981 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
10982 UNSPEC_VBPERMQ. Sort other unspecs.
10984 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
10986 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
10988 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
10990 PR middle-end/83055
10991 * predict.c (drop_profile): Do not push/pop cfun; update also
10993 (handle_missing_profiles): Fix logic looking for zero profiles.
10995 2018-01-25 Jakub Jelinek <jakub@redhat.com>
10997 PR middle-end/83977
10998 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
10999 on functions with #pragma omp declare simd or functions with simd
11001 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
11002 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
11003 Remove trailing \n from warning_at calls.
11005 2018-01-25 Tom de Vries <tom@codesourcery.com>
11008 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11009 for neutered workers.
11011 2018-01-24 Joseph Myers <joseph@codesourcery.com>
11014 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
11015 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11017 2018-01-24 Jeff Law <law@redhat.com>
11020 * i386.c (get_probe_interval): Move to earlier point.
11021 (ix86_compute_frame_layout): If -fstack-clash-protection and
11022 the frame is larger than the probe interval, then use pushes
11023 to save registers rather than reg->mem moves.
11024 (ix86_expand_prologue): Remove conditional for int_registers_saved
11027 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
11030 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11031 min/max for never referenced object.
11033 2018-01-24 Jakub Jelinek <jakub@redhat.com>
11035 PR middle-end/83977
11036 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11038 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11039 attributes from DECL_ATTRIBUTES (decl) without affecting
11040 DECL_ATTRIBUTES (current_function_decl).
11041 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11042 functions with non-NULL DECL_ABSTRACT_ORIGIN.
11044 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
11046 PR tree-optimization/83979
11047 * fold-const.c (fold_comparison): Use constant_boolean_node
11048 instead of boolean_{true,false}_node.
11050 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
11052 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11055 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11057 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11058 Simplify the clause that sets the length attribute.
11059 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11060 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11061 clause that sets the length attribute.
11062 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11064 2018-01-24 Tom de Vries <tom@codesourcery.com>
11067 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11068 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11069 Add strict parameter.
11070 (prevent_branch_around_nothing): Insert dummy insn between branch to
11071 label and label with no ptx insn inbetween.
11072 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11074 2018-01-24 Tom de Vries <tom@codesourcery.com>
11077 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11078 for neutered threads in warp.
11079 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11081 2018-01-24 Richard Biener <rguenther@suse.de>
11083 PR tree-optimization/83176
11084 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11087 2018-01-24 Richard Biener <rguenther@suse.de>
11089 PR tree-optimization/82819
11090 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11091 code generating pluses that are no-ops in the target precision.
11093 2018-01-24 Richard Biener <rguenther@suse.de>
11095 PR middle-end/84000
11096 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11098 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11100 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11101 to merge probabilities.
11102 * predict.c (probably_never_executed): Also mark as cold functions
11103 with global 0 profile and guessed local profile.
11104 * profile-count.c (profile_probability::combine_with_count): New
11106 * profile-count.h (profile_probability::operator*,
11107 profile_probability::operator*=, profile_probability::operator/,
11108 profile_probability::operator/=): Reduce precision to adjusted
11109 and set value to guessed on contradictory divisions.
11110 (profile_probability::combine_with_freq): Remove.
11111 (profile_probability::combine_wiht_count): Declare.
11112 (profile_count::force_nonzero):: Set to adjusted.
11113 (profile_count::probability_in):: Set quality to adjusted.
11114 * tree-ssa-tail-merge.c (replace_block_by): Use
11115 combine_with_count.
11117 2018-01-23 Andrew Waterman <andrew@sifive.com>
11118 Jim Wilson <jimw@sifive.com>
11120 * config/riscv/riscv.c (riscv_stack_boundary): New.
11121 (riscv_option_override): Set riscv_stack_boundary. Handle
11122 riscv_preferred_stack_boundary_arg.
11123 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11124 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11125 (STACK_BOUNDARY): Set to riscv_stack_boundary.
11126 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11127 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11128 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11130 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
11133 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11134 of struct ix86_frame.
11135 (ix86_expand_epilogue): Likewise. Add a local variable for
11136 the reg_save_offset field in struct ix86_frame.
11138 2018-01-23 Bin Cheng <bin.cheng@arm.com>
11140 PR tree-optimization/82604
11141 * tree-loop-distribution.c (enum partition_kind): New enum item
11142 PKIND_PARTIAL_MEMSET.
11143 (partition_builtin_p): Support above new enum item.
11144 (generate_code_for_partition): Ditto.
11145 (compute_access_range): Differentiate cases that equality can be
11146 proven at all loops, the innermost loops or no loops.
11147 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11148 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
11149 (finalize_partitions, distribute_loop): Don't fuse partition of
11150 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11151 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11152 parloop is enabled.
11154 2018-01-23 Martin Liska <mliska@suse.cz>
11156 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11157 order to ignore the predictor.
11158 (PRED_POLYMORPHIC_CALL): Likewise.
11159 (PRED_RECURSIVE_CALL): Likewise.
11161 2018-01-23 Martin Liska <mliska@suse.cz>
11163 * tree-profile.c (tree_profiling): Print function header to
11164 aware reader which function we are working on.
11165 * value-prof.c (gimple_find_values_to_profile): Do not print
11166 not interesting value histograms.
11168 2018-01-23 Martin Liska <mliska@suse.cz>
11170 * profile-count.h (enum profile_quality): Add
11171 profile_uninitialized as the first value. Do not number values
11172 as they are zero based.
11173 (profile_count::verify): Update sanity check.
11174 (profile_probability::verify): Likewise.
11176 2018-01-23 Nathan Sidwell <nathan@acm.org>
11178 * doc/invoke.texi (ffor-scope): Deprecate.
11180 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11182 PR tree-optimization/83510
11183 * domwalk.c (set_all_edges_as_executable): New function.
11184 (dom_walker::dom_walker): Convert bool param
11185 "skip_unreachable_blocks" to enum reachability. Move setup of
11186 edge flags to set_all_edges_as_executable and only do it when
11187 reachability is REACHABLE_BLOCKS.
11188 * domwalk.h (enum dom_walker::reachability): New enum.
11189 (dom_walker::dom_walker): Convert bool param
11190 "skip_unreachable_blocks" to enum reachability.
11191 (set_all_edges_as_executable): New decl.
11192 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
11193 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
11195 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
11196 but converting true to REACHABLE_BLOCKS.
11197 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
11199 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
11200 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
11201 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
11203 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
11204 if check_all_array_refs will be called.
11206 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11208 * tree.c (selftest::test_location_wrappers): Add more test
11211 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11213 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
11214 (selftest::test_bit_in_range): Likewise.
11216 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11219 * doc/sourcebuild.texi (vect_float): Say that the selector
11220 only describes the situation when -funsafe-math-optimizations is on.
11221 (vect_float_strict): Document.
11223 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11225 PR tree-optimization/83965
11226 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11227 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
11228 instead of checking only for a reduction.
11229 (vect_recog_widen_sum_pattern): Likewise.
11231 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11233 * predict.c (probably_never_executed): Only use precise profile info.
11234 (compute_function_frequency): Skip after inlining hack since we now
11235 have quality checking.
11237 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11239 * profile-count.h (profile_probability::very_unlikely,
11240 profile_probability::unlikely, profile_probability::even): Set
11241 precision to guessed.
11243 2018-01-23 Richard Biener <rguenther@suse.de>
11245 PR tree-optimization/83963
11246 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
11247 Properly terminate dominator walk when crossing the exit edge not
11248 when visiting its source block.
11250 2018-01-23 Jakub Jelinek <jakub@redhat.com>
11253 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
11254 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
11256 2018-01-22 Jakub Jelinek <jakub@redhat.com>
11258 PR tree-optimization/83957
11259 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
11260 semicolon after for body surrounded by braces.
11262 PR tree-optimization/83081
11263 * profile-count.h (profile_probability::split): New method.
11264 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
11265 Use profile_probability::split.
11266 (do_compare_rtx_and_jump): Fix adjustment of probabilities
11267 when splitting a single conditional jump into 2.
11269 2018-01-22 David Malcolm <dmalcolm@redhat.com>
11271 PR tree-optimization/69452
11272 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
11275 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11277 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
11278 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
11279 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
11281 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11283 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
11285 * config/rl78/rl78.md (movdi): New define_expand.
11286 * config/rl78/rl78.c (rl78_split_movdi): New function.
11288 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
11291 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
11293 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
11294 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
11295 128-bit to produce an UNSPEC move to get the double word with the
11296 signbit and then a shift directly to do signbit.
11297 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
11298 implementation with a new version that just does either a direct
11299 move or a regular move. Move memory interface to separate insns.
11300 Move insns so they are next to the expander.
11301 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
11302 with signbit move. Split big and little endian case.
11303 (signbit<mode>2_dm_mem_le): Likewise.
11304 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
11305 (signbit<mode>2_dm2): Likewise.
11307 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11309 * config/rl78/rl78.md (anddi3): New define_expand.
11311 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11313 * config/rl78/rl78.md (umindi3): New define_expand.
11315 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11317 * config/rl78/rl78.md (smindi3): New define_expand.
11319 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11321 * config/rl78/rl78.md (smaxdi3): New define_expand.
11323 2018-01-22 Carl Love <cel@us.ibm.com>
11325 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
11326 LVX_V1TI): Add macro expansion.
11327 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
11328 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
11329 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
11330 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
11331 Change check to determine if the instruction is a byte reversing
11332 entry. Fix typo in comment.
11333 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
11334 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
11335 Add def_builtin calls for new builtins.
11336 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
11337 Add define_insn expansion.
11339 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11341 * config/rl78/rl78.md (umaxdi3): New define_expand.
11343 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11345 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
11346 for non-QImode registers.
11348 2018-01-22 Richard Biener <rguenther@suse.de>
11350 PR tree-optimization/83963
11351 * graphite-scop-detection.c (scop_detection::get_sese): Delay
11352 including the loop exit block.
11353 (scop_detection::merge_sese): Likewise.
11354 (scop_detection::add_scop): Do it here instead.
11356 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11358 * doc/sourcebuild.texi (arm_softfloat): Document.
11360 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
11363 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
11364 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
11365 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
11367 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11368 David Edelsohn <dje.gcc@gmail.com>
11371 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11372 Change "crset eq" to "crset 2".
11373 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11374 (*call_indirect_aix<mode>_nospec): Likewise.
11375 (*call_value_indirect_aix<mode>_nospec): Likewise.
11376 (*call_indirect_elfv2<mode>_nospec): Likewise.
11377 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
11378 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
11379 change assembly output from . to $.
11380 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11381 (indirect_jump<mode>_nospec): Change assembly output from . to $.
11382 (*tablejump<mode>_internal1_nospec): Likewise.
11384 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
11387 * config/sh/sh_optimize_sett_clrt.cc:
11388 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
11390 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
11392 PR tree-optimization/83940
11393 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
11394 offset_dt to vect_constant_def rather than vect_unknown_def_type.
11395 (vect_check_load_store_mask): Add a mask_dt_out parameter and
11396 use it to pass back the definition type.
11397 (vect_check_store_rhs): Likewise rhs_dt_out.
11398 (vect_build_gather_load_calls): Add a mask_dt argument and use
11399 it instead of a call to vect_is_simple_use.
11400 (vectorizable_store): Update calls to vect_check_load_store_mask
11401 and vect_check_store_rhs. Use the dt returned by the latter instead
11402 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
11403 instead of calls to vect_is_simple_use. Pass the scalar rather
11404 than the vector operand to vect_is_simple_use when handling
11405 second and subsequent copies of an rhs value.
11406 (vectorizable_load): Update calls to vect_check_load_store_mask
11407 and vect_build_gather_load_calls. Use the cached mask_dt and
11408 gs_info.offset_dt instead of calls to vect_is_simple_use.
11410 2018-01-20 Jakub Jelinek <jakub@redhat.com>
11412 PR middle-end/83945
11413 * tree-emutls.c: Include gimplify.h.
11414 (lower_emutls_2): New function.
11415 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
11416 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
11417 it before further processing.
11420 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
11421 UINTVAL (trueop1) instead of INTVAL (op1).
11423 2018-01-19 Jakub Jelinek <jakub@redhat.com>
11427 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
11428 INCOMING_FRAME_SP_OFFSET if not defined.
11429 (scan_trace): Add ENTRY argument. If true and
11430 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
11431 emit a note to adjust the CFA offset.
11432 (create_cfi_notes): Adjust scan_trace callers.
11433 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
11434 INCOMING_FRAME_SP_OFFSET in the CIE.
11435 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
11436 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
11438 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
11439 * doc/tm.texi: Regenerated.
11441 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11443 PR rtl-optimization/83147
11444 * lra-constraints.c (remove_inheritance_pseudos): Use
11445 lra_substitute_pseudo_within_insn.
11447 2018-01-19 Tom de Vries <tom@codesourcery.com>
11448 Cesar Philippidis <cesar@codesourcery.com>
11451 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
11453 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
11456 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
11457 spaces for function labels.
11459 2018-01-19 Martin Liska <mliska@suse.cz>
11461 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
11462 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
11463 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
11464 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
11465 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
11466 (PRED_CONST_RETURN): Change from 69 to 65.
11467 (PRED_NULL_RETURN): Change from 91 to 71.
11468 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
11469 (PRED_LOOP_GUARD): Change from 66 to 73.
11471 2018-01-19 Martin Liska <mliska@suse.cz>
11473 * predict.c (predict_insn_def): Add new assert.
11474 (struct branch_predictor): Change type to signed integer.
11475 (test_prediction_value_range): Amend test to cover
11476 PROB_UNINITIALIZED.
11477 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
11478 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
11479 (PRED_LOOP_ITERATIONS_MAX): Likewise.
11480 (PRED_LOOP_IV_COMPARE): Likewise.
11481 * predict.h (PROB_UNINITIALIZED): Define new constant.
11483 2018-01-19 Martin Liska <mliska@suse.cz>
11485 * predict.c (dump_prediction): Add new format for
11486 analyze_brprob.py script which is enabled with -details
11488 * profile-count.h (precise_p): New function.
11490 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11492 PR tree-optimization/83922
11493 * tree-vect-loop.c (vect_verify_full_masking): Return false if
11494 there are no statements that need masking.
11495 (vect_active_double_reduction_p): New function.
11496 (vect_analyze_loop_operations): Use it when handling phis that
11497 are not in the loop header.
11499 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11501 PR tree-optimization/83914
11502 * tree-vect-loop.c (vectorizable_induction): Don't convert
11503 init_expr or apply the peeling adjustment for inductions
11504 that are nested within the vectorized loop.
11506 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11508 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
11511 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11515 * function.h (gimplify_parameters): Add gimple_seq * argument.
11516 * function.c: Include gimple.h and options.h.
11517 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
11518 for the added local temporaries if needed.
11519 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
11520 if there are any parameter cleanups, wrap whole body into a
11521 try/finally with the cleanups.
11523 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
11526 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
11527 Use GET_MODE_CLASS for scalar floating point.
11529 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11533 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
11534 Fix call of call_cgraph_insertion_hooks.
11536 2018-01-18 Martin Sebor <msebor@redhat.com>
11538 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
11540 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11543 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
11546 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
11549 * common.opt: (-ffile-prefix-map): New option.
11550 * opts.c (common_handle_option): Defer it.
11551 * opts-global.c (handle_common_deferred_options): Handle it.
11552 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
11553 * file-prefix-map.h: New file.
11554 (remap_debug_filename, add_debug_prefix_map): ...here.
11555 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
11556 * final.c (debug_prefix_map, add_debug_prefix_map
11557 remap_debug_filename): Move to...
11558 * file-prefix-map.c: New file.
11559 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
11560 generalize, get rid of alloca(), use strrchr() instead of strchr().
11561 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
11562 Implement in terms of add_prefix_map().
11563 (remap_macro_filename, remap_debug_filename): Implement in term of
11565 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
11566 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
11567 * dbxout.c: Include file-prefix-map.h.
11568 * varasm.c: Likewise.
11569 * vmsdbgout.c: Likewise.
11570 * xcoffout.c: Likewise.
11571 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
11572 * doc/cppopts.texi (-fmacro-prefix-map): Document.
11573 * doc/invoke.texi (-ffile-prefix-map): Document.
11574 (-fdebug-prefix-map): Update description.
11576 2018-01-18 Martin Liska <mliska@suse.cz>
11578 * config/i386/i386.c (indirect_thunk_name): Document that also
11580 (output_indirect_thunk): Document why both instructions
11581 (pause and lfence) are generated.
11583 2018-01-18 Richard Biener <rguenther@suse.de>
11585 PR tree-optimization/83887
11586 * graphite-scop-detection.c
11587 (scop_detection::get_nearest_dom_with_single_entry): Remove.
11588 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
11589 (scop_detection::merge_sese): Re-implement with a flood-fill
11590 algorithm that properly finds a SESE region if it exists.
11592 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11595 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
11596 pointer_diff optimizations use view_convert instead of convert.
11598 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11600 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11601 Generate different code for -mno-speculate-indirect-jumps.
11602 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11603 (*call_indirect_aix<mode>): Disable for
11604 -mno-speculate-indirect-jumps.
11605 (*call_indirect_aix<mode>_nospec): New define_insn.
11606 (*call_value_indirect_aix<mode>): Disable for
11607 -mno-speculate-indirect-jumps.
11608 (*call_value_indirect_aix<mode>_nospec): New define_insn.
11609 (*sibcall_nonlocal_sysv<mode>): Generate different code for
11610 -mno-speculate-indirect-jumps.
11611 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11613 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
11615 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
11616 long double type, set the flags for noting the default long double
11617 type, even if we don't pass or return a long double type.
11619 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
11622 * ipa-inline.c (flatten_function): Do not overwrite final inlining
11625 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
11627 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
11628 support for merge[hl].
11629 (fold_mergehl_helper): New helper function.
11630 (tree-vector-builder.h): New #include for tree_vector_builder usage.
11631 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
11632 (altivec_vmrglw_direct): Add xxmrglw insn.
11634 2018-01-17 Andrew Waterman <andrew@sifive.com>
11636 * config/riscv/riscv.c (riscv_conditional_register_usage): If
11637 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
11639 2018-01-17 David Malcolm <dmalcolm@redhat.com>
11642 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
11643 call the lto_location_cache before reading the
11644 DECL_SOURCE_LOCATION of the types.
11646 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
11647 Richard Sandiford <richard.sandiford@linaro.org>
11649 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
11650 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
11651 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
11652 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
11653 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
11655 * config/aarch64/constraints.md (aarch64_movti_operand):
11657 * config/aarch64/predicates.md (Uti): Add new constraint.
11659 2018-01-17 Carl Love <cel@us.ibm.com>
11661 * config/rs6000/vsx.md (define_expand xl_len_r,
11662 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
11663 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
11665 (define_expand, define_insn): Move the shift left from the
11666 define_insn to the define_expand for lxvl and stxvl instructions.
11667 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
11668 and XL_LEN_R definitions to PURE.
11670 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
11672 * config/i386/i386.c (indirect_thunk_name): Declare regno
11673 as unsigned int. Compare regno with INVALID_REGNUM.
11674 (output_indirect_thunk): Ditto.
11675 (output_indirect_thunk_function): Ditto.
11676 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
11677 in the call to output_indirect_thunk_function.
11679 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
11681 PR middle-end/83884
11682 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
11683 rather than the size of inner_type to determine the stack slot size
11684 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
11686 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
11689 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
11692 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
11694 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
11695 endian Linux systems to optionally enable multilibs for selecting
11696 the long double type if the user configured an explicit type.
11697 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
11698 have no long double multilibs if not defined.
11699 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
11700 warn if the user used -mabi={ieee,ibm}longdouble and we built
11701 multilibs for long double.
11702 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
11703 appropriate multilib option.
11704 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
11706 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
11707 for building long double multilibs.
11708 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
11710 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
11712 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
11715 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
11717 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
11720 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
11723 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
11726 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11728 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
11729 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
11731 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
11733 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
11734 different rtl trees depending on TARGET_64BIT.
11735 (rs6000_gen_lvx): Likewise.
11737 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11739 * config/visium/visium.md (nop): Tweak comment.
11740 (hazard_nop): Likewise.
11742 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11744 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
11745 -mspeculate-indirect-jumps.
11746 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
11747 for -mno-speculate-indirect-jumps.
11748 (*call_indirect_elfv2<mode>_nospec): New define_insn.
11749 (*call_value_indirect_elfv2<mode>): Disable for
11750 -mno-speculate-indirect-jumps.
11751 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
11752 (indirect_jump): Emit different RTL for
11753 -mno-speculate-indirect-jumps.
11754 (*indirect_jump<mode>): Disable for
11755 -mno-speculate-indirect-jumps.
11756 (*indirect_jump<mode>_nospec): New define_insn.
11757 (tablejump): Emit different RTL for
11758 -mno-speculate-indirect-jumps.
11759 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
11760 (tablejumpsi_nospec): New define_expand.
11761 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
11762 (tablejumpdi_nospec): New define_expand.
11763 (*tablejump<mode>_internal1): Disable for
11764 -mno-speculate-indirect-jumps.
11765 (*tablejump<mode>_internal1_nospec): New define_insn.
11766 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
11769 2018-01-16 Artyom Skrobov tyomitch@gmail.com
11771 * caller-save.c (insert_save): Drop unnecessary parameter. All
11774 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11775 Richard Biener <rguenth@suse.de>
11778 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
11779 return early, inline manually is_gimple_sizepos. Make sure if we
11780 call gimplify_expr we don't end up with a gimple constant.
11781 * tree.c (variably_modified_type_p): Don't return true for
11782 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
11783 * gimplify.h (is_gimple_sizepos): Remove.
11785 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11787 PR tree-optimization/83857
11788 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
11789 vectorizable_live_operation for pure SLP statements.
11790 (vectorizable_live_operation): Handle PHIs.
11792 2018-01-16 Richard Biener <rguenther@suse.de>
11794 PR tree-optimization/83867
11795 * tree-vect-stmts.c (vect_transform_stmt): Precompute
11796 nested_in_vect_loop_p since the scalar stmt may get invalidated.
11798 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11801 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
11802 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
11803 If off is not INTEGER_CST, issue a may not be aligned warning
11804 rather than isn't aligned. Use isn%'t rather than isn't.
11805 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
11807 <case MULT_EXPR>: Improve the case when bottom and one of the
11808 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
11809 operand, in that case check if the other operand is multiple of
11810 bottom divided by the INTEGER_CST operand.
11812 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11815 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
11816 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
11817 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
11818 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
11819 * config/pa/pa.c (pa_function_arg_advance): Likewise.
11820 (pa_function_arg, pa_arg_partial_bytes): Likewise.
11821 (pa_function_arg_size): New function.
11823 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11825 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
11826 in a separate statement.
11828 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11830 PR tree-optimization/83847
11831 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
11832 group gathers and scatters.
11834 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11836 PR rtl-optimization/86620
11837 * params.def (max-sched-ready-insns): Bump minimum value to 1.
11839 PR rtl-optimization/83213
11840 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
11841 to last if both are JUMP_INSNs.
11843 PR tree-optimization/83843
11844 * gimple-ssa-store-merging.c
11845 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
11846 store_immediate_info for bswap/nop orig_stores.
11848 2018-01-15 Andrew Waterman <andrew@sifive.com>
11850 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
11852 <UDIV>: Increase cost if !TARGET_DIV.
11854 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
11856 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
11857 (define_attr "cr_logical_3op"): New.
11858 (cceq_ior_compare): Adjust.
11859 (cceq_ior_compare_complement): Adjust.
11860 (*cceq_rev_compare): Adjust.
11861 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
11862 (is_cracked_insn): Adjust.
11863 (insn_must_be_first_in_group): Adjust.
11864 * config/rs6000/40x.md: Adjust.
11865 * config/rs6000/440.md: Adjust.
11866 * config/rs6000/476.md: Adjust.
11867 * config/rs6000/601.md: Adjust.
11868 * config/rs6000/603.md: Adjust.
11869 * config/rs6000/6xx.md: Adjust.
11870 * config/rs6000/7450.md: Adjust.
11871 * config/rs6000/7xx.md: Adjust.
11872 * config/rs6000/8540.md: Adjust.
11873 * config/rs6000/cell.md: Adjust.
11874 * config/rs6000/e300c2c3.md: Adjust.
11875 * config/rs6000/e500mc.md: Adjust.
11876 * config/rs6000/e500mc64.md: Adjust.
11877 * config/rs6000/e5500.md: Adjust.
11878 * config/rs6000/e6500.md: Adjust.
11879 * config/rs6000/mpc.md: Adjust.
11880 * config/rs6000/power4.md: Adjust.
11881 * config/rs6000/power5.md: Adjust.
11882 * config/rs6000/power6.md: Adjust.
11883 * config/rs6000/power7.md: Adjust.
11884 * config/rs6000/power8.md: Adjust.
11885 * config/rs6000/power9.md: Adjust.
11886 * config/rs6000/rs64.md: Adjust.
11887 * config/rs6000/titan.md: Adjust.
11889 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11891 * config/i386/predicates.md (indirect_branch_operand): Rewrite
11892 ix86_indirect_branch_register logic.
11894 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11896 * config/i386/constraints.md (Bs): Update
11897 ix86_indirect_branch_register check. Don't check
11898 ix86_indirect_branch_register with GOT_memory_operand.
11900 * config/i386/predicates.md (GOT_memory_operand): Don't check
11901 ix86_indirect_branch_register here.
11902 (GOT32_symbol_operand): Likewise.
11904 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11906 * config/i386/predicates.md (constant_call_address_operand):
11907 Rewrite ix86_indirect_branch_register logic.
11908 (sibcall_insn_operand): Likewise.
11910 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11912 * config/i386/constraints.md (Bs): Replace
11913 ix86_indirect_branch_thunk_register with
11914 ix86_indirect_branch_register.
11916 * config/i386/i386.md (indirect_jump): Likewise.
11917 (tablejump): Likewise.
11918 (*sibcall_memory): Likewise.
11919 (*sibcall_value_memory): Likewise.
11920 Peepholes of indirect call and jump via memory: Likewise.
11921 * config/i386/i386.opt: Likewise.
11922 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11923 (GOT_memory_operand): Likewise.
11924 (call_insn_operand): Likewise.
11925 (sibcall_insn_operand): Likewise.
11926 (GOT32_symbol_operand): Likewise.
11928 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11930 PR middle-end/83837
11931 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
11932 type rather than type addr's type points to.
11933 (expand_omp_atomic_mutex): Likewise.
11934 (expand_omp_atomic): Likewise.
11936 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11939 * config/i386/i386.c (output_indirect_thunk_function): Use
11940 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
11941 for __x86_return_thunk.
11943 2018-01-15 Richard Biener <rguenther@suse.de>
11945 PR middle-end/83850
11946 * expmed.c (extract_bit_field_1): Fix typo.
11948 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11951 * config/arm/iterators.md (VF): New mode iterator.
11952 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
11953 Remove integer-related logic from pattern.
11954 (neon_vabd<mode>_3): Likewise.
11956 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11958 PR middle-end/82694
11959 * common.opt (fstrict-overflow): No longer an alias.
11960 (fwrapv-pointer): New option.
11961 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
11962 also for pointer types based on flag_wrapv_pointer.
11963 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
11964 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
11965 opts->x_flag_wrapv got set.
11966 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
11967 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
11968 POINTER_TYPE_OVERFLOW_UNDEFINED.
11969 * match.pd: Likewise in address comparison pattern.
11970 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
11972 2018-01-15 Richard Biener <rguenther@suse.de>
11975 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
11976 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
11977 Reset type names to their identifier if their TYPE_DECL doesn't
11978 have linkage (and thus is used for ODR and devirt).
11979 (save_debug_info_for_decl): Remove.
11980 (save_debug_info_for_type): Likewise.
11981 (add_tree_to_fld_list): Adjust.
11982 * tree-pretty-print.c (dump_generic_node): Make dumping of
11983 type names more robust.
11985 2018-01-15 Richard Biener <rguenther@suse.de>
11987 * BASE-VER: Bump to 8.0.1.
11989 2018-01-14 Martin Sebor <msebor@redhat.com>
11992 * builtins.c (check_access): Avoid warning when the no-warning bit
11995 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
11997 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
11998 * ira-color (allocno_hard_regs_compare): Likewise.
12000 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
12003 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
12004 Use .pushsection/.popsection.
12006 2018-01-14 Martin Sebor <msebor@redhat.com>
12009 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
12011 2018-01-14 Jakub Jelinek <jakub@redhat.com>
12013 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
12014 entry from extra_headers.
12015 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12016 extra_headers, make the list bitwise identical to the i?86-*-* one.
12018 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12020 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12021 -mcmodel=large with -mindirect-branch=thunk,
12022 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12023 -mfunction-return=thunk-extern.
12024 * doc/invoke.texi: Document -mcmodel=large is incompatible with
12025 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12026 -mfunction-return=thunk and -mfunction-return=thunk-extern.
12028 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12030 * config/i386/i386.c (print_reg): Print the name of the full
12031 integer register without '%'.
12032 (ix86_print_operand): Handle 'V'.
12033 * doc/extend.texi: Document 'V' modifier.
12035 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12037 * config/i386/constraints.md (Bs): Disallow memory operand for
12038 -mindirect-branch-register.
12040 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12041 (GOT_memory_operand): Likewise.
12042 (call_insn_operand): Likewise.
12043 (sibcall_insn_operand): Likewise.
12044 (GOT32_symbol_operand): Likewise.
12045 * config/i386/i386.md (indirect_jump): Call convert_memory_address
12046 for -mindirect-branch-register.
12047 (tablejump): Likewise.
12048 (*sibcall_memory): Likewise.
12049 (*sibcall_value_memory): Likewise.
12050 Disallow peepholes of indirect call and jump via memory for
12051 -mindirect-branch-register.
12052 (*call_pop): Replace m with Bw.
12053 (*call_value_pop): Likewise.
12054 (*sibcall_pop_memory): Replace m with Bs.
12055 * config/i386/i386.opt (mindirect-branch-register): New option.
12056 * doc/invoke.texi: Document -mindirect-branch-register option.
12058 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12060 * config/i386/i386-protos.h (ix86_output_function_return): New.
12061 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12062 set function_return_type.
12063 (indirect_thunk_name): Add ret_p to indicate thunk for function
12065 (output_indirect_thunk_function): Pass false to
12066 indirect_thunk_name.
12067 (ix86_output_indirect_branch_via_reg): Likewise.
12068 (ix86_output_indirect_branch_via_push): Likewise.
12069 (output_indirect_thunk_function): Create alias for function
12070 return thunk if regno < 0.
12071 (ix86_output_function_return): New function.
12072 (ix86_handle_fndecl_attribute): Handle function_return.
12073 (ix86_attribute_table): Add function_return.
12074 * config/i386/i386.h (machine_function): Add
12075 function_return_type.
12076 * config/i386/i386.md (simple_return_internal): Use
12077 ix86_output_function_return.
12078 (simple_return_internal_long): Likewise.
12079 * config/i386/i386.opt (mfunction-return=): New option.
12080 (indirect_branch): Mention -mfunction-return=.
12081 * doc/extend.texi: Document function_return function attribute.
12082 * doc/invoke.texi: Document -mfunction-return= option.
12084 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12086 * config/i386/i386-opts.h (indirect_branch): New.
12087 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12088 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12089 with local indirect jump when converting indirect call and jump.
12090 (ix86_set_indirect_branch_type): New.
12091 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12092 (indirectlabelno): New.
12093 (indirect_thunk_needed): Likewise.
12094 (indirect_thunk_bnd_needed): Likewise.
12095 (indirect_thunks_used): Likewise.
12096 (indirect_thunks_bnd_used): Likewise.
12097 (INDIRECT_LABEL): Likewise.
12098 (indirect_thunk_name): Likewise.
12099 (output_indirect_thunk): Likewise.
12100 (output_indirect_thunk_function): Likewise.
12101 (ix86_output_indirect_branch_via_reg): Likewise.
12102 (ix86_output_indirect_branch_via_push): Likewise.
12103 (ix86_output_indirect_branch): Likewise.
12104 (ix86_output_indirect_jmp): Likewise.
12105 (ix86_code_end): Call output_indirect_thunk_function if needed.
12106 (ix86_output_call_insn): Call ix86_output_indirect_branch if
12108 (ix86_handle_fndecl_attribute): Handle indirect_branch.
12109 (ix86_attribute_table): Add indirect_branch.
12110 * config/i386/i386.h (machine_function): Add indirect_branch_type
12111 and has_local_indirect_jump.
12112 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12114 (tablejump): Likewise.
12115 (*indirect_jump): Use ix86_output_indirect_jmp.
12116 (*tablejump_1): Likewise.
12117 (simple_return_indirect_internal): Likewise.
12118 * config/i386/i386.opt (mindirect-branch=): New option.
12119 (indirect_branch): New.
12122 (thunk-inline): Likewise.
12123 (thunk-extern): Likewise.
12124 * doc/extend.texi: Document indirect_branch function attribute.
12125 * doc/invoke.texi: Document -mindirect-branch= option.
12127 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
12130 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12132 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
12134 * ipa-inline.c (want_inline_small_function_p): Return false if
12135 inlining has already failed with CIF_FINAL_ERROR.
12136 (update_caller_keys): Call want_inline_small_function_p before
12138 (update_callee_keys): Likewise.
12140 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12142 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12144 (rs6000_quadword_masked_address_p): Likewise.
12145 (quad_aligned_load_p): Likewise.
12146 (quad_aligned_store_p): Likewise.
12147 (const_load_sequence_p): Add comment to describe the outer-most loop.
12148 (mimic_memory_attributes_and_flags): New function.
12149 (rs6000_gen_stvx): Likewise.
12150 (replace_swapped_aligned_store): Likewise.
12151 (rs6000_gen_lvx): Likewise.
12152 (replace_swapped_aligned_load): Likewise.
12153 (replace_swapped_load_constant): Capitalize argument name in
12154 comment describing this function.
12155 (rs6000_analyze_swaps): Add a third pass to search for vector loads
12156 and stores that access quad-word aligned addresses and replace
12157 with stvx or lvx instructions when appropriate.
12158 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12159 New function prototype.
12160 (rs6000_quadword_masked_address_p): Likewise.
12161 (rs6000_gen_lvx): Likewise.
12162 (rs6000_gen_stvx): Likewise.
12163 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12164 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12165 when memory address is aligned.
12166 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12167 this split to select lvx instruction when memory address is aligned.
12168 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12169 instruction when memory address is aligned.
12170 (*vsx_le_perm_load_v16qi): Likewise.
12171 (four unnamed splitters): Modify to select the stvx instruction
12172 when memory is aligned.
12174 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
12176 * predict.c (determine_unlikely_bbs): Handle correctly BBs
12177 which appears in the queue multiple times.
12179 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12180 Alan Hayward <alan.hayward@arm.com>
12181 David Sherwood <david.sherwood@arm.com>
12183 * tree-vectorizer.h (vec_lower_bound): New structure.
12184 (_loop_vec_info): Add check_nonzero and lower_bounds.
12185 (LOOP_VINFO_CHECK_NONZERO): New macro.
12186 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
12187 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
12188 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
12189 fields. Make seg_len the distance travelled, not including the
12191 (dr_direction_indicator): Declare.
12192 (dr_zero_step_indicator): Likewise.
12193 (dr_known_forward_stride_p): Likewise.
12194 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
12196 (runtime_alias_check_p): Allow runtime alias checks with
12198 (operator ==): Compare access_size and align.
12199 (prune_runtime_alias_test_list): Rework for new distinction between
12200 the access_size and seg_len.
12201 (create_intersect_range_checks_index): Likewise. Cope with polynomial
12203 (get_segment_min_max): New function.
12204 (create_intersect_range_checks): Use it.
12205 (dr_step_indicator): New function.
12206 (dr_direction_indicator): Likewise.
12207 (dr_zero_step_indicator): Likewise.
12208 (dr_known_forward_stride_p): Likewise.
12209 * tree-loop-distribution.c (data_ref_segment_size): Return
12210 DR_STEP * (niters - 1).
12211 (compute_alias_check_pairs): Update call to the dr_with_seg_len
12213 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
12214 (vect_preserves_scalar_order_p): New function, split out from...
12215 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
12216 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
12217 (vect_vfa_access_size): New function.
12218 (vect_vfa_align): Likewise.
12219 (vect_compile_time_alias): Take access_size_a and access_b arguments.
12220 (dump_lower_bound): New function.
12221 (vect_check_lower_bound): Likewise.
12222 (vect_small_gap_p): Likewise.
12223 (vectorizable_with_step_bound_p): Likewise.
12224 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
12225 depencies if the vectorization factor is 1. Convert the checks
12226 for nonzero steps into checks on the bounds of DR_STEP. Try using
12227 a bunds check for variable steps if the minimum required step is
12228 relatively small. Update calls to the dr_with_seg_len
12229 constructor and to vect_compile_time_alias.
12230 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
12232 (vect_loop_versioning): Call it.
12233 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
12235 (vect_estimate_min_profitable_iters): Account for any bounds checks.
12237 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12238 Alan Hayward <alan.hayward@arm.com>
12239 David Sherwood <david.sherwood@arm.com>
12241 * doc/sourcebuild.texi (vect_scatter_store): Document.
12242 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
12244 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
12246 * genopinit.c (main): Add supports_vec_scatter_store and
12247 supports_vec_scatter_store_cached to target_optabs.
12248 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
12249 IFN_MASK_SCATTER_STORE.
12250 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
12252 * internal-fn.h (internal_store_fn_p): Declare.
12253 (internal_fn_stored_value_index): Likewise.
12254 * internal-fn.c (scatter_store_direct): New macro.
12255 (expand_scatter_store_optab_fn): New function.
12256 (direct_scatter_store_optab_supported_p): New macro.
12257 (internal_store_fn_p): New function.
12258 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
12259 IFN_MASK_SCATTER_STORE.
12260 (internal_fn_mask_index): Likewise.
12261 (internal_fn_stored_value_index): New function.
12262 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
12263 for scatter stores.
12264 * optabs-query.h (supports_vec_scatter_store_p): Declare.
12265 * optabs-query.c (supports_vec_scatter_store_p): New function.
12266 * tree-vectorizer.h (vect_get_store_rhs): Declare.
12267 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
12268 true for scatter stores.
12269 (vect_gather_scatter_fn_p): Handle scatter stores too.
12270 (vect_check_gather_scatter): Consider using scatter stores if
12271 supports_vec_scatter_store_p.
12272 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
12273 scatter stores too.
12274 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12275 internal_fn_stored_value_index.
12276 (check_load_store_masking): Handle scatter stores too.
12277 (vect_get_store_rhs): Make public.
12278 (vectorizable_call): Use internal_store_fn_p.
12279 (vectorizable_store): Handle scatter store internal functions.
12280 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
12281 when deciding whether the end of the group has been reached.
12282 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
12283 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
12284 (mask_scatter_store<mode>): New insns.
12286 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12287 Alan Hayward <alan.hayward@arm.com>
12288 David Sherwood <david.sherwood@arm.com>
12290 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
12291 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
12292 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
12294 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
12295 Use vect_truncate_gather_scatter_offset if we can't treat the
12296 operation as a normal gather load or scatter store.
12297 (get_group_load_store_type): Take the gather_scatter_info
12298 as argument. Try using a gather load or scatter store for
12299 single-element groups.
12300 (get_load_store_type): Update calls to get_group_load_store_type
12301 and vect_use_strided_gather_scatters_p.
12303 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12304 Alan Hayward <alan.hayward@arm.com>
12305 David Sherwood <david.sherwood@arm.com>
12307 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
12308 optional tree argument.
12309 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
12311 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
12312 but continue to use the current value as a fallback.
12313 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
12314 to compare the updates.
12315 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
12316 (get_load_store_type): Use it when handling a strided access.
12317 (vect_get_strided_load_store_ops): New function.
12318 (vect_get_data_ptr_increment): Likewise.
12319 (vectorizable_load): Handle strided gather loads. Always pass
12320 a step to vect_create_data_ref_ptr and bump_vector_ptr.
12322 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12323 Alan Hayward <alan.hayward@arm.com>
12324 David Sherwood <david.sherwood@arm.com>
12326 * doc/md.texi (gather_load@var{m}): Document.
12327 (mask_gather_load@var{m}): Likewise.
12328 * genopinit.c (main): Add supports_vec_gather_load and
12329 supports_vec_gather_load_cached to target_optabs.
12330 * optabs-tree.c (init_tree_optimization_optabs): Use
12331 ggc_cleared_alloc to allocate target_optabs.
12332 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
12333 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
12335 * internal-fn.h (internal_load_fn_p): Declare.
12336 (internal_gather_scatter_fn_p): Likewise.
12337 (internal_fn_mask_index): Likewise.
12338 (internal_gather_scatter_fn_supported_p): Likewise.
12339 * internal-fn.c (gather_load_direct): New macro.
12340 (expand_gather_load_optab_fn): New function.
12341 (direct_gather_load_optab_supported_p): New macro.
12342 (direct_internal_fn_optab): New function.
12343 (internal_load_fn_p): Likewise.
12344 (internal_gather_scatter_fn_p): Likewise.
12345 (internal_fn_mask_index): Likewise.
12346 (internal_gather_scatter_fn_supported_p): Likewise.
12347 * optabs-query.c (supports_at_least_one_mode_p): New function.
12348 (supports_vec_gather_load_p): Likewise.
12349 * optabs-query.h (supports_vec_gather_load_p): Declare.
12350 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
12351 and memory_type field.
12352 (NUM_PATTERNS): Bump to 15.
12353 * tree-vect-data-refs.c: Include internal-fn.h.
12354 (vect_gather_scatter_fn_p): New function.
12355 (vect_describe_gather_scatter_call): Likewise.
12356 (vect_check_gather_scatter): Try using internal functions for
12357 gather loads. Recognize existing calls to a gather load function.
12358 (vect_analyze_data_refs): Consider using gather loads if
12359 supports_vec_gather_load_p.
12360 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
12361 (vect_get_gather_scatter_offset_type): Likewise.
12362 (vect_convert_mask_for_vectype): Likewise.
12363 (vect_add_conversion_to_patterm): Likewise.
12364 (vect_try_gather_scatter_pattern): Likewise.
12365 (vect_recog_gather_scatter_pattern): New pattern recognizer.
12366 (vect_vect_recog_func_ptrs): Add it.
12367 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12368 internal_fn_mask_index and internal_gather_scatter_fn_p.
12369 (check_load_store_masking): Take the gather_scatter_info as an
12370 argument and handle gather loads.
12371 (vect_get_gather_scatter_ops): New function.
12372 (vectorizable_call): Check internal_load_fn_p.
12373 (vectorizable_load): Likewise. Handle gather load internal
12375 (vectorizable_store): Update call to check_load_store_masking.
12376 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
12377 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
12378 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
12379 (aarch64_gather_scale_operand_d): New predicates.
12380 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
12381 (mask_gather_load<mode>): New insns.
12383 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12384 Alan Hayward <alan.hayward@arm.com>
12385 David Sherwood <david.sherwood@arm.com>
12387 * optabs.def (fold_left_plus_optab): New optab.
12388 * doc/md.texi (fold_left_plus_@var{m}): Document.
12389 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
12390 * internal-fn.c (fold_left_direct): Define.
12391 (expand_fold_left_optab_fn): Likewise.
12392 (direct_fold_left_optab_supported_p): Likewise.
12393 * fold-const-call.c (fold_const_fold_left): New function.
12394 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
12395 * tree-parloops.c (valid_reduction_p): New function.
12396 (gather_scalar_reductions): Use it.
12397 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
12398 (vect_finish_replace_stmt): Declare.
12399 * tree-vect-loop.c (fold_left_reduction_fn): New function.
12400 (needs_fold_left_reduction_p): New function, split out from...
12401 (vect_is_simple_reduction): ...here. Accept reductions that
12402 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
12403 (vect_force_simple_reduction): Also store the reduction type in
12404 the assignment's STMT_VINFO_REDUC_TYPE.
12405 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
12406 (merge_with_identity): New function.
12407 (vect_expand_fold_left): Likewise.
12408 (vectorize_fold_left_reduction): Likewise.
12409 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
12410 scalar phi in place for it. Check for target support and reject
12411 cases that would reassociate the operation. Defer the transform
12412 phase to vectorize_fold_left_reduction.
12413 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
12414 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
12415 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
12417 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12419 * tree-if-conv.c (predicate_mem_writes): Remove redundant
12420 call to ifc_temp_var.
12422 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12423 Alan Hayward <alan.hayward@arm.com>
12424 David Sherwood <david.sherwood@arm.com>
12426 * target.def (legitimize_address_displacement): Take the original
12427 offset as a poly_int.
12428 * targhooks.h (default_legitimize_address_displacement): Update
12430 * targhooks.c (default_legitimize_address_displacement): Likewise.
12431 * doc/tm.texi: Regenerate.
12432 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
12433 as an argument, moving assert of ad->disp == ad->disp_term to...
12434 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
12435 Try calling targetm.legitimize_address_displacement before expanding
12436 the address rather than afterwards, and adjust for the new interface.
12437 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
12438 Match the new hook interface. Handle SVE addresses.
12439 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
12440 new hook interface.
12442 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12444 * Makefile.in (OBJS): Add early-remat.o.
12445 * target.def (select_early_remat_modes): New hook.
12446 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
12447 * doc/tm.texi: Regenerate.
12448 * targhooks.h (default_select_early_remat_modes): Declare.
12449 * targhooks.c (default_select_early_remat_modes): New function.
12450 * timevar.def (TV_EARLY_REMAT): New timevar.
12451 * passes.def (pass_early_remat): New pass.
12452 * tree-pass.h (make_pass_early_remat): Declare.
12453 * early-remat.c: New file.
12454 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
12456 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
12458 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12459 Alan Hayward <alan.hayward@arm.com>
12460 David Sherwood <david.sherwood@arm.com>
12462 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
12463 vfm1 with a bound_epilog parameter.
12464 (vect_do_peeling): Update calls accordingly, and move the prologue
12465 call earlier in the function. Treat the base bound_epilog as 0 for
12466 fully-masked loops and retain vf - 1 for other loops. Add 1 to
12467 this base when peeling for gaps.
12468 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
12469 with fully-masked loops.
12470 (vect_estimate_min_profitable_iters): Handle the single peeled
12471 iteration in that case.
12473 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12474 Alan Hayward <alan.hayward@arm.com>
12475 David Sherwood <david.sherwood@arm.com>
12477 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
12478 single-element interleaving even if the size is not a power of 2.
12479 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
12480 accesses for single-element interleaving if the group size is
12483 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12484 Alan Hayward <alan.hayward@arm.com>
12485 David Sherwood <david.sherwood@arm.com>
12487 * doc/md.texi (fold_extract_last_@var{m}): Document.
12488 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
12489 * optabs.def (fold_extract_last_optab): New optab.
12490 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
12491 * internal-fn.c (fold_extract_direct): New macro.
12492 (expand_fold_extract_optab_fn): Likewise.
12493 (direct_fold_extract_optab_supported_p): Likewise.
12494 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
12495 * tree-vect-loop.c (vect_model_reduction_cost): Handle
12496 EXTRACT_LAST_REDUCTION.
12497 (get_initial_def_for_reduction): Do not create an initial vector
12498 for EXTRACT_LAST_REDUCTION reductions.
12499 (vectorizable_reduction): Leave the scalar phi in place for
12500 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
12501 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
12502 epilogue code for EXTRACT_LAST_REDUCTION and defer the
12503 transform phase to vectorizable_condition.
12504 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
12506 (vect_finish_stmt_generation): ...here.
12507 (vect_finish_replace_stmt): New function.
12508 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
12509 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
12511 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
12513 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12514 Alan Hayward <alan.hayward@arm.com>
12515 David Sherwood <david.sherwood@arm.com>
12517 * doc/md.texi (extract_last_@var{m}): Document.
12518 * optabs.def (extract_last_optab): New optab.
12519 * internal-fn.def (EXTRACT_LAST): New internal function.
12520 * internal-fn.c (cond_unary_direct): New macro.
12521 (expand_cond_unary_optab_fn): Likewise.
12522 (direct_cond_unary_optab_supported_p): Likewise.
12523 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
12524 loops using EXTRACT_LAST.
12525 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
12526 (extract_last_<mode>): ...this optab.
12527 (vec_extract<mode><Vel>): Update accordingly.
12529 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12530 Alan Hayward <alan.hayward@arm.com>
12531 David Sherwood <david.sherwood@arm.com>
12533 * target.def (empty_mask_is_expensive): New hook.
12534 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
12535 * doc/tm.texi: Regenerate.
12536 * targhooks.h (default_empty_mask_is_expensive): Declare.
12537 * targhooks.c (default_empty_mask_is_expensive): New function.
12538 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
12539 if the target says that empty masks are expensive.
12540 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
12542 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
12544 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12545 Alan Hayward <alan.hayward@arm.com>
12546 David Sherwood <david.sherwood@arm.com>
12548 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
12549 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
12550 (vect_use_loop_mask_for_alignment_p): New function.
12551 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
12552 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
12553 niters_skip argument. Make sure that the first niters_skip elements
12554 of the first iteration are inactive.
12555 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
12556 Update call to vect_set_loop_masks_directly.
12557 (get_misalign_in_elems): New function, split out from...
12558 (vect_gen_prolog_loop_niters): ...here.
12559 (vect_update_init_of_dr): Take a code argument that specifies whether
12560 the adjustment should be added or subtracted.
12561 (vect_update_init_of_drs): Likewise.
12562 (vect_prepare_for_masked_peels): New function.
12563 (vect_do_peeling): Skip prologue peeling if we're using a mask
12564 instead. Update call to vect_update_inits_of_drs.
12565 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12567 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
12568 alignment. Do not include the number of peeled iterations in
12569 the minimum threshold in that case.
12570 (vectorizable_induction): Adjust the start value down by
12571 LOOP_VINFO_MASK_SKIP_NITERS iterations.
12572 (vect_transform_loop): Call vect_prepare_for_masked_peels.
12573 Take the number of skipped iterations into account when calculating
12575 * tree-vect-stmts.c (vect_gen_while_not): New function.
12577 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12578 Alan Hayward <alan.hayward@arm.com>
12579 David Sherwood <david.sherwood@arm.com>
12581 * doc/sourcebuild.texi (vect_fully_masked): Document.
12582 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
12583 default value to 0.
12584 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
12586 (vect_analyze_loop_2): ...here. Don't check the vectorization
12587 factor against the number of loop iterations if the loop is
12590 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12591 Alan Hayward <alan.hayward@arm.com>
12592 David Sherwood <david.sherwood@arm.com>
12594 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
12595 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
12596 (dump_groups): Update accordingly.
12597 (iv_use::mem_type): New member variable.
12598 (address_p): New function.
12599 (record_use): Add a mem_type argument and initialize the new
12601 (record_group_use): Add a mem_type argument. Use address_p.
12602 Remove obsolete null checks of base_object. Update call to record_use.
12603 (find_interesting_uses_op): Update call to record_group_use.
12604 (find_interesting_uses_cond): Likewise.
12605 (find_interesting_uses_address): Likewise.
12606 (get_mem_type_for_internal_fn): New function.
12607 (find_address_like_use): Likewise.
12608 (find_interesting_uses_stmt): Try find_address_like_use before
12609 calling find_interesting_uses_op.
12610 (addr_offset_valid_p): Use the iv mem_type field as the type
12611 of the addressed memory.
12612 (add_autoinc_candidates): Likewise.
12613 (get_address_cost): Likewise.
12614 (split_small_address_groups_p): Use address_p.
12615 (split_address_groups): Likewise.
12616 (add_iv_candidate_for_use): Likewise.
12617 (autoinc_possible_for_pair): Likewise.
12618 (rewrite_groups): Likewise.
12619 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
12620 (determine_group_iv_cost): Update after split of USE_ADDRESS.
12621 (get_alias_ptr_type_for_ptr_address): New function.
12622 (rewrite_use_address): Rewrite address uses in calls that were
12623 identified by find_address_like_use.
12625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12626 Alan Hayward <alan.hayward@arm.com>
12627 David Sherwood <david.sherwood@arm.com>
12629 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
12631 * gimple-expr.h (is_gimple_addressable: Likewise.
12632 * gimple-expr.c (is_gimple_address): Likewise.
12633 * internal-fn.c (expand_call_mem_ref): New function.
12634 (expand_mask_load_optab_fn): Use it.
12635 (expand_mask_store_optab_fn): Likewise.
12637 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12638 Alan Hayward <alan.hayward@arm.com>
12639 David Sherwood <david.sherwood@arm.com>
12641 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
12642 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
12643 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
12644 (cond_umax@var{mode}): Document.
12645 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
12646 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
12647 (cond_umin_optab, cond_umax_optab): New optabs.
12648 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
12649 (COND_IOR, COND_XOR): New internal functions.
12650 * internal-fn.h (get_conditional_internal_fn): Declare.
12651 * internal-fn.c (cond_binary_direct): New macro.
12652 (expand_cond_binary_optab_fn): Likewise.
12653 (direct_cond_binary_optab_supported_p): Likewise.
12654 (get_conditional_internal_fn): New function.
12655 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
12656 Cope with reduction statements that are vectorized as calls rather
12658 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
12659 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
12660 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
12661 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12662 (UNSPEC_COND_EOR): New unspecs.
12663 (optab): Add mappings for them.
12664 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
12665 (sve_int_op, sve_fp_op): New int attributes.
12667 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12668 Alan Hayward <alan.hayward@arm.com>
12669 David Sherwood <david.sherwood@arm.com>
12671 * optabs.def (while_ult_optab): New optab.
12672 * doc/md.texi (while_ult@var{m}@var{n}): Document.
12673 * internal-fn.def (WHILE_ULT): New internal function.
12674 * internal-fn.h (direct_internal_fn_supported_p): New override
12675 that takes two types as argument.
12676 * internal-fn.c (while_direct): New macro.
12677 (expand_while_optab_fn): New function.
12678 (convert_optab_supported_p): Likewise.
12679 (direct_while_optab_supported_p): New macro.
12680 * wide-int.h (wi::udiv_ceil): New function.
12681 * tree-vectorizer.h (rgroup_masks): New structure.
12682 (vec_loop_masks): New typedef.
12683 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
12684 and fully_masked_p.
12685 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
12686 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
12687 (vect_max_vf): New function.
12688 (slpeel_make_loop_iterate_ntimes): Delete.
12689 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
12690 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
12691 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
12692 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
12693 internal-fn.h, stor-layout.h and optabs-query.h.
12694 (vect_set_loop_mask): New function.
12695 (add_preheader_seq): Likewise.
12696 (add_header_seq): Likewise.
12697 (interleave_supported_p): Likewise.
12698 (vect_maybe_permute_loop_masks): Likewise.
12699 (vect_set_loop_masks_directly): Likewise.
12700 (vect_set_loop_condition_masked): Likewise.
12701 (vect_set_loop_condition_unmasked): New function, split out from
12702 slpeel_make_loop_iterate_ntimes.
12703 (slpeel_make_loop_iterate_ntimes): Rename to..
12704 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
12705 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
12706 (vect_do_peeling): Update call accordingly.
12707 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
12709 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12710 mask_compare_type, can_fully_mask_p and fully_masked_p.
12711 (release_vec_loop_masks): New function.
12712 (_loop_vec_info): Use it to free the loop masks.
12713 (can_produce_all_loop_masks_p): New function.
12714 (vect_get_max_nscalars_per_iter): Likewise.
12715 (vect_verify_full_masking): Likewise.
12716 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
12717 retries, and free the mask rgroups before retrying. Check loop-wide
12718 reasons for disallowing fully-masked loops. Make the final decision
12719 about whether use a fully-masked loop or not.
12720 (vect_estimate_min_profitable_iters): Do not assume that peeling
12721 for the number of iterations will be needed for fully-masked loops.
12722 (vectorizable_reduction): Disable fully-masked loops.
12723 (vectorizable_live_operation): Likewise.
12724 (vect_halve_mask_nunits): New function.
12725 (vect_double_mask_nunits): Likewise.
12726 (vect_record_loop_mask): Likewise.
12727 (vect_get_loop_mask): Likewise.
12728 (vect_transform_loop): Handle the case in which the final loop
12729 iteration might handle a partial vector. Call vect_set_loop_condition
12730 instead of slpeel_make_loop_iterate_ntimes.
12731 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
12732 (check_load_store_masking): New function.
12733 (prepare_load_store_mask): Likewise.
12734 (vectorizable_store): Handle fully-masked loops.
12735 (vectorizable_load): Likewise.
12736 (supportable_widening_operation): Use vect_halve_mask_nunits for
12738 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
12739 (vect_gen_while): New function.
12740 * config/aarch64/aarch64.md (umax<mode>3): New expander.
12741 (aarch64_uqdec<mode>): New insn.
12743 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12744 Alan Hayward <alan.hayward@arm.com>
12745 David Sherwood <david.sherwood@arm.com>
12747 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
12748 (reduc_xor_scal_optab): New optabs.
12749 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
12750 (reduc_xor_scal_@var{m}): Document.
12751 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
12752 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
12753 internal functions.
12754 * fold-const-call.c (fold_const_call): Handle them.
12755 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
12756 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
12757 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
12758 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
12759 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
12760 (UNSPEC_XORV): New unspecs.
12761 (optab): Add entries for them.
12762 (BITWISEV): New int iterator.
12763 (bit_reduc_op): New int attributes.
12765 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12766 Alan Hayward <alan.hayward@arm.com>
12767 David Sherwood <david.sherwood@arm.com>
12769 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
12770 * internal-fn.def (VEC_SHL_INSERT): New internal function.
12771 * optabs.def (vec_shl_insert_optab): New optab.
12772 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
12773 (duplicate_and_interleave): Likewise.
12774 * tree-vect-loop.c: Include internal-fn.h.
12775 (neutral_op_for_slp_reduction): New function, split out from
12776 get_initial_defs_for_reduction.
12777 (get_initial_def_for_reduction): Handle option 2 for variable-length
12778 vectors by loading the neutral value into a vector and then shifting
12779 the initial value into element 0.
12780 (get_initial_defs_for_reduction): Replace the code argument with
12781 the neutral value calculated by neutral_op_for_slp_reduction.
12782 Use gimple_build_vector for constant-length vectors.
12783 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
12784 but the first group_size elements have a neutral value.
12785 Use duplicate_and_interleave otherwise.
12786 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
12787 Update call to get_initial_defs_for_reduction. Handle SLP
12788 reductions for variable-length vectors by creating one vector
12789 result for each scalar result, with the elements associated
12790 with other scalar results stubbed out with the neutral value.
12791 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
12792 Require IFN_VEC_SHL_INSERT for double reductions on
12793 variable-length vectors, or SLP reductions that have
12794 a neutral value. Require can_duplicate_and_interleave_p
12795 support for variable-length unchained SLP reductions if there
12796 is no neutral value, such as for MIN/MAX reductions. Also require
12797 the number of vector elements to be a multiple of the number of
12798 SLP statements when doing variable-length unchained SLP reductions.
12799 Update call to vect_create_epilog_for_reduction.
12800 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
12801 and remove initial values.
12802 (duplicate_and_interleave): Make public.
12803 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
12804 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
12806 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12807 Alan Hayward <alan.hayward@arm.com>
12808 David Sherwood <david.sherwood@arm.com>
12810 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
12811 (can_duplicate_and_interleave_p): New function.
12812 (vect_get_and_check_slp_defs): Take the vector of statements
12813 rather than just the current one. Remove excess parentheses.
12814 Restriction rejectinon of vect_constant_def and vect_external_def
12815 for variable-length vectors to boolean types, or types for which
12816 can_duplicate_and_interleave_p is false.
12817 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
12818 (duplicate_and_interleave): New function.
12819 (vect_get_constant_vectors): Use gimple_build_vector for
12820 constant-length vectors and suitable variable-length constant
12821 vectors. Use duplicate_and_interleave for other variable-length
12822 vectors. Don't defer the update when inserting new statements.
12824 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12825 Alan Hayward <alan.hayward@arm.com>
12826 David Sherwood <david.sherwood@arm.com>
12828 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
12829 min_profitable_iters doesn't go negative.
12831 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12832 Alan Hayward <alan.hayward@arm.com>
12833 David Sherwood <david.sherwood@arm.com>
12835 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
12836 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
12837 * optabs.def (vec_mask_load_lanes_optab): New optab.
12838 (vec_mask_store_lanes_optab): Likewise.
12839 * internal-fn.def (MASK_LOAD_LANES): New internal function.
12840 (MASK_STORE_LANES): Likewise.
12841 * internal-fn.c (mask_load_lanes_direct): New macro.
12842 (mask_store_lanes_direct): Likewise.
12843 (expand_mask_load_optab_fn): Handle masked operations.
12844 (expand_mask_load_lanes_optab_fn): New macro.
12845 (expand_mask_store_optab_fn): Handle masked operations.
12846 (expand_mask_store_lanes_optab_fn): New macro.
12847 (direct_mask_load_lanes_optab_supported_p): Likewise.
12848 (direct_mask_store_lanes_optab_supported_p): Likewise.
12849 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
12851 (vect_load_lanes_supported): Likewise.
12852 * tree-vect-data-refs.c (strip_conversion): New function.
12853 (can_group_stmts_p): Likewise.
12854 (vect_analyze_data_ref_accesses): Use it instead of checking
12855 for a pair of assignments.
12856 (vect_store_lanes_supported): Take a masked_p parameter.
12857 (vect_load_lanes_supported): Likewise.
12858 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
12859 vect_store_lanes_supported and vect_load_lanes_supported.
12860 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12861 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
12862 parameter. Don't allow gaps for masked accesses.
12863 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
12864 and vect_load_lanes_supported.
12865 (get_load_store_type): Take a masked_p parameter and update
12866 call to get_group_load_store_type.
12867 (vectorizable_store): Update call to get_load_store_type.
12868 Handle IFN_MASK_STORE_LANES.
12869 (vectorizable_load): Update call to get_load_store_type.
12870 Handle IFN_MASK_LOAD_LANES.
12872 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12873 Alan Hayward <alan.hayward@arm.com>
12874 David Sherwood <david.sherwood@arm.com>
12876 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
12878 * config/aarch64/aarch64-protos.h
12879 (aarch64_sve_struct_memory_operand_p): Declare.
12880 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
12881 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
12882 (VPRED, vpred): Handle SVE structure modes.
12883 * config/aarch64/constraints.md (Utx): New constraint.
12884 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
12885 (aarch64_sve_struct_nonimmediate_operand): New predicates.
12886 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
12887 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
12888 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
12889 structure modes. Split into pieces after RA.
12890 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
12891 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
12893 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
12894 SVE structure modes.
12895 (aarch64_classify_address): Likewise.
12896 (sizetochar): Move earlier in file.
12897 (aarch64_print_operand): Handle SVE register lists.
12898 (aarch64_array_mode): New function.
12899 (aarch64_sve_struct_memory_operand_p): Likewise.
12900 (TARGET_ARRAY_MODE): Redefine.
12902 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12903 Alan Hayward <alan.hayward@arm.com>
12904 David Sherwood <david.sherwood@arm.com>
12906 * target.def (array_mode): New target hook.
12907 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
12908 * doc/tm.texi: Regenerate.
12909 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
12910 * hooks.c (hook_optmode_mode_uhwi_none): New function.
12911 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
12912 targetm.array_mode.
12913 * stor-layout.c (mode_for_array): Likewise. Support polynomial
12916 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12917 Alan Hayward <alan.hayward@arm.com>
12918 David Sherwood <david.sherwood@arm.com>
12920 * fold-const.c (fold_binary_loc): Check the argument types
12921 rather than the result type when testing for a vector operation.
12923 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12925 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
12926 * doc/tm.texi: Regenerate.
12928 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12929 Alan Hayward <alan.hayward@arm.com>
12930 David Sherwood <david.sherwood@arm.com>
12932 * doc/invoke.texi (-msve-vector-bits=): Document new option.
12933 (sve): Document new AArch64 extension.
12934 * doc/md.texi (w): Extend the description of the AArch64
12935 constraint to include SVE vectors.
12936 (Upl, Upa): Document new AArch64 predicate constraints.
12937 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
12939 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
12940 (msve-vector-bits=): New option.
12941 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
12942 SVE when these are disabled.
12943 (sve): New extension.
12944 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
12945 modes. Adjust their number of units based on aarch64_sve_vg.
12946 (MAX_BITSIZE_MODE_ANY_MODE): Define.
12947 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
12948 aarch64_addr_query_type.
12949 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
12950 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
12951 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
12952 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
12953 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
12954 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
12955 (aarch64_simd_imm_zero_p): Delete.
12956 (aarch64_check_zero_based_sve_index_immediate): Declare.
12957 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12958 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12959 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12960 (aarch64_sve_float_mul_immediate_p): Likewise.
12961 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12962 rather than an rtx.
12963 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
12964 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
12965 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
12966 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
12967 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
12968 (aarch64_regmode_natural_size): Likewise.
12969 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
12970 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
12972 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
12973 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
12974 for VG and the SVE predicate registers.
12975 (V_ALIASES): Add a "z"-prefixed alias.
12976 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
12977 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
12978 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
12979 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
12980 (REG_CLASS_NAMES): Add entries for them.
12981 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
12982 and the predicate registers.
12983 (aarch64_sve_vg): Declare.
12984 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
12985 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
12986 (REGMODE_NATURAL_SIZE): Define.
12987 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
12989 * config/aarch64/aarch64.c: Include cfgrtl.h.
12990 (simd_immediate_info): Add a constructor for series vectors,
12991 and an associated step field.
12992 (aarch64_sve_vg): New variable.
12993 (aarch64_dbx_register_number): Handle VG and the predicate registers.
12994 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
12995 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
12996 (VEC_ANY_DATA, VEC_STRUCT): New constants.
12997 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
12998 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
12999 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
13000 (aarch64_get_mask_mode): New functions.
13001 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
13002 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
13003 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
13004 predicate modes and predicate registers. Explicitly restrict
13005 GPRs to modes of 16 bytes or smaller. Only allow FP registers
13006 to store a vector mode if it is recognized by
13007 aarch64_classify_vector_mode.
13008 (aarch64_regmode_natural_size): New function.
13009 (aarch64_hard_regno_caller_save_mode): Return the original mode
13011 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
13012 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
13013 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
13014 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
13016 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
13017 does not overlap dest if the function is frame-related. Handle
13019 (aarch64_split_add_offset): New function.
13020 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13021 them aarch64_add_offset.
13022 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13023 and update call to aarch64_sub_sp.
13024 (aarch64_add_cfa_expression): New function.
13025 (aarch64_expand_prologue): Pass extra temporary registers to the
13026 functions above. Handle the case in which we need to emit new
13027 DW_CFA_expressions for registers that were originally saved
13028 relative to the stack pointer, but now have to be expressed
13029 relative to the frame pointer.
13030 (aarch64_output_mi_thunk): Pass extra temporary registers to the
13032 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
13033 IP0 and IP1 values for SVE frames.
13034 (aarch64_expand_vec_series): New function.
13035 (aarch64_expand_sve_widened_duplicate): Likewise.
13036 (aarch64_expand_sve_const_vector): Likewise.
13037 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13038 Handle SVE constants. Use emit_move_insn to move a force_const_mem
13039 into the register, rather than emitting a SET directly.
13040 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13041 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13042 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13043 (offset_9bit_signed_scaled_p): New functions.
13044 (aarch64_replicate_bitmask_imm): New function.
13045 (aarch64_bitmask_imm): Use it.
13046 (aarch64_cannot_force_const_mem): Reject expressions involving
13047 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
13048 (aarch64_classify_index): Handle SVE indices, by requiring
13049 a plain register index with a scale that matches the element size.
13050 (aarch64_classify_address): Handle SVE addresses. Assert that
13051 the mode of the address is VOIDmode or an integer mode.
13052 Update call to aarch64_classify_symbol.
13053 (aarch64_classify_symbolic_expression): Update call to
13054 aarch64_classify_symbol.
13055 (aarch64_const_vec_all_in_range_p): New function.
13056 (aarch64_print_vector_float_operand): Likewise.
13057 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
13058 "vN" for FP registers with SVE modes. Handle (const ...) vectors
13059 and the FP immediates 1.0 and 0.5.
13060 (aarch64_print_address_internal): Handle SVE addresses.
13061 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13062 (aarch64_regno_regclass): Handle predicate registers.
13063 (aarch64_secondary_reload): Handle big-endian reloads of SVE
13065 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13066 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13067 (aarch64_convert_sve_vector_bits): New function.
13068 (aarch64_override_options): Use it to handle -msve-vector-bits=.
13069 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13070 rather than an rtx.
13071 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13072 Handle SVE vector and predicate modes. Accept VL-based constants
13073 that need only one temporary register, and VL offsets that require
13074 no temporary registers.
13075 (aarch64_conditional_register_usage): Mark the predicate registers
13076 as fixed if SVE isn't available.
13077 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13078 Return true for SVE vector and predicate modes.
13079 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13080 rather than an unsigned int. Handle SVE modes.
13081 (aarch64_preferred_simd_mode): Update call accordingly. Handle
13083 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13085 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13086 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13087 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13088 (aarch64_sve_float_mul_immediate_p): New functions.
13089 (aarch64_sve_valid_immediate): New function.
13090 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13091 Explicitly reject structure modes. Check for INDEX constants.
13092 Handle PTRUE and PFALSE constants.
13093 (aarch64_check_zero_based_sve_index_immediate): New function.
13094 (aarch64_simd_imm_zero_p): Delete.
13095 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13096 vector modes. Accept constants in the range of CNT[BHWD].
13097 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13098 ask for an Advanced SIMD mode.
13099 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13100 (aarch64_simd_vector_alignment): Handle SVE predicates.
13101 (aarch64_vectorize_preferred_vector_alignment): New function.
13102 (aarch64_simd_vector_alignment_reachable): Use it instead of
13104 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13105 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13107 (MAX_VECT_LEN): Delete.
13108 (expand_vec_perm_d): Add a vec_flags field.
13109 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13110 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13111 (aarch64_evpc_ext): Don't apply a big-endian lane correction
13113 (aarch64_evpc_rev): Rename to...
13114 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
13115 (aarch64_evpc_rev_global): New function.
13116 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13117 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13119 (aarch64_evpc_sve_tbl): New function.
13120 (aarch64_expand_vec_perm_const_1): Update after rename of
13121 aarch64_evpc_rev. Handle SVE permutes too, trying
13122 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13123 than aarch64_evpc_tbl.
13124 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13125 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13126 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13127 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13128 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13129 (aarch64_expand_sve_vcond): New functions.
13130 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13131 of aarch64_vector_mode_p.
13132 (aarch64_dwarf_poly_indeterminate_value): New function.
13133 (aarch64_compute_pressure_classes): Likewise.
13134 (aarch64_can_change_mode_class): Likewise.
13135 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13136 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13137 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13138 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13139 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13140 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13141 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13142 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13144 (Dn, Dl, Dr): Accept const as well as const_vector.
13145 (Dz): Likewise. Compare against CONST0_RTX.
13146 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13147 of "vector" where appropriate.
13148 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13149 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13150 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13151 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13152 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13153 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13154 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13155 (v_int_equiv): Extend to SVE modes.
13156 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13158 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13159 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13160 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13161 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13162 (SVE_COND_FP_CMP): New int iterators.
13163 (perm_hilo): Handle the new unpack unspecs.
13164 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13166 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13167 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13168 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13169 (aarch64_equality_operator, aarch64_constant_vector_operand)
13170 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13171 (aarch64_sve_nonimmediate_operand): Likewise.
13172 (aarch64_sve_general_operand): Likewise.
13173 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13174 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13175 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13176 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
13177 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
13178 (aarch64_sve_float_arith_immediate): Likewise.
13179 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
13180 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
13181 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
13182 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
13183 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
13184 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
13185 (aarch64_sve_float_arith_operand): Likewise.
13186 (aarch64_sve_float_arith_with_sub_operand): Likewise.
13187 (aarch64_sve_float_mul_operand): Likewise.
13188 (aarch64_sve_vec_perm_operand): Likewise.
13189 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
13190 (aarch64_mov_operand): Accept const_poly_int and const_vector.
13191 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
13192 as well as const_vector.
13193 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
13194 in file. Use CONST0_RTX and CONSTM1_RTX.
13195 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
13196 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
13197 Use aarch64_simd_imm_zero.
13198 * config/aarch64/aarch64-sve.md: New file.
13199 * config/aarch64/aarch64.md: Include it.
13200 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
13201 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
13202 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
13203 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
13204 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
13205 (sve): New attribute.
13206 (enabled): Disable instructions with the sve attribute unless
13208 (movqi, movhi): Pass CONST_POLY_INT operaneds through
13209 aarch64_expand_mov_immediate.
13210 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
13211 CNT[BHSD] immediates.
13212 (movti): Split CONST_POLY_INT moves into two halves.
13213 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
13214 Split additions that need a temporary here if the destination
13215 is the stack pointer.
13216 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
13217 (*add<mode>3_poly_1): New instruction.
13218 (set_clobber_cc): New expander.
13220 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13222 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
13223 parameter and use it instead of GET_MODE_SIZE (innermode). Use
13224 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
13225 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
13226 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
13227 Change innermode from fixed_mode_size to machine_mode.
13228 (simplify_subreg): Update call accordingly. Handle a constant-sized
13229 subreg of a variable-length CONST_VECTOR.
13231 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13232 Alan Hayward <alan.hayward@arm.com>
13233 David Sherwood <david.sherwood@arm.com>
13235 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
13236 (add_offset_to_base): New function, split out from...
13237 (create_mem_ref): ...here. When handling a scale other than 1,
13238 check first whether the address is valid without the offset.
13239 Add it into the base if so, leaving the index and scale as-is.
13241 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13244 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
13245 fold_for_warn before checking if arg2 is INTEGER_CST.
13247 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
13249 * config/rs6000/predicates.md (load_multiple_operation): Delete.
13250 (store_multiple_operation): Delete.
13251 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
13252 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
13253 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
13254 guarded by TARGET_STRING.
13255 (rs6000_output_load_multiple): Delete.
13256 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
13257 OPTION_MASK_STRING / TARGET_STRING handling.
13258 (print_operand) <'N', 'O'>: Add comment that these are unused now.
13259 (const rs6000_opt_masks) <"string">: Change mask to 0.
13260 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
13261 (MASK_STRING): Delete.
13262 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
13264 (load_multiple): Delete.
13271 (store_multiple): Delete.
13278 (movmemsi_8reg): Delete.
13279 (corresponding unnamed define_insn): Delete.
13280 (movmemsi_6reg): Delete.
13281 (corresponding unnamed define_insn): Delete.
13282 (movmemsi_4reg): Delete.
13283 (corresponding unnamed define_insn): Delete.
13284 (movmemsi_2reg): Delete.
13285 (corresponding unnamed define_insn): Delete.
13286 (movmemsi_1reg): Delete.
13287 (corresponding unnamed define_insn): Delete.
13288 * config/rs6000/rs6000.opt (mno-string): New.
13289 (mstring): Replace by deprecation warning stub.
13290 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
13292 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13294 * regrename.c (regrename_do_replace): If replacing the same
13295 reg multiple times, try to reuse last created gen_raw_REG.
13298 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
13299 main to workaround a bug in GDB.
13301 2018-01-12 Tom de Vries <tom@codesourcery.com>
13304 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
13306 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
13308 PR rtl-optimization/80481
13309 * ira-color.c (get_cap_member): New function.
13310 (allocnos_conflict_by_live_ranges_p): Use it.
13311 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
13312 (setup_slot_coalesced_allocno_live_ranges): Ditto.
13314 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
13317 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
13318 (*saddl_se_1): Ditto.
13319 (*ssubsi_1): Ditto.
13320 (*ssubl_se_1): Ditto.
13322 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13324 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
13325 rather than wi::to_widest for DR_INITs.
13326 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
13327 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
13328 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
13330 (vect_analyze_group_access_1): Note that here.
13332 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13334 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
13335 polynomial type sizes.
13337 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13339 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
13340 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
13341 (gimple_add_tmp_var): Likewise.
13343 2018-01-12 Martin Liska <mliska@suse.cz>
13345 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
13346 (gimple_alloc_sizes): Likewise.
13347 (dump_gimple_statistics): Use PRIu64 in printf format.
13348 * gimple.h: Change uint64_t to int.
13350 2018-01-12 Martin Liska <mliska@suse.cz>
13352 * tree-core.h: Use uint64_t instead of int.
13353 * tree.c (tree_node_counts): Likewise.
13354 (tree_node_sizes): Likewise.
13355 (dump_tree_statistics): Use PRIu64 in printf format.
13357 2018-01-12 Martin Liska <mliska@suse.cz>
13359 * Makefile.in: As qsort_chk is implemented in vec.c, add
13360 vec.o to linkage of gencfn-macros.
13361 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
13362 passing the info to record_node_allocation_statistics.
13363 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
13365 * ggc-common.c (struct ggc_usage): Add operator== and use
13366 it in operator< and compare function.
13367 * mem-stats.h (struct mem_usage): Likewise.
13368 * vec.c (struct vec_usage): Remove operator< and compare
13369 function. Can be simply inherited.
13371 2018-01-12 Martin Jambor <mjambor@suse.cz>
13374 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
13375 * tree-ssa-math-opts.c: Include domwalk.h.
13376 (convert_mult_to_fma_1): New function.
13377 (fma_transformation_info): New type.
13378 (fma_deferring_state): Likewise.
13379 (cancel_fma_deferring): New function.
13380 (result_of_phi): Likewise.
13381 (last_fma_candidate_feeds_initial_phi): Likewise.
13382 (convert_mult_to_fma): Added deferring logic, split actual
13383 transformation to convert_mult_to_fma_1.
13384 (math_opts_dom_walker): New type.
13385 (math_opts_dom_walker::after_dom_children): New method, body moved
13386 here from pass_optimize_widening_mul::execute, added deferring logic
13388 (pass_optimize_widening_mul::execute): Moved most of code to
13389 math_opts_dom_walker::after_dom_children.
13390 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
13391 * config/i386/i386.c (ix86_option_override_internal): Added
13392 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
13394 2018-01-12 Richard Biener <rguenther@suse.de>
13397 * dwarf2out.c (gen_variable_die): Do not reset old_die for
13398 inline instance vars.
13400 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
13403 * config/rx/rx.c (rx_is_restricted_memory_address):
13404 Handle SUBREG case.
13406 2018-01-12 Richard Biener <rguenther@suse.de>
13408 PR tree-optimization/80846
13409 * target.def (split_reduction): New target hook.
13410 * targhooks.c (default_split_reduction): New function.
13411 * targhooks.h (default_split_reduction): Declare.
13412 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
13413 target requests first reduce vectors by combining low and high
13415 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
13416 (get_vectype_for_scalar_type_and_size): Export.
13417 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
13418 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
13419 * doc/tm.texi: Regenerate.
13420 * config/i386/i386.c (ix86_split_reduction): Implement
13421 TARGET_VECTORIZE_SPLIT_REDUCTION.
13423 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13426 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
13427 in PIC mode except for TARGET_VXWORKS_RTP.
13428 * config/sparc/sparc.c: Include cfgrtl.h.
13429 (TARGET_INIT_PIC_REG): Define.
13430 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
13431 (sparc_pic_register_p): New predicate.
13432 (sparc_legitimate_address_p): Use it.
13433 (sparc_legitimize_pic_address): Likewise.
13434 (sparc_delegitimize_address): Likewise.
13435 (sparc_mode_dependent_address_p): Likewise.
13436 (gen_load_pcrel_sym): Remove 4th parameter.
13437 (load_got_register): Adjust call to above. Remove obsolete stuff.
13438 (sparc_expand_prologue): Do not call load_got_register here.
13439 (sparc_flat_expand_prologue): Likewise.
13440 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
13441 (sparc_use_pseudo_pic_reg): New function.
13442 (sparc_init_pic_reg): Likewise.
13443 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13444 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
13446 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
13448 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
13449 Add item for branch_cost.
13451 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13453 PR rtl-optimization/83565
13454 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
13455 not extend the result to a larger mode for rotate operations.
13456 (num_sign_bit_copies1): Likewise.
13458 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13461 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
13463 Use values-Xc.o for -pedantic.
13464 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
13466 2018-01-12 Martin Liska <mliska@suse.cz>
13469 * ipa-devirt.c (final_warning_record::grow_type_warnings):
13471 (possible_polymorphic_call_targets): Use it.
13472 (ipa_devirt): Likewise.
13474 2018-01-12 Martin Liska <mliska@suse.cz>
13476 * profile-count.h (enum profile_quality): Use 0 as invalid
13477 enum value of profile_quality.
13479 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
13481 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
13482 -mext-string options.
13484 2018-01-12 Richard Biener <rguenther@suse.de>
13486 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
13487 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
13488 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
13490 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
13492 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13494 * configure.ac (--with-long-double-format): Add support for the
13495 configuration option to change the default long double format on
13497 * config.gcc (powerpc*-linux*-*): Likewise.
13498 * configure: Regenerate.
13499 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
13500 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
13501 used without modification.
13503 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13505 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
13506 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
13507 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
13508 MISC_BUILTIN_SPEC_BARRIER.
13509 (rs6000_init_builtins): Likewise.
13510 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
13512 (speculation_barrier): New define_insn.
13513 * doc/extend.texi: Document __builtin_speculation_barrier.
13515 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13518 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
13519 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
13520 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
13522 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
13523 integral modes instead of "ss" and "sd".
13524 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
13525 vectors with 32-bit and 64-bit elements.
13526 (vecdupssescalarmodesuffix): New mode attribute.
13527 (vec_dup<mode>): Use it.
13529 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
13532 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
13533 frame if argument is passed on stack.
13535 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13538 * ree.c (combine_reaching_defs): Optimize also
13539 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
13540 reg2=any_extend(exp); reg1=reg2;, formatting fix.
13542 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13544 PR middle-end/83189
13545 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
13547 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13549 PR middle-end/83718
13550 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
13551 after they are computed.
13553 2018-01-11 Bin Cheng <bin.cheng@arm.com>
13555 PR tree-optimization/83695
13556 * gimple-loop-linterchange.cc
13557 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
13558 reset cached scev information after interchange.
13559 (pass_linterchange::execute): Remove call to scev_reset_htab.
13561 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13563 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
13564 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
13565 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
13566 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
13567 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
13568 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
13569 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
13570 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
13571 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
13572 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
13573 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
13574 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
13575 (V_lane_reg): Likewise.
13576 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
13578 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
13579 (vfmal_lane_low<mode>_intrinsic,
13580 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
13581 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
13582 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
13583 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
13584 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
13585 vfmsl_lane_high<mode>_intrinsic): New define_insns.
13587 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13589 * config/arm/arm-cpus.in (fp16fml): New feature.
13590 (ALL_SIMD): Add fp16fml.
13591 (armv8.2-a): Add fp16fml as an option.
13592 (armv8.3-a): Likewise.
13593 (armv8.4-a): Add fp16fml as part of fp16.
13594 * config/arm/arm.h (TARGET_FP16FML): Define.
13595 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
13597 * config/arm/arm-modes.def (V2HF): Define.
13598 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
13599 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
13600 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
13601 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
13602 vfmsl_low, vfmsl_high): New set of builtins.
13603 * config/arm/iterators.md (PLUSMINUS): New code iterator.
13604 (vfml_op): New code attribute.
13605 (VFMLHALVES): New int iterator.
13606 (VFML, VFMLSEL): New mode attributes.
13607 (V_reg): Define mapping for V2HF.
13608 (V_hi, V_lo): New mode attributes.
13609 (VF_constraint): Likewise.
13610 (vfml_half, vfml_half_selector): New int attributes.
13611 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
13613 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
13614 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
13616 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
13617 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
13618 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
13619 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
13621 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
13622 Document new effective target and option set.
13624 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13626 * config/arm/arm-cpus.in (armv8_4): New feature.
13627 (ARMv8_4a): New fgroup.
13628 (armv8.4-a): New arch.
13629 * config/arm/arm-tables.opt: Regenerate.
13630 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
13631 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
13632 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
13633 Add matching rules for -march=armv8.4-a and extensions.
13634 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
13636 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
13639 * config/rx/rx.md (BW): New mode attribute.
13640 (sync_lock_test_and_setsi): Add mode suffix to insn output.
13642 2018-01-11 Richard Biener <rguenther@suse.de>
13644 PR tree-optimization/83435
13645 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
13646 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
13647 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
13649 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13650 Alan Hayward <alan.hayward@arm.com>
13651 David Sherwood <david.sherwood@arm.com>
13653 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
13655 (aarch64_classify_address): Initialize it. Track polynomial offsets.
13656 (aarch64_print_address_internal): Use it to check for a zero offset.
13658 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13659 Alan Hayward <alan.hayward@arm.com>
13660 David Sherwood <david.sherwood@arm.com>
13662 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
13663 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
13664 Return a poly_int64 rather than a HOST_WIDE_INT.
13665 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
13666 rather than a HOST_WIDE_INT.
13667 * config/aarch64/aarch64.h (aarch64_frame): Protect with
13668 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
13669 hard_fp_offset, frame_size, initial_adjust, callee_offset and
13670 final_offset from HOST_WIDE_INT to poly_int64.
13671 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
13672 to_constant when getting the number of units in an Advanced SIMD
13674 (aarch64_builtin_vectorized_function): Check for a constant number
13676 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
13678 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
13679 attribute instead of GET_MODE_NUNITS.
13680 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
13681 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
13682 GET_MODE_SIZE for fixed-size registers.
13683 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
13684 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
13685 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
13686 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
13687 (aarch64_print_operand, aarch64_print_address_internal)
13688 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
13689 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
13690 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
13691 Handle polynomial GET_MODE_SIZE.
13692 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
13693 wider than SImode without modification.
13694 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
13695 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
13696 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
13697 passing and returning SVE modes.
13698 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
13699 rather than GEN_INT.
13700 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
13701 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
13702 (aarch64_allocate_and_probe_stack_space): Likewise.
13703 (aarch64_layout_frame): Cope with polynomial offsets.
13704 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
13705 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
13706 polynomial offsets.
13707 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
13708 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
13709 poly_int64 rather than a HOST_WIDE_INT.
13710 (aarch64_get_separate_components, aarch64_process_components)
13711 (aarch64_expand_prologue, aarch64_expand_epilogue)
13712 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
13713 (aarch64_anchor_offset): New function, split out from...
13714 (aarch64_legitimize_address): ...here.
13715 (aarch64_builtin_vectorization_cost): Handle polynomial
13716 TYPE_VECTOR_SUBPARTS.
13717 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
13719 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
13720 number of elements from the PARALLEL rather than the mode.
13721 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
13722 rather than GET_MODE_BITSIZE.
13723 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
13724 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
13725 (aarch64_expand_vec_perm_const_1): Handle polynomial
13726 d->perm.length () and d->perm elements.
13727 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
13728 Apply to_constant to d->perm elements.
13729 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
13730 polynomial CONST_VECTOR_NUNITS.
13731 (aarch64_move_pointer): Take amount as a poly_int64 rather
13733 (aarch64_progress_pointer): Avoid temporary variable.
13734 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
13735 the mode attribute instead of GET_MODE.
13737 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13738 Alan Hayward <alan.hayward@arm.com>
13739 David Sherwood <david.sherwood@arm.com>
13741 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
13742 x exists before using it.
13743 (aarch64_add_constant_internal): Rename to...
13744 (aarch64_add_offset_1): ...this. Replace regnum with separate
13745 src and dest rtxes. Handle the case in which they're different,
13746 including when the offset is zero. Replace scratchreg with an rtx.
13747 Use 2 additions if there is no spare register into which we can
13748 move a 16-bit constant.
13749 (aarch64_add_constant): Delete.
13750 (aarch64_add_offset): Replace reg with separate src and dest
13751 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
13752 Use aarch64_add_offset_1.
13753 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
13754 an rtx rather than an int. Take the delta as a poly_int64
13755 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
13756 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
13757 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
13758 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
13759 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
13760 and aarch64_add_sp.
13761 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
13762 aarch64_add_constant.
13764 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13766 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
13767 Use scalar_float_mode.
13769 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13771 * config/aarch64/aarch64-simd.md
13772 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
13773 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
13774 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
13775 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
13776 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
13777 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
13778 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
13779 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
13780 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
13781 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
13783 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13786 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
13787 targ_options->x_arm_arch_string is non NULL.
13789 2018-01-11 Tamar Christina <tamar.christina@arm.com>
13791 * config/aarch64/aarch64.h
13792 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
13794 2018-01-11 Sudakshina Das <sudi.das@arm.com>
13797 * expmed.c (emit_store_flag_force): Swap if const op0
13798 and change VOIDmode to mode of op0.
13800 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13802 PR rtl-optimization/83761
13803 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
13804 than bytes to mode_for_size.
13806 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13808 PR middle-end/83189
13809 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
13810 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
13813 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13815 PR middle-end/83575
13816 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
13817 when in layout mode.
13818 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
13819 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
13822 2018-01-10 Michael Collison <michael.collison@arm.com>
13824 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
13825 * config/aarch64/aarch64-option-extension.def: Add
13826 AARCH64_OPT_EXTENSION of 'fp16fml'.
13827 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13828 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
13829 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
13830 * config/aarch64/constraints.md (Ui7): New constraint.
13831 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
13832 (VFMLA_SEL_W): Ditto.
13835 (VFMLA16_LOW): New int iterator.
13836 (VFMLA16_HIGH): Ditto.
13837 (UNSPEC_FMLAL): New unspec.
13838 (UNSPEC_FMLSL): Ditto.
13839 (UNSPEC_FMLAL2): Ditto.
13840 (UNSPEC_FMLSL2): Ditto.
13841 (f16mac): New code attribute.
13842 * config/aarch64/aarch64-simd-builtins.def
13843 (aarch64_fmlal_lowv2sf): Ditto.
13844 (aarch64_fmlsl_lowv2sf): Ditto.
13845 (aarch64_fmlalq_lowv4sf): Ditto.
13846 (aarch64_fmlslq_lowv4sf): Ditto.
13847 (aarch64_fmlal_highv2sf): Ditto.
13848 (aarch64_fmlsl_highv2sf): Ditto.
13849 (aarch64_fmlalq_highv4sf): Ditto.
13850 (aarch64_fmlslq_highv4sf): Ditto.
13851 (aarch64_fmlal_lane_lowv2sf): Ditto.
13852 (aarch64_fmlsl_lane_lowv2sf): Ditto.
13853 (aarch64_fmlal_laneq_lowv2sf): Ditto.
13854 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
13855 (aarch64_fmlalq_lane_lowv4sf): Ditto.
13856 (aarch64_fmlsl_lane_lowv4sf): Ditto.
13857 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
13858 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
13859 (aarch64_fmlal_lane_highv2sf): Ditto.
13860 (aarch64_fmlsl_lane_highv2sf): Ditto.
13861 (aarch64_fmlal_laneq_highv2sf): Ditto.
13862 (aarch64_fmlsl_laneq_highv2sf): Ditto.
13863 (aarch64_fmlalq_lane_highv4sf): Ditto.
13864 (aarch64_fmlsl_lane_highv4sf): Ditto.
13865 (aarch64_fmlalq_laneq_highv4sf): Ditto.
13866 (aarch64_fmlsl_laneq_highv4sf): Ditto.
13867 * config/aarch64/aarch64-simd.md:
13868 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
13869 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13870 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
13871 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13872 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
13873 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
13874 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
13875 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
13876 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
13877 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
13878 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
13879 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
13880 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
13881 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
13882 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
13883 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
13884 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
13885 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
13886 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
13887 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
13888 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
13889 (vfmlsl_low_u32): Ditto.
13890 (vfmlalq_low_u32): Ditto.
13891 (vfmlslq_low_u32): Ditto.
13892 (vfmlal_high_u32): Ditto.
13893 (vfmlsl_high_u32): Ditto.
13894 (vfmlalq_high_u32): Ditto.
13895 (vfmlslq_high_u32): Ditto.
13896 (vfmlal_lane_low_u32): Ditto.
13897 (vfmlsl_lane_low_u32): Ditto.
13898 (vfmlal_laneq_low_u32): Ditto.
13899 (vfmlsl_laneq_low_u32): Ditto.
13900 (vfmlalq_lane_low_u32): Ditto.
13901 (vfmlslq_lane_low_u32): Ditto.
13902 (vfmlalq_laneq_low_u32): Ditto.
13903 (vfmlslq_laneq_low_u32): Ditto.
13904 (vfmlal_lane_high_u32): Ditto.
13905 (vfmlsl_lane_high_u32): Ditto.
13906 (vfmlal_laneq_high_u32): Ditto.
13907 (vfmlsl_laneq_high_u32): Ditto.
13908 (vfmlalq_lane_high_u32): Ditto.
13909 (vfmlslq_lane_high_u32): Ditto.
13910 (vfmlalq_laneq_high_u32): Ditto.
13911 (vfmlslq_laneq_high_u32): Ditto.
13912 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
13913 (AARCH64_FL_FOR_ARCH8_4): New.
13914 (AARCH64_ISA_F16FML): New ISA flag.
13915 (TARGET_F16FML): New feature flag for fp16fml.
13916 (doc/invoke.texi): Document new fp16fml option.
13918 2018-01-10 Michael Collison <michael.collison@arm.com>
13920 * config/aarch64/aarch64-builtins.c:
13921 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
13922 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13923 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
13924 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
13925 (AARCH64_ISA_SHA3): New ISA flag.
13926 (TARGET_SHA3): New feature flag for sha3.
13927 * config/aarch64/iterators.md (sha512_op): New int attribute.
13928 (CRYPTO_SHA512): New int iterator.
13929 (UNSPEC_SHA512H): New unspec.
13930 (UNSPEC_SHA512H2): Ditto.
13931 (UNSPEC_SHA512SU0): Ditto.
13932 (UNSPEC_SHA512SU1): Ditto.
13933 * config/aarch64/aarch64-simd-builtins.def
13934 (aarch64_crypto_sha512hqv2di): New builtin.
13935 (aarch64_crypto_sha512h2qv2di): Ditto.
13936 (aarch64_crypto_sha512su0qv2di): Ditto.
13937 (aarch64_crypto_sha512su1qv2di): Ditto.
13938 (aarch64_eor3qv8hi): Ditto.
13939 (aarch64_rax1qv2di): Ditto.
13940 (aarch64_xarqv2di): Ditto.
13941 (aarch64_bcaxqv8hi): Ditto.
13942 * config/aarch64/aarch64-simd.md:
13943 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
13944 (aarch64_crypto_sha512su0qv2di): Ditto.
13945 (aarch64_crypto_sha512su1qv2di): Ditto.
13946 (aarch64_eor3qv8hi): Ditto.
13947 (aarch64_rax1qv2di): Ditto.
13948 (aarch64_xarqv2di): Ditto.
13949 (aarch64_bcaxqv8hi): Ditto.
13950 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
13951 (vsha512h2q_u64): Ditto.
13952 (vsha512su0q_u64): Ditto.
13953 (vsha512su1q_u64): Ditto.
13954 (veor3q_u16): Ditto.
13955 (vrax1q_u64): Ditto.
13956 (vxarq_u64): Ditto.
13957 (vbcaxq_u16): Ditto.
13958 * config/arm/types.md (crypto_sha512): New type attribute.
13959 (crypto_sha3): Ditto.
13960 (doc/invoke.texi): Document new sha3 option.
13962 2018-01-10 Michael Collison <michael.collison@arm.com>
13964 * config/aarch64/aarch64-builtins.c:
13965 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
13966 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13967 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
13968 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
13969 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
13970 (AARCH64_ISA_SM4): New ISA flag.
13971 (TARGET_SM4): New feature flag for sm4.
13972 * config/aarch64/aarch64-simd-builtins.def
13973 (aarch64_sm3ss1qv4si): Ditto.
13974 (aarch64_sm3tt1aq4si): Ditto.
13975 (aarch64_sm3tt1bq4si): Ditto.
13976 (aarch64_sm3tt2aq4si): Ditto.
13977 (aarch64_sm3tt2bq4si): Ditto.
13978 (aarch64_sm3partw1qv4si): Ditto.
13979 (aarch64_sm3partw2qv4si): Ditto.
13980 (aarch64_sm4eqv4si): Ditto.
13981 (aarch64_sm4ekeyqv4si): Ditto.
13982 * config/aarch64/aarch64-simd.md:
13983 (aarch64_sm3ss1qv4si): Ditto.
13984 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
13985 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
13986 (aarch64_sm4eqv4si): Ditto.
13987 (aarch64_sm4ekeyqv4si): Ditto.
13988 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
13989 (sm3part_op): Ditto.
13990 (CRYPTO_SM3TT): Ditto.
13991 (CRYPTO_SM3PART): Ditto.
13992 (UNSPEC_SM3SS1): New unspec.
13993 (UNSPEC_SM3TT1A): Ditto.
13994 (UNSPEC_SM3TT1B): Ditto.
13995 (UNSPEC_SM3TT2A): Ditto.
13996 (UNSPEC_SM3TT2B): Ditto.
13997 (UNSPEC_SM3PARTW1): Ditto.
13998 (UNSPEC_SM3PARTW2): Ditto.
13999 (UNSPEC_SM4E): Ditto.
14000 (UNSPEC_SM4EKEY): Ditto.
14001 * config/aarch64/constraints.md (Ui2): New constraint.
14002 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
14003 * config/arm/types.md (crypto_sm3): New type attribute.
14004 (crypto_sm4): Ditto.
14005 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
14006 (vsm3tt1aq_u32): Ditto.
14007 (vsm3tt1bq_u32): Ditto.
14008 (vsm3tt2aq_u32): Ditto.
14009 (vsm3tt2bq_u32): Ditto.
14010 (vsm3partw1q_u32): Ditto.
14011 (vsm3partw2q_u32): Ditto.
14012 (vsm4eq_u32): Ditto.
14013 (vsm4ekeyq_u32): Ditto.
14014 (doc/invoke.texi): Document new sm4 option.
14016 2018-01-10 Michael Collison <michael.collison@arm.com>
14018 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14019 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14020 (AARCH64_FL_FOR_ARCH8_4): New.
14021 (AARCH64_FL_V8_4): New flag.
14022 (doc/invoke.texi): Document new armv8.4-a option.
14024 2018-01-10 Michael Collison <michael.collison@arm.com>
14026 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14027 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14028 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14029 * config/aarch64/aarch64-option-extension.def: Add
14030 AARCH64_OPT_EXTENSION of 'sha2'.
14031 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14032 (crypto): Disable sha2 and aes if crypto disabled.
14033 (crypto): Enable aes and sha2 if enabled.
14034 (simd): Disable sha2 and aes if simd disabled.
14035 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14037 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14038 (TARGET_SHA2): New feature flag for sha2.
14039 (TARGET_AES): New feature flag for aes.
14040 * config/aarch64/aarch64-simd.md:
14041 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14042 conditional on TARGET_AES.
14043 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14044 (aarch64_crypto_sha1hsi): Make pattern conditional
14046 (aarch64_crypto_sha1hv4si): Ditto.
14047 (aarch64_be_crypto_sha1hv4si): Ditto.
14048 (aarch64_crypto_sha1su1v4si): Ditto.
14049 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14050 (aarch64_crypto_sha1su0v4si): Ditto.
14051 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14052 (aarch64_crypto_sha256su0v4si): Ditto.
14053 (aarch64_crypto_sha256su1v4si): Ditto.
14054 (doc/invoke.texi): Document new aes and sha2 options.
14056 2018-01-10 Martin Sebor <msebor@redhat.com>
14058 PR tree-optimization/83781
14059 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14062 2018-01-11 Martin Sebor <msebor@gmail.com>
14063 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14065 PR tree-optimization/83501
14066 PR tree-optimization/81703
14068 * tree-ssa-strlen.c (get_string_cst): Rename...
14069 (get_string_len): ...to this. Handle global constants.
14070 (handle_char_store): Adjust.
14072 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
14073 Jim Wilson <jimw@sifive.com>
14075 * config/riscv/riscv-protos.h (riscv_output_return): New.
14076 * config/riscv/riscv.c (struct machine_function): New naked_p field.
14077 (riscv_attribute_table, riscv_output_return),
14078 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14079 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14080 (riscv_compute_frame_info): Only compute frame->mask if not a naked
14082 (riscv_expand_prologue): Add early return for naked function.
14083 (riscv_expand_epilogue): Likewise.
14084 (riscv_function_ok_for_sibcall): Return false for naked function.
14085 (riscv_set_current_function): New.
14086 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14087 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14088 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14089 * doc/extend.texi (RISC-V Function Attributes): New.
14091 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14093 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14094 check for 128-bit long double before checking TCmode.
14095 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14096 128-bit long doubles before checking TFmode or TCmode.
14097 (FLOAT128_IBM_P): Likewise.
14099 2018-01-10 Martin Sebor <msebor@redhat.com>
14101 PR tree-optimization/83671
14102 * builtins.c (c_strlen): Unconditionally return zero for the empty
14104 Use -Warray-bounds for warnings.
14105 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14106 for non-constant array indices with COMPONENT_REF, arrays of
14107 arrays, and pointers to arrays.
14108 (gimple_fold_builtin_strlen): Determine and set length range for
14109 non-constant character arrays.
14111 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
14113 PR middle-end/81897
14114 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14117 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
14119 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14121 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14124 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14125 VECTOR_MEM_ALTIVEC_OR_VSX_P.
14126 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14127 indexed_or_indirect_operand predicate.
14128 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14129 (*vsx_le_perm_load_v8hi): Likewise.
14130 (*vsx_le_perm_load_v16qi): Likewise.
14131 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14132 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14133 (*vsx_le_perm_store_v8hi): Likewise.
14134 (*vsx_le_perm_store_v16qi): Likewise.
14135 (eight unnamed splitters): Likewise.
14137 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14139 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14140 * config/rs6000/emmintrin.h: Likewise.
14141 * config/rs6000/mmintrin.h: Likewise.
14142 * config/rs6000/xmmintrin.h: Likewise.
14144 2018-01-10 David Malcolm <dmalcolm@redhat.com>
14147 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14149 * tree.c (tree_nop_conversion): Return true for location wrapper
14151 (maybe_wrap_with_location): New function.
14152 (selftest::check_strip_nops): New function.
14153 (selftest::test_location_wrappers): New function.
14154 (selftest::tree_c_tests): Call it.
14155 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14156 (maybe_wrap_with_location): New decl.
14157 (EXPR_LOCATION_WRAPPER_P): New macro.
14158 (location_wrapper_p): New inline function.
14159 (tree_strip_any_location_wrapper): New inline function.
14161 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
14164 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14165 stack_realign_offset for the largest alignment of stack slot
14167 (ix86_find_max_used_stack_alignment): New function.
14168 (ix86_finalize_stack_frame_flags): Use it. Set
14169 max_used_stack_alignment if we don't realign stack.
14170 * config/i386/i386.h (machine_function): Add
14171 max_used_stack_alignment.
14173 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
14175 * config/arm/arm.opt (-mbranch-cost): New option.
14176 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
14179 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
14182 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
14183 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
14185 2018-01-10 Richard Biener <rguenther@suse.de>
14188 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
14189 early out so it also covers the case where we have a non-NULL
14192 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14194 PR tree-optimization/83753
14195 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
14196 for non-strided grouped accesses if the number of elements is 1.
14198 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14201 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
14202 * i386.h (TARGET_USE_GATHER): Define.
14203 * x86-tune.def (X86_TUNE_USE_GATHER): New.
14205 2018-01-10 Martin Liska <mliska@suse.cz>
14208 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
14209 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
14211 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
14212 CLEANUP_NO_PARTITIONING is not set.
14214 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14216 * doc/rtl.texi: Remove documentation of (const ...) wrappers
14217 for vectors, as a partial revert of r254296.
14218 * rtl.h (const_vec_p): Delete.
14219 (const_vec_duplicate_p): Don't test for vector CONSTs.
14220 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
14221 * expmed.c (make_tree): Likewise.
14224 * common.md (E, F): Use CONSTANT_P instead of checking for
14226 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
14227 checking for CONST_VECTOR.
14229 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14231 PR middle-end/83575
14232 * predict.c (force_edge_cold): Handle in more sane way edges
14233 with no prediction.
14235 2018-01-09 Carl Love <cel@us.ibm.com>
14237 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
14239 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
14240 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
14241 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
14242 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
14243 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
14244 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
14245 * config/rs6000/rs6000-protos.h: Add extern defition for
14246 rs6000_generate_float2_double_code.
14247 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
14249 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
14250 (float2_v2df): Add define_expand.
14252 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
14255 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
14256 op_mode in the force_to_mode call.
14258 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14260 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
14261 instead of checking each element individually.
14262 (aarch64_evpc_uzp): Likewise.
14263 (aarch64_evpc_zip): Likewise.
14264 (aarch64_evpc_ext): Likewise.
14265 (aarch64_evpc_rev): Likewise.
14266 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
14267 instead of checking each element individually. Return true without
14269 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
14270 whether all selected elements come from the same input, instead of
14271 checking each element individually. Remove calls to gen_rtx_REG,
14272 start_sequence and end_sequence and instead assert that no rtl is
14275 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14277 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
14278 order of HIGH and CONST checks.
14280 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14282 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
14283 if the destination isn't an SSA_NAME.
14285 2018-01-09 Richard Biener <rguenther@suse.de>
14287 PR tree-optimization/83668
14288 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
14290 (canonicalize_loop_form): ... here, renamed from ...
14291 (canonicalize_loop_closed_ssa_form): ... this and amended to
14292 swap successor edges for loop exit blocks to make us use
14293 the RPO order we need for initial schedule generation.
14295 2018-01-09 Joseph Myers <joseph@codesourcery.com>
14297 PR tree-optimization/64811
14298 * match.pd: When optimizing comparisons with Inf, avoid
14299 introducing or losing exceptions from comparisons with NaN.
14301 2018-01-09 Martin Liska <mliska@suse.cz>
14304 * asan.c (shadow_mem_size): Add gcc_assert.
14306 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
14308 Don't save registers in main().
14311 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
14312 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
14313 * config/avr/avr.c (avr_set_current_function): Don't error if
14314 naked, OS_task or OS_main are specified at the same time.
14315 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
14317 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
14319 * common/config/avr/avr-common.c (avr_option_optimization_table):
14320 Switch on -mmain-is-OS_task for optimizing compilations.
14322 2018-01-09 Richard Biener <rguenther@suse.de>
14324 PR tree-optimization/83572
14325 * graphite.c: Include cfganal.h.
14326 (graphite_transform_loops): Connect infinite loops to exit
14327 and remove fake edges at the end.
14329 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14331 * ipa-inline.c (edge_badness): Revert accidental checkin.
14333 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14336 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
14337 symbols; not inline clones.
14339 2018-01-09 Jakub Jelinek <jakub@redhat.com>
14342 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
14343 hard registers. Formatting fixes.
14345 PR preprocessor/83722
14346 * gcc.c (try_generate_repro): Pass
14347 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
14348 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
14351 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
14352 Kito Cheng <kito.cheng@gmail.com>
14354 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
14355 (riscv_leaf_function_p): Delete.
14356 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
14358 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14360 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
14362 (do_ifelse): New function.
14363 (do_isel): New function.
14364 (do_sub3): New function.
14365 (do_add3): New function.
14366 (do_load_mask_compare): New function.
14367 (do_overlap_load_compare): New function.
14368 (expand_compare_loop): New function.
14369 (expand_block_compare): Call expand_compare_loop() when appropriate.
14370 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
14371 option description.
14372 (-mblock-compare-inline-loop-limit): New option.
14374 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14377 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
14378 Reverse order of second and third operands in first alternative.
14379 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
14380 of first and second elements in UNSPEC_VPERMR vector.
14381 (altivec_expand_vec_perm_le): Likewise.
14383 2018-01-08 Jeff Law <law@redhat.com>
14385 PR rtl-optimizatin/81308
14386 * tree-switch-conversion.c (cfg_altered): New file scoped static.
14387 (process_switch): If group_case_labels makes a change, then set
14389 (pass_convert_switch::execute): If a switch is converted, then
14390 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
14392 PR rtl-optimization/81308
14393 * recog.c (split_all_insns): Conditionally cleanup the CFG after
14396 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
14398 PR target/83663 - Revert r255946
14399 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
14400 generation for cases where splatting a value is not useful.
14401 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
14402 across a vec_duplicate and a paradoxical subreg forming a vector
14403 mode to a vec_concat.
14405 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14407 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
14408 -march=armv8.3-a variants.
14409 * config/arm/t-multilib: Likewise.
14410 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
14412 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14414 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
14416 (cceq_ior_compare_complement): Give it a name so I can use it, and
14417 change boolean_or_operator predicate to boolean_operator so it can
14418 be used to generate a crand.
14419 (eqne): New code iterator.
14420 (bd/bd_neg): New code_attrs.
14421 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
14422 a single define_insn.
14423 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
14424 decrement (bdnzt/bdnzf/bdzt/bdzf).
14425 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
14426 with the new names of the branch decrement patterns, and added the
14427 names of the branch decrement conditional patterns.
14429 2018-01-08 Richard Biener <rguenther@suse.de>
14431 PR tree-optimization/83563
14432 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
14435 2018-01-08 Richard Biener <rguenther@suse.de>
14437 PR middle-end/83713
14438 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
14440 2018-01-08 Richard Biener <rguenther@suse.de>
14442 PR tree-optimization/83685
14443 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
14444 references to abnormals.
14446 2018-01-08 Richard Biener <rguenther@suse.de>
14449 * dwarf2out.c (output_indirect_strings): Handle empty
14450 skeleton_debug_str_hash.
14451 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
14453 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14455 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
14456 (emit_store_direct): Likewise.
14457 (arc_trampoline_adjust_address): Likewise.
14458 (arc_asm_trampoline_template): New function.
14459 (arc_initialize_trampoline): Use asm_trampoline_template.
14460 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
14461 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
14462 * config/arc/arc.md (flush_icache): Delete pattern.
14464 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14466 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
14467 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
14470 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14473 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
14474 by not USED_FOR_TARGET.
14475 (make_pass_resolve_sw_modes): Likewise.
14477 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14479 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
14482 2018-01-08 Richard Biener <rguenther@suse.de>
14484 PR middle-end/83580
14485 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
14487 2018-01-08 Richard Biener <rguenther@suse.de>
14489 PR middle-end/83517
14490 * match.pd ((t * 2) / 2) -> t): Add missing :c.
14492 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
14494 PR middle-end/81897
14495 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
14496 basic blocks with a small number of successors.
14497 (convert_control_dep_chain_into_preds): Improve handling of
14499 (dump_predicates): Split apart into...
14500 (dump_pred_chain): ...here...
14501 (dump_pred_info): ...and here.
14502 (can_one_predicate_be_invalidated_p): Add debugging printfs.
14503 (can_chain_union_be_invalidated_p): Improve check for invalidation
14505 (uninit_uses_cannot_happen): Avoid unnecessary if
14506 convert_control_dep_chain_into_preds yielded nothing.
14508 2018-01-06 Martin Sebor <msebor@redhat.com>
14510 PR tree-optimization/83640
14511 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
14512 subtracting negative offset from size.
14513 (builtin_access::overlap): Adjust offset bounds of the access to fall
14514 within the size of the object if possible.
14516 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
14518 PR rtl-optimization/83699
14519 * expmed.c (extract_bit_field_1): Restrict the vector usage of
14520 extract_bit_field_as_subreg to cases in which the extracted
14521 value is also a vector.
14523 * lra-constraints.c (process_alt_operands): Test for the equivalence
14524 substitutions when detecting a possible reload cycle.
14526 2018-01-06 Jakub Jelinek <jakub@redhat.com>
14529 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
14530 by default if flag_selective_schedling{,2}. Formatting fixes.
14532 PR rtl-optimization/83682
14533 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
14534 if it has non-VECTOR_MODE element mode.
14535 (vec_duplicate_p): Likewise.
14537 PR middle-end/83694
14538 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
14539 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
14541 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14544 * config/i386/i386-builtin.def
14545 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
14546 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
14547 Require also OPTION_MASK_ISA_AVX512F in addition to
14548 OPTION_MASK_ISA_GFNI.
14549 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
14550 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
14551 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
14552 to OPTION_MASK_ISA_GFNI.
14553 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
14554 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
14555 OPTION_MASK_ISA_AVX512BW.
14556 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
14557 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
14558 addition to OPTION_MASK_ISA_GFNI.
14559 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
14560 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
14561 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
14562 to OPTION_MASK_ISA_GFNI.
14563 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
14564 a requirement for all ISAs rather than any of them with a few
14566 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
14568 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
14569 bitmasks to be enabled with 3 exceptions, instead of requiring any
14570 enabled ISA with lots of exceptions.
14571 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
14572 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
14573 Change avx512bw in isa attribute to avx512f.
14574 * config/i386/sgxintrin.h: Add license boilerplate.
14575 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
14576 to __AVX512F__ and __AVX512VL to __AVX512VL__.
14577 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
14578 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
14580 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
14581 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
14582 temporarily sse2 rather than sse if not enabled already.
14585 * config/i386/sse.md (VI248_VLBW): Rename to ...
14586 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
14587 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
14588 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
14589 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
14590 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
14591 mode iterator instead of VI248_VLBW.
14593 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
14595 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
14596 (record_modified): Skip clobbers; add debug output.
14597 (param_change_prob): Use sreal frequencies.
14599 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14601 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
14602 punt for user-aligned variables.
14604 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14606 * tree-chrec.c (chrec_contains_symbols): Return true for
14609 2018-01-05 Sudakshina Das <sudi.das@arm.com>
14612 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
14613 of (x|y) == x for BICS pattern.
14615 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14617 PR tree-optimization/83605
14618 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
14619 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
14622 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
14624 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
14625 * config/epiphany/rtems.h: New file.
14627 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14628 Uros Bizjak <ubizjak@gmail.com>
14631 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
14632 QIreg_operand instead of register_operand predicate.
14633 * config/i386/i386.c (ix86_rop_should_change_byte_p,
14634 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
14635 comments instead of -fmitigate[-_]rop.
14637 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14640 * cgraphunit.c (symbol_table::compile): Switch to text_section
14641 before calling assembly_start debug hook.
14642 * run-rtl-passes.c (run_rtl_passes): Likewise.
14645 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14647 * tree-vrp.c (extract_range_from_binary_expr_1): Check
14648 range_int_cst_p rather than !symbolic_range_p before calling
14649 extract_range_from_multiplicative_op_1.
14651 2018-01-04 Jeff Law <law@redhat.com>
14653 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
14654 redundant test in assertion.
14656 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14658 * doc/rtl.texi: Document machine_mode wrapper classes.
14660 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14662 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
14663 using tree_to_uhwi.
14665 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14667 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
14668 the VEC_PERM_EXPR fold to fail.
14670 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14673 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
14674 to switched_sections.
14676 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14679 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
14680 test for d.testing.
14682 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
14685 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
14686 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
14688 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14691 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
14692 is BLKmode and bitpos not zero or mode change is needed.
14694 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14697 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
14700 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
14703 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
14704 instead of MULT rtx. Update all corresponding splitters.
14705 (*saddl_se): Ditto.
14706 (*ssub<modesuffix>): Ditto.
14707 (*ssubl_se): Ditto.
14708 (*cmp_sadd_di): Update split patterns.
14709 (*cmp_sadd_si): Ditto.
14710 (*cmp_sadd_sidi): Ditto.
14711 (*cmp_ssub_di): Ditto.
14712 (*cmp_ssub_si): Ditto.
14713 (*cmp_ssub_sidi): Ditto.
14714 * config/alpha/predicates.md (const23_operand): New predicate.
14715 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
14716 Look for ASHIFT, not MULT inner operand.
14717 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
14719 2018-01-04 Martin Liska <mliska@suse.cz>
14721 PR gcov-profile/83669
14722 * gcov.c (output_intermediate_file): Add version to intermediate
14724 * doc/gcov.texi: Document new field 'version' in intermediate
14725 file format. Fix location of '-k' option of gcov command.
14727 2018-01-04 Martin Liska <mliska@suse.cz>
14730 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
14732 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14734 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
14736 2018-01-03 Martin Sebor <msebor@redhat.com>
14738 PR tree-optimization/83655
14739 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
14740 checking calls with invalid arguments.
14742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14744 * tree-vect-stmts.c (vect_get_store_rhs): New function.
14745 (vectorizable_mask_load_store): Delete.
14746 (vectorizable_call): Return false for masked loads and stores.
14747 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
14748 instead of gimple_assign_rhs1.
14749 (vectorizable_load): Handle IFN_MASK_LOAD.
14750 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
14752 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14754 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
14756 (vectorizable_mask_load_store): ...here.
14757 (vectorizable_load): ...and here.
14759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14761 * tree-vect-stmts.c (vect_build_all_ones_mask)
14762 (vect_build_zero_merge_argument): New functions, split out from...
14763 (vectorizable_load): ...here.
14765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14767 * tree-vect-stmts.c (vect_check_store_rhs): New function,
14769 (vectorizable_mask_load_store): ...here.
14770 (vectorizable_store): ...and here.
14772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14774 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
14776 (vectorizable_mask_load_store): ...here.
14778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14780 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
14781 (vect_model_store_cost): Take a vec_load_store_type instead of a
14783 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
14784 (vect_model_store_cost): Take a vec_load_store_type instead of a
14786 (vectorizable_mask_load_store): Update accordingly.
14787 (vectorizable_store): Likewise.
14788 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
14790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14792 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
14793 IFN_MASK_LOAD calls here rather than...
14794 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
14796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14797 Alan Hayward <alan.hayward@arm.com>
14798 David Sherwood <david.sherwood@arm.com>
14800 * expmed.c (extract_bit_field_1): For vector extracts,
14801 fall back to extract_bit_field_as_subreg if vec_extract
14804 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14805 Alan Hayward <alan.hayward@arm.com>
14806 David Sherwood <david.sherwood@arm.com>
14808 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
14809 they are variable or constant sized.
14810 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
14811 slots for constant-sized data.
14813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14814 Alan Hayward <alan.hayward@arm.com>
14815 David Sherwood <david.sherwood@arm.com>
14817 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
14818 handling COND_EXPRs with boolean comparisons, try to find a better
14819 basis for the mask type than the boolean itself.
14821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14823 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
14824 is calculated and how it can be overridden.
14825 * genmodes.c (max_bitsize_mode_any_mode): New variable.
14826 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
14828 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
14831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14832 Alan Hayward <alan.hayward@arm.com>
14833 David Sherwood <david.sherwood@arm.com>
14835 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
14836 Remove the mode argument.
14837 (aarch64_simd_valid_immediate): Remove the mode and inverse
14839 * config/aarch64/iterators.md (bitsize): New iterator.
14840 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
14841 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
14842 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
14843 aarch64_simd_valid_immediate.
14844 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
14845 (aarch64_reg_or_bic_imm): Likewise.
14846 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
14847 with an insn_type enum and msl with a modifier_type enum.
14848 Replace element_width with a scalar_mode. Change the shift
14849 to unsigned int. Add constructors for scalar_float_mode and
14850 scalar_int_mode elements.
14851 (aarch64_vect_float_const_representable_p): Delete.
14852 (aarch64_can_const_movi_rtx_p)
14853 (aarch64_simd_scalar_immediate_valid_for_move)
14854 (aarch64_simd_make_constant): Update call to
14855 aarch64_simd_valid_immediate.
14856 (aarch64_advsimd_valid_immediate_hs): New function.
14857 (aarch64_advsimd_valid_immediate): Likewise.
14858 (aarch64_simd_valid_immediate): Remove mode and inverse
14859 arguments. Rewrite to use the above. Use const_vec_duplicate_p
14860 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
14861 and aarch64_float_const_representable_p on the result.
14862 (aarch64_output_simd_mov_immediate): Remove mode argument.
14863 Update call to aarch64_simd_valid_immediate and use of
14864 simd_immediate_info.
14865 (aarch64_output_scalar_simd_mov_immediate): Update call
14868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14869 Alan Hayward <alan.hayward@arm.com>
14870 David Sherwood <david.sherwood@arm.com>
14872 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
14873 (mode_nunits): Likewise CONST_MODE_NUNITS.
14874 * machmode.def (ADJUST_NUNITS): Document.
14875 * genmodes.c (mode_data::need_nunits_adj): New field.
14876 (blank_mode): Update accordingly.
14877 (adj_nunits): New variable.
14878 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
14880 (emit_mode_size_inline): Set need_bytesize_adj for all modes
14881 listed in adj_nunits.
14882 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
14883 listed in adj_nunits. Don't emit case statements for such modes.
14884 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
14885 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
14886 nothing if adj_nunits is nonnull.
14887 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
14888 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
14889 (emit_mode_fbit): Update use of print_maybe_const_decl.
14890 (emit_move_size): Likewise. Treat the array as non-const
14892 (emit_mode_adjustments): Handle adj_nunits.
14894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14896 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
14897 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
14898 (VECTOR_MODES): Use it.
14899 (make_vector_modes): Take the prefix as an argument.
14901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14902 Alan Hayward <alan.hayward@arm.com>
14903 David Sherwood <david.sherwood@arm.com>
14905 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
14906 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
14907 for MODE_VECTOR_BOOL.
14908 * machmode.def (VECTOR_BOOL_MODE): Document.
14909 * genmodes.c (VECTOR_BOOL_MODE): New macro.
14910 (make_vector_bool_mode): New function.
14911 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
14913 * lto-streamer-in.c (lto_input_mode_table): Likewise.
14914 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
14916 * stor-layout.c (int_mode_for_mode): Likewise.
14917 * tree.c (build_vector_type_for_mode): Likewise.
14918 * varasm.c (output_constant_pool_2): Likewise.
14919 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
14920 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
14921 for MODE_VECTOR_BOOL.
14922 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
14923 of mode class checks.
14924 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
14925 instead of a list of mode class checks.
14926 (expand_vector_scalar_condition): Likewise.
14927 (type_for_widest_vector_mode): Handle BImode as an inner mode.
14929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14930 Alan Hayward <alan.hayward@arm.com>
14931 David Sherwood <david.sherwood@arm.com>
14933 * machmode.h (mode_size): Change from unsigned short to
14935 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
14936 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14937 or if measurement_type is not polynomial.
14938 (fixed_size_mode::includes_p): Check for constant-sized modes.
14939 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
14940 return a poly_uint16 rather than an unsigned short.
14941 (emit_mode_size): Change the type of mode_size from unsigned short
14942 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
14943 (emit_mode_adjustments): Cope with polynomial vector sizes.
14944 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14946 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14948 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
14949 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
14950 * caller-save.c (setup_save_areas): Likewise.
14951 (replace_reg_with_saved_mem): Likewise.
14952 * calls.c (emit_library_call_value_1): Likewise.
14953 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
14954 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
14955 (gen_lowpart_for_combine): Likewise.
14956 * convert.c (convert_to_integer_1): Likewise.
14957 * cse.c (equiv_constant, cse_insn): Likewise.
14958 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
14959 (cselib_subst_to_values): Likewise.
14960 * dce.c (word_dce_process_block): Likewise.
14961 * df-problems.c (df_word_lr_mark_ref): Likewise.
14962 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
14963 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
14964 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
14965 (rtl_for_decl_location): Likewise.
14966 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
14967 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
14968 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
14969 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
14970 (expand_expr_real_1): Likewise.
14971 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
14972 (pad_below): Likewise.
14973 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14974 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
14975 * ira.c (get_subreg_tracking_sizes): Likewise.
14976 * ira-build.c (ira_create_allocno_objects): Likewise.
14977 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
14978 (ira_sort_regnos_for_alter_reg): Likewise.
14979 * ira-costs.c (record_operand_costs): Likewise.
14980 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
14981 (resolve_simple_move): Likewise.
14982 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
14983 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
14984 (lra_constraints): Likewise.
14985 (CONST_POOL_OK_P): Reject variable-sized modes.
14986 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
14987 (add_pseudo_to_slot, lra_spill): Likewise.
14988 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14989 * optabs-query.c (get_best_extraction_insn): Likewise.
14990 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14991 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
14992 (expand_mult_highpart, valid_multiword_target_p): Likewise.
14993 * recog.c (offsettable_address_addr_space_p): Likewise.
14994 * regcprop.c (maybe_mode_change): Likewise.
14995 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
14996 * regrename.c (build_def_use): Likewise.
14997 * regstat.c (dump_reg_info): Likewise.
14998 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
14999 (find_reloads, find_reloads_subreg_address): Likewise.
15000 * reload1.c (eliminate_regs_1): Likewise.
15001 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
15002 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
15003 (simplify_binary_operation_1, simplify_subreg): Likewise.
15004 * targhooks.c (default_function_arg_padding): Likewise.
15005 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
15006 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
15007 (verify_gimple_assign_ternary): Likewise.
15008 * tree-inline.c (estimate_move_cost): Likewise.
15009 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15010 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
15011 (get_address_cost_ainc): Likewise.
15012 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
15013 (vect_supportable_dr_alignment): Likewise.
15014 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15015 (vectorizable_reduction): Likewise.
15016 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15017 (vectorizable_operation, vectorizable_load): Likewise.
15018 * tree.c (build_same_sized_truth_vector_type): Likewise.
15019 * valtrack.c (cleanup_auto_inc_dec): Likewise.
15020 * var-tracking.c (emit_note_insn_var_location): Likewise.
15021 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15022 (ADDR_VEC_ALIGN): Likewise.
15024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15025 Alan Hayward <alan.hayward@arm.com>
15026 David Sherwood <david.sherwood@arm.com>
15028 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
15030 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15031 or if measurement_type is polynomial.
15032 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15033 * combine.c (make_extraction): Likewise.
15034 * dse.c (find_shift_sequence): Likewise.
15035 * dwarf2out.c (mem_loc_descriptor): Likewise.
15036 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15037 (extract_bit_field, extract_low_bits): Likewise.
15038 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15039 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15040 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15041 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15042 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15043 * reload.c (find_reloads): Likewise.
15044 * reload1.c (alter_reg): Likewise.
15045 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15046 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15047 * tree-if-conv.c (predicate_mem_writes): Likewise.
15048 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15049 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15050 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15051 * valtrack.c (dead_debug_insert_temp): Likewise.
15052 * varasm.c (mergeable_constant_section): Likewise.
15053 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15056 Alan Hayward <alan.hayward@arm.com>
15057 David Sherwood <david.sherwood@arm.com>
15059 * expr.c (expand_assignment): Cope with polynomial mode sizes
15060 when assigning to a CONCAT.
15062 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15063 Alan Hayward <alan.hayward@arm.com>
15064 David Sherwood <david.sherwood@arm.com>
15066 * machmode.h (mode_precision): Change from unsigned short to
15068 (mode_to_precision): Return a poly_uint16 rather than an unsigned
15070 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15071 or if measurement_type is not polynomial.
15072 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
15073 in which the mode is already known to be a scalar_int_mode.
15074 * genmodes.c (emit_mode_precision): Change the type of mode_precision
15075 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
15077 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15078 for GET_MODE_PRECISION.
15079 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15080 for GET_MODE_PRECISION.
15081 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15083 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15084 (expand_field_assignment, make_extraction): Likewise.
15085 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15086 (get_last_value): Likewise.
15087 * convert.c (convert_to_integer_1): Likewise.
15088 * cse.c (cse_insn): Likewise.
15089 * expr.c (expand_expr_real_1): Likewise.
15090 * lra-constraints.c (simplify_operand_subreg): Likewise.
15091 * optabs-query.c (can_atomic_load_p): Likewise.
15092 * optabs.c (expand_atomic_load): Likewise.
15093 (expand_atomic_store): Likewise.
15094 * ree.c (combine_reaching_defs): Likewise.
15095 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15096 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15097 * tree.h (type_has_mode_precision_p): Likewise.
15098 * ubsan.c (instrument_si_overflow): Likewise.
15100 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15101 Alan Hayward <alan.hayward@arm.com>
15102 David Sherwood <david.sherwood@arm.com>
15104 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15105 polynomial numbers of units.
15106 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15107 (valid_vector_subparts_p): New function.
15108 (build_vector_type): Remove temporary shim and take the number
15109 of units as a poly_uint64 rather than an int.
15110 (build_opaque_vector_type): Take the number of units as a
15111 poly_uint64 rather than an int.
15112 * tree.c (build_vector_from_ctor): Handle polynomial
15113 TYPE_VECTOR_SUBPARTS.
15114 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15115 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15116 (build_vector_from_val): If the number of units is variable,
15117 use build_vec_duplicate_cst for constant operands and
15118 VEC_DUPLICATE_EXPR otherwise.
15119 (make_vector_type): Remove temporary is_constant ().
15120 (build_vector_type, build_opaque_vector_type): Take the number of
15121 units as a poly_uint64 rather than an int.
15122 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15124 * cfgexpand.c (expand_debug_expr): Likewise.
15125 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15126 (store_constructor, expand_expr_real_1): Likewise.
15127 (const_scalar_mask_from_tree): Likewise.
15128 * fold-const-call.c (fold_const_reduction): Likewise.
15129 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15130 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15131 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15132 (fold_relational_const): Likewise.
15133 (native_interpret_vector): Likewise. Change the size from an
15134 int to an unsigned int.
15135 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15136 TYPE_VECTOR_SUBPARTS.
15137 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15138 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15139 duplicating a non-constant operand into a variable-length vector.
15140 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15141 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15142 * ipa-icf.c (sem_variable::equals): Likewise.
15143 * match.pd: Likewise.
15144 * omp-simd-clone.c (simd_clone_subparts): Likewise.
15145 * print-tree.c (print_node): Likewise.
15146 * stor-layout.c (layout_type): Likewise.
15147 * targhooks.c (default_builtin_vectorization_cost): Likewise.
15148 * tree-cfg.c (verify_gimple_comparison): Likewise.
15149 (verify_gimple_assign_binary): Likewise.
15150 (verify_gimple_assign_ternary): Likewise.
15151 (verify_gimple_assign_single): Likewise.
15152 * tree-pretty-print.c (dump_generic_node): Likewise.
15153 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15154 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15155 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15156 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15157 (vect_shift_permute_load_chain): Likewise.
15158 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15159 (expand_vector_condition, optimize_vector_constructor): Likewise.
15160 (lower_vec_perm, get_compute_type): Likewise.
15161 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15162 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15163 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15164 (vect_recog_mask_conversion_pattern): Likewise.
15165 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15166 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15167 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15168 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15169 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15170 (vectorizable_shift, vectorizable_operation, vectorizable_store)
15171 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15172 (supportable_widening_operation): Likewise.
15173 (supportable_narrowing_operation): Likewise.
15174 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15176 * varasm.c (output_constant): Likewise.
15178 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15179 Alan Hayward <alan.hayward@arm.com>
15180 David Sherwood <david.sherwood@arm.com>
15182 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
15183 so that both the length == 3 and length != 3 cases set up their
15184 own permute vectors. Add comments explaining why we know the
15185 number of elements is constant.
15186 (vect_permute_load_chain): Likewise.
15188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15189 Alan Hayward <alan.hayward@arm.com>
15190 David Sherwood <david.sherwood@arm.com>
15192 * machmode.h (mode_nunits): Change from unsigned char to
15194 (ONLY_FIXED_SIZE_MODES): New macro.
15195 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
15196 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
15197 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
15199 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
15200 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
15201 or if measurement_type is not polynomial.
15202 * genmodes.c (ZERO_COEFFS): New macro.
15203 (emit_mode_nunits_inline): Make mode_nunits_inline return a
15205 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
15206 Use ZERO_COEFFS when emitting initializers.
15207 * data-streamer.h (bp_pack_poly_value): New function.
15208 (bp_unpack_poly_value): Likewise.
15209 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15210 for GET_MODE_NUNITS.
15211 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15212 for GET_MODE_NUNITS.
15213 * tree.c (make_vector_type): Remove temporary shim and make
15214 the real function take the number of units as a poly_uint64
15215 rather than an int.
15216 (build_vector_type_for_mode): Handle polynomial nunits.
15217 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
15218 * emit-rtl.c (const_vec_series_p_1): Likewise.
15219 (gen_rtx_CONST_VECTOR): Likewise.
15220 * fold-const.c (test_vec_duplicate_folding): Likewise.
15221 * genrecog.c (validate_pattern): Likewise.
15222 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
15223 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15224 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
15225 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
15226 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
15227 * rtlanal.c (subreg_get_info): Likewise.
15228 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15229 (vect_grouped_load_supported): Likewise.
15230 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
15231 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
15232 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15233 (simplify_const_unary_operation, simplify_binary_operation_1)
15234 (simplify_const_binary_operation, simplify_ternary_operation)
15235 (test_vector_ops_duplicate, test_vector_ops): Likewise.
15236 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
15237 instead of CONST_VECTOR_NUNITS.
15238 * varasm.c (output_constant_pool_2): Likewise.
15239 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
15240 explicit-encoded elements in the XVEC for variable-length vectors.
15242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15244 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
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 * coretypes.h (fixed_size_mode): Declare.
15251 (fixed_size_mode_pod): New typedef.
15252 * builtins.h (target_builtins::x_apply_args_mode)
15253 (target_builtins::x_apply_result_mode): Change type to
15254 fixed_size_mode_pod.
15255 * builtins.c (apply_args_size, apply_result_size, result_vector)
15256 (expand_builtin_apply_args_1, expand_builtin_apply)
15257 (expand_builtin_return): Update accordingly.
15259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15261 * cse.c (hash_rtx_cb): Hash only the encoded elements.
15262 * cselib.c (cselib_hash_rtx): Likewise.
15263 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
15264 CONST_VECTOR encoding.
15266 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15267 Jeff Law <law@redhat.com>
15270 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
15271 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
15272 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
15273 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
15276 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
15277 explicitly probe *sp in a noreturn function if there were any callee
15278 register saves or frame pointer is needed.
15280 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15283 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
15284 BLKmode for ternary, binary or unary expressions.
15287 * var-tracking.c (delete_vta_debug_insn): New inline function.
15288 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
15289 insns from get_insns () to NULL instead of each bb separately.
15290 Use delete_vta_debug_insn. No longer static.
15291 (vt_debug_insns_local, variable_tracking_main_1): Adjust
15292 delete_vta_debug_insns callers.
15293 * rtl.h (delete_vta_debug_insns): Declare.
15294 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
15295 instead of variable_tracking_main.
15297 2018-01-03 Martin Sebor <msebor@redhat.com>
15299 PR tree-optimization/83603
15300 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
15301 arguments past the endof the argument list in functions declared
15302 without a prototype.
15303 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
15304 Avoid checking when arguments are null.
15306 2018-01-03 Martin Sebor <msebor@redhat.com>
15309 * doc/extend.texi (attribute const): Fix a typo.
15310 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
15311 issuing -Wsuggest-attribute for void functions.
15313 2018-01-03 Martin Sebor <msebor@redhat.com>
15315 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
15316 offset_int::from instead of wide_int::to_shwi.
15317 (maybe_diag_overlap): Remove assertion.
15318 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
15319 * gimple-ssa-sprintf.c (format_directive): Same.
15320 (parse_directive): Same.
15321 (sprintf_dom_walker::compute_format_length): Same.
15322 (try_substitute_return_value): Same.
15324 2018-01-03 Jeff Law <law@redhat.com>
15326 PR middle-end/83654
15327 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
15328 non-constant residual for zero at runtime and avoid probing in
15329 that case. Reorganize code for trailing problem to mirror handling
15332 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15334 PR tree-optimization/83501
15335 * tree-ssa-strlen.c (get_string_cst): New.
15336 (handle_char_store): Call get_string_cst.
15338 2018-01-03 Martin Liska <mliska@suse.cz>
15340 PR tree-optimization/83593
15341 * tree-ssa-strlen.c: Include tree-cfg.h.
15342 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
15343 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
15344 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
15346 (strlen_dom_walker::before_dom_children): Call
15347 gimple_purge_dead_eh_edges. Dump tranformation with details
15349 (strlen_dom_walker::before_dom_children): Update call by adding
15350 new argument cleanup_eh.
15351 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
15353 2018-01-03 Martin Liska <mliska@suse.cz>
15356 * cif-code.def (VARIADIC_THUNK): New enum value.
15357 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
15360 2018-01-03 Jan Beulich <jbeulich@suse.com>
15362 * sse.md (mov<mode>_internal): Tighten condition for when to use
15363 vmovdqu<ssescalarsize> for TI and OI modes.
15365 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15367 Update copyright years.
15369 2018-01-03 Martin Liska <mliska@suse.cz>
15372 * ipa-visibility.c (function_and_variable_visibility): Skip
15373 functions with noipa attribure.
15375 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15377 * gcc.c (process_command): Update copyright notice dates.
15378 * gcov-dump.c (print_version): Ditto.
15379 * gcov.c (print_version): Ditto.
15380 * gcov-tool.c (print_version): Ditto.
15381 * gengtype.c (create_file): Ditto.
15382 * doc/cpp.texi: Bump @copying's copyright year.
15383 * doc/cppinternals.texi: Ditto.
15384 * doc/gcc.texi: Ditto.
15385 * doc/gccint.texi: Ditto.
15386 * doc/gcov.texi: Ditto.
15387 * doc/install.texi: Ditto.
15388 * doc/invoke.texi: Ditto.
15390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15392 * vector-builder.h (vector_builder::m_full_nelts): Change from
15393 unsigned int to poly_uint64.
15394 (vector_builder::full_nelts): Update prototype accordingly.
15395 (vector_builder::new_vector): Likewise.
15396 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
15397 (vector_builder::operator ==): Likewise.
15398 (vector_builder::finalize): Likewise.
15399 * int-vector-builder.h (int_vector_builder::int_vector_builder):
15400 Take the number of elements as a poly_uint64 rather than an
15402 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
15403 from unsigned int to poly_uint64.
15404 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
15405 (vec_perm_indices::new_vector): Likewise.
15406 (vec_perm_indices::length): Likewise.
15407 (vec_perm_indices::nelts_per_input): Likewise.
15408 (vec_perm_indices::input_nelts): Likewise.
15409 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
15410 number of elements per input as a poly_uint64 rather than an
15411 unsigned int. Use the original encoding for variable-length
15412 vectors, rather than clamping each individual element.
15413 For the second and subsequent elements in each pattern,
15414 clamp the step and base before clamping their sum.
15415 (vec_perm_indices::series_p): Handle polynomial element counts.
15416 (vec_perm_indices::all_in_range_p): Likewise.
15417 (vec_perm_indices_to_tree): Likewise.
15418 (vec_perm_indices_to_rtx): Likewise.
15419 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
15420 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
15421 (tree_vector_builder::new_binary_operation): Handle polynomial
15422 element counts. Return false if we need to know the number
15423 of elements at compile time.
15424 * fold-const.c (fold_vec_perm): Punt if the number of elements
15425 isn't known at compile time.
15427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15429 * vec-perm-indices.h (vec_perm_builder): Change element type
15430 from HOST_WIDE_INT to poly_int64.
15431 (vec_perm_indices::element_type): Update accordingly.
15432 (vec_perm_indices::clamp): Handle polynomial element_types.
15433 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15434 (vec_perm_indices::all_in_range_p): Likewise.
15435 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
15437 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
15438 polynomial vec_perm_indices element types.
15439 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
15440 * fold-const.c (fold_vec_perm): Likewise.
15441 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
15442 * tree-vect-generic.c (lower_vec_perm): Likewise.
15443 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15444 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
15445 element type to HOST_WIDE_INT.
15447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15448 Alan Hayward <alan.hayward@arm.com>
15449 David Sherwood <david.sherwood@arm.com>
15451 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
15452 rather than an int. Use plus_constant.
15453 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
15454 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
15456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15457 Alan Hayward <alan.hayward@arm.com>
15458 David Sherwood <david.sherwood@arm.com>
15460 * calls.c (emit_call_1, expand_call): Change struct_value_size from
15461 a HOST_WIDE_INT to a poly_int64.
15463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15464 Alan Hayward <alan.hayward@arm.com>
15465 David Sherwood <david.sherwood@arm.com>
15467 * calls.c (load_register_parameters): Cope with polynomial
15468 mode sizes. Require a constant size for BLKmode parameters
15469 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
15470 forces a parameter to be padded at the lsb end in order to
15471 fill a complete number of words, require the parameter size
15472 to be ordered wrt UNITS_PER_WORD.
15474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15475 Alan Hayward <alan.hayward@arm.com>
15476 David Sherwood <david.sherwood@arm.com>
15478 * reload1.c (spill_stack_slot_width): Change element type
15479 from unsigned int to poly_uint64_pod.
15480 (alter_reg): Treat mode sizes as polynomial.
15482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15483 Alan Hayward <alan.hayward@arm.com>
15484 David Sherwood <david.sherwood@arm.com>
15486 * reload.c (complex_word_subreg_p): New function.
15487 (reload_inner_reg_of_subreg, push_reload): Use it.
15489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15490 Alan Hayward <alan.hayward@arm.com>
15491 David Sherwood <david.sherwood@arm.com>
15493 * lra-constraints.c (process_alt_operands): Reject matched
15494 operands whose sizes aren't ordered.
15495 (match_reload): Refer to this check here.
15497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15498 Alan Hayward <alan.hayward@arm.com>
15499 David Sherwood <david.sherwood@arm.com>
15501 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
15502 that the mode size is in the set {1, 2, 4, 8, 16}.
15504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15505 Alan Hayward <alan.hayward@arm.com>
15506 David Sherwood <david.sherwood@arm.com>
15508 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
15509 Use plus_constant instead of gen_rtx_PLUS.
15511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15512 Alan Hayward <alan.hayward@arm.com>
15513 David Sherwood <david.sherwood@arm.com>
15515 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
15516 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
15517 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
15518 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
15519 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
15520 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
15521 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
15522 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
15523 * config/i386/i386.c (ix86_push_rounding): ...this new function.
15524 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
15526 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
15527 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
15528 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
15529 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
15530 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
15531 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
15532 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
15533 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
15534 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
15535 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
15537 * expr.c (emit_move_resolve_push): Treat the input and result
15538 of PUSH_ROUNDING as a poly_int64.
15539 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
15540 (emit_push_insn): Likewise.
15541 * lra-eliminations.c (mark_not_eliminable): Likewise.
15542 * recog.c (push_operand): Likewise.
15543 * reload1.c (elimination_effects): Likewise.
15544 * rtlanal.c (nonzero_bits1): Likewise.
15545 * calls.c (store_one_arg): Likewise. Require the padding to be
15546 known at compile time.
15548 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15549 Alan Hayward <alan.hayward@arm.com>
15550 David Sherwood <david.sherwood@arm.com>
15552 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
15553 Use plus_constant instead of gen_rtx_PLUS.
15555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15556 Alan Hayward <alan.hayward@arm.com>
15557 David Sherwood <david.sherwood@arm.com>
15559 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
15560 rather than an int.
15562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15563 Alan Hayward <alan.hayward@arm.com>
15564 David Sherwood <david.sherwood@arm.com>
15566 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
15567 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
15568 via stack temporaries. Treat the mode size as polynomial too.
15570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15571 Alan Hayward <alan.hayward@arm.com>
15572 David Sherwood <david.sherwood@arm.com>
15574 * expr.c (expand_expr_real_2): When handling conversions involving
15575 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
15576 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
15577 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
15585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15586 Alan Hayward <alan.hayward@arm.com>
15587 David Sherwood <david.sherwood@arm.com>
15589 * combine.c (can_change_dest_mode): Handle polynomial
15590 REGMODE_NATURAL_SIZE.
15591 * expmed.c (store_bit_field_1): Likewise.
15592 * expr.c (store_constructor): Likewise.
15593 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
15594 and polynomial REGMODE_NATURAL_SIZE.
15595 (gen_lowpart_common): Likewise.
15596 * reginfo.c (record_subregs_of_mode): Likewise.
15597 * rtlanal.c (read_modify_subreg_p): Likewise.
15599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15600 Alan Hayward <alan.hayward@arm.com>
15601 David Sherwood <david.sherwood@arm.com>
15603 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
15604 numbers of elements.
15606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15607 Alan Hayward <alan.hayward@arm.com>
15608 David Sherwood <david.sherwood@arm.com>
15610 * match.pd: Cope with polynomial numbers of vector elements.
15612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15613 Alan Hayward <alan.hayward@arm.com>
15614 David Sherwood <david.sherwood@arm.com>
15616 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
15617 in a POINTER_PLUS_EXPR.
15619 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15620 Alan Hayward <alan.hayward@arm.com>
15621 David Sherwood <david.sherwood@arm.com>
15623 * omp-simd-clone.c (simd_clone_subparts): New function.
15624 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
15625 (ipa_simd_modify_function_body): Likewise.
15627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15628 Alan Hayward <alan.hayward@arm.com>
15629 David Sherwood <david.sherwood@arm.com>
15631 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
15632 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
15633 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
15634 (expand_vector_condition, vector_element): Likewise.
15635 (subparts_gt): New function.
15636 (get_compute_type): Use subparts_gt.
15637 (count_type_subparts): Delete.
15638 (expand_vector_operations_1): Use subparts_gt instead of
15639 count_type_subparts.
15641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15642 Alan Hayward <alan.hayward@arm.com>
15643 David Sherwood <david.sherwood@arm.com>
15645 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
15646 (vect_compile_time_alias): ...this new function. Do the calculation
15647 on poly_ints rather than trees.
15648 (vect_prune_runtime_alias_test_list): Update call accordingly.
15650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15651 Alan Hayward <alan.hayward@arm.com>
15652 David Sherwood <david.sherwood@arm.com>
15654 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
15656 (vect_schedule_slp_instance): Likewise.
15658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15659 Alan Hayward <alan.hayward@arm.com>
15660 David Sherwood <david.sherwood@arm.com>
15662 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
15663 constant and extern definitions for variable-length vectors.
15664 (vect_get_constant_vectors): Note that the number of units
15665 is known to be constant.
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 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
15672 of units as polynomial. Choose between WIDE and NARROW based
15675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15676 Alan Hayward <alan.hayward@arm.com>
15677 David Sherwood <david.sherwood@arm.com>
15679 * tree-vect-stmts.c (simd_clone_subparts): New function.
15680 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
15682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15683 Alan Hayward <alan.hayward@arm.com>
15684 David Sherwood <david.sherwood@arm.com>
15686 * tree-vect-stmts.c (vectorizable_call): Treat the number of
15687 vectors as polynomial. Use build_index_vector for
15688 IFN_GOMP_SIMD_LANE.
15690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15691 Alan Hayward <alan.hayward@arm.com>
15692 David Sherwood <david.sherwood@arm.com>
15694 * tree-vect-stmts.c (get_load_store_type): Treat the number of
15695 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
15696 for variable-length vectors.
15697 (vectorizable_mask_load_store): Treat the number of units as
15698 polynomial, asserting that it is constant if the condition has
15699 already been enforced.
15700 (vectorizable_store, vectorizable_load): Likewise.
15702 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15703 Alan Hayward <alan.hayward@arm.com>
15704 David Sherwood <david.sherwood@arm.com>
15706 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
15707 of units as polynomial. Punt if we can't tell at compile time
15708 which vector contains the final result.
15710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15711 Alan Hayward <alan.hayward@arm.com>
15712 David Sherwood <david.sherwood@arm.com>
15714 * tree-vect-loop.c (vectorizable_induction): Treat the number
15715 of units as polynomial. Punt on SLP inductions. Use an integer
15716 VEC_SERIES_EXPR for variable-length integer reductions. Use a
15717 cast of such a series for variable-length floating-point
15720 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15721 Alan Hayward <alan.hayward@arm.com>
15722 David Sherwood <david.sherwood@arm.com>
15724 * tree.h (build_index_vector): Declare.
15725 * tree.c (build_index_vector): New function.
15726 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
15727 of units as polynomial, forcibly converting it to a constant if
15728 vectorizable_reduction has already enforced the condition.
15729 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
15730 to create a {1,2,3,...} vector.
15731 (vectorizable_reduction): Treat the number of units as polynomial.
15732 Choose vectype_in based on the largest scalar element size rather
15733 than the smallest number of units. Enforce the restrictions
15736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15737 Alan Hayward <alan.hayward@arm.com>
15738 David Sherwood <david.sherwood@arm.com>
15740 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
15741 number of units as polynomial.
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 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
15748 * target.def (autovectorize_vector_sizes): Return the vector sizes
15749 by pointer, using vector_sizes rather than a bitmask.
15750 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
15751 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
15752 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
15754 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
15755 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
15756 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
15757 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
15758 * omp-general.c (omp_max_vf): Likewise.
15759 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15760 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
15761 * tree-vect-loop.c (vect_analyze_loop): Likewise.
15762 * tree-vect-slp.c (vect_slp_bb): Likewise.
15763 * doc/tm.texi: Regenerate.
15764 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
15766 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
15767 the vector size as a poly_uint64 rather than an unsigned int.
15768 (current_vector_size): Change from an unsigned int to a poly_uint64.
15769 (get_vectype_for_scalar_type): Update accordingly.
15770 * tree.h (build_truth_vector_type): Take the size and number of
15771 units as a poly_uint64 rather than an unsigned int.
15772 (build_vector_type): Add a temporary overload that takes
15773 the number of units as a poly_uint64 rather than an unsigned int.
15774 * tree.c (make_vector_type): Likewise.
15775 (build_truth_vector_type): Take the number of units as a poly_uint64
15776 rather than an unsigned int.
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 * target.def (get_mask_mode): Take the number of units and length
15783 as poly_uint64s rather than unsigned ints.
15784 * targhooks.h (default_get_mask_mode): Update accordingly.
15785 * targhooks.c (default_get_mask_mode): Likewise.
15786 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
15787 * doc/tm.texi: Regenerate.
15789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15790 Alan Hayward <alan.hayward@arm.com>
15791 David Sherwood <david.sherwood@arm.com>
15793 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
15794 * omp-general.c (omp_max_vf): Likewise.
15795 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
15796 (expand_omp_simd): Handle polynomial safelen.
15797 * omp-low.c (omplow_simd_context): Add a default constructor.
15798 (omplow_simd_context::max_vf): Change from int to poly_uint64.
15799 (lower_rec_simd_input_clauses): Update accordingly.
15800 (lower_rec_input_clauses): Likewise.
15802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15803 Alan Hayward <alan.hayward@arm.com>
15804 David Sherwood <david.sherwood@arm.com>
15806 * tree-vectorizer.h (vect_nunits_for_cost): New function.
15807 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
15808 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
15809 (vect_analyze_slp_cost): Likewise.
15810 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
15811 (vect_model_load_cost): Likewise.
15813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15814 Alan Hayward <alan.hayward@arm.com>
15815 David Sherwood <david.sherwood@arm.com>
15817 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
15818 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
15819 from an unsigned int * to a poly_uint64_pod *.
15820 (calculate_unrolling_factor): New function.
15821 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
15823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15824 Alan Hayward <alan.hayward@arm.com>
15825 David Sherwood <david.sherwood@arm.com>
15827 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
15828 from an unsigned int to a poly_uint64.
15829 (_loop_vec_info::slp_unrolling_factor): Likewise.
15830 (_loop_vec_info::vectorization_factor): Change from an int
15832 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
15833 (vect_get_num_vectors): New function.
15834 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
15835 (vect_get_num_copies): Use vect_get_num_vectors.
15836 (vect_analyze_data_ref_dependences): Change max_vf from an int *
15837 to an unsigned int *.
15838 (vect_analyze_data_refs): Change min_vf from an int * to a
15840 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15841 than an unsigned HOST_WIDE_INT.
15842 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
15843 (vect_analyze_data_ref_dependence): Change max_vf from an int *
15844 to an unsigned int *.
15845 (vect_analyze_data_ref_dependences): Likewise.
15846 (vect_compute_data_ref_alignment): Handle polynomial vf.
15847 (vect_enhance_data_refs_alignment): Likewise.
15848 (vect_prune_runtime_alias_test_list): Likewise.
15849 (vect_shift_permute_load_chain): Likewise.
15850 (vect_supportable_dr_alignment): Likewise.
15851 (dependence_distance_ge_vf): Take the vectorization factor as a
15852 poly_uint64 rather than an unsigned HOST_WIDE_INT.
15853 (vect_analyze_data_refs): Change min_vf from an int * to a
15855 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
15856 vfm1 as a poly_uint64 rather than an int. Make the same change
15857 for the returned bound_scalar.
15858 (vect_gen_vector_loop_niters): Handle polynomial vf.
15859 (vect_do_peeling): Likewise. Update call to
15860 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
15861 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
15863 * tree-vect-loop.c (vect_determine_vectorization_factor)
15864 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
15865 (vect_get_known_peeling_cost): Likewise.
15866 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
15867 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
15868 (vect_transform_loop): Likewise. Use the lowest possible VF when
15869 updating the upper bounds of the loop.
15870 (vect_min_worthwhile_factor): Make static. Return an unsigned int
15871 rather than an int.
15872 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
15873 polynomial unroll factors.
15874 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
15875 (vect_make_slp_decision): Likewise.
15876 (vect_supported_load_permutation_p): Likewise, and polynomial
15878 (vect_analyze_slp_cost): Handle polynomial vf.
15879 (vect_slp_analyze_node_operations): Likewise.
15880 (vect_slp_analyze_bb_1): Likewise.
15881 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15882 than an unsigned HOST_WIDE_INT.
15883 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
15884 (vectorizable_load): Handle polynomial vf.
15885 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
15887 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
15889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15890 Alan Hayward <alan.hayward@arm.com>
15891 David Sherwood <david.sherwood@arm.com>
15893 * match.pd: Handle bit operations involving three constants
15894 and try to fold one pair.
15896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15898 * tree-vect-loop-manip.c: Include gimple-fold.h.
15899 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
15900 niters_maybe_zero parameters. Handle other cases besides a step of 1.
15901 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
15902 Add a path that uses a step of VF instead of 1, but disable it
15904 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
15905 and niters_no_overflow parameters. Update calls to
15906 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
15907 Create a new SSA name if the latter choses to use a ste other
15908 than zero, and return it via niters_vector_mult_vf_var.
15909 * tree-vect-loop.c (vect_transform_loop): Update calls to
15910 vect_do_peeling, vect_gen_vector_loop_niters and
15911 slpeel_make_loop_iterate_ntimes.
15912 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
15913 (vect_gen_vector_loop_niters): Update declarations after above changes.
15915 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
15917 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
15918 128-bit round to integer instructions.
15919 (ceil<mode>2): Likewise.
15920 (btrunc<mode>2): Likewise.
15921 (round<mode>2): Likewise.
15923 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15925 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
15926 unaligned VSX load/store on P8/P9.
15927 (expand_block_clear): Allow the use of unaligned VSX
15928 load/store on P8/P9.
15930 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15932 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
15934 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
15935 swap associated with both a load and a store.
15937 2018-01-02 Andrew Waterman <andrew@sifive.com>
15939 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
15940 * config/riscv/riscv.md (clear_cache): Use it.
15942 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
15944 * web.c: Remove out-of-date comment.
15946 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15948 * expr.c (fixup_args_size_notes): Check that any existing
15949 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
15950 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
15951 (emit_single_push_insn): ...here.
15953 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15955 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
15956 (const_vector_encoded_nelts): New function.
15957 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
15958 (const_vector_int_elt, const_vector_elt): Declare.
15959 * emit-rtl.c (const_vector_int_elt_1): New function.
15960 (const_vector_elt): Likewise.
15961 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
15962 of CONST_VECTOR_ELT.
15964 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15966 * expr.c: Include rtx-vector-builder.h.
15967 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
15968 directly on the tree encoding.
15969 (const_vector_from_tree): Likewise.
15970 * optabs.c: Include rtx-vector-builder.h.
15971 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
15972 sequence of "u" values.
15973 * vec-perm-indices.c: Include rtx-vector-builder.h.
15974 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
15975 directly on the vec_perm_indices encoding.
15977 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15979 * doc/rtl.texi (const_vector): Describe new encoding scheme.
15980 * Makefile.in (OBJS): Add rtx-vector-builder.o.
15981 * rtx-vector-builder.h: New file.
15982 * rtx-vector-builder.c: Likewise.
15983 * rtl.h (rtx_def::u2): Add a const_vector field.
15984 (CONST_VECTOR_NPATTERNS): New macro.
15985 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
15986 (CONST_VECTOR_DUPLICATE_P): Likewise.
15987 (CONST_VECTOR_STEPPED_P): Likewise.
15988 (CONST_VECTOR_ENCODED_ELT): Likewise.
15989 (const_vec_duplicate_p): Check for a duplicated vector encoding.
15990 (unwrap_const_vec_duplicate): Likewise.
15991 (const_vec_series_p): Check for a non-duplicated vector encoding.
15992 Say that the function only returns true for integer vectors.
15993 * emit-rtl.c: Include rtx-vector-builder.h.
15994 (gen_const_vec_duplicate_1): Delete.
15995 (gen_const_vector): Call gen_const_vec_duplicate instead of
15996 gen_const_vec_duplicate_1.
15997 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
15998 (gen_const_vec_duplicate): Use rtx_vector_builder.
15999 (gen_const_vec_series): Likewise.
16000 (gen_rtx_CONST_VECTOR): Likewise.
16001 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
16002 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16003 Build a new vector rather than modifying a CONST_VECTOR in-place.
16004 (handle_special_swappables): Update call accordingly.
16005 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
16006 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16007 Build a new vector rather than modifying a CONST_VECTOR in-place.
16008 (handle_special_swappables): Update call accordingly.
16010 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16012 * simplify-rtx.c (simplify_const_binary_operation): Use
16013 CONST_VECTOR_ELT instead of XVECEXP.
16015 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16017 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16018 the selector elements to be different from the data elements
16019 if the selector is a VECTOR_CST.
16020 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16021 ssizetype for the selector.
16023 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16025 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16026 before testing each element individually.
16027 * tree-vect-generic.c (lower_vec_perm): Likewise.
16029 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16031 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16032 * selftest-run-tests.c (selftest::run_tests): Call it.
16033 * vector-builder.h (vector_builder::operator ==): New function.
16034 (vector_builder::operator !=): Likewise.
16035 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16036 (vec_perm_indices::all_from_input_p): New function.
16037 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16038 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16039 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16040 instead of reading the VECTOR_CST directly. Detect whether both
16041 vector inputs are the same before constructing the vec_perm_indices,
16042 and update the number of inputs argument accordingly. Use the
16043 utility functions added above. Only construct sel2 if we need to.
16045 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16047 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16048 the broadcast of the low byte.
16049 (expand_mult_highpart): Use an explicit encoding for the permutes.
16050 * optabs-query.c (can_mult_highpart_p): Likewise.
16051 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16052 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16053 (vectorizable_bswap): Likewise.
16054 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16055 explicit encoding for the power-of-2 permutes.
16056 (vect_permute_store_chain): Likewise.
16057 (vect_grouped_load_supported): Likewise.
16058 (vect_permute_load_chain): Likewise.
16060 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16062 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16063 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16064 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16065 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16066 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16067 (vect_gen_perm_mask_any): Likewise.
16069 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16071 * int-vector-builder.h: New file.
16072 * vec-perm-indices.h: Include int-vector-builder.h.
16073 (vec_perm_indices): Redefine as an int_vector_builder.
16074 (auto_vec_perm_indices): Delete.
16075 (vec_perm_builder): Redefine as a stand-alone class.
16076 (vec_perm_indices::vec_perm_indices): New function.
16077 (vec_perm_indices::clamp): Likewise.
16078 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16079 (vec_perm_indices::new_vector): New function.
16080 (vec_perm_indices::new_expanded_vector): Update for new
16081 vec_perm_indices class.
16082 (vec_perm_indices::rotate_inputs): New function.
16083 (vec_perm_indices::all_in_range_p): Operate directly on the
16084 encoded form, without computing elided elements.
16085 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16086 encoding. Update for new vec_perm_indices class.
16087 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16088 the given vec_perm_builder.
16089 (expand_vec_perm_var): Update vec_perm_builder constructor.
16090 (expand_mult_highpart): Use vec_perm_builder instead of
16091 auto_vec_perm_indices.
16092 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16093 vec_perm_indices instead of auto_vec_perm_indices. Use a single
16094 or double series encoding as appropriate.
16095 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16096 vec_perm_indices instead of auto_vec_perm_indices.
16097 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16098 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16099 (vect_permute_store_chain): Likewise.
16100 (vect_grouped_load_supported): Likewise.
16101 (vect_permute_load_chain): Likewise.
16102 (vect_shift_permute_load_chain): Likewise.
16103 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16104 (vect_transform_slp_perm_load): Likewise.
16105 (vect_schedule_slp_instance): Likewise.
16106 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16107 (vectorizable_mask_load_store): Likewise.
16108 (vectorizable_bswap): Likewise.
16109 (vectorizable_store): Likewise.
16110 (vectorizable_load): Likewise.
16111 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16112 vec_perm_indices instead of auto_vec_perm_indices. Use
16113 tree_to_vec_perm_builder to read the vector from a tree.
16114 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16115 vec_perm_builder instead of a vec_perm_indices.
16116 (have_whole_vector_shift): Use vec_perm_builder and
16117 vec_perm_indices instead of auto_vec_perm_indices. Leave the
16118 truncation to calc_vec_perm_mask_for_shift.
16119 (vect_create_epilog_for_reduction): Likewise.
16120 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16121 from auto_vec_perm_indices to vec_perm_indices.
16122 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16123 instead of changing individual elements.
16124 (aarch64_vectorize_vec_perm_const): Use new_vector to install
16125 the vector in d.perm.
16126 * config/arm/arm.c (expand_vec_perm_d::perm): Change
16127 from auto_vec_perm_indices to vec_perm_indices.
16128 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16129 instead of changing individual elements.
16130 (arm_vectorize_vec_perm_const): Use new_vector to install
16131 the vector in d.perm.
16132 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16133 Update vec_perm_builder constructor.
16134 (rs6000_expand_interleave): Likewise.
16135 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16136 (rs6000_expand_interleave): Likewise.
16138 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16140 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16141 to qimode could truncate the indices.
16142 * optabs.c (expand_vec_perm_var): Likewise.
16144 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16146 * Makefile.in (OBJS): Add vec-perm-indices.o.
16147 * vec-perm-indices.h: New file.
16148 * vec-perm-indices.c: Likewise.
16149 * target.h (vec_perm_indices): Replace with a forward class
16151 (auto_vec_perm_indices): Move to vec-perm-indices.h.
16152 * optabs.h: Include vec-perm-indices.h.
16153 (expand_vec_perm): Delete.
16154 (selector_fits_mode_p, expand_vec_perm_var): Declare.
16155 (expand_vec_perm_const): Declare.
16156 * target.def (vec_perm_const_ok): Replace with...
16157 (vec_perm_const): ...this new hook.
16158 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16159 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16160 * doc/tm.texi: Regenerate.
16161 * optabs.def (vec_perm_const): Delete.
16162 * doc/md.texi (vec_perm_const): Likewise.
16163 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16164 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16165 expand_vec_perm for constant permutation vectors. Assert that
16166 the mode of variable permutation vectors is the integer equivalent
16167 of the mode that is being permuted.
16168 * optabs-query.h (selector_fits_mode_p): Declare.
16169 * optabs-query.c: Include vec-perm-indices.h.
16170 (selector_fits_mode_p): New function.
16171 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16172 is defined, instead of checking whether the vec_perm_const_optab
16173 exists. Use targetm.vectorize.vec_perm_const instead of
16174 targetm.vectorize.vec_perm_const_ok. Check whether the indices
16175 fit in the vector mode before using a variable permute.
16176 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
16177 vec_perm_indices instead of an rtx.
16178 (expand_vec_perm): Replace with...
16179 (expand_vec_perm_const): ...this new function. Take the selector
16180 as a vec_perm_indices rather than an rtx. Also take the mode of
16181 the selector. Update call to shift_amt_for_vec_perm_mask.
16182 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
16183 Use vec_perm_indices::new_expanded_vector to expand the original
16184 selector into bytes. Check whether the indices fit in the vector
16185 mode before using a variable permute.
16186 (expand_vec_perm_var): Make global.
16187 (expand_mult_highpart): Use expand_vec_perm_const.
16188 * fold-const.c: Includes vec-perm-indices.h.
16189 * tree-ssa-forwprop.c: Likewise.
16190 * tree-vect-data-refs.c: Likewise.
16191 * tree-vect-generic.c: Likewise.
16192 * tree-vect-loop.c: Likewise.
16193 * tree-vect-slp.c: Likewise.
16194 * tree-vect-stmts.c: Likewise.
16195 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
16197 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
16198 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
16199 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
16200 (aarch64_vectorize_vec_perm_const): ...this new function.
16201 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16202 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16203 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
16204 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
16205 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16206 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16207 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
16209 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
16210 check for NEON modes.
16211 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
16212 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
16213 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
16214 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
16216 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
16217 the old VEC_PERM_CONST conditions.
16218 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
16219 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
16220 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
16221 (ia64_vectorize_vec_perm_const_ok): Merge into...
16222 (ia64_vectorize_vec_perm_const): ...this new function.
16223 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
16224 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
16225 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
16226 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
16227 * config/mips/mips.c (mips_expand_vec_perm_const)
16228 (mips_vectorize_vec_perm_const_ok): Merge into...
16229 (mips_vectorize_vec_perm_const): ...this new function.
16230 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
16231 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
16232 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
16233 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
16234 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
16235 (rs6000_expand_vec_perm_const): Delete.
16236 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
16238 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16239 (altivec_expand_vec_perm_const_le): Take each operand individually.
16240 Operate on constant selectors rather than rtxes.
16241 (altivec_expand_vec_perm_const): Likewise. Update call to
16242 altivec_expand_vec_perm_const_le.
16243 (rs6000_expand_vec_perm_const): Delete.
16244 (rs6000_vectorize_vec_perm_const_ok): Delete.
16245 (rs6000_vectorize_vec_perm_const): New function.
16246 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16247 an element count and rtx array.
16248 (rs6000_expand_extract_even): Update call accordingly.
16249 (rs6000_expand_interleave): Likewise.
16250 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
16251 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
16252 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
16253 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
16254 (rs6000_expand_vec_perm_const): Delete.
16255 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16256 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16257 (altivec_expand_vec_perm_const_le): Take each operand individually.
16258 Operate on constant selectors rather than rtxes.
16259 (altivec_expand_vec_perm_const): Likewise. Update call to
16260 altivec_expand_vec_perm_const_le.
16261 (rs6000_expand_vec_perm_const): Delete.
16262 (rs6000_vectorize_vec_perm_const_ok): Delete.
16263 (rs6000_vectorize_vec_perm_const): New function. Remove stray
16264 reference to the SPE evmerge intructions.
16265 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16266 an element count and rtx array.
16267 (rs6000_expand_extract_even): Update call accordingly.
16268 (rs6000_expand_interleave): Likewise.
16269 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
16270 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
16272 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16274 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16276 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
16277 vector mode and that that mode matches the mode of the data
16279 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
16280 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
16281 directly using expand_vec_perm_1 when forcing selectors into
16283 (expand_vec_perm_var): New function, split out from expand_vec_perm.
16285 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16287 * optabs-query.h (can_vec_perm_p): Delete.
16288 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
16289 * optabs-query.c (can_vec_perm_p): Split into...
16290 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
16291 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
16292 particular selector is valid.
16293 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16294 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16295 (vect_grouped_load_supported): Likewise.
16296 (vect_shift_permute_load_chain): Likewise.
16297 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16298 (vect_transform_slp_perm_load): Likewise.
16299 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16300 (vectorizable_bswap): Likewise.
16301 (vect_gen_perm_mask_checked): Likewise.
16302 * fold-const.c (fold_ternary_loc): Likewise. Don't take
16303 implementations of variable permutation vectors into account
16304 when deciding which selector to use.
16305 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
16306 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
16307 with a false third argument.
16308 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
16309 to test whether the constant selector is valid and can_vec_perm_var_p
16310 to test whether a variable selector is valid.
16312 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16314 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
16315 * optabs-query.c (can_vec_perm_p): Likewise.
16316 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
16317 instead of vec_perm_indices.
16318 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
16319 (vect_gen_perm_mask_checked): Likewise,
16320 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
16321 (vect_gen_perm_mask_checked): Likewise,
16323 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16325 * optabs-query.h (qimode_for_vec_perm): Declare.
16326 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
16327 (qimode_for_vec_perm): ...this new function.
16328 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
16330 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16332 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
16333 does not have a conditional at the top.
16335 2018-01-02 Richard Biener <rguenther@suse.de>
16337 * ipa-inline.c (big_speedup_p): Fix expression.
16339 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16342 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
16345 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16349 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
16350 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
16351 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
16352 cond_taken_branch_cost 3->4.
16354 2018-01-01 Jakub Jelinek <jakub@redhat.com>
16356 PR tree-optimization/83581
16357 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
16358 TODO_cleanup_cfg if any changes have been made.
16360 PR middle-end/83608
16361 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
16362 convert_modes if target mode has the right side, but different mode
16365 PR middle-end/83609
16366 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
16367 last argument when extracting from CONCAT. If either from_real or
16368 from_imag is NULL, use expansion through memory. If result is not
16369 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
16370 the parts directly to inner mode, if even that fails, use expansion
16373 PR middle-end/83623
16374 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
16375 check for bswap in mode rather than HImode and use that in expand_unop
16378 Copyright (C) 2018 Free Software Foundation, Inc.
16380 Copying and distribution of this file, with or without modification,
16381 are permitted in any medium without royalty provided the copyright
16382 notice and this notice are preserved.