1 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
3 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
4 BUILTIN_VEC_XST entries for pointer to double and long long.
6 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
9 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
14 2018-06-12 Martin Liska <mliska@suse.cz>
16 * doc/options.texi: Document IntegerRange.
18 2018-06-12 Martin Liska <mliska@suse.cz>
20 * config/i386/i386.opt: Make MPX-related options as Deprecated.
21 * opt-functions.awk: Handle Deprecated flag.
22 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
24 (read_cmdline_option): Report warning for a deprecated option.
25 * opts.h (struct cl_option): Add new field cl_deprecated.
26 (CL_ERR_DEPRECATED): New.
28 2018-06-12 Martin Liska <mliska@suse.cz>
30 * doc/options.texi: Document Deprecated option flag.
32 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
34 * config/arc/arc-arch.h (arc_extras): New enum.
35 (arc_cpu_t):Add field extra.
36 (arc_cpu_types): Consider the extras.
37 * config/arc/arc-cpus.def: Add extras info.
38 * config/arc/arc-opts.h (processor_type): Consider extra field.
39 * config/arc/arc.c (arc_override_options): Handle extra field.
41 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
43 * config/arc/arc-arch.h: Update ARC_OPTX macro.
44 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
46 * config/arc/arc.c (arc_init): Update pic warning.
47 (irq_range): Update irq range parsing warnings.
48 (arc_override_options): Update various warning messages.
49 (arc_handle_aux_attribute): Likewise.
51 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
53 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
55 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
57 * doc/sourcebuild.texi: Document usage of line number 0 in verify
58 compiler messages directives.
60 2018-06-12 Matthew Fortune <mfortune@gmail.com>
62 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
63 * config/mips/mips-tables.opt: Regenerate.
64 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
66 * doc/invoke.texi: Document -march=i6500.
68 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
70 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
71 (i6400_gpmul): Add cpu_unit.
72 (i6400_gpdiv): Likewise.
73 (i6400_msa_add_d): Update reservations.
74 (i6400_msa_int_add) Likewise.
75 (i6400_msa_short_logic3) Likewise.
76 (i6400_msa_short_logic2) Likewise.
77 (i6400_msa_short_logic) Likewise.
78 (i6400_msa_move) Likewise.
79 (i6400_msa_cmp) Likewise.
80 (i6400_msa_short_float2) Likewise.
81 (i6400_msa_div_d) Likewise.
82 (i6400_msa_long_logic1) Likewise.
83 (i6400_msa_long_logic2) Likewise.
84 (i6400_msa_mult) Likewise.
85 (i6400_msa_long_float2) Likewise.
86 (i6400_msa_long_float4) Likewise.
87 (i6400_msa_long_float5) Likewise.
88 (i6400_msa_long_float8) Likewise.
89 (i6400_fpu_fadd): Include frint type.
90 (i6400_fpu_store): New define_insn_reservation.
91 (i6400_fpu_load): Likewise.
92 (i6400_fpu_move): Likewise.
93 (i6400_fpu_fcmp): Likewise.
94 (i6400_fpu_fmadd): Likewise.
95 (i6400_int_mult): Include imul3nc type and update reservation.
96 (i6400_int_div): Include idiv3 type and update reservation.
97 (i6400_int_load): Update to check type not move_type.
98 (i6400_int_store): Likewise.
99 (i6400_int_prefetch): Set zero latency.
101 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
103 * gcc.c: Document new %@{...} sequence.
104 (LINK_COMMAND_SPEC): Use it for the -L switches.
105 (cpp_unique_options): Use it for the -I switches.
106 (at_file_argbuf): New global variable.
107 (in_at_file): Likewise.
108 (alloc_args): Create at_file_argbuf.
109 (clear_args): Truncate at_file_argbuf.
110 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
111 (open_at_file): New function.
112 (close_at_file): Likewise.
113 (create_at_file): Delete.
114 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
117 (validate_switches_from_spec): Deal with %@{...} sequence.
118 (validate_switches): Likewise.
119 (driver::finalize): Call clear_args.
121 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
123 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
125 2018-06-11 Martin Sebor <msebor@redhat.com>
127 * doc/invoke.texi (-Wall): List -Wc++17-compat.
128 (Wno-class-memaccess): Add @opindex.
129 (Wno-templates, Wno-multiple-inheritance): Same.
130 (Wno-virtual-inheritance, Wno-namespaces): Same.
131 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
132 (Wno-format-overflow, Wno-format-truncation): Same.
133 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
134 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
135 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
136 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
137 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
138 (Wno-misspelled-isr): Same.
140 2018-06-11 Martin Sebor <msebor@redhat.com>
142 * PR tree-optimization/86083
143 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
145 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
147 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
149 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
152 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
153 on the correct operand.
154 (*movdi_internal64): Ditto.
156 2018-06-11 Martin Liska <mliska@suse.cz>
158 PR tree-optimization/86089
159 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
161 2018-06-11 Julia Koval <julia.koval@intel.com>
163 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
164 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
165 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
167 2018-06-11 Olivier Hainque <hainque@adacore.com>
169 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
170 for Ada with strict dwarf2.
172 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
175 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
178 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
180 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
182 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
184 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
187 2018-06-08 Martin Liska <mliska@suse.cz>
189 * tree-cfg.h (debug_function): Fix argument type to match
192 2018-06-08 Martin Liska <mliska@suse.cz>
194 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
195 Remove usage of MPX-related (and removed) fields.
196 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
198 2018-06-08 David Malcolm <dmalcolm@redhat.com>
200 * cfg.c (debug): Use TDF_NONE rather than 0.
201 * cfghooks.c (debug): Likewise.
202 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
203 (struct dump_option_value_info): Convert to...
204 (struct kv_pair): ...this template type.
205 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
207 (optinfo_verbosity_options): Likewise.
208 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
210 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
211 than int for "optgroup_flags" param.
212 (dump_generic_expr_loc): Use dump_flags_t rather than int for
214 (dump_dec): Likewise.
215 (dump_finish): Use TDF_NONE rather than 0.
216 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
217 rather than int for "optgroup_flags" param. Use TDF_NONE rather
218 than 0. Update for change to option_ptr.
219 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
220 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
221 0. Update for changes to optinfo_verbosity_options and
223 (opt_info_switch_p): Convert optgroup_flags from int to
225 (dump_basic_block): Use dump_flags_t rather than int
226 for "dump_kind" param.
227 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
228 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
229 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
230 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
231 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
232 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
233 TDF_NONE): Convert from macros to...
234 (enum dump_flag): ...this new enum.
235 (dump_flags_t): Update to use enum.
236 (operator|, operator&, operator~, operator|=, operator&=):
237 Implement for dump_flags_t.
238 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
239 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
240 Convert from macros to...
241 (enum optgroup_flag): ...this new enum.
242 (optgroup_flags_t): New typedef.
243 (operator|, operator|=): Implement for optgroup_flags_t.
244 (struct dump_file_info): Convert field "alt_flags" to
245 dump_flags_t. Convert field "optgroup_flags" to
247 (dump_basic_block): Use dump_flags_t rather than int for param.
248 (dump_generic_expr_loc): Likewise.
249 (dump_dec): Likewise.
250 (dump_register): Convert param "optgroup_flags" to
252 (opt_info_enable_passes): Likewise.
253 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
255 * gimple-pretty-print.c (debug): Likewise.
256 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
257 (merged_store_group::apply_stores): Likewise.
258 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
259 * gimple.c (verify_gimple_pp): Likewise.
260 * graphite-poly.c (print_pbb_body): Likewise.
261 * passes.c (pass_manager::register_one_dump_file): Convert
262 local "optgroup_flags" to optgroup_flags_t.
263 * print-tree.c (print_node): Use TDF_NONE rather than 0.
265 (debug_body): Likewise.
266 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
268 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
270 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
271 (convert_mult_to_fma): Likewise.
272 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
273 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
274 * tree-vect-data-refs.c (dump_lower_bound): Convert param
275 "dump_kind" to dump_flags_t.
277 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
279 * config/rs6000/rs6000.c (min, max): Delete.
281 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
283 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
284 -mabi=spe and -mabi=no-spe.
286 2018-06-08 Martin Liska <mliska@suse.cz>
288 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
289 where we expect an existing summary.
291 2018-06-08 Martin Liska <mliska@suse.cz>
293 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
294 * ipa-inline.h (estimate_edge_growth): Likewise.
296 2018-06-08 Martin Liska <mliska@suse.cz>
298 * cgraph.c (function_version_hasher::hash): Use
299 cgraph_node::get_uid ().
300 (function_version_hasher::equal):
301 * cgraph.h (cgraph_node::get_uid): New method.
302 * ipa-inline.c (update_caller_keys): Use
303 cgraph_node::get_uid ().
304 (update_callee_keys): Likewise.
305 * ipa-utils.c (searchc): Likewise.
306 (ipa_reduced_postorder): Likewise.
307 * lto-cgraph.c (input_node): Likewise.
308 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
309 * symbol-summary.h (symtab_insertion): Likewise.
310 (symtab_removal): Likewise.
311 (symtab_duplication): Likewise.
312 * tree-pretty-print.c (dump_function_header): Likewise.
313 * tree-sra.c (convert_callers_for_node): Likewise.
315 2018-06-08 Martin Liska <mliska@suse.cz>
317 * cgraph.c (symbol_table::create_edge): Always assign a new
319 (symbol_table::free_edge): Do not recycle numbers.
320 * cgraph.h (cgraph_edge::get): New method.
321 * symbol-summary.h (symtab_removal): Use it.
322 (symtab_duplication): Likewise.
323 (call_summary::hashable_uid): Remove.
325 2018-06-08 Martin Liska <mliska@suse.cz>
327 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
328 (initialize_growth_caches): Remove.
329 (free_growth_caches): Likewise.
330 (do_estimate_edge_time): Use edge_growth_cache.
331 (do_estimate_edge_size): Likewise.
332 (do_estimate_edge_hints): Likewise.
333 * ipa-inline.c (reset_edge_caches): Likewise.
334 (recursive_inlining): Likewise.
335 (inline_small_functions): Likewise.
336 * ipa-inline.h (initialize_growth_caches): Remove.
337 (estimate_edge_size): Likewise.
338 (estimate_edge_time): Likewise.
339 (estimate_edge_hints): Likewise.
340 (reset_edge_growth_cache): Likewise.
341 * symbol-summary.h (call_summary::remove): New method.
343 2018-06-08 Martin Liska <mliska@suse.cz>
345 * ipa-cp.c (class edge_clone_summary): New summary.
346 (grow_edge_clone_vectors): Remove.
347 (ipcp_edge_duplication_hook): Remove.
348 (class edge_clone_summary_t): New call_summary class.
349 (ipcp_edge_removal_hook): Remove.
350 (edge_clone_summary_t::duplicate): New function.
351 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
352 (create_specialized_node): Likewise.
353 (ipcp_driver): Initialize edge_clone_summaries and do not
356 2018-06-08 Martin Liska <mliska@suse.cz>
358 * symbol-summary.h (get): New function.
359 (call_summary::m_initialize_when_cloning): New class member.
361 2018-06-08 Martin Liska <mliska@suse.cz>
363 * cgraph.c (cgraph_node::remove): Do not recycle uid.
364 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
365 (symbol_table::allocate_cgraph_symbol): Do not set uid.
366 * passes.c (uid_hash_t): Record removed_nodes by their uids.
367 (remove_cgraph_node_from_order): Use the removed_nodes set.
368 (do_per_function_toporder): Likwise.
369 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
370 instead of summary_uid.
371 (symtab_removal): Likewise.
372 (symtab_duplication): Likewise.
374 2018-06-08 Martin Liska <mliska@suse.cz>
376 * ipa-cp.c (ipcp_store_bits_results): Use
377 ipcp_transformation_sum.
378 (ipcp_store_vr_results): Likewise.
379 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
381 (ipcp_transformation_initialize): ... this.
382 (ipa_set_node_agg_value_chain):
383 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
384 (write_ipcp_transformation_info): Likewise.
385 (read_ipcp_transformation_info): Likewise.
386 (ipcp_update_bits): Likewise.
387 (ipcp_update_vr): Likewise.
388 (ipcp_transform_function): Likewise.
389 * ipa-prop.h: Rename ipcp_transformation_summary to
391 (class ipcp_transformation_t): New function summary.
392 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
393 (ipa_get_agg_replacements_for_node): Likewise.
395 2018-06-08 Martin Liska <mliska@suse.cz>
397 * ipa-pure-const.c (struct funct_state_d): Do it class instead
399 (class funct_state_summary_t): New function_summary class.
400 (has_function_state): Remove.
401 (get_function_state): Likewise.
402 (set_function_state): Likewise.
403 (add_new_function): Likewise.
404 (funct_state_summary_t::insert): New function.
405 (duplicate_node_data): Remove.
406 (remove_node_data): Remove.
407 (funct_state_summary_t::duplicate): New function.
408 (register_hooks): Create new funct_state_summaries.
409 (pure_const_generate_summary): Use it.
410 (pure_const_write_summary): Likewise.
411 (pure_const_read_summary): Likewise.
412 (propagate_pure_const): Likewise.
413 (propagate_nothrow): Likewise.
414 (dump_malloc_lattice): Likewise.
415 (propagate_malloc): Likewise.
416 (execute): Do not register hooks, just remove summary
418 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
421 2018-06-08 Martin Liska <mliska@suse.cz>
423 * ipa-reference.c (remove_node_data): Remove.
424 (duplicate_node_data): Likewise.
425 (class ipa_ref_var_info_summary_t): New class.
426 (class ipa_ref_opt_summary_t): Likewise.
427 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
428 (get_reference_optimization_summary): Use
429 ipa_ref_opt_sum_summaries.
430 (set_reference_vars_info): Remove.
431 (set_reference_optimization_summary): Likewise.
432 (ipa_init): Create summaries.
433 (init_function_info): Use function summary.
434 (ipa_ref_opt_summary_t::duplicate): New function.
435 (ipa_ref_opt_summary_t::remove): New function.
436 (get_read_write_all_from_node): Fix GNU coding style.
437 (propagate): Use function summary.
438 (write_node_summary_p): Fix GNU coding style.
439 (stream_out_bitmap): Likewise.
440 (ipa_reference_read_optimization_summary): Use function summary.
441 (ipa_reference_c_finalize): Do not release hooks.
443 2018-06-08 Martin Liska <mliska@suse.cz>
445 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
446 (analyze_function_body): Extract multiple calls of get_create.
447 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
448 * ipa-inline.c (recursive_inlining): Use ::get method.
449 * ipa-inline.h (estimate_edge_growth): Likewise.
451 2018-06-08 Martin Liska <mliska@suse.cz>
453 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
455 (hsa_function_summary::hsa_function_summary): Use the new enum
457 (hsa_gpu_implementation_p): Use hsa_summaries::get.
458 * hsa-gen.c (hsa_get_host_function): Likewise.
459 (get_brig_function_name): Likewise.
460 * ipa-hsa.c (process_hsa_functions): Likewise.
461 (ipa_hsa_write_summary): Likewise.
462 * symbol-summary.h (symtab_duplication): Use ::get function/
465 2018-06-08 Martin Liska <mliska@suse.cz>
467 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
469 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
470 (hsa_register_kernel): Likewise.
471 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
472 * hsa-gen.c (hsa_get_host_function): Likewise.
473 (get_brig_function_name): Likewise.
474 (generate_hsa): Likewise.
475 (pass_gen_hsail::execute): Likewise.
476 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
477 (devirtualization_time_bonus): Likewise.
478 (ipcp_propagate_stage): Likewise.
479 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
480 (edge_set_predicate): Likewise.
481 (evaluate_conditions_for_known_args): Likewise.
482 (evaluate_properties_for_edge): Likewise.
483 (ipa_fn_summary::reset): Likewise.
484 (ipa_fn_summary_t::duplicate): Likewise.
485 (dump_ipa_call_summary): Likewise.
486 (ipa_dump_fn_summary): Likewise.
487 (analyze_function_body): Likewise.
488 (compute_fn_summary): Likewise.
489 (estimate_edge_devirt_benefit): Likewise.
490 (estimate_edge_size_and_time): Likewise.
491 (estimate_calls_size_and_time): Likewise.
492 (estimate_node_size_and_time): Likewise.
493 (inline_update_callee_summaries): Likewise.
494 (remap_edge_change_prob): Likewise.
495 (remap_edge_summaries): Likewise.
496 (ipa_merge_fn_summary_after_inlining): Likewise.
497 (ipa_update_overall_fn_summary): Likewise.
498 (read_ipa_call_summary): Likewise.
499 (inline_read_section): Likewise.
500 (write_ipa_call_summary): Likewise.
501 (ipa_fn_summary_write): Likewise.
502 (ipa_free_fn_summary): Likewise.
503 * ipa-hsa.c (process_hsa_functions): Likewise.
504 (ipa_hsa_write_summary): Likewise.
505 (ipa_hsa_read_section): Likewise.
506 * ipa-icf.c (sem_function::merge): Likewise.
507 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
508 (do_estimate_edge_time): Likewise.
509 (estimate_size_after_inlining): Likewise.
510 (estimate_growth): Likewise.
511 (growth_likely_positive): Likewise.
512 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
513 (inline_call): Likewise.
514 * ipa-inline.c (caller_growth_limits): Likewise.
515 (can_inline_edge_p): Likewise.
516 (can_inline_edge_by_limits_p): Likewise.
517 (compute_uninlined_call_time): Likewise.
518 (compute_inlined_call_time): Likewise.
519 (want_inline_small_function_p): Likewise.
520 (edge_badness): Likewise.
521 (update_caller_keys): Likewise.
522 (update_callee_keys): Likewise.
523 (recursive_inlining): Likewise.
524 (inline_small_functions): Likewise.
525 (inline_to_all_callers_1): Likewise.
526 (dump_overall_stats): Likewise.
527 (early_inline_small_functions): Likewise.
528 (early_inliner): Likewise.
529 * ipa-inline.h (estimate_edge_growth): Likewise.
530 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
531 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
532 * ipa-prop.h (IPA_NODE_REF): Likewise.
533 (IPA_EDGE_REF): Likewise.
534 * ipa-pure-const.c (malloc_candidate_p): Likewise.
535 (propagate_malloc): Likewise.
536 * ipa-split.c (execute_split_functions): Likewise.
537 * symbol-summary.h: Rename get to get_create.
539 (get_create): Likewise.
540 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
542 2018-06-08 Martin Liska <mliska@suse.cz>
544 * symbol-summary.h (release): Move definition out of class
546 (symtab_removal): Likewise.
547 (symtab_duplication): Likewise.
549 2018-06-08 Martin Liska <mliska@suse.cz>
551 * symbol-summary.h (function_summary): Move constructor
552 implementation out of class declaration.
554 (symtab_insertion): Likewise.
555 (symtab_removal): Likewise.
556 (symtab_duplication): Likewise.
559 2018-06-08 Martin Liska <mliska@suse.cz>
561 * Makefile.in: Remove support for MPX (macros, related functions,
562 fields in cgraph_node, ...).
563 * builtin-types.def (BT_BND): Likewise.
564 (BT_FN_BND_CONST_PTR): Likewise.
565 (BT_FN_CONST_PTR_BND): Likewise.
566 (BT_FN_VOID_PTR_BND): Likewise.
567 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
568 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
569 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
570 (expand_builtin_mempcpy_with_bounds): Likewise.
571 (expand_builtin_memset_with_bounds): Likewise.
572 (expand_builtin_memset_args): Likewise.
573 (std_expand_builtin_va_start): Likewise.
574 (expand_builtin): Likewise.
575 (expand_builtin_with_bounds): Likewise.
576 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
577 (DEF_LIB_BUILTIN_CHKP): Likewise.
578 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
579 (DEF_CHKP_BUILTIN): Likewise.
580 (BUILT_IN_MEMCPY): Likewise.
581 (BUILT_IN_MEMMOVE): Likewise.
582 (BUILT_IN_MEMPCPY): Likewise.
583 (BUILT_IN_MEMSET): Likewise.
584 (BUILT_IN_STPCPY): Likewise.
585 (BUILT_IN_STRCAT): Likewise.
586 (BUILT_IN_STRCHR): Likewise.
587 (BUILT_IN_STRCPY): Likewise.
588 (BUILT_IN_STRLEN): Likewise.
589 (BUILT_IN_MEMCPY_CHK): Likewise.
590 (BUILT_IN_MEMMOVE_CHK): Likewise.
591 (BUILT_IN_MEMPCPY_CHK): Likewise.
592 (BUILT_IN_MEMSET_CHK): Likewise.
593 (BUILT_IN_STPCPY_CHK): Likewise.
594 (BUILT_IN_STRCAT_CHK): Likewise.
595 (BUILT_IN_STRCPY_CHK): Likewise.
596 * calls.c (store_bounds): Likewise.
597 (emit_call_1): Likewise.
598 (special_function_p): Likewise.
599 (maybe_warn_nonstring_arg): Likewise.
600 (initialize_argument_information): Likewise.
601 (finalize_must_preallocate): Likewise.
602 (compute_argument_addresses): Likewise.
603 (expand_call): Likewise.
604 * cfgexpand.c (expand_call_stmt): Likewise.
605 (expand_return): Likewise.
606 (expand_gimple_stmt_1): Likewise.
607 (pass_expand::execute): Likewise.
608 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
609 (cgraph_node::remove): Likewise.
610 (cgraph_node::dump): Likewise.
611 (cgraph_node::verify_node): Likewise.
612 * cgraph.h (chkp_function_instrumented_p): Likewise.
613 (symtab_node::get_alias_target): Likewise.
614 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
615 (cgraph_local_p): Likewise.
616 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
617 (cgraph_edge::rebuild_references): Likewise.
618 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
619 (walk_polymorphic_call_targets): Likewise.
620 (cgraph_node::expand_thunk): Likewise.
621 (symbol_table::output_weakrefs): Likewise.
622 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
623 (ix86_handle_option): Likewise.
624 * config/i386/constraints.md: Likewise.
625 * config/i386/i386-builtin-types.def (BND): Likewise.
629 * config/i386/i386-builtin.def (BDESC_END): Likewise.
630 (BDESC_FIRST): Likewise.
632 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
633 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
634 * config/i386/i386.c (enum reg_class): Likewise.
635 (ix86_target_string): Likewise.
636 (ix86_option_override_internal): Likewise.
637 (ix86_conditional_register_usage): Likewise.
638 (ix86_valid_target_attribute_inner_p): Likewise.
639 (ix86_set_indirect_branch_type): Likewise.
640 (ix86_set_current_function): Likewise.
641 (ix86_function_arg_regno_p): Likewise.
642 (init_cumulative_args): Likewise.
643 (ix86_function_arg_advance): Likewise.
644 (ix86_function_arg): Likewise.
645 (ix86_pass_by_reference): Likewise.
646 (ix86_function_value_regno_p): Likewise.
647 (ix86_function_value_1): Likewise.
648 (ix86_function_value_bounds): Likewise.
649 (ix86_return_in_memory): Likewise.
650 (ix86_setup_incoming_vararg_bounds): Likewise.
651 (ix86_va_start): Likewise.
652 (indirect_thunk_need_prefix): Likewise.
653 (print_reg): Likewise.
654 (ix86_print_operand): Likewise.
655 (ix86_expand_call): Likewise.
656 (ix86_output_function_return): Likewise.
657 (reg_encoded_number): Likewise.
658 (BDESC_VERIFYS): Likewise.
659 (ix86_init_mpx_builtins): Likewise.
660 (ix86_init_builtins): Likewise.
661 (ix86_emit_cmove): Likewise.
662 (ix86_emit_move_max): Likewise.
663 (ix86_expand_builtin): Likewise.
664 (ix86_builtin_mpx_function): Likewise.
665 (ix86_get_arg_address_for_bt): Likewise.
666 (ix86_load_bounds): Likewise.
667 (ix86_store_bounds): Likewise.
668 (ix86_load_returned_bounds): Likewise.
669 (ix86_store_returned_bounds): Likewise.
670 (ix86_class_likely_spilled_p): Likewise.
671 (ix86_hard_regno_mode_ok): Likewise.
672 (x86_order_regs_for_local_alloc): Likewise.
673 (ix86_mitigate_rop): Likewise.
674 (ix86_bnd_prefixed_insn_p): Likewise.
675 (ix86_mpx_bound_mode): Likewise.
676 (ix86_make_bounds_constant): Likewise.
677 (ix86_initialize_bounds): Likewise.
678 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
679 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
680 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
681 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
682 (TARGET_CHKP_BOUND_MODE): Likewise.
683 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
684 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
685 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
686 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
687 * config/i386/i386.h (TARGET_MPX): Likewise.
688 (TARGET_MPX_P): Likewise.
689 (VALID_BND_REG_MODE): Likewise.
690 (FIRST_BND_REG): Likewise.
691 (LAST_BND_REG): Likewise.
692 (enum reg_class): Likewise.
693 (BND_REG_P): Likewise.
694 (BND_REGNO_P): Likewise.
696 (ADJUST_INSN_LENGTH): Likewise.
697 * config/i386/i386.md: Likewise.
698 * config/i386/i386.opt: Likewise.
699 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
701 (LINK_MPX): Likewise.
702 (MPX_SPEC): Likewise.
703 (LIBMPX_SPEC): Likewise.
704 (LIBMPXWRAPPERS_SPEC): Likewise.
705 (CHKP_SPEC): Likewise.
706 * config/i386/predicates.md: Likewise.
707 * dbxout.c (dbxout_type): Likewise.
708 * doc/extend.texi: Likewise.
709 * doc/invoke.texi: Likewise.
710 * doc/md.texi: Likewise.
711 * doc/tm.texi: Likewise.
712 * doc/tm.texi.in: Likewise.
713 * dwarf2out.c (is_base_type): Likewise.
714 (gen_formal_types_die): Likewise.
715 (gen_subprogram_die): Likewise.
716 (gen_type_die_with_usage): Likewise.
717 (gen_decl_die): Likewise.
718 (dwarf2out_late_global_decl): Likewise.
719 * expr.c (expand_assignment): Likewise.
720 (emit_storent_insn): Likewise.
721 (store_expr_with_bounds): Likewise.
722 (store_expr): Likewise.
723 (expand_expr_real_1): Likewise.
724 * expr.h (store_expr_with_bounds): Likewise.
725 * function.c (use_register_for_decl): Likewise.
726 (struct bounds_parm_data): Likewise.
727 (assign_parms_augmented_arg_list): Likewise.
728 (assign_parm_find_entry_rtl): Likewise.
729 (assign_parm_is_stack_parm): Likewise.
730 (assign_parm_load_bounds): Likewise.
731 (assign_bounds): Likewise.
732 (assign_parms): Likewise.
733 (expand_function_start): Likewise.
734 * gcc.c (CHKP_SPEC): Likewise.
735 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
736 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
737 (wrestrict_dom_walker::check_call): Likewise.
738 * gimple.c (gimple_build_call_from_tree): Likewise.
739 * gimple.h (enum gf_mask): Likewise.
740 (gimple_call_with_bounds_p): Likewise.
741 (gimple_call_set_with_bounds): Likewise.
742 * gimplify.c (gimplify_init_constructor): Likewise.
743 * ipa-cp.c (initialize_node_lattices): Likewise.
744 (propagate_constants_across_call): Likewise.
745 (find_more_scalar_values_for_callers_subset): Likewise.
746 * ipa-hsa.c (process_hsa_functions): Likewise.
747 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
748 * ipa-icf.c (sem_function::merge): Likewise.
749 * ipa-inline.c (early_inliner): Likewise.
750 * ipa-pure-const.c (warn_function_noreturn): Likewise.
751 (warn_function_cold): Likewise.
752 (propagate_pure_const): Likewise.
753 * ipa-ref.h (enum GTY): Likewise.
754 * ipa-split.c (find_retbnd): Likewise.
755 (consider_split): Likewise.
756 (split_function): Likewise.
757 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
758 * ipa.c (walk_polymorphic_call_targets): Likewise.
759 (symbol_table::remove_unreachable_nodes): Likewise.
760 (process_references): Likewise.
761 (cgraph_build_static_cdtor_1): Likewise.
762 * lto-cgraph.c (lto_output_node): Likewise.
763 (output_refs): Likewise.
764 (compute_ltrans_boundary): Likewise.
765 (input_overwrite_node): Likewise.
766 (input_node): Likewise.
767 (input_cgraph_1): Likewise.
768 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
769 * passes.c (pass_manager::execute_early_local_passes): Likewise.
770 (class pass_chkp_instrumentation_passes): Likewise.
771 (make_pass_chkp_instrumentation_passes): Likewise.
772 * passes.def: Likewise.
773 * rtl.h (struct GTY): Likewise.
774 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
775 * stor-layout.c (layout_type): Likewise.
776 * symtab.c: Likewise.
777 * target.def: Likewise.
778 * targhooks.c (default_chkp_bound_type): Likewise.
779 (default_chkp_bound_mode): Likewise.
780 (default_builtin_chkp_function): Likewise.
781 (default_chkp_function_value_bounds): Likewise.
782 (default_chkp_make_bounds_constant): Likewise.
783 (default_chkp_initialize_bounds): Likewise.
784 * targhooks.h (default_chkp_bound_type): Likewise.
785 (default_chkp_bound_mode): Likewise.
786 (default_builtin_chkp_function): Likewise.
787 (default_chkp_function_value_bounds): Likewise.
788 (default_chkp_make_bounds_constant): Likewise.
789 (default_chkp_initialize_bounds): Likewise.
790 * toplev.c (compile_file): Likewise.
791 (process_options): Likewise.
792 * tree-core.h (DEF_BUILTIN): Likewise.
793 (DEF_BUILTIN_CHKP): Likewise.
794 * tree-inline.c (declare_return_variable): Likewise.
795 (remap_gimple_stmt): Likewise.
797 (initialize_inlined_parameters): Likewise.
798 (expand_call_inline): Likewise.
799 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
800 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
801 (make_pass_ipa_chkp_produce_thunks): Likewise.
802 (make_pass_chkp): Likewise.
803 (make_pass_chkp_opt): Likewise.
804 (make_pass_chkp_instrumentation_passes): Likewise.
805 * tree-pretty-print.c (dump_generic_node): Likewise.
806 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
807 * tree-ssa-dce.c (propagate_necessity): Likewise.
808 (eliminate_unnecessary_stmts): Likewise.
809 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
810 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
811 * tree-ssa-sccvn.h: Likewise.
812 * tree-ssa-strlen.c (get_string_length): Likewise.
813 (valid_builtin_call): Likewise.
814 (adjust_last_stmt): Likewise.
815 (handle_builtin_strchr): Likewise.
816 (handle_builtin_strcpy): Likewise.
817 (handle_builtin_stxncpy): Likewise.
818 (handle_builtin_memcpy): Likewise.
819 (handle_builtin_strcat): Likewise.
820 (strlen_check_and_optimize_stmt): Likewise.
821 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
822 * tree-streamer-in.c: Likewise.
823 * tree-streamer.c (record_common_node): Likewise.
824 * tree.c (tree_code_size): Likewise.
825 (wide_int_to_tree_1): Likewise.
826 (type_contains_placeholder_1): Likewise.
827 (build_common_tree_nodes): Likewise.
828 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
829 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
830 (POINTER_BOUNDS_P): Likewise.
831 (BOUNDED_TYPE_P): Likewise.
832 (BOUNDED_P): Likewise.
833 (CALL_WITH_BOUNDS_P): Likewise.
834 (pointer_bounds_type_node): Likewise.
835 * value-prof.c (gimple_ic): Likewise.
836 * var-tracking.c (vt_add_function_parameters): Likewise.
837 * varasm.c (make_decl_rtl): Likewise.
838 (assemble_start_function): Likewise.
839 (output_constant): Likewise.
840 (maybe_assemble_visibility): Likewise.
841 * varpool.c (ctor_for_folding): Likewise.
842 * chkp-builtins.def: Remove.
843 * ipa-chkp.c: Remove.
844 * ipa-chkp.h: Remove.
845 * rtl-chkp.c: Remove.
846 * rtl-chkp.h: Remove.
847 * tree-chkp-opt.c: Remove.
848 * tree-chkp.c: Remove.
849 * tree-chkp.h: Remove.
851 2018-06-07 Carl Love <cel@us.ibm.com>
853 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
854 vextract_fp_from_shortl): Add BE support.
856 2018-06-07 Paul Koning <ni1d@arrl.net>
858 * compare-elim.c (try_merge_compare): Don't merge compare if
859 address contains a side effect.
860 (try_eliminate_compare): Likewise.
862 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
864 * config.gcc: Support "tremont".
865 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
866 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
868 * config/i386/i386.c (m_TREMONT): Define.
869 (processor_target_table): Add "tremont".
870 (PTA_TREMONT): Define.
871 (ix86_lea_outperforms): Add TARGET_TREMONT.
872 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
873 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
874 and M_INTEL_GOLDMONT_PLUS.
875 (fold_builtin_cpu): Add "tremont".
876 (ix86_add_stmt_cost): Add TARGET_TREMONT.
877 (ix86_option_override_internal): Add "tremont".
878 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
879 (processor_type): Add PROCESSOR_TREMONT.
880 * config/i386/x86-tune.def: Add m_TREMONT.
881 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
883 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
885 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
886 symbol defined for msp430i* devices to be lower case.
888 gcc/testsuite/gcc.target/msp430/
889 * msp430i-device-symbol.c: New test.
890 * msp430f-device-symbol.c: New test.
891 * msp430.h: New test header file.
893 2018-06-07 Richard Biener <rguenther@suse.de>
895 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
896 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
897 Properly wrap signed arithmetic if overflow wraps.
899 2018-06-07 Jakub Jelinek <jakub@redhat.com>
901 PR tree-optimization/69615
902 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
903 of a cast from a same precision integral SSA_NAME in a bb dominated
904 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
905 cast to utype if rhs2 has already a compatible type.
907 2018-06-07 Richard Biener <rguenther@suse.de>
909 PR tree-optimization/85935
910 * graphite-scop-detection.c (find_params_in_bb): Analyze
911 condition operands with respect to the correct loop. Assert
912 the analysis doesn't fail.
914 2018-06-04 Carl Love <cel@us.ibm.com>
916 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
917 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
918 as it is slightly cheaper.
919 (first_match_or_eos_index_<mode>):
920 Calculate index using natural element order.
921 (first_match_index_<mode>):
922 Calculate index using natural element order.
923 (first_match_or_eos_index_<mode>):
924 Calculate index using natural order.
925 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
927 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
928 P9V_BUILTIN_VCLZLSBB_V16QI.
929 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
932 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
934 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
935 indentation and line wrap for many prototypes. Add missing
936 @smallexample directives around block of prototypes for vec_xl and
939 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
941 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
942 track if we pass or return IEEE 128-bit floating point.
943 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
944 C++ mangling that is compatible with GCC 8.1.
945 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
946 (init_cumulative_args): Note if we pass or return IEEE 128-bit
947 floating point types.
948 (rs6000_function_arg_advance_1): Likewise.
949 (rs6000_mangle_type): Optionally generate mangled names that match
950 what GCC 8.1 generated for IEEE 128-bit floating point types.
951 (rs6000_globalize_decl_name): If we have an external function that
952 passes or returns IEEE 128-bit types, generate a weak reference
953 from the mangled name used in GCC 8.1 to the current mangled
955 (rs6000_init_builtins): Make __ibm128 use the long double type if
956 long double is IBM extended double. Make __float128 use the long
957 double type if long double is IEEE 128-bit.
960 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
961 macro for __ibm128 built-in functions.
962 (PACK_IF): Add __ibm128 pack/unpack functions.
963 (UNPACK_IF): Likewise.
964 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
965 enable long double built-in functions if long double is IEEE
966 128-bit floating point.
967 (rs6000_invalid_builtin): Update long double built-in function
969 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
970 functions, adjust the built-in function to use the long double
971 built-in function if __ibm128 and long double are the same type.
972 * doc/extend.texi (PowerPC builtins): Update documention for
973 __builtin_{,un}pack_longdouble. Add documentation for
974 __builtin_{,un}pack_ibm128.
976 2018-06-06 Jim Wilson <jimw@sifive.com>
978 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
979 (struct machine_function): New field interrupt_mode.
980 (riscv_handle_type_attribute): New function. Add forward declaration.
981 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
982 (riscv_expand_epilogue): Check interrupt_mode field.
983 (riscv_set_current_function): Check interrupt attribute args and
984 set interrupt_mode field.
985 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
986 (riscv_sret, riscv_uret): New.
987 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
988 new arguments to interrupt attribute.
990 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
993 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
994 Don't handle -mcpu=power8 if -mpower9-vector is also used.
996 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
998 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
999 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1000 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1001 several redundant entries.
1003 2018-06-06 David Malcolm <dmalcolm@redhat.com>
1005 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1006 type from "rtx" to "rtx_insn *".
1007 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1008 for local "call_insn", removing cast.
1009 (ix86_expand_call): Likewise, introducing a "call_insn" local.
1011 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
1013 PR tree-optimization/86066
1014 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1015 for BIT_INSERT_EXPR stores.
1017 2018-06-06 Richard Biener <rguenther@suse.de>
1019 PR tree-optimization/86062
1020 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1021 component refs ontop
1022 of to be offsetted base.
1024 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1026 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1027 to be static and remove check on interrupt attribute name.
1029 gcc/testsuite/gcc.target/msp430/
1030 * function-attributes-4.c: New test.
1031 * static-interrupts.c: New test.
1033 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1035 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1036 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1038 2018-06-05 Steve Ellcey <sellcey@cavium.com>
1041 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1043 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1044 Remove second argument, change how error is called.
1045 (aarch64_layout_arg): Remove second argument from
1046 aarch64_err_no_fpadvsimd call.
1047 (aarch64_init_cumulative_args): Ditto.
1048 (aarch64_gimplify_va_arg_expr): Ditto.
1049 * config/aarch64/aarch64.md (mov<mode>): Ditto.
1051 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
1053 * config/i386/i386.md (simple_return_indirect_internal): New expander.
1054 (*simple_return_indirect_internal<mode>): Rename from
1055 simple_return_indirect_internal. Use W mode iterator.
1056 (rstorssp): New expander.
1057 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
1058 (clrssbsy): New expander.
1059 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
1061 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
1063 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1065 (cmse_check_pointed_object): Likewise.
1067 2018-06-05 Martin Liska <mliska@suse.cz>
1069 PR gcov-profile/47618
1070 * doc/invoke.texi: Document how -fprofile-dir format
1073 2018-06-05 Richard Biener <rguenther@suse.de>
1075 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1076 removal pretend DOM info isn't available so we do not update
1077 it and only remove edges, not dominated blocks. Actually free
1078 DOM info in case we removed something. Remove unreachable blocks.
1079 (mfb_keep_latches): Work with either DOM info or marked backedges.
1080 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1081 first. Mark backedges if DOM info isn't available.
1082 (Re-)compute DOM info after cleanup_control_flow_pre.
1084 2018-06-05 Richard Biener <rguenther@suse.de>
1086 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1087 (locus_discrim_hasher::hash): Adjust.
1088 (locus_discrim_hasher::equal): Likewise.
1089 (next_discriminator_for_locus): Work on line directly.
1090 (same_line_p): Pass in expanded locus1 as well.
1091 (assign_discriminators): Avoid redundant location expansions.
1093 2018-06-05 Richard Biener <rguenther@suse.de>
1095 PR tree-optimization/86046
1096 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1097 if required after clearing TREE_ADDRESSABLE.
1099 2018-06-05 Richard Biener <rguenther@suse.de>
1101 PR tree-optimization/86047
1102 * tree-ssa-loop.c (for_each_index): Glob handling of all
1103 decls and constants and really handle all of them.
1105 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1108 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1109 qualifier_void_pointer and qualifier_const_void_pointer.
1110 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1111 (arm_init_builtins): Handle the above.
1112 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1113 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1114 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1117 2018-06-05 Martin Liska <mliska@suse.cz>
1119 * auto-profile.c (read_autofdo_file): Do not use
1120 gcov_ctr_summary struct.
1121 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1122 * coverage.c (struct counts_entry): Likewise.
1123 (read_counts_file): Read just single summary entry.
1124 (get_coverage_counts): Use gcov_summary struct.
1125 * coverage.h (get_coverage_counts): Likewise.
1126 * gcov-dump.c (dump_working_sets): Likewise.
1127 (tag_summary): Dump just single summary.
1128 * gcov-io.c (gcov_write_summary): Write just histogram
1130 (gcov_read_summary): Read just single summary.
1131 (compute_working_sets): Use gcov_summary struct.
1132 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1133 of GCOV_COUNTERS_SUMMABLE.
1134 (GCOV_COUNTERS_SUMMABLE): Remove.
1135 (GCOV_FIRST_VALUE_COUNTER): Replace with
1136 GCOV_COUNTER_V_INTERVAL.
1137 (struct gcov_ctr_summary): Remove.
1138 (struct gcov_summary): Directly use fields of former
1140 (compute_working_sets): Use gcov_summary struct.
1141 * gcov.c (read_count_file): Do not use ctrs fields.
1142 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1144 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1146 * profile.c: Likewise.
1147 * profile.h: Likewise.
1149 2018-06-05 Martin Liska <mliska@suse.cz>
1151 PR gcov-profile/84846
1152 * gcov.c (output_lines): Print working directory only
1153 in intermediate format.
1155 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
1157 * config/s390/s390-builtin-types.def: Add void function type.
1158 * config/s390/s390-builtins.def: Use the function type for the
1161 2018-06-04 Jim Wilson <jimw@sifive.com>
1163 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
1165 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
1166 and maybe_eh_return. Change regno to unsigned int. Use new args to
1167 handle EH_RETURN_DATA_REGNO registers properly.
1168 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
1169 (riscv_expand_epilogue): Update comment. Change argument name and
1170 type. Update code to use new name and type. Pass new args to
1171 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
1173 * config/riscv/riscv.md (NORMAL_RETURN): New.
1174 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
1175 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
1176 (eh_return): Call gen_eh_return_internal and emit barrier.
1177 (eh_return_internal): Call riscv_expand_epilogue.
1179 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
1181 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
1182 bit_insertion field and declare can_be_merged_into method.
1183 (merged_store_group::can_be_merged_into): New method.
1184 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
1185 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
1186 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
1188 2018-06-04 Richard Biener <rguenther@suse.de>
1190 PR tree-optimization/85955
1191 * builtins.c (fold_builtin_sincos): Convert pointers to
1192 destination to appropriate type before dereferencing.
1194 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
1196 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
1198 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
1200 * expr.c (expand_expr_real_1): Force the operand into memory if
1201 its TYPE_MODE is BLKmode and if there is no integer mode for
1202 the number of bits being extracted.
1204 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1208 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1209 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
1211 2018-06-04 Richard Biener <rguenther@suse.de>
1213 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
1214 (cleanup_tree_cfg_noloop): ... single caller. Do
1215 start_recording_case_labels later.
1217 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
1219 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
1220 to _IMMINTRIN_H_INCLUDED.
1221 * config/i386/pconfigintrin.h: Ditto.
1222 * config/i386/waitpkgintrin.h: Ditto.
1223 * config/i386/immintrin.h: Add includes for sgxintrin.h,
1224 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
1225 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
1226 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
1227 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
1228 waitpkgintrin.h and cldemoteintrin.h.
1230 2018-06-04 Richard Biener <rguenther@suse.de>
1232 PR tree-optimization/86038
1233 * tracer.c (find_best_successor): Check probability for
1234 being initialized, bail out if not.
1236 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
1239 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
1240 of bits to ignore when comparing architectures.
1242 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1244 PR tree-optimization/69615
1245 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
1246 maximum or minimum of the type, try to merge it also as if
1247 range1 is + [-, x - 1] or + [x + 1, -].
1250 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
1252 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
1254 PR tree-optimization/86034
1255 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
1256 the unsigned bitfield type in a bit insertion sequence if it does not
1257 have a larger precision than the bitfield size.
1258 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
1260 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
1262 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
1264 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1266 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
1267 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
1268 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
1269 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
1271 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1273 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
1274 Disable -fdelete-null-pointer-checks for ELF toolchain.
1276 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1277 Kito Cheng <kito.cheng@gmail.com>
1279 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
1280 (nds32le-*-*, nds32be-*-*): Integrate checking process.
1281 (nds32*-*-*): Add glibc and uclibc conditions.
1282 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
1283 (TARGET_EXCEPT_UNWIND_INFO): Define.
1284 * config/nds32/elf.h: New file.
1285 * config/nds32/linux.h: New file.
1286 * config/nds32/nds32-elf.opt: New file.
1287 * config/nds32/nds32-linux.opt: New file.
1288 * config/nds32/nds32-fp-as-gp.c
1289 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
1290 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
1292 (nds32_asm_file_end): Ditto.
1293 (nds32_print_operand): Ditto.
1294 (nds32_insert_attributes): Ditto.
1295 (nds32_init_libfuncs): New function.
1296 (TARGET_HAVE_TLS): Define.
1297 (TARGET_INIT_LIBFUNCS): Define.
1298 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
1300 (TARGET_ELF): Apply different mcmodel setting.
1301 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
1302 been migrated into elf.h and linux.h files.
1303 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
1304 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
1305 (mcmodel): The content has been migrated into nds32-elf.opt and
1306 nds32-linux.opt files.
1307 * config/nds32/t-elf: New file.
1308 * config/nds32/t-linux: New file.
1310 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1311 Shiva Chen <shiva0217@gmail.com>
1313 * config/nds32/constants.md (unspec_volatile_element): Add
1314 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
1315 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
1317 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
1318 (make_pass_nds32_fp_as_gp): Declare.
1319 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
1321 (nds32_asm_function_end_prologue): Remove unused asm output.
1322 (nds32_asm_function_begin_epilogue): Remove unused asm output.
1323 (nds32_asm_file_start): Output necessary fp_as_gp information.
1324 (nds32_option_override): Adjust register usage.
1325 (nds32_expand_prologue): Consider fp_as_gp situation.
1326 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
1327 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
1330 (simple_return): Ditto.
1331 (omit_fp_begin): Output special directive for fp_as_gp.
1332 (omit_fp_end): Output special directive for fp_as_gp.
1333 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
1334 mforbid-fp-as-gp): New options.
1336 2018-06-01 Mark Wielaard <mark@klomp.org>
1338 * dwarf2out.c (dwarf2out_finish): Remove generation of
1339 DW_AT_loclists_base.
1341 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
1343 * gimple-ssa-store-merging.c: Include gimple-fold.h.
1344 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
1345 (struct merged_store_group): Add bit_insertion field.
1346 (dump_char_array): Use standard hexadecimal format.
1347 (merged_store_group::merged_store_group): Set bit_insertion to false.
1348 (merged_store_group::apply_stores): Use optimal buffer size. Deal
1349 with BIT_INSERT_EXPR stores. Move up code updating the mask and
1350 also print the mask in the dump file.
1351 (pass_store_merging::gate): Minor tweak.
1352 (imm_store_chain_info::coalesce_immediate): Fix wrong association
1353 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
1354 stores with INTEGER_CST stores.
1355 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
1356 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
1357 and use it throughout. Generate bit insertion sequences if need be.
1358 (pass_store_merging::process_store): Remove redundant condition.
1359 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
1361 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
1363 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
1364 the 128-bit floating point types. Fix function comment.
1366 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1368 * config/aarch64/aarch64-simd.md
1369 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
1371 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
1374 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1376 PR tree-optimization/85989
1377 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
1379 (backprop::intersect_uses): Check it when deciding whether this
1380 is a backedge reference.
1381 (backprop::process_block): Add each phi to m_visited_phis
1382 after visiting it, then clear it at the end.
1384 2018-06-01 Richard Biener <rguenther@suse.de>
1386 * tree-vectorizer.h (vect_dr_stmt): New function.
1387 (vect_get_load_cost): Adjust.
1388 (vect_get_store_cost): Likewise.
1389 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
1390 Use vect_dr_stmt instead of DR_SMTT.
1391 (vect_record_base_alignments): Likewise.
1392 (vect_calculate_target_alignment): Likewise.
1393 (vect_compute_data_ref_alignment): Likewise and make static.
1394 (vect_update_misalignment_for_peel): Likewise.
1395 (vect_verify_datarefs_alignment): Likewise.
1396 (vector_alignment_reachable_p): Likewise.
1397 (vect_get_data_access_cost): Likewise. Pass down
1398 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
1399 (vect_get_peeling_costs_all_drs): Likewise.
1400 (vect_peeling_hash_get_lowest_cost): Likewise.
1401 (vect_enhance_data_refs_alignment): Likewise.
1402 (vect_find_same_alignment_drs): Likewise.
1403 (vect_analyze_data_refs_alignment): Likewise.
1404 (vect_analyze_group_access_1): Likewise.
1405 (vect_analyze_group_access): Likewise.
1406 (vect_analyze_data_ref_access): Likewise.
1407 (vect_analyze_data_ref_accesses): Likewise.
1408 (vect_vfa_segment_size): Likewise.
1409 (vect_small_gap_p): Likewise.
1410 (vectorizable_with_step_bound_p): Likewise.
1411 (vect_prune_runtime_alias_test_list): Likewise.
1412 (vect_analyze_data_refs): Likewise.
1413 (vect_supportable_dr_alignment): Likewise.
1414 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1415 (vect_gen_prolog_loop_niters): Likewise.
1416 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1417 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
1419 (vect_recog_mask_conversion_pattern): Likewise.
1420 (vect_try_gather_scatter_pattern): Likewise.
1421 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
1422 to vect_get_store_cost.
1423 (vect_get_store_cost): Get stmt_info instead of DR.
1424 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
1425 (vect_get_load_cost): Get stmt_info instead of DR.
1427 2018-06-01 Richard Biener <rguenther@suse.de>
1430 * gimple-fold.c (var_decl_component_p): Also allow offsetted
1431 vars wrapped in MEM_REFs.
1433 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1435 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1436 Fix subreg tests so that we only return a choice between
1437 GENERAL_REGS and FP_REGS if the original classes included both.
1439 2018-06-01 Richard Biener <rguenther@suse.de>
1442 * tree-ssa-structalias.c (get_function_part_constraint):
1443 Handle NULL fi->decl.
1444 (find_func_aliases_for_call): Properly handle indirect
1445 fi from direct call.
1446 (find_func_clobbers): Likewise.
1447 (ipa_pta_execute): Likewise.
1448 (create_variable_info_for): For functions that are ifunc_resolver
1449 resolve to a varinfo that contains the result of the resolver call.
1450 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
1453 2018-05-31 Michael Collison <michael.collison@arm.com>
1455 * config/aarch64/aarch64.md:
1456 (*fix_to_zero_extenddfdi2): New pattern.
1457 * gcc.target/aarch64/fix_extend1.c: New testcase.
1459 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
1463 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
1464 and BUILT_IN_STRNCMP_EQ.
1465 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
1466 BUILT_IN_STRNCMP_EQ.
1467 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
1468 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1469 (gimple_fold_builtin): Likewise.
1470 * tree-ssa-strlen.c (compute_string_length): New function.
1471 (determine_min_obsize): New function.
1472 (handle_builtin_string_cmp): New function to handle calls to
1473 string compare functions.
1474 (strlen_optimize_stmt): Add handling to builtin string compare
1476 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1477 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1478 * tree.c (build_common_builtin_nodes): Add new defines of
1479 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
1481 2018-05-31 Jakub Jelinek <jakub@redhat.com>
1484 * bb-reorder.c (pass_partition_blocks::gate): Return false for
1485 functions with naked attribute.
1487 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
1489 * config/i386/sse.md (avx_vec_concat<mode>):
1490 Substitute concat_tg_mode mode attribute with xtg_mode.
1491 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
1492 (concat_tg_mode): Remove mode attribute.
1494 2018-05-31 Martin Sebor <msebor@redhat.com>
1497 * calls.c (alloc_max_size): Correct a logic error/typo.
1498 Treat excessive arguments as infinite. Warn for invalid arguments.
1499 * doc/invoke.texi (-Walloc-size-larger-than): Update.
1501 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
1504 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
1505 and movx for Haswell.
1507 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
1508 Cesar Philippidis <cesar@codesourcery.com>
1511 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
1512 when emitting error on private/firstprivate reductions.
1513 * omp-low.c (lower_omp_target): Avoid reference-type processing
1514 on pointers for firstprivate clause.
1516 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1518 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
1521 * config/aarch64/aarch64-simd.md
1522 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
1523 (aarch64_ld1_x3_<mode>): Likewise
1524 (aarch64_st1x2<VALLDIF:mode>): Likewise
1525 (aarch64_st1_x2_<mode>): Likewise
1526 (aarch64_st1x3<VALLDIF:mode>): Likewise
1527 (aarch64_st1_x3_<mode>): Likewise
1528 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
1529 (vld1_s8_x3): Likewise.
1530 (vld1_u16_x3): Likewise.
1531 (vld1_s16_x3): Likewise.
1532 (vld1_u32_x3): Likewise.
1533 (vld1_s32_x3): Likewise.
1534 (vld1_u64_x3): Likewise.
1535 (vld1_s64_x3): Likewise.
1536 (vld1_f16_x3): Likewise.
1537 (vld1_f32_x3): Likewise.
1538 (vld1_f64_x3): Likewise.
1539 (vld1_p8_x3): Likewise.
1540 (vld1_p16_x3): Likewise.
1541 (vld1_p64_x3): Likewise.
1542 (vld1q_u8_x3): Likewise.
1543 (vld1q_s8_x3): Likewise.
1544 (vld1q_u16_x3): Likewise.
1545 (vld1q_s16_x3): Likewise.
1546 (vld1q_u32_x3): Likewise.
1547 (vld1q_s32_x3): Likewise.
1548 (vld1q_u64_x3): Likewise.
1549 (vld1q_s64_x3): Likewise.
1550 (vld1q_f16_x3): Likewise.
1551 (vld1q_f32_x3): Likewise.
1552 (vld1q_f64_x3): Likewise.
1553 (vld1q_p8_x3): Likewise.
1554 (vld1q_p16_x3): Likewise.
1555 (vld1q_p64_x3): Likewise.
1556 (vst1_s64_x2): Likewise.
1557 (vst1_u64_x2): Likewise.
1558 (vst1_f64_x2): Likewise.
1559 (vst1_s8_x2): Likewise.
1560 (vst1_p8_x2): Likewise.
1561 (vst1_s16_x2): Likewise.
1562 (vst1_p16_x2): Likewise.
1563 (vst1_s32_x2): Likewise.
1564 (vst1_u8_x2): Likewise.
1565 (vst1_u16_x2): Likewise.
1566 (vst1_u32_x2): Likewise.
1567 (vst1_f16_x2): Likewise.
1568 (vst1_f32_x2): Likewise.
1569 (vst1_p64_x2): Likewise.
1570 (vst1q_s8_x2): Likewise.
1571 (vst1q_p8_x2): Likewise.
1572 (vst1q_s16_x2): Likewise.
1573 (vst1q_p16_x2): Likewise.
1574 (vst1q_s32_x2): Likewise.
1575 (vst1q_s64_x2): Likewise.
1576 (vst1q_u8_x2): Likewise.
1577 (vst1q_u16_x2): Likewise.
1578 (vst1q_u32_x2): Likewise.
1579 (vst1q_u64_x2): Likewise.
1580 (vst1q_f16_x2): Likewise.
1581 (vst1q_f32_x2): Likewise.
1582 (vst1q_f64_x2): Likewise.
1583 (vst1q_p64_x2): Likewise.
1584 (vst1_s64_x3): Likewise.
1585 (vst1_u64_x3): Likewise.
1586 (vst1_f64_x3): Likewise.
1587 (vst1_s8_x3): Likewise.
1588 (vst1_p8_x3): Likewise.
1589 (vst1_s16_x3): Likewise.
1590 (vst1_p16_x3): Likewise.
1591 (vst1_s32_x3): Likewise.
1592 (vst1_u8_x3): Likewise.
1593 (vst1_u16_x3): Likewise.
1594 (vst1_u32_x3): Likewise.
1595 (vst1_f16_x3): Likewise.
1596 (vst1_f32_x3): Likewise.
1597 (vst1_p64_x3): Likewise.
1598 (vst1q_s8_x3): Likewise.
1599 (vst1q_p8_x3): Likewise.
1600 (vst1q_s16_x3): Likewise.
1601 (vst1q_p16_x3): Likewise.
1602 (vst1q_s32_x3): Likewise.
1603 (vst1q_s64_x3): Likewise.
1604 (vst1q_u8_x3): Likewise.
1605 (vst1q_u16_x3): Likewise.
1606 (vst1q_u32_x3): Likewise.
1607 (vst1q_u64_x3): Likewise.
1608 (vst1q_f16_x3): Likewise.
1609 (vst1q_f32_x3): Likewise.
1610 (vst1q_f64_x3): Likewise.
1611 (vst1q_p64_x3): Likewise.
1613 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1615 * config/msp430/msp430.c (msp430_output_labelref): Prepend
1616 user_label_prefix to name.
1618 * tree-core.h: Update comment about the format of NAME string
1619 passed to handler in attribute_spec.
1621 * config/msp430/msp430.md: Remove erroneous subreg expression from
1622 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
1623 zero_extend{q,h}isi2.
1625 2018-05-30 Borislav Petkov <bp@suse.de>
1627 * doc/extend.texi: Document some architecture specific
1628 constraints and sort entries.
1630 2018-05-30 Martin Sebor <msebor@redhat.com>
1633 * builtins.c (expand_builtin_stpcpy_1): New function.
1634 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
1635 only if the former succeeds.
1637 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1639 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
1642 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1644 * doc/invoke.texi (-flinker-output): Document
1646 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1648 * passes.c (ipa_write_summaries): Only modify statements if body
1650 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
1651 incrementally linking.
1652 (ipa_passes): Likewise.
1653 * lto-cgraph.c (lto_output_node): When incrementally linking do not
1654 pass down resolution info.
1655 * common.opt (flag_incremental_link): Update info.
1656 * gcc.c (plugin specs): Turn flinker-output=* to
1657 -plugin-opt=-linker-output-known
1658 * toplev.c (compile_file): Also cut compilation when doing incremental
1660 * flag-types. (enum lto_partition_model): Add
1661 LTO_LINKER_OUTPUT_NOLTOREL.
1662 (invoke.texi): Add -flinker-output docs.
1663 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
1664 link same way as WPA; do not stream in dead initializers.
1666 * dwarf2out.c (dwarf2out_die_ref_for_decl,
1667 darf2out_register_external_decl): Support incremental link.
1669 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1671 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
1673 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1675 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
1676 it down to simple_object_copy_lto_debug_sections.
1677 (run_gcc): Determine incremental LTO link time and configure
1678 lto1 into non-wpa mode, disable renaming of debug sections.
1680 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
1682 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1683 descriptions of various incorrectly documented functions.
1685 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
1688 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1689 address check not strict.
1691 2018-05-30 Richard Biener <rguenther@suse.de>
1693 PR tree-optimization/85964
1694 * tracer.c (better_p): Drop initialized count check, we only
1695 call the function with initialized counts now.
1696 (find_best_successor): Do find a best edge if one
1697 has uninitialized count.
1698 (find_best_predecessor): Likewise. Do BB frequency check only
1699 if count is initialized.
1701 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
1703 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
1704 (aarch64_ldrstr_offset_compare): New.
1705 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
1706 load/store orderings.
1707 (aarch64_gen_adjusted_ldpstp): Likewise.
1709 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
1711 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1712 Check for subset of GENERAL_REGS and FP_REGS.
1713 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
1716 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
1718 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
1719 and wi::to_poly_offset. Add the current offset and then check
1720 whether the sum fits, rather than using an unchecked addition of
1721 a checked term. Check for a shwi rather than a uhwi.
1722 * expr.c (get_bit_range): Use tree_to_poly_uint64.
1723 (store_constructor): Use poly_int_tree_p.
1724 (expand_expr_real_1): Likewise.
1725 * function.c (assign_temp): Likewise.
1726 * fold-const.c (const_binop): Use poly_int_tree_p and
1728 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
1730 * ipa-icf-gimple.c (func_checker::compare_operand): Use
1731 to_poly_offset for MEM offsets.
1732 * ipa-icf.c (sem_variable::equals): Likewise.
1733 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
1734 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
1735 wi::to_poly_offset for BIT_FIELD_REF offsets.
1736 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
1738 * var-tracking.c (emit_note_insn_var_location): Use
1739 tree_to_poly_uint64.
1741 2018-05-29 Jim Wilson <jimw@sifive.com>
1743 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
1745 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
1748 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
1749 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
1751 (sse4_1_round<mode>2): Use nonimmediate_operand
1752 for operand 1 predicate.
1754 2018-05-29 Martin Sebor <msebor@redhat.com>
1755 Richard Biener <rguenther@suse.de>
1758 * calls.c (get_size_range): Call determine_value_range instead
1759 of get_value_range..
1760 * tree-vrp.h (determine_value_range): Declared new function.
1761 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
1763 2018-05-29 Richard Biener <rguenther@suse.de>
1765 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
1766 sure to use non-pattern stmts for get_earlier_stmt arguments.
1767 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
1768 called on pattern stmts.
1769 (get_later_stmt): Likewise.
1771 2018-05-29 Martin Liska <mliska@suse.cz>
1773 PR gcov-profile/85759
1774 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
1777 2018-05-29 Jakub Jelinek <jakub@redhat.com>
1779 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
1781 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
1785 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
1786 VEC_PACK_FLOAT_EXPR): New tree codes.
1787 * tree-pretty-print.c (op_code_prio): Handle
1788 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
1789 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
1790 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
1791 * tree-inline.c (estimate_operator_cost): Likewise.
1792 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
1793 * fold-const.c (const_binop): Likewise.
1794 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
1795 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
1796 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1797 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
1798 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
1799 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
1800 * expr.c (expand_expr_real_2): Likewise.
1801 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
1802 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
1803 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
1805 * optabs.c (expand_widen_pattern_expr): For
1806 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
1807 sign from result type rather than operand's type.
1808 (expand_binop_directly): For vec_packu_float_optab and
1809 vec_packs_float_optab allow result type to be different from operand's
1811 * optabs-tree.c (optab_for_tree_code): Handle
1812 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
1813 VEC_PACK_FLOAT_EXPR. Formatting fixes.
1814 * tree-vect-generic.c (expand_vector_operations_1): Handle
1815 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
1816 VEC_PACK_FLOAT_EXPR.
1817 * tree-vect-stmts.c (supportable_widening_operation): Handle
1819 (supportable_narrowing_operation): Handle FLOAT_EXPR.
1820 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
1821 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
1822 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
1823 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
1825 (vec_pack<floatprefix>_float_<mode>): New expander.
1826 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
1828 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
1829 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
1830 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
1831 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
1832 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
1834 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
1835 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
1836 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
1837 VEC_PACK_FLOAT_EXPR): Document.
1839 2018-05-29 Richard Biener <rguenther@suse.de>
1841 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
1843 (stmt_vec_info_vec): Make pointer.
1844 (init_stmt_vec_info_vec): Remove.
1845 (free_stmt_vec_info_vec): Likewise.
1846 (set_stmt_vec_info_vec): New function.
1847 (free_stmt_vec_infos): Likewise.
1848 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
1849 (set_vinfo_for_stmt): Likewise.
1850 (get_earlier_stmt): Likewise.
1851 (get_later_stmt): Likewise.
1852 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
1853 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
1854 (vec_info::~vec_info): Free stmt_vec_infos.
1855 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
1856 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
1857 (pass_slp_vectorize::execute): Likewise.
1858 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
1859 (free_stmt_vec_info_vec): Likewise.
1860 (set_stmt_vec_info_vec): New function.
1861 (free_stmt_vec_infos): Likewise.
1862 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
1863 the global stmt_vec_info_vec.
1864 * tree-parloops.c (gather_scalar_reductions): Use
1865 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
1868 2018-05-29 Richard Biener <rguenther@suse.de>
1870 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
1872 2018-05-29 Martin Liska <mliska@suse.cz>
1873 David Malcolm <dmalcolm@redhat.com>
1875 * vec.c (test_reverse): New.
1876 (vec_c_tests): Add new test.
1877 * vec.h (vl_ptr>::reverse): New function.
1879 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
1881 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
1883 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
1886 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
1888 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
1890 2018-05-28 Richard Biener <rguenther@suse.de>
1892 PR tree-optimization/85933
1893 * tree-vect-data-refs.c (vect_record_base_alignments): Only
1894 look at stmts marked as vectorizable.
1896 2018-05-28 Richard Biener <rguenther@suse.de>
1898 PR tree-optimization/85934
1899 * tree-vect-generic.c (expand_vector_operations_1): Hoist
1900 vector boolean check before scalar optimization.
1902 2018-05-28 Jakub Jelinek <jakub@redhat.com>
1904 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
1907 2018-05-28 Mark Wielaard <mark@klomp.org>
1909 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
1910 if it is an expression containing a minus sign.
1912 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
1914 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1916 2018-05-27 Paul Koning <ni1d@arrl.net>
1918 * config/pdp11/pdp11.md (truncsihi2): Remove.
1920 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
1921 Chung-Ju Wu <jasonwucj@gmail.com>
1923 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
1925 (unaligned_store_dw): Ditto.
1926 * config/nds32/nds32-memory-manipulation.c
1927 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
1928 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
1929 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
1930 (emit_setmem_word_loop): Rename to ...
1931 (emit_setmem_doubleword_loop): ... this.
1932 (nds32_gen_dup_4_byte_to_word_value): New function.
1933 (nds32_gen_dup_8_byte_to_double_word_value): New function.
1934 (nds32_expand_setmem_loop): Refine implementation.
1935 (nds32_expand_setmem_loop_v3m): Ditto.
1936 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
1939 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
1941 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
1943 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
1945 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
1946 (nds32_init_machine_status): Initialize machine->attr_naked_p and
1947 machine->attr_no_prologue_p.
1948 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
1949 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
1950 (nds32_expand_epilogue): Consider attr_naked_p.
1951 (nds32_expand_epilogue_v3pop): Likewise.
1952 (nds32_can_use_return_insn): Likewise.
1953 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
1954 attr_no_prologue_p fields.
1955 * config/nds32/nds32.opt (mret-in-naked-func): New option.
1957 2018-05-27 Jakub Jelinek <jakub@redhat.com>
1960 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
1962 * config/i386/sse.md
1963 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
1965 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
1967 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
1969 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
1971 (*<floatsuffix>floatv2div2sf2): Rename to ...
1972 (*float<floatunssuffix>v2div2sf2): ... this.
1973 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
1974 (float<floatunssuffix>v2div2sf2_mask): ... this.
1975 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
1976 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
1977 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
1979 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
1981 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1983 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1985 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1987 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1989 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
1990 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
1991 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
1992 gen_ufix_truncv8dfv8si2.
1993 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
1994 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
1995 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
1996 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
1997 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
1998 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
1999 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2000 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2002 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2006 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2008 2018-05-25 Jim Wilson <jimw@sifive.com>
2010 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2011 * config/riscv/riscv.c (struct machine_function): Add
2012 interrupt_handler_p and attribute_checked_p fields.
2013 (riscv_attribute_table): Add interrupt.
2014 (riscv_interrupt_type_p): New.
2015 (riscv_save_reg_p): Save extra regs for interrupt handler.
2016 (riscv_use_save_libcall): Return false for interrupt handler.
2017 (riscv_first_stack_step): Add forward declaration.
2018 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
2019 for interrupt handler with large frame. Use it for saved reg list.
2020 (riscv_expand_prologue): Move flag_stack_usage_info support to
2021 eliminate duplication.
2022 (riscv_expand_epilogue): Generate mret for interrupt handler.
2023 (riscv_epilogue_uses): New.
2024 (riscv_can_use_return_insn): Return false for interrupt handler.
2025 (riscv_function_ok_for_sibcall): Likewise.
2026 (riscv_set_current_function): Add interrupt handler support.
2027 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2028 * config/riscv/riscv.md (UNSPECV_MRET): New.
2030 (riscv_frflags, riscv_fsflags): Use tab after opcode.
2032 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2034 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
2036 PR tree-optimization/85712
2037 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2038 this candidate has already been replaced in-situ by a copy.
2040 2018-05-25 Jason Merrill <jason@redhat.com>
2042 PR c++/80485 - inline function non-zero address.
2043 * symtab.c (nonzero_address): Check DECL_COMDAT.
2045 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
2048 * config/alpha/alpha.md (ashlsi3): New insn pattern.
2049 (*ashlsi_se): Rename from *ashldi_se. Define as sign
2050 extension of SImode operation. Use const123_operand predicate.
2051 (*saddsi_1): Remove.
2052 (*saddl_se_1): Ditto.
2054 (*ssubl_se_1): Ditto.
2055 * config/alpha/predicates.md (const123_operand): New predicate.
2056 * config/alpha/constraints.md (P): Use IN_RANGE.
2058 2018-05-25 Richard Biener <rguenther@suse.de>
2060 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2062 (ref_maybe_used_by_stmt_p): Likewise.
2063 (stmt_may_clobber_ref_p): Likewise.
2064 (stmt_may_clobber_ref_p_1): Likewise.
2065 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2067 (ref_maybe_used_by_stmt_p): Likewise.
2068 (stmt_may_clobber_ref_p): Likewise.
2069 (stmt_may_clobber_ref_p_1): Likewise.
2070 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2071 the alias oracle to disambiguate DRs with stmts DR analysis
2073 (vect_analyze_data_refs): Do not give up on not analyzable
2074 DRs for BB vectorization. Remove code truncating the dataref
2077 2018-05-25 Jakub Jelinek <jakub@redhat.com>
2080 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2081 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
2082 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2084 2018-05-25 Richard Biener <rguenther@suse.de>
2086 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2087 function, combining stmt data ref gathering and fatal analysis
2089 (vect_analyze_data_refs): Remove now redudnant code and simplify.
2090 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2091 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2092 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2093 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2095 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2097 PR tree-optimization/85720
2098 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2099 SCC if all partitions are builtins.
2100 (version_loop_by_alias_check): New parameter. Generate cancelable
2101 runtime alias check if all partitions are builtins.
2102 (distribute_loop): Update call to above function.
2104 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2106 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2107 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2108 (parm_default_def_partition_arg): Ditto.
2109 (set_parm_default_def_partition): Ditto.
2110 (get_parm_default_def_partitions): Ditto and make it static.
2111 (get_undefined_value_partitions): Ditto and make it static.
2112 (remove_ssa_form): Refactor call to init_var_map here.
2113 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2114 computation for loop region.
2115 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2116 (register_default_def): Delete.
2117 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2118 (parm_default_def_partition_arg): Ditto.
2119 (set_parm_default_def_partition): Ditto.
2120 (get_parm_default_def_partitions): Ditto and make it static.
2121 (get_undefined_value_partitions): Ditto and make it static.
2122 (coalesce_with_default, coalesce_with_default): Update comment.
2123 (create_coalesce_list_for_region): New func factored out from
2124 create_outofssa_var_map.
2125 (populate_coalesce_list_for_outofssa): New func factored out from
2126 create_outofssa_var_map and coalesce_ssa_name.
2127 (create_outofssa_var_map): Delete.
2128 (coalesce_ssa_name): Refactor to support live range computation.
2129 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2130 (get_parm_default_def_partitions): Delete.
2131 (get_undefined_value_partitions): Ditto.
2132 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2133 computation for loop region.
2134 (new_tree_live_info, loe_visit_block): Ditto.
2135 (live_worklist, set_var_live_on_entry): Ditto.
2136 (calculate_live_on_exit, verify_live_on_entry): Ditto.
2137 * tree-ssa-live.h (struct _var_map): New fields.
2138 (init_var_map): Change decl.
2139 (region_contains_p): New.
2141 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2143 * tree-ssa-live.h (live_merge_and_clear): Delete.
2145 2018-05-25 Richard Biener <rguenther@suse.de>
2148 * tree-dump.c (dequeue_and_dump): Remove access to removed
2149 operand 2 of a SWITCH_EXPR.
2151 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2153 * doc/sourcebuild.texi (vect_double_cond_arith): Include
2154 multiplication and division.
2155 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2156 (cond_udiv@var{m}, cond_umod@var{m}): Document.
2157 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2158 (cond_udiv_optab, cond_umod_optab): New optabs.
2159 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2160 (IFN_COND_RDIV): New internal functions.
2161 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
2162 TRUNC_MOD_EXPR and RDIV_EXPR.
2163 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
2164 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
2166 (SVE_INT_BINARY): Include mult.
2167 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
2168 (optab, sve_int_op): Handle mult.
2169 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
2171 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
2172 for SVE_INT_BINARY_SD.
2174 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2176 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
2177 (optab, sve_int_op): Handle div and udiv.
2178 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
2179 for SVE_INT_BINARY_SD.
2180 (*<optab><mode>3): New insn for the same.
2182 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2184 * tree-vect-patterns.c: Include predict.h.
2185 (vect_recog_divmod_pattern): Restrict check for division support
2186 to when optimizing for size.
2188 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2190 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
2191 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
2192 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
2193 (gimple_match_op::set_op): Likewise.
2194 (gimple_resimplify4): Declare.
2195 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
2196 (expr::gen_transform): Likewise.
2197 (decision_tree::gen): Generate a simplification routine for 4 operands.
2198 * gimple-match-head.c (gimple_simplify): Add an overload for
2199 4 operands. In the top-level function, handle up to 4 call
2200 arguments and call gimple_resimplify4.
2201 (gimple_resimplify4): New function.
2202 (build_call_internal): Pass a fourth operand.
2203 (maybe_push_to_seq): Likewise.
2204 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
2205 Fold VEC_COND_EXPRs of an operation and a default value into
2206 an IFN_COND_* function if possible.
2207 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
2209 (SVE_COND_FP_BINARY): Include them.
2210 (optab, sve_fp_op): Handle them.
2211 (SVE_INT_BINARY_REV): New code iterator.
2212 (SVE_COND_FP_BINARY_REV): New int iterator.
2213 (commutative): New int attribute.
2214 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
2216 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
2218 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
2219 (*cond_<optab><mode>): New patterns for reversed operands.
2221 2018-05-25 Richard Biener <rguenther@suse.de>
2223 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
2224 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
2225 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
2226 (STMT_VINFO_GROUPED_ACCESS): Adjust.
2227 * tree-vect-data-refs.c (everywhere): Adjust users.
2228 * tree-vect-loop.c (everywhere): Likewise.
2229 * tree-vect-slp.c (everywhere): Likewise.
2230 * tree-vect-stmts.c (everywhere): Likewise.
2231 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2233 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2235 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
2237 (gcc_cv_as_section_exclude): ... this.
2238 Try Solaris as #exclude syntax.
2239 * configure: Regenerate.
2240 * config.in: Regenerate.
2241 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
2243 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
2244 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
2246 * varasm.c (default_elf_asm_named_section): Don't check if
2247 HAVE_GAS_SECTION_EXCLUDE is defined.
2249 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2251 * doc/md.texi: Update the documentation of the cond_* optabs
2252 to mention the new final operand. Fix GET_MODE_NUNITS call.
2253 Describe the scalar case too.
2254 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
2255 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
2257 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
2258 (get_conditional_internal_fn): Update comment.
2259 * tree-vect-loop.c (vectorizable_reduction): Pass the original
2260 accumulator value as a final argument to conditional functions.
2261 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
2262 a define_expand and add an "else" operand. Assert for now that
2263 the else operand is equal to operand 2. Use SVE_INT_BINARY and
2264 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
2265 (*cond_<optab><mode>): New patterns.
2266 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
2267 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
2268 (UNSPEC_COND_EOR): Delete.
2269 (optab): Remove associated mappings.
2270 (SVE_INT_BINARY): New code iterator.
2271 (sve_int_op): Remove int attribute and add "minus" to the code
2273 (SVE_COND_INT_OP): Delete.
2274 (SVE_COND_FP_OP): Rename to...
2275 (SVE_COND_FP_BINARY): ...this.
2277 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2279 * optabs.c (can_reuse_operands_p): New function.
2280 (maybe_legitimize_operands): Try to reuse the results for
2283 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2285 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
2286 Add {q} suffix to insn mnemonic.
2288 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2290 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
2291 (msp430_warn_func_return): New.
2293 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
2295 * fold-const.c (tree_nonzero_bits): New function.
2296 * fold-const.h (tree_nonzero_bits): Likewise.
2297 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
2298 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
2300 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2304 * varasm.c (assemble_alias): Check ifunc_resolver only on
2307 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2310 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
2311 when memory input operand is handled.
2313 2018-05-24 Luis Machado <luis.machado@linaro.org>
2315 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
2317 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
2319 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2321 * match.pd: Delay FMA folds until after vectorization.
2323 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
2326 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2327 address check not strict.
2329 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2331 * gimple-match.h (gimple_match_op): New class.
2332 (mprts_hook): Replace parameters with a gimple_match_op *.
2333 (maybe_build_generic_op): Likewise.
2334 (gimple_simplified_result_is_gimple_val): Replace parameters with
2335 a const gimple_match_op *.
2336 (gimple_simplify): Replace code_helper * and tree * parameters with
2337 a gimple_match_op * parameter.
2338 (gimple_resimplify1): Replace code_helper *, tree and tree *
2339 parameters with a gimple_match_op * parameter.
2340 (gimple_resimplify2): Likewise.
2341 (gimple_resimplify3): Likewise.
2342 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
2343 parameters with a gimple_match_op * parameter.
2344 * gimple-match-head.c (gimple_simplify): Change prototypes of
2345 auto-generated functions to take a gimple_match_op * instead of
2346 separate code_helper * and tree * parameters. Make the same
2347 change in the top-level overload and update calls to the
2348 gimple_resimplify routines. Update calls to the auto-generated
2349 functions and to maybe_push_res_to_seq in the publicly-facing
2350 operation-specific gimple_simplify overloads.
2351 (gimple_match_op::MAX_NUM_OPS): Define.
2352 (gimple_resimplify1): Replace rcode and ops with a single res_op
2353 parameter. Update call to gimple_simplify.
2354 (gimple_resimplify2): Likewise.
2355 (gimple_resimplify3): Likewise.
2356 (mprts_hook): Replace parameters with a gimple_match_op *.
2357 (maybe_build_generic_op): Likewise.
2358 (build_call_internal): Replace type, nargs and ops with
2359 a gimple_match_op *.
2360 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
2361 with a single gimple_match_op *. Update calls to mprts_hook,
2362 build_call_internal and gimple_simplified_result_is_gimple_val.
2363 Factor out code that is common to the tree_code and combined_fn cases.
2364 * genmatch.c (expr::gen_transform): Replace tem_code and
2365 tem_ops with a gimple_match_op called tem_op. Update calls
2366 to the gimple_resimplify functions and maybe_push_res_to_seq.
2367 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
2368 res_ops. Update call to the gimple_resimplify functions.
2369 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
2370 (decision_tree::gen): Make the functions take a gimple_match_op *
2371 called res_op instead of separate res_code and res_ops parameters.
2372 Update call accordingly.
2373 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
2374 and ops with a single res_op parameter. Update calls to
2375 maybe_build_generic_op and maybe_push_res_to_seq.
2376 (fold_stmt_1): Update calls to gimple_simplify and
2377 replace_stmt_with_simplification.
2378 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
2379 and gimple_simplified_result_is_gimple_val.
2380 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
2382 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
2383 with a gimple_match_op *.
2384 (vn_nary_build_or_lookup): Likewise. Update call to
2385 vn_nary_build_or_lookup_1.
2386 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
2387 gimple_match_op *. Update calls to the gimple_resimplify routines
2388 and to gimple_simplified_result_is_gimple_val.
2389 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
2390 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
2391 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
2392 (visit_nary_op): Likewise.
2393 (visit_reference_op_load): Likewise.
2395 2018-05-23 Luis Machado <luis.machado@linaro.org>
2397 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
2398 modifier for printing the step amount.
2400 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
2403 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
2406 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
2408 * doc/sourcebuild.texi (Endianness): New subsubsection.
2410 2018-05-23 Luis Machado <luis.machado@linaro.org>
2412 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2413 <prefetch_dynamic_strides>: New const bool field.
2414 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2415 prefetch_dynamic_strides.
2416 (exynosm1_prefetch_tune): Likewise.
2417 (thunderxt88_prefetch_tune): Likewise.
2418 (thunderx_prefetch_tune): Likewise.
2419 (thunderx2t99_prefetch_tune): Likewise.
2420 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
2422 (aarch64_override_options_internal): Update to set
2423 PARAM_PREFETCH_DYNAMIC_STRIDES.
2424 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2425 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2426 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2427 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2428 prefetch-dynamic-strides setting.
2430 2018-05-23 Luis Machado <luis.machado@linaro.org>
2432 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2433 <minimum_stride>: New const int field.
2434 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2435 minimum_stride field defaulting to -1.
2436 (exynosm1_prefetch_tune): Likewise.
2437 (thunderxt88_prefetch_tune): Likewise.
2438 (thunderx_prefetch_tune): Likewise.
2439 (thunderx2t99_prefetch_tune): Likewise.
2440 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
2441 <default_opt_level>: Set to 3.
2442 (aarch64_override_options_internal): Update to set
2443 PARAM_PREFETCH_MINIMUM_STRIDE.
2444 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2445 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2446 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2447 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2448 stride is constant and is below the minimum stride threshold.
2450 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2452 * config/arm/arm-cpus.in (mode26): Delete.
2453 (armv4): Delete mode26 reference.
2454 * config/arm/arm.c (arm_configure_build_target): Delete use of
2457 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2459 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
2461 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
2462 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
2464 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
2465 (floatunsdidf2): Ditto.
2467 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2469 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
2470 (fixuns_trunc<mode>si2_avx512f): Ditto.
2471 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
2472 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
2473 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
2475 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2477 PR rtl-optimization/79985
2478 * df-scan.c (df_insn_refs_collect): Remove special case for
2479 global registers and asm statements.
2481 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2483 * extend.texi (Global Register Variables): Rewrite the bullet list.
2484 Note that the register is available for allocation. Note that access
2485 via inline asm must use constraints. Add note about async-signal
2486 handlers. Remove paragraph about automagic register selection.
2488 2018-05-23 Richard Biener <rguenther@suse.de>
2490 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
2491 of fixed offset from memset VN.
2493 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
2495 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
2497 (alloc_cand_and_find_basis): Initialize first_interp field.
2498 (slsr_process_mul): Modify first_interp field.
2499 (slsr_process_add): Likewise.
2500 (slsr_process_cast): Modify first_interp field for each new
2502 (slsr_process_copy): Likewise.
2503 (dump_candidate): Dump first_interp field.
2504 (replace_mult_candidate): Process all interpretations, not just
2506 (replace_rhs_if_not_dup): Likewise.
2507 (replace_one_candidate): Likewise.
2509 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
2511 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
2513 (aarch64_needs_frame_chain): New function.
2514 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
2516 2018-05-23 Sudakshina Das <sudi.das@arm.com>
2519 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
2520 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
2521 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
2522 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
2523 as true for strict-align.
2524 (aarch64_can_inline_p): Perform checks even when callee has no
2525 attributes to check for strict alignment.
2526 * doc/extend.texi (AArch64 Function Attributes): Document
2528 * doc/invoke.texi: (AArch64 Options): Likewise.
2530 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
2532 PR tree-optimization/85853
2533 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
2534 the handling of the root of the node to...
2535 (vect_slp_analyze_node_operations_1): ...this new function,
2536 and run the whole thing with the child nodes' def types
2537 set according to their SLP node's def type.
2539 2018-05-23 Richard Biener <rguenther@suse.de>
2542 * tree-data-ref.c (create_runtime_alias_checks): Defer
2543 and ignore overflow warnings.
2545 2018-05-23 Yury Gribov <tetra2005@gmail.com>
2547 PR tree-optimization/85822
2548 * tree-vrp.c (is_masked_range_test): Fix handling of negative
2551 2018-05-23 Richard Biener <rguenther@suse.de>
2553 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2554 memset constants via native_interpret_expr.
2556 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
2559 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
2561 (cgraph_node::create_alias): Likewise.
2562 (cgraph_node::get_availability): Check ifunc_resolver instead
2563 of looking up ifunc attribute.
2564 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2565 * varasm.c (do_assemble_alias): Likewise.
2566 (assemble_alias): Likewise.
2567 (default_binds_local_p_3): Likewise.
2568 * cgraph.h (cgraph_node): Add ifunc_resolver.
2569 (cgraph_node::only_called_directly_or_aliased_p): Return false
2571 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
2573 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
2574 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
2575 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
2576 instead of looking up ifunc attribute.
2578 2018-05-22 Luis Machado <luis.machado@linaro.org>
2580 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
2582 2018-05-22 Martin Sebor <msebor@redhat.com>
2585 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
2586 only when expasion succeeds.
2587 (expand_builtin_strcmp): Same.
2588 (expand_builtin_strncmp): Same.
2590 2018-05-22 Martin Sebor <msebor@redhat.com>
2592 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
2594 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2595 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2597 * config/aarch64/aarch64-ldpstp.md: Replace uses of
2598 aarch64_mem_pair_operand with memory_operand and delete operand swapping
2600 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2601 Add check for legitimate_address.
2602 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
2603 (aarch64_swap_ldrstr_operands): New.
2604 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
2607 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2608 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2610 * config/aarch64/aarch64.md: New patterns to generate stp
2612 (store_pair_sw, store_pair_dw): New patterns to generate stp for
2613 single words and double words.
2614 (load_pair_sw, load_pair_dw): Likewise.
2615 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
2617 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
2619 * config/aarch64/aarch64-ldpstp.md: Modify peephole
2620 for different mode ldpstp and add peephole for merged zero stores.
2622 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2624 (aarch64_gen_store_pair): Rename calls to match new patterns.
2625 (aarch64_gen_load_pair): Rename calls to match new patterns.
2626 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
2627 (load_pair<DREG:mode><DREG2:mode>): ... This.
2628 (store_pair<mode>): Rename to...
2629 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
2630 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
2632 (V_INT_EQUIV): Handle SImode.
2633 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
2636 2018-05-22 Martin Sebor <msebor@redhat.com>
2639 * calls.c (maybe_warn_nonstring_arg): Use string length to set
2640 or ajust the presumed bound on an operation to avoid unnecessary
2643 2018-05-22 Martin Sebor <msebor@redhat.com>
2645 PR tree-optimization/85826
2646 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
2647 assuming that a DECL necesarily has a constant size.
2649 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
2652 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
2654 2018-05-22 Richard Biener <rguenther@suse.de>
2656 PR tree-optimization/85834
2657 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
2658 non-constant and non-zero memset arguments.
2660 2018-05-22 Martin Liska <mliska@suse.cz>
2663 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
2665 2018-05-22 Richard Biener <rguenther@suse.de>
2667 PR tree-optimization/85863
2668 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
2669 comparisons when vectype is specified.
2670 (vectorizable_condition): Do not specify vectype for
2671 vect_is_simple_cond when SLP vectorizing.
2673 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
2676 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
2677 define __ibm128 as long double.
2678 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
2679 as a distinct type when IEEE 128-bit support is enabled.
2680 (init_float128_ieee): Fix up conversions between IFmode and IEEE
2681 128-bit types to use the correct functions.
2682 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
2683 convert between 128-bit floating point types that have different
2684 modes but the same representation, instead of using gen_lowpart to
2686 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
2688 (IFKF_reg): New attributes to give the register constraints for
2690 (extend<mode>tf2_internal): New insns to mark an explicit
2691 conversion between 128-bit floating point types that have a
2692 different mode but share the same representation.
2694 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
2696 PR tree-optimization/85814
2697 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
2698 a null return from get_strinfo when unsharing the next
2699 strinfo in the chain.
2701 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2704 * varasm.c (weak_finish): Clean up weak_decls.
2706 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2708 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
2709 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
2710 UNSPEC_UADALP values.
2711 * config/aarch64/iterators.md (ABAL): New int iterator.
2714 (sur): Add mappings for the above.
2715 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
2717 (aarch64_<sur>abal<mode>_4): Likewise.
2718 (aarch64_<sur>adalp<mode>_3): Likewise.
2719 (<sur>sadv16qi): New define_expand.
2721 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
2723 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
2724 (*movdf_internal): Ditto.
2725 (*rcpsf2_sse): Ditto.
2726 (*rsqrtsf2_sse): Ditto.
2727 (*sqrt<mode>2_sse): Ditto.
2729 2018-05-21 Tamar Christina <tamar.christina@arm.com>
2731 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
2733 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
2734 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
2735 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
2736 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
2738 * config/aarch64/arm_neon.h: Likewise.
2739 * config/aarch64/iterators.md (VQ_I): New.
2741 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
2743 * config.gcc: Add arc/t-multilib-linux to tmake_file for
2745 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
2748 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
2750 * config/nds32/constraints.md (S): New constraint.
2751 * config/nds32/nds32.md (call_internal): Use constraint S.
2752 (call_value_internal): Likewise.
2753 (sibcall_internal): Likewise.
2754 (sibcall_value_internal): Likewise.
2756 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
2757 Chung-Ju Wu <jasonwucj@gmail.com>
2759 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
2762 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
2763 Chung-Ju Wu <jasonwucj@gmail.com>
2765 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
2766 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
2767 (nds32_rtx_costs_impl): Simplify.
2768 (nds32_address_cost_impl): Simplify.
2769 (nds32_init_rtx_costs): New function.
2770 (nds32_rtx_costs_speed_prefer): Likewise.
2771 (nds32_rtx_costs_size_prefer): Likewise.
2772 (nds32_address_cost_speed_prefer): Likewise.
2773 (nds32_address_cost_speed_fwprop): Likewise.
2774 (nds32_address_cost_size_prefer): Likewise.
2775 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
2776 * config/nds32/nds32.c (nds32_option_override): Use
2777 nds32_init_rtx_costs function.
2779 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
2781 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
2782 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
2783 (TARGET_PIPELINE_N8): Likewise.
2784 (TARGET_PIPELINE_N10): Likewise.
2785 (TARGET_PIPELINE_N13): Likewise.
2786 (TARGET_PIPELINE_GRAYWOLF): Likewise.
2788 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
2790 * config/nds32/nds32-fpu.md: Update copyright year.
2792 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2794 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
2796 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2798 * config/nds32/nds32.c
2799 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
2800 * config/nds32/nds32.opt (minline-asm-r15): New option.
2802 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2804 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
2806 * config/nds32/nds32.md (abssi2): New pattern.
2808 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
2810 * config/i386/i386.md (rex64namesuffix): New mode attribute.
2811 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
2812 Merge insn pattern from sse_cvtsi2ss<round_name> and
2813 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
2814 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
2815 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
2816 using SWI48 mode iterator.
2817 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
2818 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
2819 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
2820 pattern from sse_cvttss2si<round_saeonly_name>
2821 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
2822 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
2823 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
2824 using SWI48 mode iterator.
2825 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
2826 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
2827 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
2828 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
2829 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
2830 using SWI48 mode iterator.
2831 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
2832 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
2833 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
2834 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
2835 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
2836 SWI48 mode iterator.
2837 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
2838 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
2839 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
2840 pattern from sse_cvttsd2si<round_saeonly_name>
2841 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
2843 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2845 * config/nds32/nds32-md-auxiliary.c
2846 (nds32_valid_smw_lwm_base_p): Refine.
2847 (nds32_output_smw_single_word): Refine.
2848 (nds32_output_smw_double_word): New.
2849 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
2851 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2853 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
2854 (nds32_output_stack_pop): Refine.
2855 (nds32_expand_unaligned_load): Refine.
2856 (nds32_expand_unaligned_store): Refine.
2858 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2859 Chung-Ju Wu <jasonwucj@gmail.com>
2861 * config/nds32/constants.md: Add TP_REGNUM constant.
2862 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
2863 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
2865 * config/nds32/nds32-doubleword.md: Consider flag_pic.
2866 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
2867 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
2868 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
2869 and PIC code generation.
2870 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
2872 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
2874 * config/nds32/nds32.md: Support TLS and PIC.
2875 * config/nds32/nds32.c: Support TLS and PIC.
2876 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
2877 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
2880 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2882 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
2883 mode with E_ prefix.
2885 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2886 Chung-Ju Wu <jasonwucj@gmail.com>
2888 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
2889 * config/nds32/nds32-md-auxiliary.c
2890 (symbolic_reference_mentioned_p): New.
2891 (nds32_legitimize_ict_address): New.
2892 (nds32_expand_ict_move): New.
2893 (nds32_indirect_call_referenced_p): New.
2894 (nds32_symbol_binds_local_p): Delete.
2895 (nds32_long_call_p): Modify.
2896 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
2897 * config/nds32/nds32-protos.h
2898 (symbolic_reference_mentioned_p): Declare.
2899 (nds32_legitimize_ict_address): Declare.
2900 (nds32_expand_ict_move): Declare.
2901 (nds32_indirect_call_referenced_p): Declare.
2902 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
2903 (nds32_relax_group): Use nds32_ict_const_p as condition.
2904 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
2905 (nds32_asm_file_start): Output ict_model directive in asm code.
2906 (nds32_legitimate_address_p): Consider indirect call.
2907 (nds32_print_operand): Consider indirect call.
2908 (nds32_print_operand_address): Consider indirect call.
2909 (nds32_insert_attributes): Handle "indirect_call" attribute.
2910 (TARGET_LEGITIMATE_ADDRESS_P): Define.
2911 (TARGET_LEGITIMATE_CONSTANT_P): Define.
2912 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
2913 (TARGET_DELEGITIMIZE_ADDRESS): Define.
2914 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
2915 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
2916 (TARGET_ICT_MODEL_SMALL): Define.
2917 (TARGET_ICT_MODEL_LARGE): Define.
2918 * config/nds32/nds32.md (movsi): Consider ict model.
2919 (call, call_value): Consider ict model.
2920 (sibcall, sibcall_value): Consider ict model.
2921 * config/nds32/nds32.opt (mict-model): New option.
2922 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
2925 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
2926 Monk Chiang <sh.chiang04@gmail.com>
2927 Jim Wilson <jimw@sifive.com>
2929 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
2930 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
2931 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
2932 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
2933 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
2934 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
2935 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
2936 compute save_libcall_adjustment properly.
2937 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
2938 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
2939 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
2940 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
2941 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
2942 (ABI_SPEC): Handle mabi=ilp32e.
2943 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
2944 (RVE): Add RVE mask.
2945 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
2946 <-march>: Add rv32e as an example.
2948 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
2951 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
2952 (intra_create_variable_infos): Handle C++ constructors.
2954 2018-05-18 Martin Liska <mliska@suse.cz>
2956 * passes.def: Remove a redundant pass.
2958 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
2961 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
2963 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2965 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
2967 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
2969 (armv2, armv2a, armv3, armv3m): Delete architectures.
2970 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
2971 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
2972 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
2974 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
2975 (*mulsidi3adddi): Likewise.
2976 (mulsidi3): Likewise.
2977 (*mulsidi3_nov6): Likewise.
2978 (umulsidi3): Likewise.
2979 (umulsidi3_nov6): Likewise.
2980 (umaddsidi4): Likewise.
2981 (*umulsidi3adddi): Likewise.
2982 (smulsi3_highpart): Likewise.
2983 (*smulsi3_highpart_nov6): Likewise.
2984 (umulsi3_highpart): Likewise.
2985 (*umulsi3_highpart_nov6): Likewise.
2986 * config/arm/arm.h (arm_arch3m): Delete.
2987 * config/arm/arm.c (arm_arch3m): Delete.
2988 (arm_option_override_internal): Update armv3-related comment.
2989 (arm_configure_build_target): Delete use of isa_bit_mode32.
2990 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
2991 (arm_rtx_costs_internal): Delete check of arm_arch3m.
2992 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
2994 (mulusa3): Likewise.
2995 * config/arm/arm-protos.h (arm_arch3m): Delete.
2996 * config/arm/arm-tables.opt: Regenerate.
2997 * config/arm/arm-tune.md: Likewise.
2998 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
2999 deleted architectures.
3001 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3003 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3004 (armv5t, armv5te): New features.
3005 (ARMv5, ARMv5e): Delete fgroups.
3006 (ARMv5t, ARMv5te): Adjust for above changes.
3008 (armv5, armv5e): Delete arches.
3009 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3011 (*call_reg_arm): Likewise.
3012 (*call_value_reg_armv5): Likewise.
3013 (*call_value_reg_arm): Likewise.
3014 (*call_symbol): Likewise.
3015 (*call_value_symbol): Likewise.
3016 (*sibcall_insn): Likewise.
3017 (*sibcall_value_insn): Likewise.
3019 (prefetch): Likewise.
3020 (define_split and define_peephole2 dependent on arm_arch5):
3022 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3024 (TARGET_ARM_QBIT): Likewise.
3025 (TARGET_DSP_MULTIPLY): Likewise.
3026 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3027 (arm_arch5, arm_arch5e): Delete.
3028 (arm_arch5t, arm_arch5te): Declare.
3029 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3030 (arm_arch5t): Declare.
3031 (arm_option_reconfigure_globals): Update for the above.
3032 (arm_options_perform_arch_sanity_checks): Update comment, replace
3033 use of arm_arch5 with arm_arch5t.
3034 (use_return_insn): Likewise.
3035 (arm_emit_call_insn): Likewise.
3036 (output_return_instruction): Likewise.
3037 (arm_final_prescan_insn): Likewise.
3038 (arm_coproc_builtin_available): Likewise.
3039 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3040 arm_arch5e with arm_arch5t and arm_arch5te.
3041 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3042 (arm_arch5t, arm_arch5te): Declare.
3043 * config/arm/arm-tables.opt: Regenerate.
3044 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3045 * config/arm/t-multilib: Likewise.
3046 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3047 instead of arm_arch5.
3048 (*call_reg_thumb1): Likewise.
3049 (*call_value_reg_thumb1_v5): Likewise.
3050 (*call_value_reg_thumb1): Likewise.
3051 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3053 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3055 2018-05-18 Martin Liska <mliska@suse.cz>
3057 PR gcov-profile/84846
3058 * doc/gcov.texi: Document -t option of gcov tool.
3060 2018-05-18 Martin Liska <mliska@suse.cz>
3062 PR gcov-profile/84846
3063 * gcov.c (print_usage): Add new -t option.
3064 (process_args): Handle the option.
3065 (generate_results): Use stdout as output when requested by
3068 2018-05-18 Martin Liska <mliska@suse.cz>
3070 PR gcov-profile/84846
3071 * coverage.c (coverage_init): Write PWD to .gcno file.
3072 * doc/gcov.texi: Document how working directory is printed.
3073 * gcov-dump.c (dump_gcov_file): Print PWD.
3074 * gcov.c (output_intermediate_file): Likewise.
3075 (read_graph_file): Read PWD string.
3076 (output_lines): Print PWD.
3078 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3081 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3082 for retval and return false if all args to phi are zero.
3084 2018-05-18 Richard Biener <rguenther@suse.de>
3086 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3088 (evrp_dom_walker::before_dom_children): Call it.
3090 2018-05-18 Richard Biener <rguenther@suse.de>
3092 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3093 results when processing array refs with variable index.
3095 2018-05-18 Toon Moene <toon@moene.org>
3097 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3098 directly after that of -floop-interchange. Indicate that both
3099 options are enabled by default when specifying -O3.
3101 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3103 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3104 iterator. Delete separate integer-mode vec_set<mode> expander.
3105 (aarch64_simd_vec_setv2di): Delete.
3106 (vec_setv2di): Delete.
3107 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3108 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
3109 the "w, r" alternative.
3111 2018-05-18 Martin Liska <mliska@suse.cz>
3113 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3114 * tree-pass.h (make_pass_lower_switch_O0): New function.
3115 * tree-switch-conversion.c (node_has_low_bound): Remove.
3116 (node_has_high_bound): Likewise.
3117 (node_is_bounded): Likewise.
3118 (class pass_lower_switch): Make it a template type and create
3120 (pass_lower_switch::execute): Add template argument.
3121 (make_pass_lower_switch): New function.
3122 (make_pass_lower_switch_O0): New function.
3123 (do_jump_if_equal): Remove.
3124 (emit_case_nodes): Simplify to just handle all 3 cases and leave
3125 all the hard work to tree optimization passes.
3127 2018-05-18 Martin Liska <mliska@suse.cz>
3129 * dbgcnt.c (limit_low): Renamed from limit.
3130 (limit_high): New variable.
3131 (dbg_cnt_is_enabled): Check for upper limit.
3132 (dbg_cnt): Adjust dumping.
3133 (dbg_cnt_set_limit_by_index): Add new argument for high
3135 (dbg_cnt_set_limit_by_name): Likewise.
3136 (dbg_cnt_process_single_pair): Parse new format.
3137 (dbg_cnt_process_opt): Use strtok.
3138 (dbg_cnt_list_all_counters): Remove 'value' and add
3140 * doc/invoke.texi: Document changes.
3142 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
3144 * doc/sourcebuild.texi (scalar_all_fma): Document.
3145 * tree.def (FMA_EXPR): Delete.
3146 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3147 * internal-fn.c (ternary_direct): New macro.
3148 (expand_ternary_optab_fn): Likewise.
3149 (direct_ternary_optab_supported_p): Likewise.
3150 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3151 * builtins.c (fold_builtin_fma): Delete.
3152 (fold_builtin_3): Don't call it.
3153 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3154 * expr.c (expand_expr_real_2): Likewise.
3155 * fold-const.c (operand_equal_p): Likewise.
3156 (fold_ternary_loc): Likewise.
3157 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3158 * gimple.c (DEFTREECODE): Likewise.
3159 * gimplify.c (gimplify_expr): Likewise.
3160 * optabs-tree.c (optab_for_tree_code): Likewise.
3161 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
3162 * tree-eh.c (operation_could_trap_p): Likewise.
3163 (stmt_could_throw_1_p): Likewise.
3164 * tree-inline.c (estimate_operator_cost): Likewise.
3165 * tree-pretty-print.c (dump_generic_node): Likewise.
3166 (op_code_prio): Likewise.
3167 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3168 * tree-ssa-operands.c (get_expr_operands): Likewise.
3169 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
3170 * fold-const-call.h (fold_fma): Delete.
3171 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
3172 CFN_FNMA and CFN_FNMS.
3174 * genmatch.c (combined_fn): New enum.
3175 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
3176 (commutative_op): New function.
3177 (commutate): Use it. Handle more than 2 operands.
3178 (dt_operand::gen_gimple_expr): Use commutative_op.
3179 (parser::parse_expr): Allow :c to be used with non-binary
3180 operators if the commutative operand is known.
3181 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
3182 CFN_FMS, CFN_FNMA and CFN_FNMS.
3183 (backprop::process_assign_use): Remove FMA_EXPR handling.
3184 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
3185 (gen_hsa_fma): New function.
3186 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
3187 IFN_FNMA and IFN_FNMS.
3188 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
3189 * gimple-fold.h (follow_all_ssa_edges): Declare.
3190 * gimple-fold.c (follow_all_ssa_edges): New function.
3191 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
3192 gimple_build interface and use follow_all_ssa_edges to fold the result.
3193 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
3194 instead of checking for optabs directly.
3195 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
3196 rather than FMA_EXPRs.
3197 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
3198 call to IFN_FMA instead of an FMA_EXPR.
3200 2018-05-17 Jim Wilson <jimw@sifive.com>
3202 * expr.c (do_tablejump): When converting index to Pmode, if we have a
3203 sign extended promoted subreg, and the range does not have the sign bit
3204 set, then do a sign extend.
3206 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
3207 test, check for sign extended subreg and/or constant operands, and
3208 do a sign extend in that case.
3210 2018-05-17 Steve Ellcey <sellcey@cavium.com>
3212 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
3213 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
3215 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
3216 Change logics_shift_reg to logics_shift_imm.
3217 (thunderx2t99_fp_loadpair_basic): Delete.
3218 (thunderx2t99_fp_storepair_basic): Delete.
3219 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
3220 (thunderx2t99_asimd_polynomial): Delete.
3221 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
3222 and neon_fp_mul_d_scalar_q.
3223 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
3224 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
3225 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
3226 (thunderx2t99_asimd_lut): Add missing tbl types.
3227 (thunderx2t99_asimd_ext): Delete.
3228 (thunderx2t99_asimd_load1_1_mult): Delete.
3229 (thunderx2t99_asimd_load1_2_mult): Delete.
3230 (thunderx2t99_asimd_load1_ldp): New.
3231 (thunderx2t99_asimd_load1): New.
3232 (thunderx2t99_asimd_load2): Add missing *load2* types.
3233 (thunderx2t99_asimd_load3): New.
3234 (thunderx2t99_asimd_load4): New.
3235 (thunderx2t99_asimd_store1_1_mult): Delete.
3236 (thunderx2t99_asimd_store1_2_mult): Delete.
3237 (thunderx2t99_asimd_store2_mult): Delete.
3238 (thunderx2t99_asimd_store2_onelane): Delete.
3239 (thunderx2t99_asimd_store_stp): New.
3240 (thunderx2t99_asimd_store1): New.
3241 (thunderx2t99_asimd_store2): New.
3242 (thunderx2t99_asimd_store3): New.
3243 (thunderx2t99_asimd_store4): New.
3245 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
3247 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
3248 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
3250 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
3251 Segher Boessenkool <segher@kernel.crashing.org>
3254 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
3257 2018-05-17 Richard Biener <rguenther@suse.de>
3259 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
3260 for pruning loop and prune defs feeding only already visited PHIs.
3262 2018-05-17 Richard Biener <rguenther@suse.de>
3264 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
3266 2018-05-17 Bin Cheng <bin.cheng@arm.com>
3267 Richard Biener <rguenther@suse.de>
3269 PR tree-optimization/85793
3270 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
3271 for VMAT_ELEMENTWISE.
3273 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3275 * internal-fn.h (lookup_internal_fn): Declare
3276 * internal-fn.c (lookup_internal_fn): New function.
3277 * gimple.c (gimple_build_call_from_tree): Handle calls to
3279 * gimple-pretty-print.c (dump_gimple_call): Print "." before
3280 internal function names.
3281 * tree-pretty-print.c (dump_generic_node): Likewise.
3282 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
3284 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3286 * gimple-fold.h (gimple_build): Make the function forms take
3287 combined_fn rather than built_in_function.
3288 (gimple_simplify): Likewise.
3289 * gimple-match-head.c (gimple_simplify): Likewise.
3290 * gimple-fold.c (gimple_build): Likewise.
3291 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
3292 rather than gimple_build_call_internal.
3293 (get_initial_defs_for_reduction): Likewise.
3294 (vect_create_epilog_for_reduction): Likewise.
3295 (vectorizable_live_operation): Likewise.
3297 2018-05-17 Martin Liska <mliska@suse.cz>
3299 * gimple-ssa-sprintf.c (format_directive): Do not use
3300 space in between 'G_' and '('.
3302 2018-05-17 Jakub Jelinek <jakub@redhat.com>
3305 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
3306 even if the mask is not all ones.
3309 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
3311 (ix86_gimple_fold_builtin): Likewise.
3314 * config/i386/i386.c: Include tree-vector-builder.h.
3315 (ix86_vector_shift_count): New function.
3316 (ix86_fold_builtin): Fold shift builtins by scalar count.
3317 (ix86_gimple_fold_builtin): Likewise.
3319 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
3320 _mm512_setzero): New intrinsics.
3322 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
3323 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3325 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
3326 code generation for cases where splatting a value is not useful.
3327 * simplify-rtx.c (simplify_ternary_operation): Simplify
3328 vec_merge across a vec_duplicate and a paradoxical subreg forming
3329 a vector mode to a vec_concat.
3331 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
3333 * config.gcc: Support "goldmont-plus".
3334 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3336 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3337 PROCESSOR_GOLDMONT_PLUS.
3338 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
3339 (processor_target_table): Add "goldmont-plus".
3340 (PTA_GOLDMONT_PLUS): Define.
3341 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
3342 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
3343 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
3344 (fold_builtin_cpu): Add "goldmont-plus".
3345 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
3346 (ix86_option_override_internal): Add "goldmont-plus".
3347 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
3348 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
3349 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
3350 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
3352 2018-05-17 Richard Biener <rguenther@suse.de>
3354 PR tree-optimization/85757
3355 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
3356 remove defs that only feed that PHI from further processing.
3358 2018-05-16 Jim Wilson <jimw@sifive.com>
3360 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
3362 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
3364 2018-05-16 Mark Wielaard <mark@klomp.org>
3366 * dwarf2out.c (count_index_strings): New function.
3367 (output_indirect_strings): Call count_index_strings and generate
3368 header for dwarf_version >= 5.
3370 2018-05-16 Mark Wielaard <mark@klomp.org>
3372 * dwarf2out.c (dwarf_FORM): New function.
3373 (set_indirect_string): Use dwarf_FORM.
3374 (reset_indirect_string): Likewise.
3375 (size_of_die): Likewise.
3376 (value_format): Likewise.
3377 (output_die): Likewise.
3378 (add_skeleton_AT_string): Likewise.
3379 (output_macinfo_op): Likewise.
3380 (index_string): Likewise.
3381 (output_index_string_offset): Likewise.
3382 (output_index_string): Likewise.
3383 (count_index_strings): Likewise.
3385 2018-05-16 Carl Love <cel@us.ibm.com>
3387 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3388 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
3390 2018-05-16 Martin Jambor <mjambor@suse.cz>
3392 * ipa-prop.c (ipa_free_all_edge_args): Remove.
3393 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
3395 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
3397 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
3398 (fnma<mode>4): Likewise.
3399 (fms<mode>4): Likewise.
3400 (fnms<mode>4): Likewise.
3401 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
3402 (aarch64_fnma<mode>4): Likewise.
3403 (aarch64_fms<mode>4): Likewise.
3404 (aarch64_fnms<mode>4): Likewise.
3405 (aarch64_fnmadd<mode>4): Likewise.
3407 2018-05-16 Jason Merrill <jason@redhat.com>
3409 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
3411 2018-05-16 Richard Biener <rguenther@suse.de>
3413 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
3414 (dump_stmt_cost): Declare.
3415 (add_stmt_cost): Dump cost we add.
3416 (add_stmt_costs): New function.
3417 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
3419 (vect_analyze_stmt): Adjust prototype.
3420 (vectorizable_condition): Likewise.
3421 (vectorizable_live_operation): Likewise.
3422 (vectorizable_reduction): Likewise.
3423 (vectorizable_induction): Likewise.
3424 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
3425 cost vector to pass to vectorizable_ and record afterwards.
3426 (vect_model_reduction_cost): Take cost vector argument and adjust.
3427 (vect_model_induction_cost): Likewise.
3428 (vectorizable_reduction): Likewise.
3429 (vectorizable_induction): Likewise.
3430 (vectorizable_live_operation): Likewise.
3431 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
3432 SLP_TREE_NUMBER_OF_VEC_STMTS.
3433 (vect_analyze_slp_cost_1): Remove.
3434 (vect_analyze_slp_cost): Likewise.
3435 (vect_slp_analyze_node_operations): Take visited args and
3436 a target cost vector. Avoid processing already visited stmt sets.
3437 (vect_slp_analyze_operations): Use a local cost vector to gather
3438 costs and register those of non-discarded instances.
3439 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
3440 (vect_schedule_slp_instance): Remove copying of
3441 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
3443 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
3444 adding cost. Record cost entry location.
3445 (vect_prologue_cost_for_slp_op): Function to compute cost of
3446 a constant or invariant generated for SLP vect in the prologue,
3447 split out from vect_analyze_slp_cost_1.
3448 (vect_model_simple_cost): Make static. Adjust for SLP costing.
3449 (vect_model_promotion_demotion_cost): Likewise.
3450 (vect_model_store_cost): Likewise, make static.
3451 (vect_model_load_cost): Likewise.
3452 (vectorizable_bswap): Add cost vector arg and adjust.
3453 (vectorizable_call): Likewise.
3454 (vectorizable_simd_clone_call): Likewise.
3455 (vectorizable_conversion): Likewise.
3456 (vectorizable_assignment): Likewise.
3457 (vectorizable_shift): Likewise.
3458 (vectorizable_operation): Likewise.
3459 (vectorizable_store): Likewise.
3460 (vectorizable_load): Likewise.
3461 (vectorizable_condition): Likewise.
3462 (vectorizable_comparison): Likewise.
3463 (can_vectorize_live_stmts): Likewise.
3464 (vect_analyze_stmt): Likewise.
3465 (vect_transform_stmt): Adjust calls to vectorizable_*.
3466 * tree-vectorizer.c: Include gimple-pretty-print.h.
3467 (dump_stmt_cost): New function.
3469 2018-05-16 Richard Biener <rguenther@suse.de>
3471 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
3472 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
3473 * tree-ssa-dse.c: Include tree-ssa-loop.h.
3474 (check_name): New callback.
3475 (dse_classify_store): Track cycles via a visited bitmap of PHI
3476 defs and simplify handling of in-loop and across loop dead stores
3477 and properly fail for loop-variant refs. Handle byte-tracking with
3478 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
3481 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
3483 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
3484 (vect_get_mask_type_for_stmt): Likewise.
3485 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
3487 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
3488 to determine the statement's vector type and the vector type that
3489 should be used for calculating nunits. Deal with cases in which
3490 the type has to be deferred.
3491 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
3492 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
3493 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
3494 (vect_determine_vf_for_stmt): New functions, split out from...
3495 (vect_determine_vectorization_factor): ...here.
3496 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
3497 (vect_get_mask_type_for_stmt): New functions, split out from
3498 vect_determine_vectorization_factor.
3500 2018-05-16 Richard Biener <rguenther@suse.de>
3502 * tree-cfg.c (verify_gimple_assign_ternary): Properly
3503 verify the [VEC_]COND_EXPR embedded comparison.
3505 2018-05-15 Martin Sebor <msebor@redhat.com>
3507 PR tree-optimization/85753
3508 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
3509 RECORD_TYPE in addition to ARRAY_TYPE.
3511 2018-05-15 Martin Sebor <msebor@redhat.com>
3514 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
3516 2018-05-15 Richard Biener <rguenther@suse.de>
3518 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
3519 add by_clobber_p one. Change algorithm to collect all defs
3520 representing uses we need to walk and try reducing them to
3521 a single one before failing.
3522 (dse_dom_walker::dse_optimize_stmt): Adjust.
3524 2018-05-13 Mark Wielaard <mark@klomp.org>
3526 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
3527 (size_of_loc_descr): Likewise.
3528 (output_loc_operands): Likewise.
3529 (output_loc_operands_raw): Likewise.
3530 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
3531 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
3532 (hash_loc_operands): Likewise.
3533 (compare_loc_operands): Likewise.
3535 2018-05-14 Mark Wielaard <mark@klomp.org>
3537 * dwarf2out.c (count_index_addrs): New function.
3538 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
3540 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3542 PR tree-optimization/83648
3543 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
3544 return value as malloc candidate.
3546 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3549 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
3550 param as true in call to suggest_attribute.
3552 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
3554 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
3555 -mreadonly-in-sdata.
3557 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3559 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
3561 (aarch64_crypto_aesd_fused): Likewise.
3563 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
3565 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
3566 (movsi_aarch64): Likewise.
3567 (load_pairsi): Likewise.
3568 (load_pairdi): Likewise.
3569 (store_pairsi): Likewise.
3570 (store_pairdi): Likewise.
3571 (load_pairsf): Likewise.
3572 (load_pairdf): Likewise.
3573 (store_pairsf): Likewise.
3574 (store_pairdf): Likewise.
3575 (zero_extend): Likewise.
3576 (trunc): Swap alternatives.
3577 (fcvt_target): Add '?' to prefer w over r.
3579 2018-05-14 Jakub Jelinek <jakub@redhat.com>
3582 * config/i386/i386.md: Disallow non-commutative arithmetics in
3583 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
3584 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
3585 in the peephole2 before it.
3587 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
3589 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
3590 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
3591 (ix86_handle_option): Handle -mcldemote.
3592 * config.gcc: New header.
3593 * config/i386/cldemoteintrin.h: New file.
3594 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
3595 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3597 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3598 OPTION_MASK_ISA_CLDEMOTE.
3599 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
3600 (ix86_valid_target_attribute_inner_p): Ditto.
3601 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
3602 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
3603 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
3604 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
3605 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
3607 * config/i386/i386.opt: Add -mcldemote.
3608 * config/i386/x86intrin.h: New header.
3609 * doc/invoke.texi: Add -mcldemote.
3611 2018-05-14 Richard Biener <rguenther@suse.de>
3613 * doc/match-and-simplify.texi: Adjust :s documentation.
3615 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
3617 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
3618 intended memcpy size.
3619 (REORDER_45): Likewise.
3621 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
3623 * sort.cc: New file.
3624 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
3625 * vec.c (qsort_chk): Use gcc_qsort.
3626 * Makefile.in (OBJS-libcommon): Add sort.o.
3627 (build/sort.o): New target. Use it...
3628 (BUILD_RTL): ... here, and...
3629 (build/gencfn-macros): ... here, and...
3630 (build/genmatch): ... here.
3632 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3633 Chung-Ju Wu <jasonwucj@gmail.com>
3635 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
3636 * config/nds32/nds32-graywolf.md: New file.
3637 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
3638 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
3640 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
3641 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
3642 * config/nds32/nds32.md (pipeline_model): Add graywolf.
3643 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
3644 * config/nds32/pipelines.md: Include n15 settings.
3646 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3647 Chung-Ju Wu <jasonwucj@gmail.com>
3649 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
3650 * config/nds32/nds32-n13.md: New file.
3651 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
3652 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
3654 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
3655 * config/nds32/nds32.md (pipeline_model): Add n13.
3656 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
3657 * config/nds32/pipelines.md: Include n13 settings.
3659 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3660 Chung-Ju Wu <jasonwucj@gmail.com>
3662 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
3663 * config/nds32/nds32-n10.md: New file.
3664 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
3665 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
3667 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
3668 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
3669 * config/nds32/nds32.md (pipeline_model): Add n10.
3670 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
3671 * config/nds32/pipelines.md: Include n10 settings.
3673 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
3674 Kito Cheng <kito.cheng@gmail.com>
3675 Chung-Ju Wu <jasonwucj@gmail.com>
3677 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
3678 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3679 Add enum values for DSP extension instructions.
3680 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
3682 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
3683 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
3685 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
3686 * config/nds32/nds32-dspext.md: New file for DSP implementation.
3687 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
3688 * config/nds32/nds32-intrinsic.md: Likewise.
3689 * config/nds32/nds32_intrinsic.h: Likewise.
3690 * config/nds32/nds32-md-auxiliary.c: Likewise.
3691 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
3692 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
3693 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
3694 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
3695 * config/nds32/nds32-protos.h: New declarations for DSP extension.
3696 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
3697 TYPE_DMAC in switch statement.
3698 * config/nds32/nds32.c: New checking and implementation for DSP
3699 extension instructions.
3700 * config/nds32/nds32.h: Likewise.
3701 * config/nds32/nds32.md: Likewise.
3702 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
3703 * config/nds32/predicates.md: Implement new predicates for DSP
3706 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
3708 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
3709 Reformat alternatives and attributes so it is easier to identify
3710 which constraints/attributes go with which instruction.
3711 (mov<mode>_hardfloat32, FMOVE64): Likewise.
3712 (mov<mode>_softfloat32, FMOVE64): Likewise.
3713 (mov<mode>_hardfloat64, FMOVE64): Likewise.
3714 (mov<mode>_softfloat64, FMOVE64): Likewise.
3716 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
3718 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3720 (Basic PowerPC Built-in Functions): The new name of the
3721 subsection previously known as "PowerPC Built-in Functions".
3722 (Basic PowerPC Built-in Functions Available on all Configurations):
3724 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
3725 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
3726 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
3727 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
3729 2018-05-11 Martin Jambor <mjambor@suse.cz>
3732 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
3735 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
3738 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
3740 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
3742 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
3743 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
3744 (ix86_handle_option): Handle -mwaitpkg.
3745 * config.gcc: New header.
3746 * config/i386/cpuid.h (bit_WAITPKG): New bit.
3747 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
3748 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
3750 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3751 OPTION_MASK_ISA_WAITPKG.
3752 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
3753 (ix86_option_override_internal): Add PTA_WAITPKG.
3754 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
3755 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
3756 IX86_BUILTIN_TPAUSE.
3757 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
3758 __builtin_ia32_umwait and __builtin_ia32_tpause.
3759 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
3760 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
3761 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
3762 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
3763 UNSPECV_TPAUSE): New.
3764 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
3765 * config/i386/i386.opt: Add -mwaitpkg.
3766 * config/i386/waitpkgintrin.h: New file.
3767 * config/i386/x86intrin.h: New header.
3768 * doc/invoke.texi: Add -mwaitpkg.
3770 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
3773 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
3775 (cortex-m0): Use armv6s-m isa.
3776 (cortex-m0plus): Likewise.
3777 (cortex-m1): Likewise.
3778 (cortex-m0.small-multiply): Likewise.
3779 (cortex-m0plus.small-multiply): Likewise.
3780 (cortex-m1.small-multiply): Likewise.
3782 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
3783 Jakub Jelinek <jakub@redhat.com>
3785 PR tree-optimization/85692
3786 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
3787 source permute as well.
3789 2018-05-11 Martin Liska <mliska@suse.cz>
3792 * doc/extend.texi: Document LLVM style format for no_sanitize
3795 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
3797 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
3798 mode_supports_vsx_dform_quad to mode_supports_dq_form.
3799 (mode_supports_vsx_dform_quad): Likewise.
3800 (mode_supports_vmx_dform): Move these functions to be next to the
3801 other mode_supports functions.
3802 (mode_supports_dq_form): Likewise.
3803 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
3804 mode_supports_dq_form.
3805 (reg_offset_addressing_ok_p): Likewise.
3806 (offsettable_ok_by_alignment): Likewise.
3807 (rs6000_legitimate_offset_address_p): Likewise.
3808 (legitimate_lo_sum_address_p): Likewise.
3809 (rs6000_legitimize_address): Likewise.
3810 (rs6000_legitimize_reload_address): Likewise.
3811 (rs6000_secondary_reload_inner): Likewise.
3812 (rs6000_preferred_reload_class): Likewise.
3813 (rs6000_output_move_128bit): Likewise.
3815 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
3817 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
3818 Generate SImode target register for null target.
3819 <case IX86_BUILTIN_XGETBV>: Ditto.
3820 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
3821 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
3823 2018-05-10 Carl Love <cel@us.ibm.com>
3825 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3826 dcbtt and dcbtstt if operands[2] is 0.
3828 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
3831 * config/i386/sse.md (usadv64qi): New expander.
3833 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
3835 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
3836 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
3837 -maltivec=be support.
3838 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
3839 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
3840 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
3841 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
3842 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
3843 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
3844 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
3845 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
3846 altivec_vsumsws): Adjust.
3847 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
3848 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
3849 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
3851 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
3852 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
3853 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
3854 (altivec_lve<VI_char>x): Delete expand.
3855 (*altivec_lve<VI_char>x_internal): Rename to...
3856 (altivec_lve<VI_char>x): ... this.
3857 (altivec_lvxl_<mode>): Delete expand.
3858 (*altivec_lvxl_<mode>_internal): Rename to ...
3859 (altivec_lvxl_<mode>): ... this.
3860 (altivec_stvxl_<mode>): Delete expand.
3861 (*altivec_stvxl_<mode>_internal): Rename to ...
3862 (altivec_stvxl_<mode>): ... this.
3863 (altivec_stve<VI_char>x): Delete expand.
3864 (*altivec_stve<VI_char>x_internal): Rename to ...
3865 (altivec_stve<VI_char>x): ... this.
3866 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
3867 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
3868 reduc_plus_scal_<mode>): Adjust.
3869 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
3871 (rs6000_cpu_cpp_builtins): Adjust.
3872 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
3873 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
3874 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
3875 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
3876 -maltivec=be support.
3877 (rs6000_split_vec_extract_var): Adjust.
3878 (rs6000_split_v4si_init): Adjust.
3879 (swap_selector_for_mode): Delete.
3880 (altivec_expand_lvx_be, altivec_expand_stvx_be,
3881 altivec_expand_stvex_be): Delete.
3882 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
3883 -maltivec=be support.
3884 (rs6000_gimple_fold_builtin): Ditto.
3885 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
3887 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
3888 (TARGET_DIRECT_MOVE_64BIT): Adjust.
3889 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
3890 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
3891 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
3892 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
3893 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
3894 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
3895 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
3896 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
3897 anonymous split): Adjust.
3898 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
3899 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
3901 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
3903 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
3904 when --with-gxx-include-dir is also specified.
3905 * configure: Regenerate.
3907 2018-05-09 Jim Wilson <jimw@sifive.com>
3910 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
3911 * config/riscv/t-withmultilib: New.
3912 * config/riscv/withmultilib.h: New.
3913 * doc/install.texi: Document RISC-V --with-multilib-list support.
3915 2018-05-09 Richard Biener <rguenther@suse.de>
3917 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
3919 (vect_bb_vectorization_profitable_p): Adjust. Compute
3920 actual scalar cost using the cost vector and the add_stmt_cost
3923 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3925 PR rtl-optimization/85645
3926 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
3927 in the REG_CFA_REGISTER note for LR, don't leave it empty.
3929 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3931 PR rtl-optimization/85645
3932 * shrink-wrap.c (spread_components): Return a boolean saying if
3933 anything was changed.
3934 (try_shrink_wrapping_separate): Iterate spread_components until
3935 nothing changes anymore.
3937 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3939 PR rtl-optimization/85645
3940 * regrename.c (build_def_use): Also kill the chains that include the
3941 destination of a REG_CFA_REGISTER note.
3943 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3945 PR rtl-optimization/85645
3946 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
3947 insn that has a REG_CFA_REGISTER note.
3949 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
3951 * cfgexpand.c (expand_clobber): New function.
3952 (expand_gimple_stmt_1): Use it.
3953 * tree-vect-stmts.c (vect_clobber_variable): New function,
3955 (vectorizable_simd_clone_call): ...here.
3956 (vectorizable_store): Emit a clobber either side of an
3957 IFN_STORE_LANES sequence.
3958 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
3960 2018-05-09 Tom de Vries <tom@codesourcery.com>
3963 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
3964 (define_insn "trap_if_false"): Add exit after trap.
3966 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
3968 PR rtl-optimization/85638
3969 * bb-reorder.c: Include common/common-target.h.
3970 (create_forwarder_block): New function extracted from...
3971 (fix_up_crossing_landing_pad): ...here. Rename into...
3972 (dw2_fix_up_crossing_landing_pad): ...this.
3973 (sjlj_fix_up_crossing_landing_pad): New function.
3974 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
3975 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
3976 from both partitions and exit the loop after one iteration.
3978 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
3981 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3983 (Basic PowerPC Built-in Functions): The new name of the
3984 subsection previously known as "PowerPC Built-in Functions".
3985 (Basic PowerPC Built-in Functions Available on all Configurations):
3987 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3989 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3991 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3993 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3996 2018-05-08 Jim Wilson <jimw@sifive.com>
3998 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
3999 (LD_EMUL_SUFFIX): New.
4000 (LINK_SPEC): Use it.
4002 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4004 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4006 (Basic PowerPC Built-in Functions): The new name of the
4007 subsection previously known as "PowerPC Built-in Functions".
4008 (Basic PowerPC Built-in Functions Available on all Configurations):
4010 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4012 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4014 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4016 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4019 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4022 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4023 after cmpelim optimization.
4025 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
4027 * config.gcc: Support "goldmont".
4028 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4029 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4031 * config/i386/i386.c (m_GOLDMONT): Define.
4032 (processor_target_table): Add "goldmont".
4033 (PTA_GOLDMONT): Define.
4034 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4035 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4036 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4037 (fold_builtin_cpu): Add "goldmont".
4038 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4039 (ix86_option_override_internal): Add "goldmont".
4040 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4041 (processor_type): Add PROCESSOR_GOLDMONT.
4042 * config/i386/i386.md: Add CPU "glm".
4043 * config/i386/glm.md: New file.
4044 * config/i386/x86-tune.def: Add m_GOLDMONT.
4045 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4047 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4050 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4052 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4053 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
4054 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
4057 * config/i386/i386.c (ix86_fold_builtin): Handle
4058 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4061 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4062 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4064 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4067 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4068 (check_arch): Likewise.
4069 (check_fpu): Return the result rather than printing it.
4070 (end arch): Fix operator precedence.
4071 (end cpu): Likewise.
4072 (END): Print the result from check_fpu.
4074 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4075 Alan Hayward <alan.hayward@arm.com>
4076 David Sherwood <david.sherwood@arm.com>
4078 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4079 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4080 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4081 (*fcmuo<mode>_and): New patterns.
4083 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4085 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4086 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4087 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4088 (cmp_op, sve_imm_con): New code attributes.
4089 (SVE_COND_INT_CMP, imm_con): Delete.
4090 (cmp_op): Remove above unspecs from int attribute.
4091 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4093 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
4094 comparison-specific unspecs.
4095 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4096 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4097 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4098 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4099 (*vec_fcm<cmp_op><mode>): Rename to...
4100 (*fcm<cmp_op><mode>): ...this and adjust likewise.
4101 (*vec_fcmuo<mode>): Rename to...
4102 (*fcmuo<mode>): ...this and adjust likewise.
4103 (*pred_fcm<cmp_op><mode>): New pattern.
4104 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4105 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4107 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4109 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4110 (aarch64_emit_sve_predicated_cond): New function.
4111 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4112 (aarch64_emit_unspec_cond_or): Replace with...
4113 (aarch64_emit_sve_or_conds): ...this new function. Use
4114 aarch64_emit_sve_ptrue_op for the individual comparisons and
4115 aarch64_emit_binop to OR them together.
4116 (aarch64_emit_inverted_unspec_cond): Replace with...
4117 (aarch64_emit_sve_inverted_cond): ...this new function. Use
4118 aarch64_emit_sve_ptrue_op for the comparison and
4119 aarch64_emit_unop to invert the result.
4120 (aarch64_expand_sve_vec_cmp_float): Update after the above
4121 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
4123 2018-05-07 Nathan Sidwell <nathan@acm.org>
4125 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4126 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4127 (Backwards Compatibility): Likewise.
4129 2018-05-07 Luis Machado <luis.machado@linaro.org>
4133 2018-05-07 Luis Machado <luis.machado@linaro.org>
4135 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4136 <prefetch_dynamic_strides>: New const bool field.
4137 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4138 prefetch_dynamic_strides.
4139 (exynosm1_prefetch_tune): Likewise.
4140 (thunderxt88_prefetch_tune): Likewise.
4141 (thunderx_prefetch_tune): Likewise.
4142 (thunderx2t99_prefetch_tune): Likewise.
4143 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4145 (aarch64_override_options_internal): Update to set
4146 PARAM_PREFETCH_DYNAMIC_STRIDES.
4147 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4148 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4149 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4150 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4151 prefetch-dynamic-strides setting.
4153 2018-05-07 Luis Machado <luis.machado@linaro.org>
4155 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4156 <minimum_stride>: New const int field.
4157 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4158 minimum_stride field.
4159 (exynosm1_prefetch_tune): Likewise.
4160 (thunderxt88_prefetch_tune): Likewise.
4161 (thunderx_prefetch_tune): Likewise.
4162 (thunderx2t99_prefetch_tune): Likewise.
4163 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4164 (aarch64_override_options_internal): Update to set
4165 PARAM_PREFETCH_MINIMUM_STRIDE.
4166 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4167 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4168 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4169 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4170 stride is constant and is below the minimum stride threshold.
4172 2018-05-07 Luis Machado <luis.machado@linaro.org>
4174 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
4177 2018-05-07 Luis Machado <luis.machado@linaro.org>
4179 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4180 <prefetch_dynamic_strides>: New const bool field.
4181 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4182 prefetch_dynamic_strides.
4183 (exynosm1_prefetch_tune): Likewise.
4184 (thunderxt88_prefetch_tune): Likewise.
4185 (thunderx_prefetch_tune): Likewise.
4186 (thunderx2t99_prefetch_tune): Likewise.
4187 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4189 (aarch64_override_options_internal): Update to set
4190 PARAM_PREFETCH_DYNAMIC_STRIDES.
4191 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4192 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4193 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4194 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4195 prefetch-dynamic-strides setting.
4197 2018-05-07 Luis Machado <luis.machado@linaro.org>
4199 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4200 <minimum_stride>: New const int field.
4201 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4202 minimum_stride field.
4203 (exynosm1_prefetch_tune): Likewise.
4204 (thunderxt88_prefetch_tune): Likewise.
4205 (thunderx_prefetch_tune): Likewise.
4206 (thunderx2t99_prefetch_tune): Likewise.
4207 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4208 (aarch64_override_options_internal): Update to set
4209 PARAM_PREFETCH_MINIMUM_STRIDE.
4210 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4211 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4212 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4213 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4214 stride is constant and is below the minimum stride threshold.
4216 2018-05-06 Jakub Jelinek <jakub@redhat.com>
4219 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4220 the type is addressable. Don't force op into register if it has
4223 2018-05-05 Roland McGrath <mcgrathr@google.com>
4226 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
4227 any section for which we don't know a specific type it should have,
4228 regardless of name. Previously this was done only for the exact
4229 names ".init_array", ".fini_array", and ".preinit_array".
4230 (default_elf_asm_named_section): Add comment about
4231 relationship with default_section_type_flags and SECTION_NOTYPE.
4232 (get_section): Don't consider it a type conflict if one side has
4233 SECTION_NOTYPE and the other doesn't, as long as neither has the
4234 SECTION_BSS et al used in the default_section_type_flags logic.
4236 2018-05-05 Tom de Vries <tom@codesourcery.com>
4239 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
4240 (workaround_barsyncs): New function.
4241 (nvptx_reorg): Use workaround_barsyncs.
4242 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4243 (define_expand "nvptx_membar_cta"): New define_expand.
4244 (define_insn "*nvptx_membar_cta"): New insn.
4246 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
4248 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
4249 To improve optimization opportunities.
4250 * builtin-types.def: The new needed builtin types for the above.
4252 2018-05-04 Richard Biener <rguenther@suse.de>
4254 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4255 * gimple-ssa-store-merging.c
4256 (imm_store_chain_info::output_merged_store): Remove redundant create,
4257 release split_store vector contents on failure.
4258 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4259 scalar stmt vector on cache hit.
4261 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
4263 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
4265 * config.gcc (powerpc-xilinx-eabi*): Remove.
4266 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
4268 (fusion_addis_mem_combo_load): Ditto.
4269 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
4271 (rs6000_cpu_cpp_builtins): Ditto.
4272 * config/rs6000/rs6000-linux.c
4273 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
4274 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
4275 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
4277 (rs6000_setup_reg_addr_masks): Ditto.
4278 (rs6000_init_hard_regno_mode_ok): Ditto.
4279 (rs6000_option_override_internal): Ditto.
4280 (legitimate_lo_sum_address_p): Ditto.
4281 (rs6000_legitimize_address): Ditto.
4282 (rs6000_legitimize_reload_address): Ditto.
4283 (rs6000_legitimate_address_p): Ditto.
4284 (abi_v4_pass_in_fpr): Ditto.
4285 (setup_incoming_varargs): Ditto.
4286 (rs6000_gimplify_va_arg): Ditto.
4287 (rs6000_split_multireg_move): Ditto.
4288 (rs6000_savres_strategy): Ditto.
4289 (rs6000_emit_prologue_components): Ditto.
4290 (rs6000_emit_epilogue_components): Ditto.
4291 (rs6000_emit_prologue): Ditto.
4292 (rs6000_emit_epilogue): Ditto.
4293 (rs6000_elf_file_end): Ditto.
4294 (rs6000_function_value): Ditto.
4295 (rs6000_libcall_value): Ditto.
4296 * config/rs6000/rs6000.h: Ditto.
4297 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
4298 (TARGET_MINMAX): ... this. New.
4299 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
4300 * config/rs6000/rs6000.md: Remove Xilinx FP support.
4301 (*movsi_internal1_single): Delete.
4302 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
4303 mfpu=, mxilinx-fpu): Delete.
4304 * config/rs6000/singlefp.h: Delete.
4305 * config/rs6000/sysv4.h: Remove Xilinx FP support.
4306 * config/rs6000/t-rs6000: Ditto.
4307 * config/rs6000/t-xilinx: Delete.
4308 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
4309 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
4310 (VStype_simple): Delete.
4311 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
4312 * config/rs6000/xfpu.h: Delete.
4313 * config/rs6000/xfpu.md: Delete.
4314 * config/rs6000/xilinx.h: Delete.
4315 * config/rs6000/xilinx.opt: Delete.
4316 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
4317 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
4319 2018-05-04 Tom de Vries <tom@codesourcery.com>
4322 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
4325 2018-05-04 Richard Biener <rguenther@suse.de>
4328 * tree-complex.c (update_complex_assignment): We are always in SSA form.
4329 (expand_complex_div_wide): Likewise.
4330 (expand_complex_operations_1): Likewise.
4331 (expand_complex_libcall): Preserve EH info of the original stmt.
4332 (tree_lower_complex): Handle removed blocks.
4333 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
4334 on complex multiplication and division libcall builtins.
4336 2018-05-04 Richard Biener <rguenther@suse.de>
4339 * fold-const.c (negate_expr_p): Restrict negation of operand
4340 zero of a division to when we know that can happen without
4342 (fold_negate_expr_1): Likewise.
4344 2018-05-04 Jakub Jelinek <jakub@redhat.com>
4347 * real.h (real_nextafter): Declare.
4348 * real.c (real_nextafter): New function.
4349 * fold-const-call.c (fold_const_nextafter): New function.
4350 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
4351 CASE_CFN_NEXTTOWARD.
4352 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
4353 even when arg1_mode is different from arg0_mode.
4355 2018-05-03 Nathan Sidwell <nathan@acm.org>
4357 * doc/extend.texi (Deprecated Features): Remove
4359 (Backwards Compatibility): Likewise.
4360 * doc/invoke.texi (C++ Language Options): Likewise.
4361 (C++ Dialect Options): Likewise.
4363 2018-05-03 Jakub Jelinek <jakub@redhat.com>
4366 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
4367 _mm512_mask_mullox_epi64): New intrinsics.
4369 2018-05-03 Tom de Vries <tom@codesourcery.com>
4372 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4373 dump files): Add offload-tree.
4375 2018-05-03 Richard Biener <rguenther@suse.de>
4377 PR tree-optimization/85615
4378 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
4379 to loops not nested in BBs loop father to avoid creating multi-entry
4382 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4384 PR tree-optimization/70291
4385 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
4386 arguments. Change return type to tree. Emit libcall as a new
4387 statement rather than replacing existing one when inplace_p is true.
4388 (expand_complex_multiplication_components): New function.
4389 (expand_complex_multiplication): Expand floating-point complex
4390 multiplication using the above.
4391 (expand_complex_division): Rename inner_type parameter to type.
4392 Update expand_complex_libcall call-site.
4393 (expand_complex_operations_1): Update expand_complex_multiplication
4394 and expand_complex_division call-sites.
4396 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4399 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4400 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4401 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
4402 the highest significant bit of the shift count mask is clear. In
4403 check whether and[sq]i3 is needed verify that all significant bits
4404 of the shift count other than the highest are set.
4406 2018-05-02 Tom de Vries <tom@codesourcery.com>
4409 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
4410 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
4411 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
4412 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
4413 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
4414 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
4415 __builtin_goacc_parlevel_size.
4417 2018-05-02 Richard Biener <rguenther@suse.de>
4419 PR tree-optimization/85597
4420 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
4421 do not use split vect_get_vec_defs call but call vect_get_slp_defs
4424 2018-05-02 Tom de Vries <tom@codesourcery.com>
4427 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4428 dump files): Add ltrans-tree.
4430 2018-05-02 Tom de Vries <tom@codesourcery.com>
4433 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4434 dump files): Add wpa-ipa.
4436 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
4438 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
4439 powerpc*-*-linux*paired* target.
4440 * config/rs6000/750cl.h: Delete.
4441 * config/rs6000/paired.h: Delete.
4442 * config/rs6000/paired.md: Delete.
4443 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
4445 * config/rs6000/rs6000-builtin.def: Remove paired float support.
4446 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
4447 comment. Remove paired float support.
4448 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
4449 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4451 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
4452 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
4454 * config/rs6000/rs6000.c: Remove paired float support.
4455 (paired_expand_vector_init, paired_expand_vector_move,
4456 paired_emit_vector_compare, paired_emit_vector_cond_expr,
4457 (paired_expand_lv_builtin, paired_expand_stv_builtin,
4458 paired_expand_builtin, paired_expand_predicate_builtin,
4459 paired_init_builtins): Delete.
4460 * config/rs6000/rs6000.h: Remove paired float support.
4461 * config/rs6000/rs6000.md: Remove paired float support.
4462 (move_from_CR_ov_bit): Delete.
4463 * config/rs6000/rs6000.opt (mpaired): Delete.
4464 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
4465 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
4467 2018-05-02 Richard Biener <rguenther@suse.de>
4470 * gimplify.c (gimplify_save_expr): When in SSA form allow
4471 SAVE_EXPRs to compute to SSA vars.
4473 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4476 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4477 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4478 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
4479 clobber operands[2], instead use a new pseudo. Formatting fixes.
4481 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
4483 PR tree-optimization/85586
4484 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
4485 exit early for statements in the same group if the accesses are
4488 2018-05-02 Tom de Vries <tom@codesourcery.com>
4491 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
4494 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
4496 PR tree-optimization/85143
4497 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
4499 2018-05-01 Tom de Vries <tom@codesourcery.com>
4502 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
4503 not found" error message.
4505 2018-05-01 Tom de Vries <tom@codesourcery.com>
4508 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
4509 * vec.c (test_ordered_remove_if): New function.
4510 (vec_c_tests): Call test_ordered_remove_if.
4511 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
4512 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
4513 * tree-vect-patterns.c (vect_pattern_recog_1): Use
4514 VEC_ORDERED_REMOVE_IF.
4516 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4518 PR tree-optimization/82665
4519 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
4520 pointer subtraction where arguments come from a memchr call.
4522 2018-05-01 Jakub Jelinek <jakub@redhat.com>
4524 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
4525 --push-state --as-needed and --pop-state instead of --as-needed and
4526 --no-as-needed if ld supports it.
4527 * configure: Regenerated.
4530 * doc/install.texi2html: Replace _002d with - and _002a with * in
4531 generated html files using sed.
4533 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4536 * gcc-rich-location.c (blank_line_before_p): New function.
4537 (use_new_line): New function.
4538 (gcc_rich_location::add_fixit_insert_formatted): New function.
4539 * gcc-rich-location.h
4540 (gcc_rich_location::add_fixit_insert_formatted): New function.
4542 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4544 * selftest.c (assert_streq): Rename "expected" and "actual" to
4545 "val1" and "val2". Extend NULL-handling to cover both inputs
4546 symmetrically, while still requiring both to be non-NULL for a pass.
4547 * selftest.h (assert_streq): Rename "expected" and "actual" to
4549 (ASSERT_EQ): Likewise.
4550 (ASSERT_EQ_AT): Likewise.
4551 (ASSERT_KNOWN_EQ): Likewise.
4552 (ASSERT_KNOWN_EQ_AT): Likewise.
4553 (ASSERT_NE): Likewise.
4554 (ASSERT_MAYBE_NE): Likewise.
4555 (ASSERT_MAYBE_NE_AT): Likewise.
4556 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
4557 the assertion to pass.
4558 (ASSERT_STREQ_AT): Likewise.
4560 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4562 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
4563 interaction with -pie.
4565 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4567 * selftest.h: Fix alphabetization of per-source-file selftest
4570 2018-04-30 Jason Merrill <jason@redhat.com>
4572 PR c++/61982 - dead stores to destroyed objects.
4573 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
4576 2018-04-30 Jason Merrill <jason@redhat.com>
4578 * tree.c (build_clobber): New.
4579 * tree.h: Declare it.
4580 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
4582 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4584 * diagnostic-show-locus.c (layout::layout): Update for
4585 location_get_source_line returning a char_span.
4586 (struct char_span): Move to input.h.
4587 (struct correction): Update for fields in char_span becoming
4589 (struct source_line): Update for location_get_source_line
4590 returning a char_span.
4591 (layout::print_line): Likewise.
4592 * edit-context.c (edited_file::print_content): Likewise.
4593 (edited_file::print_diff_hunk): Likewise.
4594 (edited_file::print_run_of_changed_lines): Likewise.
4595 (edited_file::get_num_lines): Likewise.
4596 (edited_line::edited_line): Likewise.
4597 * final.c (asm_show_source): Likewise.
4598 * input.c (location_get_source_line): Convert return type
4599 from const char * to char_span, losing the final "line_len"
4601 (dump_location_info): Update for the above.
4602 (get_substring_ranges_for_loc): Likewise. Use a char_span
4603 when handling the literal within the line.
4604 (test_reading_source_line): Update for location_get_source_line
4605 returning a char_span.
4606 * input.h (class char_span): Move here from
4607 diagnostic-show-locus.c, converting from a struct to a class.
4608 Make data members private.
4609 (char_span::operator bool): New.
4610 (char_span::length): New.
4611 (char_span::get_buffer): New.
4612 (char_span::operator[]): New.
4613 (char_span::subspan): Make const.
4614 (char_span::xstrdup): New.
4615 (location_get_source_line): Convert return type from const char *
4616 to char_span, losing the final "line_size" param.
4618 2018-04-30 Jan Hubicka <jh@suse.cz>
4620 * lto-wrapper.c (ltrans_priorities): New static var.
4621 (cmp_priority): New.
4622 (run_gcc): Read priorities and if doing parallel build order
4623 the Makefile by them.
4625 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4627 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
4629 2018-04-30 Richard Biener <rguenther@suse.de>
4631 * tree-cfg.c (verify_address): Remove base argument, add
4632 flag whether to check TREE_ADDRESSABLE and do that.
4633 (verify_expr): Remove.
4634 (verify_types_in_gimple_reference): Add pieces from verify_expr.
4635 (verify_gimple_assign_single): Likewise.
4636 (verify_gimple_switch): Likewise.
4637 (verify_expr_location_1): Dereference tp once. Add (disabled)
4638 piece from verify_expr.
4639 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
4641 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4643 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
4645 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4647 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
4648 (small_data_pattern): Likewise.
4649 (arc_rewrite_small_data): Likewise.
4650 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
4651 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
4652 (get_symbol_alignment): New function.
4653 (legitimate_small_data_address_p): Likewise.
4654 (legitimate_scaled_address): Update, call
4655 legitimate_small_data_address_p.
4656 (output_sdata): New static variable.
4657 (arc_print_operand): Update how we handle small data operands.
4658 (arc_print_operand_address): Likewise.
4659 (arc_legitimate_address_p): Update, use
4660 legitimate_small_data_address_p.
4661 (arc_rewrite_small_data_p): Remove.
4662 (arc_rewrite_small_data_1): Likewise.
4663 (arc_rewrite_small_data): Likewise.
4664 (small_data_pattern): Likewise.
4665 (compact_sda_memory_operand): Update to use
4666 legitimate_small_data_address_p and get_symbol_alignment.
4667 (prepare_move_operands): Don't rewite sdata pattern.
4668 (prepare_extend_operands): Remove.
4669 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
4671 (zero_extendqisi2): Likewise.
4672 (zero_extendhisi2): Likewise.
4673 (extendqihi2): Likewise.
4674 (extendqisi2): Likewise.
4675 (extendhisi2): Likewise.
4679 * config/arc/constraints.md (Usd): Change it to memory constraint.
4681 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4683 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
4684 as source of std instructions.
4685 * config/arc/arc.md (movsi_insn): Update pattern predicate to
4686 allow 6-bit constants as source for store instructions.
4687 (movdi_insn): Update instruction pattern to allow 6-bit constants
4688 as source for store instructions.
4690 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4692 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
4694 2018-04-30 Nathan Sidwell <nathan@acm.org>
4695 Sandra Loosemore <sandra@codesourcery.com>
4697 * dumpfile.c (dump_open): Allow '-' for stdout.
4698 * doc/invoke.texi (Developer Options): Document dump filename
4699 determination early. Document stdin/stdout selection.
4701 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
4703 Microblaze Target: PIC data text relative
4705 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
4706 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
4708 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
4709 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
4710 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
4711 New addressing mode for data-text relative position indepenedent code.
4712 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
4713 'ADDRESS_SYMBOLIC_TXT_REL'.
4714 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
4715 (microblaze_legitimate_pic_operand): Exclude function calls from
4716 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
4717 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
4719 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4720 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4721 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
4722 for 'address + offset'.
4723 (microblaze_expand_prologue): Add new function prologue call for
4725 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
4726 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
4727 table in case of TARGET_PIC_DATA_TEXT_REL.
4728 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
4729 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
4730 Add new macros 'UNSPEC_TEXT',
4731 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
4732 + exclude function calls from 'UNSPEC_PLT' in case of data text
4734 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4735 new target hook for generating address diff vector tables in case of
4737 * doc/tm.texi : Regenerate.
4738 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
4739 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
4740 of addr diff vector generation.
4741 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4742 target hook definition.
4743 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
4744 Add default function for generate_pic_addr_diff_vec -> flag_pic.
4745 * doc/invoke.texi (Add new pic option): Add new microblaze pic
4746 option for data text relative.
4748 2018-04-30 Richard Biener <rguenther@suse.de>
4750 * tree-chrec.h (evolution_function_is_constant_p): Remove
4752 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
4754 2018-04-30 Richard Biener <rguenther@suse.de>
4757 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
4759 2018-04-30 Richard Biener <rguenther@suse.de>
4761 PR tree-optimization/28364
4762 PR tree-optimization/85275
4763 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
4764 copying first exit test.
4766 2018-04-28 Mark Wielaard <mark@klomp.org>
4768 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
4770 (dwarf_AT): Handle DW_AT_addr_base.
4771 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
4773 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
4776 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
4777 (*ashl<dwi>3_doubleword_mask_1): Ditto.
4778 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
4779 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
4781 2018-04-28 Richard Biener <rguenther@suse.de>
4783 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
4784 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
4785 to reflect use. Only add interesting stmts.
4787 2018-04-27 Martin Jambor <mjambor@suse.cz>
4790 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
4791 the jump function allows for passing through aggregate values.
4793 2018-04-27 David Malcolm <dmalcolm@redhat.com>
4795 * input.h (in_system_header_at): Convert from macro to inline
4797 (from_macro_expansion_at): Likewise.
4798 (from_macro_definition_at): Likewise.
4800 2018-04-27 Jeff Law <law@redhat.com>
4802 * config.gcc: Mark tile* targets as deprecated/obsolete.
4804 2018-04-27 Richard Biener <rguenther@suse.de>
4806 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
4809 2018-04-27 Richard Biener <rguenther@suse.de>
4811 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
4813 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
4815 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
4816 with Yd constraint. Set "preferred_for_speed" attribute from
4817 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
4819 (*movdi_internal): Ditto.
4820 (movti_interunit splitters): Remove
4821 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
4822 (movdi_interunit splitters): Ditto.
4823 * config/i386/constraints.md (Ye): Remove.
4824 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
4826 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4829 * config/aarch64/constraints.md (Usg): Limit to 31.
4832 2018-04-27 Jakub Jelinek <jakub@redhat.com>
4834 PR tree-optimization/85529
4835 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
4836 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
4837 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
4838 zero extension or masking of the MSB bit.
4839 (optimize_range_tests): Add FIRST_BB argument, pass it through
4840 to optimize_range_tests_var_bound.
4841 (maybe_optimize_range_tests, reassociate_bb): Adjust
4842 optimize_range_tests callers.
4844 2018-04-26 Richard Biener <rguenther@suse.de>
4845 Jakub Jelinek <jakub@redhat.com>
4847 * cgraph.h (symbol_table): Just declare debug method here.
4848 * symtab.c (symbol_table::debug): Define.
4850 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
4852 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
4854 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
4856 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
4857 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
4858 (*movdi_internal): Substitute Yi and Yj constraint with x
4859 and Ym and Yn constraint with y constraint. Update "isa"
4860 attribute and set "preferred_for_speed" attribute from
4861 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
4862 (*movsi_internal): Ditto.
4863 (*movdf_internal): Ditto.
4864 (*movsf_internal): Ditto.
4865 (*zero_extendsidi2): Ditto.
4866 * config/i386/sse.md (vec_set<mode>_0): Ditto.
4867 (sse2_loadld): Ditto.
4868 (*vec_extract<ssevecmodelower>_0): Ditto.
4869 (*vec_extractv4si_0_zext_sse4): Ditto.
4870 (vec_concatv2di): Ditto.
4871 (*vec_dup<mode>): Ditto.
4872 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
4873 * config/i386/constraints.md (Yi): Remove.
4878 2018-04-26 Nathan Sidwell <nathan@acm.org>
4880 * dumpfile.c (dump_open): New.
4881 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
4882 (dump_finish): Detect stdio/stderr by value not name.
4884 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
4886 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
4888 2018-04-26 Tom de Vries <tom@codesourcery.com>
4891 * config/nvptx/nvptx.c (verify_neutering_jumps)
4892 (verify_neutering_labels): New function
4893 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
4895 2018-04-26 Tom de Vries <tom@codesourcery.com>
4898 * config/nvptx/nvptx.c (needs_neutering_p): New function.
4899 (nvptx_single): Use needs_neutering_p to skip over insns that do not
4902 2018-04-26 Richard Biener <rguenther@suse.de>
4903 Tom de Vries <tom@codesourcery.com>
4906 * lto-streamer-out.c (output_function): Fixup loops if required to match
4907 discovery done in the reader.
4909 2018-04-26 Richard Biener <rguenther@suse.de>
4911 PR tree-optimization/85116
4912 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
4913 have a loop exit from the single latch predecessor. Remove
4914 case of header with just condition.
4915 (ch_base::copy_headers): Exclude infinite loops from any
4917 (pass_ch::execute): Record exits.
4919 2018-04-26 Richard Biener <rguenther@suse.de>
4921 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
4922 prologue cost vector and pass it to vect_get_load_cost.
4923 (vect_get_peeling_costs_all_drs): Likewise.
4924 (vect_peeling_hash_get_lowest_cost): Likewise.
4925 (vect_enhance_data_refs_alignment): Likewise.
4927 2018-04-26 Richard Biener <rguenther@suse.de>
4930 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
4931 checking of integer<->pointer conversions.
4932 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
4933 sign-/zero-extending pointer types.
4934 (expand_omp_for_static_chunk): Likewise.
4936 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
4937 Jean Lee <xiaoyur347@gmail.com>
4939 * config/mips/mips.c (mips_asan_shadow_offset): New function.
4940 (TARGET_ASAN_SHADOW_OFFSET): Define.
4941 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
4942 true for -fsanitize=address.
4944 2018-04-25 Mark Wielaard <mark@klomp.org>
4946 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
4949 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4951 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
4952 than "alu", remove explicit "memory" and "imm_disp" attributes.
4953 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
4956 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
4957 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
4958 gen_lowpart_no_emit.
4960 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
4963 * config/i386/i386.c (ix86_expand_builtin): Change memory
4964 operand to XI, extend p0 to Pmode.
4965 * config/i386/i386.md: Change unspec volatile and operand
4966 1 mode to XI, change operand 0 mode to P.
4968 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4970 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
4971 GET_MODE_MASK before any checking.
4972 (nds32_can_use_bset_p): Likewise.
4973 (nds32_can_use_btgl_p): Likewise.
4975 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4977 * config/nds32/nds32-doubleword.md: New define_split pattern for
4978 illegal register number.
4980 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4982 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
4984 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4986 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
4988 2018-04-25 Richard Biener <rguenther@suse.de>
4990 * lto-streamer.h (LTO_major_version): Bump to 8.
4992 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4994 * BASE-VER: Set to 9.0.0.
4996 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
4998 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
4999 in __abskf2 and __powikf2.
5001 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5004 * config/aarch64/constraints.md (Usg, Usj): New constraints.
5005 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5006 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5007 Use the above on operand 2. Reindent.
5008 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5010 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5013 * common/config/i386/i386-common.c (ix86_handle_option): Don't
5015 * config/i386/i386.opt (mcet): Removed.
5016 * doc/install.texi: Remove -mcet documentation.
5017 * doc/invoke.texi: Likewise.
5019 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5022 * doc/install.texi: Remove -mcet from bootstrap-cet.
5024 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5027 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5028 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5032 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5033 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5034 containing a CONST_VECTOR.
5036 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
5038 * doc/install.texi: Update newlib dependency for nvptx.
5040 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5043 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5044 instead of INTVAL when shifting x left.
5046 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
5048 PR tree-optimization/85478
5049 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5050 vect_grouped_store_supported for single element vectors.
5052 2018-04-24 Richard Biener <rguenther@suse.de>
5055 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5056 load cost increase to the case of non-constant step.
5058 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5061 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5062 destination if any_malformed_asm.
5064 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
5067 * expr.c (store_field): In the bitfield case, if the value comes from
5068 a function call and is returned in registers by means of a PARALLEL,
5069 do not change the mode of the temporary unless BLKmode and VOIDmode.
5071 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
5073 PR rtl-optimization/85423
5074 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5075 dependencies to debug insns when the previous insn is non-debug.
5077 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5079 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5080 enums into a single definition.
5081 (fls): Fix predicates and printing.
5084 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5086 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5087 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5088 and short u6 immediate.
5089 (check_if_valid_sleep_operand): Remove.
5090 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5092 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5094 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5095 flag_always_save_lp condition.
5096 * config/nds32/nds32.opt (malways-save-lp): New option.
5098 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5100 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5101 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5102 * config/nds32/nds32.h
5103 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5104 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5106 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5108 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5109 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5111 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5112 Chung-Ju Wu <jasonwucj@gmail.com>
5114 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5116 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5117 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5119 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5121 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5123 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5125 * config/nds32/nds32-protos.h (nds32_data_alignment,
5126 nds32_local_alignment): Declare.
5127 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5128 nds32_local_alignment): New functions.
5129 (TARGET_CONSTANT_ALIGNMENT): Define.
5130 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5132 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5134 * config/nds32/nds32.c
5135 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5136 (TARGET_MODES_TIEABLE_P): Likewise.
5138 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5140 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5143 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
5144 Chung-Ju Wu <jasonwucj@gmail.com>
5146 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5147 for unaligned access.
5148 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5150 * config/nds32/nds32-intrinsic.md: Likewise.
5151 * config/nds32/nds32_intrinsic.h: Likewise.
5152 * config/nds32/nds32.h (nds32_builtins): Likewise.
5153 * config/nds32/nds32.opt (munaligned-access): New option.
5154 * config/nds32/nds32.c (nds32_asm_file_start): Display
5155 flag_unaligned_access status.
5157 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5159 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5160 -mno-relax is present.
5161 * config/riscv/linux.h (LINK_SPEC): Ditto.
5163 2018-04-20 Martin Sebor <msebor@redhat.com>
5166 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
5168 (gimple_fold_builtin_stxcpy_chk): Same.
5169 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
5171 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
5174 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
5175 __powikf2 when long double is IEEE 128-bit.
5177 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5179 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
5180 step to make sure stack always aligned.
5182 2018-04-20 Carl Love <cel@us.ibm.com>
5185 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
5186 size check for arg0.
5188 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
5189 Tom de Vries <tom@codesourcery.com>
5192 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
5193 Emit insns for calls too.
5194 (nvptx_find_par): Always look for worker-level predecessor insn.
5195 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
5197 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
5198 (nvptx_process_pars): Propagate frames for calls.
5200 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
5203 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
5205 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
5206 (ix86_handle_option): Don't handle OPT_mibt.
5207 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
5209 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
5211 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
5212 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
5213 (ix86_target_macros): Define __CET__ with flag_cf_protection
5214 for -fcf-protection.
5215 * config/i386/i386.c (isa2_opts): Remove -mibt.
5216 * config/i386/i386.h (TARGET_IBT): Removed.
5217 (TARGET_IBT_P): Likewise.
5218 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
5219 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
5220 * config/i386/i386.opt (mcet): Update help message.
5223 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
5224 -mcet as an alias for -mshstk.
5226 2018-04-20 Richard Biener <rguenther@suse.de>
5229 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
5230 complexity by forcing a single use of the multiply operand.
5232 2018-04-20 Martin Jambor <mjambor@suse.cz>
5235 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
5236 recursion dependency to only apply to non-clones.
5238 2018-04-20 Martin Jambor <mjambor@suse.cz>
5241 * ipa-cp.c (create_specialized_node): Check that clones of
5242 self-recursive edges exist during IPA-CP.
5244 2018-04-19 Toon Moene <toon@moene.org>
5246 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
5249 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5251 PR tree-optimization/85467
5252 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
5253 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
5254 VECTOR_CST element to type.
5256 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5259 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
5260 * config/i386/i386.md (builtin_setjmp_setup): Removed.
5261 (builtin_longjmp): Likewise.
5262 (save_stack_nonlocal): New pattern.
5263 (restore_stack_nonlocal): Likewise.
5265 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5268 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5269 Replace ASM_OUTPUT_LABEL with fprintf.
5271 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5274 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5275 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
5276 * config/i386/i386-c.c (ix86_target_macros_internal): Also
5277 define __IBT__ and __SHSTK__ for -fcf-protection.
5278 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
5280 (ix86_trampoline_init): Likewise.
5281 (x86_output_mi_thunk): Likewise.
5282 (ix86_notrack_prefixed_insn_p): Likewise.
5283 (ix86_option_override_internal): Don't disallow -fcf-protection.
5284 * config/i386/i386.md (rdssp<mode>): Also enable for
5286 (incssp<mode>): Likewise.
5287 (nop_endbr): Likewise.
5288 * config/i386/i386.opt (mcet): Change help message to built-in
5292 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
5293 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
5294 enable CET built-in functions.
5296 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
5298 * common/config/i386/i386-common.c
5299 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
5300 OPTION_MASK_ISA_MOVDIRI_UNSET,
5301 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
5302 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
5303 * config.gcc (movdirintrin.h): New header.
5304 * config/i386/cpuid.h (bit_MOVDIRI,
5305 bit_MOVDIR64B): New bits.
5306 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
5308 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
5309 (VOID, PVOID, PCVOID)): New function types.
5310 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
5311 __builtin_ia32_directstoreu_u64,
5312 __builtin_ia32_movdir64b): New builtins.
5313 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
5314 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
5316 (ix86_valid_target_attribute_inner_p): Ditto.
5317 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
5318 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
5319 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
5320 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
5321 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
5322 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
5323 (movdiri<mode>, movdir64b_<mode>): New.
5324 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
5325 * config/i386/immintrin.h: Include movdirintrin.h.
5326 * config/i386/movdirintrin.h: New file.
5327 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
5329 2018-04-19 Richard Biener <rguenther@suse.de>
5332 * cfg.c (clear_bb_flags): When loop state says we have
5333 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
5335 2018-04-19 Richard Biener <rguenther@suse.de>
5337 PR tree-optimization/84737
5338 * tree-vect-data-refs.c (vect_copy_ref_info): New function
5339 copying restrict info.
5340 (vect_setup_realignment): Use it.
5341 * tree-vectorizer.h (vect_copy_ref_info): Declare.
5342 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
5343 the first DR to all generated stores.
5344 (vectorizable_load): Likewise for loads.
5346 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5348 PR tree-optimization/85446
5349 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
5350 the integral and pointer types to have the same precision.
5352 * doc/install.texi: Document --disable-cet being the default and
5355 2018-04-18 Martin Liska <mliska@suse.cz>
5357 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
5360 2018-04-18 Martin Liska <mliska@suse.cz>
5363 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5366 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5367 arguments if they are comparable.
5369 2018-04-18 Martin Liska <mliska@suse.cz>
5372 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5375 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5378 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
5381 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
5382 ENDBR after calling __morestack.
5384 2018-04-18 David Malcolm <dmalcolm@redhat.com>
5387 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
5388 by using gcc_base_ver to generate a gcc_driver_version, and use
5389 it when generating GCC_DRIVER_NAME.
5390 * configure: Regenerate.
5392 2018-04-18 Jakub Jelinek <jakub@redhat.com>
5395 * config.gcc: Obsolete powerpc*-*-*spe*.
5397 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5400 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
5401 (stabstr_D): Change type of unum from unsigned int to
5402 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
5405 2018-04-17 Jim Wilson <jimw@sifive.com>
5408 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
5409 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
5410 Set arg_pointer_offset after using pretend_args_size.
5412 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5414 PR rtl-optimization/85431
5415 * dse.c (record_store): Ignore zero width stores.
5418 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
5419 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
5420 __builtin_stack_restore rather than after it.
5421 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
5422 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
5423 argument instead of virtual_dynamic_stack_rtx.
5425 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
5427 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
5429 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5430 Add note to error message to explain internal mapping of overloaded
5431 built-in function name to non-overloaded built-in function name.
5432 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
5435 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5438 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
5439 where the inputs overlap with the output.
5441 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5443 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
5444 (=v, v) alternative and explicit "memory" attribute.
5445 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
5446 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5448 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
5449 "sselog1" type instead of "sselog".
5450 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
5451 "sselog". Remove explicit "memory" attribute.
5452 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
5453 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5455 (vec_extract_hi_v32hi): Merge all alternatives into one, use
5456 "sselog1" type instead of "sselog". Remove explicit "memory"
5458 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
5459 use "sselog1" type instead of "sselog". Remove explicit "memory"
5461 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
5462 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5464 (vec_extract_hi_v64qi): Merge all alternatives into one, use
5465 "sselog1" type instead of "sselog". Remove explicit "memory"
5467 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
5468 use "sselog1" type instead of "sselog". Remove explicit "memory"
5472 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
5475 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
5478 2018-04-17 Martin Jambor <mjambor@suse.cz>
5481 * ipa-cp.c (create_specialized_node): Call
5482 expand_all_artificial_thunks if necessary.
5484 2018-04-17 Martin Liska <mliska@suse.cz>
5487 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
5488 in message, remote space in between '_G' and '('.
5490 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5493 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
5494 avx512f_vmcmp<mode>3<round_saeonly_name>,
5495 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
5496 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
5497 avx512f_rndscale<mode><round_saeonly_name>,
5498 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
5499 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
5500 Use %<iptr>2 instead of %2 for -masm=intel.
5501 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
5502 avx512f_vcvttss2usi<round_saeonly_name>,
5503 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
5505 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
5506 avx512f_vcvttsd2usi<round_saeonly_name>,
5507 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
5508 Use %q1 instead of %1 for -masm=intel.
5509 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
5510 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
5511 of %3 for -masm=intel.
5512 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
5514 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
5516 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
5518 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
5519 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
5521 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
5523 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
5524 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
5525 %g1 and one with %0 and %1.
5526 (avx512er_vmrcp28<mode><round_saeonly_name>,
5527 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
5529 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
5530 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
5531 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
5532 of %0 and %{%4%} for -masm=intel.
5533 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
5534 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
5535 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
5536 order of %0 and %{%5%}%{z%} for -masm=intel.
5538 2018-04-17 Jan Hubicka <jh@suse.cz>
5541 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
5543 2018-04-17 Martin Liska <mliska@suse.cz>
5546 * multiple_target.c (create_dispatcher_calls): Set apostrophes
5547 for target_clone error message. Make default implementation
5548 clone to be a local declaration.
5549 (separate_attrs): Add new argument and check for an empty
5551 (expand_target_clones): Handle it.
5552 (ipa_target_clone): Make redirection just for target_clones
5555 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
5556 Tom de Vries <tom@codesourcery.com>
5559 * omp-expand.c (expand_oacc_for): Add dummy false branch for
5560 tiled basic blocks without omp continue statements.
5562 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
5565 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
5566 vec_extract expression as having side effects to make sure it gets
5569 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
5572 * config/i386/i386.c (get_builtin_code_for_version): Check
5575 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
5578 * gcc/config.gcc: Support "skylake".
5579 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
5581 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
5582 (processor_target_table): Add "skylake".
5583 (ix86_option_override_internal): Add "skylake".
5584 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
5585 PROCESSOR_CANNONLAKE.
5586 (get_builtin_code_for_version): Fix priority for
5587 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
5588 PROCESSOR_SKYLAKE-AVX512.
5589 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
5590 (processor_type): Add PROCESSOR_SKYLAKE.
5592 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
5593 Jason Merrill <jason@redhat.com>
5596 * convert.c (convert_to_integer_1): Use direct recursion for
5597 enumeral types and types with a precision less than the number
5598 of bits in their mode.
5600 2018-04-16 Julia Koval <julia.koval@intel.com>
5603 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
5604 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
5606 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
5609 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
5610 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
5611 and -mno-direct-move.
5613 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
5616 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
5617 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
5618 Ensure negative shifts result in {0}.
5620 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
5622 PR rtl-optimization/79916
5623 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
5624 regs (if any) to define how to gnerate SD moves when LRA is in
5627 2018-04-13 Jakub Jelinek <jakub@redhat.com>
5629 PR rtl-optimization/85393
5630 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
5631 * except.c (expand_dw2_landing_pad_for_region): Make static.
5632 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
5633 a label and unconditional jump to old_bb, rather than
5634 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
5637 PR rtl-optimization/85376
5638 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
5639 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
5640 instead of a specific value.
5642 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5643 Bin Cheng <bin.cheng@arm.com>
5645 PR tree-optimization/82965
5646 PR tree-optimization/83991
5647 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
5648 by_profile_only parameter.
5649 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
5650 information if the loop was predicted to iterate too many times.
5651 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
5653 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5656 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
5659 2018-04-13 Martin Liska <mliska@suse.cz>
5660 Jakub Jelinek <jakub@redhat.com>
5663 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
5664 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
5665 * builtins.c (expand_builtin_memory_copy_args): Use
5666 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
5667 handle dest_addr == pc_rtx.
5669 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
5672 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
5673 asked to not generate direct moves.
5674 (fix_trunc<mode>si2_stfiwx): Similar.
5675 (fix_trunc<mode>si2_internal): Similar.
5677 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5680 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
5681 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
5682 lookup if dest in some wider mode is known to be const0_rtx and
5683 if so, record permanent equivalence for it to be ZERO_EXTEND of
5684 the narrower mode destination.
5686 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5688 * lto-streamer-out.c (output_function): Revert 259346.
5689 * omp-expand.c (expand_oacc_for): Likewise.
5691 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
5693 PR rtl-optimization/85354
5694 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
5695 * sel-sched.c (sel_global_init): ... here.
5697 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
5700 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
5701 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
5702 mode for PE-COFF targets.
5703 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
5704 (i386_pe_asm_lto_end): Likewise.
5705 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
5706 (TARGET_ASM_LTO_END): Likewise.
5707 * config/i386/winnt.c (saved_debug_info_level): New static variable.
5708 (i386_pe_asm_lto_start): New function.
5709 (i386_pe_asm_lto_end): Likewise.
5711 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5712 Richard Biener <rguenther@suse.de>
5715 * lto-streamer-out.c (output_function): Fix CFG loop state before
5717 * omp-expand.c (expand_oacc_for): Handle calls to internal
5718 functions like regular functions.
5720 2018-04-12 Richard Biener <rguenther@suse.de>
5723 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
5724 for the early LTO debug to properly generate references to it
5725 during DIE emission. Do not re-use that for the skeleton for
5727 (dwarf2out_early_finish): Likewise.
5729 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5732 * config/i386/sse.md
5733 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
5734 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
5735 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
5736 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
5737 and output is a reg, avoid creating invalid lowpart subreg, but
5738 instead split into a 512-bit move. Don't split if not AVX512VL,
5739 input is xmm16+ reg and output is a mem.
5740 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
5741 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
5742 xmm16+ reg and output is a mem.
5744 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5746 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
5747 also for flag_dwarf2_cfi_asm.
5749 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5751 PR rtl-optimization/85342
5752 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
5753 a bool scalar var inside of the loop instead. Don't try to update
5754 recog_data.operand after failed apply_change_group.
5756 2018-04-12 Tom de Vries <tom@codesourcery.com>
5759 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
5760 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
5761 array with flexible array member as array without given dimension.
5762 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
5763 argument for undefined param to true.
5765 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
5768 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5769 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
5770 from PowerPC section.
5771 * config/rs6000/sysv4.opt (mcall-): Improve help text.
5772 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
5773 help text that is too long.
5774 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
5775 help text that is too long.
5776 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
5777 help text that is too long.
5779 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
5781 * config/alpha/alpha.md (stack_probe_internal): Rename
5782 from "probe_stack". Update all callers.
5784 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5786 PR rtl-optimization/84566
5787 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
5788 sched_macro_fuse_insns.
5790 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5793 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
5794 (compute_block_dependences): ... from here.
5796 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5798 PR tree-optimization/85331
5799 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
5800 from int to HOST_WIDE_INT.
5802 2018-04-11 Martin Jambor <mjambor@suse.cz>
5805 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
5806 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
5807 not the same as the source val.
5808 (cgraph_edge_brings_value_p): New parameter.
5809 (gather_edges_for_value): Pass destination value to
5810 cgraph_edge_brings_value_p.
5811 (perhaps_add_new_callers): Likewise.
5812 (get_info_about_necessary_edges): Likewise and exclude values brought
5813 only by self-recursive edges.
5814 (create_specialized_node): Redirect only clones of self-calling edges.
5815 (+self_recursive_pass_through_p): New function.
5816 (find_more_scalar_values_for_callers_subset): Use it.
5817 (find_aggregate_values_for_callers_subset): Likewise.
5818 (known_aggs_to_agg_replacement_list): Removed.
5819 (decide_whether_version_node): Re-calculate known constants for all
5820 remaining context clones.
5822 2018-04-11 Richard Biener <rguenther@suse.de>
5825 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
5826 from early DWARF output.
5827 (dwarf2out_early_finish): Output line info unconditionally into
5828 early DWARF and add reference to it.
5830 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5833 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
5834 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
5835 other than V2DFmode using iptr mode attribute.
5836 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
5838 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5840 PR rtl-optimization/84659
5841 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
5843 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5846 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
5848 (output_loc_list): Pass address of a dummy size variable even in the
5849 locview handling loop.
5850 (index_location_lists): Add comment on why skip_loc_list_entry can't
5853 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
5856 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
5859 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
5862 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5863 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
5864 and -mstring-compare-inline-limit.
5866 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5869 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
5870 for stack clash protection in a register whenever we need it to be in
5873 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5875 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
5876 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
5878 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
5881 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
5883 (mlong-double-): Ditto.
5884 * config/rs6000/sysv4.opt (msdata=): Ditto.
5885 (mtls-size=): Ditto.
5887 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5889 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5890 erroneous entries for
5891 "vector int vec_ldl (int, long int *)", and
5892 "vector unsigned int vec_ldl (int, unsigned long int *)".
5893 Add comments and entries for
5894 "vector bool char vec_ldl (int, bool char *)",
5895 "vector bool short vec_ldl (int, bool short *)",
5896 "vector bool int vec_ldl (int, bool int *)",
5897 "vector bool long long vec_ldl (int, bool long long *)",
5898 "vector pixel vec_ldl (int, pixel *)",
5899 "vector long long vec_ldl (int, long long *)",
5900 "vector unsigned long long vec_ldl (int, unsigned long long *)".
5901 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
5902 type tree bool_long_long_type_node and correct definition of
5903 bool_V2DI_type_node to make reference to this new type tree.
5904 (rs6000_mangle_type): Replace erroneous reference to
5905 bool_long_type_node with bool_long_long_type_node.
5906 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
5907 comments to emphasize sign distinctions for char and int types and
5908 replace RS6000_BTI_bool_long constant with
5909 RS6000_BTI_bool_long_long constant. Also add comment to restrict
5910 use of RS6000_BTI_pixel.
5911 (bool_long_type_node): Remove this macro definition.
5912 (bool_long_long_type_node): New macro definition
5914 2018-04-10 Jakub Jelinek <jakub@redhat.com>
5916 PR rtl-optimization/85300
5917 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
5918 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
5919 simplify_unary_operation fails.
5921 2018-04-10 Martin Liska <mliska@suse.cz>
5923 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
5924 cgraph_edge and ipa_ref.
5926 2018-04-10 Jakub Jelinek <jakub@redhat.com>
5930 * config/i386/sse.md
5931 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
5932 computation of the VEC_MERGE selector from mask.
5933 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
5934 Fix decoding of the VEC_MERGE selector into mask.
5936 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5938 PR tree-optimization/85286
5939 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
5941 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5943 * final.c (final_1): Set insn_last_address as well as
5944 insn_current_address.
5946 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5949 * explow.c (emit_stack_probe): Call validize_mem on memory location
5950 before passing it to gen_probe_stack. Create address operand and
5951 legitimize it for the probe_stack_address case.
5953 2018-04-09 Jan Hubicka <jh@suse.cz>
5956 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
5957 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
5958 * tree.c (free_lang_data_in_type): Fix handling of binfos;
5960 (free_lang_data): Rebuild type inheritance graph.
5962 2018-04-09 Martin Sebor <msebor@redhat.com>
5964 * invoke.texi (-finline-small-functions): Mention other optimization
5966 (-findirect-inlining, -fpartial-inlining): Same.
5967 (-finline-functions-called-once): Same.
5968 (-freorder-blocks-and-partition): Same.
5970 2018-04-09 Jan Hubicka <jh@suse.cz>
5973 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
5974 jumps; choose last target that matches the criteria (i.e.
5975 no partition changes for non-crossing jumps).
5976 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
5977 support for redirecting crossing jumps to non-crossing.
5979 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
5981 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
5982 also for naked functions.
5984 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
5986 * config/arc/arc.md (add_shift): New pattern.
5987 (add_shift2): Likewise.
5988 (sub_shift): Likewise.
5989 (sub_shift_cmp0_noout): Likewise.
5990 (compare_si_ashiftsi): Likewise.
5991 (xbfu_cmp0_noout): New combine pattern.
5992 (xbfu_cmp0"): Likewise.
5993 (movsi_set_cc_insn): Place the predicable variant first.
5994 (commutative_binary_cmp0_noout): Remove clobber.
5995 (commutative_binary_cmp0): New pattern.
5996 (noncommutative_binary_cmp0): Likewise.
5997 (noncommutative_binary_cmp0_noout): Likewise.
5998 (noncommutative_binary_comparison_result_used): Removed.
5999 (rsub_cmp0): New pattern.
6000 (rsub_cmp0_noout): Likewise.
6001 (extzvsi): Changed, keep only meaningful variants.
6002 (SQH, SEZ): New iterators.
6003 (SQH_postfix): New mode attribute.
6004 (SEZ_prefix): New code attribute.
6005 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6006 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6007 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6009 (noncommutative_operator): Check the availability of barrel
6012 2018-04-09 Richard Biener <rguenther@suse.de>
6014 PR tree-optimization/85284
6015 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6016 Only use the niter constraining form of simple_iv when the exit
6019 2018-04-09 Tom de Vries <tom@codesourcery.com>
6022 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6023 (define_expand "*memory_barrier"): New define_expand.
6024 (define_insn "memory_barrier"): New insn.
6026 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6028 PR rtl-optimization/80463
6029 PR rtl-optimization/83972
6030 PR rtl-optimization/83480
6032 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6033 correct producer for the insn.
6034 (tidy_control_flow): Fixup seqnos in case of debug insns.
6036 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6038 PR rtl-optimization/83913
6040 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6041 different sched-times when merging exprs.
6043 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6045 PR rtl-optimization/83962
6047 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6048 tidy_fallthru_edge and tidy_control_flow.
6050 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6052 PR rtl-optimization/83530
6054 * sel-sched.c (force_next_insn): New global variable.
6055 (remove_insn_for_debug): When force_next_insn is true, also leave only
6056 next insn in the ready list.
6057 (sel_sched_region): When the region wasn't scheduled, make another pass
6058 over it with force_next_insn set to 1.
6060 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6062 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6064 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6065 for interrupt control.
6066 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6067 functions for interrupt control.
6068 * config/nds32/nds32-intrinsic.md: Likewise.
6069 * config/nds32/nds32_intrinsic.h: Likewise.
6070 * config/nds32/nds32.h (nds32_builtins): Likewise.
6072 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6074 * config/nds32/nds32.c (nds32_init_machine_status,
6075 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6076 strict_aligned_p field.
6077 (nds32_expand_to_rtl_hook): New function.
6078 (TARGET_EXPAND_TO_RTL_HOOK): Define.
6079 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6081 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6082 Chung-Ju Wu <jasonwucj@gmail.com>
6084 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6085 * config/nds32/nds32-n7.md: New file.
6086 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6087 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6089 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6090 * config/nds32/nds32.md (pipeline_model): Add n7.
6091 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6092 * config/nds32/pipelines.md: Include n7 settings.
6094 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6095 Chung-Ju Wu <jasonwucj@gmail.com>
6097 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6098 * config/nds32/nds32-e8.md: New file.
6099 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6100 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6102 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6103 * config/nds32/nds32.md (pipeline_model): Add e8.
6104 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6105 * config/nds32/pipelines.md: Include e8 settings.
6107 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6108 Chung-Ju Wu <jasonwucj@gmail.com>
6110 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6111 * config/nds32/nds32-n8.md: New file.
6112 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6113 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6115 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6116 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6117 * config/nds32/nds32.md (pipeline_model): Add n8.
6118 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6119 * config/nds32/pipelines.md: Include n8 settings.
6121 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6122 Chung-Ju Wu <jasonwucj@gmail.com>
6124 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6125 * config/nds32/nds32-n9-2r1w.md: New file.
6126 * config/nds32/nds32-n9-3r2w.md: New file.
6127 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6128 nds32_register_ports): New or modify for cpu n9.
6129 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6131 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6132 * config/nds32/nds32-utils.c: New file.
6133 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6134 TARGET_MUL_SLOW): Define.
6135 * config/nds32/nds32.md (pipeline_model): New attribute.
6136 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6137 New options that support cpu n9.
6138 * config/nds32/pipelines.md: Include n9 settings.
6139 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6141 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6143 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6144 information if necessary.
6145 (output_cond_branch_compare_zero): Likewise.
6146 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6147 (nds32_target_alignment): Refine for alignment.
6148 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6149 (FUNCTION_BOUNDARY): Modify.
6150 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6152 * config/nds32/nds32.opt (malways-align, malign-functions): New.
6154 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6156 * config/nds32/constants.md (unspec_volatile_element): Add values for
6157 TLB operation and data prefetch.
6158 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6159 functions for TLB operation and data prefetch.
6160 * config/nds32/nds32-intrinsic.md: Likewise.
6161 * config/nds32/nds32_intrinsic.h: Likewise.
6162 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
6163 (nds32_print_operand): Likewise.
6164 * config/nds32/nds32.h (nds32_builtins): Likewise.
6166 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
6167 Andrew Pinski <pinsika@gcc.gnu.org>
6170 * match.pd: Use constant_boolean_node of correct type instead of
6171 boolean_true_node or boolean_false_node for simplifying
6172 pointer comparisons to zero.
6174 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6176 PR tree-optimization/80021
6177 * tree.c (verify_type_variant): Make error call in verify_variant_match
6178 translatable and remove final full stop.
6180 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6182 * config/nds32/constants.md (unspec_volatile_element): Add
6183 UNSPEC_VOLATILE_EH_RETURN.
6184 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
6185 nds32_output_stack_pop): Support dwarf exception handling process.
6186 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
6187 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
6188 exception handling process.
6189 (nds32_compute_stack_frame): Likewise.
6190 (nds32_return_addr_rtx): Likewise.
6191 (nds32_initial_elimination_offset): Likewise.
6192 (nds32_expand_prologue): Likewise.
6193 (nds32_expand_epilogue): Likewise.
6194 (nds32_dynamic_chain_address): New function.
6195 * config/nds32/nds32.h (machine_function): Add fields for dwarf
6197 (DYNAMIC_CHAIN_ADDRESS): Define.
6198 (EH_RETURN_DATA_REGNO): Define.
6199 (EH_RETURN_STACKADJ_RTX): Define.
6200 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
6201 patterns for dwarf exception handling.
6203 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6205 * config/nds32/nds32.h: Clean up obsolete macros.
6207 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6209 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6210 Add enum values for particular instructions.
6211 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6212 particular intrinsic functions.
6213 * config/nds32/nds32-intrinsic.md: Likewise.
6214 * config/nds32/nds32_intrinsic.h: Likewise.
6215 * config/nds32/nds32.h (nds32_builtins): Likewise.
6216 * config/nds32/nds32.md (type): Add pbsad and pbsada.
6217 (btst, ave): New patterns for particular instructions.
6219 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6221 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6222 Add enum values for atomic load/store and memory sync.
6223 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
6225 * config/nds32/nds32-intrinsic.md: Likewise.
6226 * config/nds32/nds32_intrinsic.h: Likewise.
6227 * config/nds32/nds32.h (nds32_builtins): Likewise.
6229 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6231 PR tree-optimization/85257
6232 * fold-const.c (native_encode_vector): If not all elts could fit
6233 and off is -1, return 0 rather than offset.
6234 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
6235 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
6236 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
6237 adjust buffer in native_interpret_expr call.
6239 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6241 * config/nds32/constants.md (unspec_volatile_element): Add cache
6242 control enum values.
6243 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
6244 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
6245 * config/nds32/nds32.c (nds32_cctl_names): New.
6246 (nds32_print_operand): Handle cache control register names.
6247 * config/nds32/nds32.h (nds32_builtins): New enum values.
6248 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
6250 * config/nds32/nds32.md (type): Add mmu.
6251 * config/nds32/pipelines.md (simple_insn): Add mmu.
6253 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6255 * config/nds32/nds32.md (type): Remove call.
6256 * config/nds32/pipelines.md (simple_insn): Likewise.
6258 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6260 * config/nds32/constants.md (unspec_volatile_element): Add
6261 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
6262 UNSPEC_VOLATILE_FMFCFG.
6263 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
6264 description for fmfcfg and fmfcsr.
6265 (bdesc_1arg): Add fmtcsr.
6266 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
6267 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
6268 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
6269 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
6270 unspec_fmfcfg): New patterns.
6271 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
6272 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
6273 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
6274 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
6275 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
6276 __nds32__fmfcfg): Define.
6278 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6280 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
6281 intrinsic register names.
6282 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
6283 intrinsic register enum values and macros.
6285 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6287 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
6288 for load/store addressing form.
6289 (nds32_print_operand_address): Likewise.
6291 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
6294 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
6295 based on LABEL_REF. Remove useless assertion.
6296 (pic_address_needs_scratch): Fix formatting.
6297 (sparc_legitimize_pic_address): Minor tweaks.
6298 (sparc_delegitimize_address): Adjust assertion accordingly.
6299 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
6300 into symbolic_operand.
6301 (movsi_high_pic_label_ref): Likewise.
6302 (movsi_lo_sum_pic_label_ref): Likewise.
6303 (movdi_pic_label_ref): Likewise.
6304 (movdi_high_pic_label_ref): Likewise.
6305 (movdi_lo_sum_pic_label_ref): Likewise.
6307 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
6309 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
6310 custom LIB_SPEC setup.
6312 2018-04-06 Ruslan Bukin <br@bsdpad.com>
6313 Kito Cheng <kito.cheng@gmail.com>
6315 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
6316 * config/riscv/freebsd.h: New.
6318 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6320 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
6321 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
6324 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6325 Kito Cheng <kito.cheng@gmail.com>
6327 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
6328 nds32_output_call, nds32_symbol_binds_local_p): New functions.
6329 * config/nds32/nds32-protos.h (nds32_output_call,
6330 nds32_output_return): Declare.
6331 * config/nds32/nds32.md: Refine all the call and return patterns.
6333 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6336 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
6337 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
6339 PR rtl-optimization/84872
6340 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
6341 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
6344 2018-04-06 Tamar Christina <tamar.christina@arm.com>
6346 * expr.c (copy_blkmode_to_reg): Revert 254862.
6347 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
6349 2018-04-06 Richard Biener <rguenther@suse.de>
6352 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
6353 after seeing a component reference with an adjacent field. Treat
6354 refs to arrays at struct end of external decls similar to
6355 refs to unconstrained commons.
6357 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6360 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
6361 look through SAVE_EXPRs with non-side-effects argument. Adjust
6363 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
6366 2018-04-06 Richard Biener <rguenther@suse.de>
6369 * alias.c (find_base_term): New wrapper around find_base_term
6370 unwinding CSELIB_VAL_PTR changes.
6371 (find_base_term): Do not restore CSELIB_VAL_PTR during the
6374 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6376 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
6378 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
6379 constant definitions.
6380 ("nop"): lr 0,0 -> nopr r0
6381 ("nop_lr0", "nop_lr1"): New insn definitions.
6383 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6385 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
6386 NDS32_V3PUSH_AVAILABLE_P macro.
6388 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
6389 Chung-Ju Wu <jasonwucj@gmail.com>
6391 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
6392 (nds32*-*-*): Add float and fpu_config into supported_defaults.
6393 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
6394 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
6395 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
6396 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
6397 * config/nds32/constraints.md: New constraints and checking for hard
6398 float configuration.
6399 * config/nds32/iterators.md: New mode iterator and attribute for hard
6400 float configuration.
6401 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
6403 * config/nds32/nds32-fpu.md: New file.
6404 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
6405 deal with hard float code generation.
6406 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
6408 (abi_type, float_reg_number): New enum type.
6409 * config/nds32/nds32-predicates.c: New predicates for hard float.
6410 * config/nds32/nds32-protos.h: Declare functions for hard float.
6411 * config/nds32/nds32.c: Implementation for hard float configuration.
6412 * config/nds32/nds32.h: Definitions for hard float configuration.
6413 * config/nds32/nds32.md: Include hard float machine description and
6414 modify patterns for hard float configuration.
6415 * config/nds32/nds32.opt: New options for hard float configuration.
6416 * config/nds32/predicates.md: New predicates for hard float
6419 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6421 * common/config/nds32/nds32-common.c
6422 (nds32_option_optimization_table): Enable -mreleax-hint by default.
6424 2018-04-05 Jakub Jelinek <jakub@redhat.com>
6427 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
6428 CONSTRUCTOR_ELT (ctor, ...)->value.
6430 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
6433 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
6435 2018-04-05 Tom de Vries <tom@codesourcery.com>
6438 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
6441 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6442 Kito Cheng <kito.cheng@gmail.com>
6444 * config/nds32/constraints.md (U33): Fine-tune checking condition.
6445 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
6446 * config/nds32/nds32.h (nds32_16bit_address_type): Add
6447 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
6449 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6450 Kito Cheng <kito.cheng@gmail.com>
6452 * config/nds32/constraints.md (Ufe): New memory constraint.
6453 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
6454 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
6455 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
6457 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
6458 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
6460 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6462 * config/nds32/nds32.md: Use optimize_size in the condition for
6463 alu-shift instructions.
6465 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6467 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
6469 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6471 * config/nds32/nds32.md (negsi2): Refine pattern.
6473 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6474 Chung-Ju Wu <jasonwucj@gmail.com>
6476 * config/nds32/iterators.md (shift_rotate): New code iterator.
6477 (shift): New code attribute.
6478 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
6479 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
6480 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
6481 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
6482 bit-wise operations.
6483 (andsi3, *andsi3): Ditto.
6484 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
6485 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
6486 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
6487 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
6488 nds32_ior_operand, nds32_xor_operand): New predicates.
6490 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6492 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
6493 (addsi3, subsi3): ... this.
6495 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6497 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
6499 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6501 * config/nds32/nds32.md: Adjust indention.
6503 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6505 * config/nds32/nds32.md (feature): New attribute.
6507 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6509 * config/nds32/nds32.md (subtype): New attribute.
6511 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
6514 * config/arm/arm-builtins.c (arm_expand_builtin): Change
6515 expansion to perform a bitwise AND of the argument followed by a
6516 boolean negation of the result.
6518 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
6520 PR rtl-optimization/84878
6521 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
6522 the basic block. Assert the use reference is not artificial and that
6523 it has an associated insn.
6525 2018-04-04 Michael Matz <matz@suse.de>
6527 * builtins.c (compute_objsize): Pass correct operand
6528 to array_at_struct_end_p.
6530 2018-04-04 Richard Biener <rguenther@suse.de>
6533 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
6534 from contexts for DINFO_LEVEL_TERSE and below.
6536 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6538 * config/nds32/nds32-doubleword.md (move_<mode>): Require
6539 resiter_operand condition.
6540 * config/nds32/nds32.md (*move<mode>): Ditto.
6542 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6543 Monk Chiang <sh.chiang04@gmail.com>
6545 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
6547 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6549 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
6551 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6552 Kito Cheng <kito.cheng@gmail.com>
6554 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
6555 nds32_cond_code_str, output_cond_branch,
6556 output_cond_branch_compare_zero, nds32_expand_cbranch,
6557 nds32_expand_cstore, nds32_expand_movcc,
6558 nds32_output_cbranchsi4_equality_zero,
6559 nds32_output_cbranchsi4_equality_reg,
6560 nds32_output_cbranchsi4_equality_reg_or_const_int,
6561 nds32_output_cbranchsi4_greater_less_zero: New functions.
6562 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
6563 nds32_expand_cstore, nds32_expand_movcc,
6564 nds32_output_cbranchsi4_equality_zero,
6565 nds32_output_cbranchsi4_equality_reg,
6566 nds32_output_cbranchsi4_equality_reg_or_const_int,
6567 nds32_output_cbranchsi4_greater_less_zero): Declare.
6568 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
6569 nds32_rimm11s_operand): New predicates.
6570 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
6571 * config/nds32/nds32.md: Rewrite all the branch and conditional move
6574 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6576 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
6577 * config/nds32/nds32.md: Ditto.
6578 * config/nds32/pipelines.md: Ditto.
6580 2018-04-04 Richard Biener <rguenther@suse.de>
6582 PR tree-optimization/85168
6583 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
6584 propagating abnormals.
6586 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6588 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
6590 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6591 Kito Cheng <kito.cheng@gmail.com>
6593 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
6594 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
6595 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
6596 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
6597 * config/nds32/nds32.md (sibcall_internal): New.
6598 (sibcall_register): Remove.
6599 (sibcall_immediate): Remove.
6600 (sibcall_value_internal): New.
6601 (sibcall_value_register): Remove.
6602 (sibcall_value_immediate): Remove.
6603 * config/nds32/predicates.md (nds32_general_register_operand): New.
6604 (nds32_call_address_operand): New.
6606 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6608 PR rtl-optimization/85167
6609 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
6610 bb_defs if *split_p, instead preinitialize it to NULL.
6612 PR tree-optimization/85156
6613 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
6614 evaluating the argument multiple times.
6616 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
6618 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
6620 (_mm_cvtpd_ps): Likewise.
6621 (_mm_cvttpd_epi32): Likewise.
6622 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
6623 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
6624 vector, pixel, and bool following altivec.h include.
6626 2018-04-03 Martin Sebor <msebor@redhat.com>
6628 * doc/extend.texi (Common Function Attributes): Clarify.
6629 (const attribute): Likewise.
6630 (pure attribute): Likewise.
6632 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6635 * config/i386/i386.c (ix86_expand_vector_set): Use
6636 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
6638 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
6640 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
6641 instructions when changing rounding bits to preserve precision bits
6642 in the x87 control word.
6644 2018-04-03 Martin Liska <mliska@suse.cz>
6646 PR tree-optimization/82491
6647 * rtl.h (strip_offset_and_add): Replace += suboffset with
6648 poly_uint64 () + suboffset.
6650 2018-03-29 Martin Liska <mliska@suse.cz>
6651 Martin Jambor <mjambor@suse.cz>
6654 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
6655 param_type is not an integral or pointer type.
6657 2018-04-03 Richard Biener <rguenther@suse.de>
6659 * sese.h (recompute_all_dominators): Remove.
6661 2018-04-02 Martin Sebor <msebor@redhat.com>
6663 * doc/invoke.texi (-Wrestrict): Fix typos.
6665 2018-04-02 Jim Wilson <jimw@sifive.com>
6667 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
6668 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
6669 (<optab>di3, <optab>si3_extend): Likewise.
6670 (<optab>si3_mask, <optab>si3_mask_1): New.
6671 (<optab>di3_mask, <optab>di3_mask_1): New.
6672 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
6673 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
6674 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
6676 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
6678 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
6681 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
6683 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
6684 (nds32_canonicalize_comparison): New function.
6686 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6687 Kito Cheng <kito.cheng@gmail.com>
6688 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6690 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
6691 * config/nds32/constants.md (unspec_volatile_element): Add
6692 UNSPEC_VOLATILE_RELAX_GROUP.
6693 * config/nds32/nds32-relax-opt.c: New file.
6694 * config/nds32/nds32-predicates.c
6695 (nds32_symbol_load_store_p): New function.
6696 * config/nds32/nds32-protos.h
6697 (nds32_symbol_load_store_p): Declare function.
6698 (make_pass_nds32_relax_opt): Declare new rtl pass function.
6699 * config/nds32/nds32.c
6700 (nds32_register_pass): New function to register pass.
6701 (nds32_register_passes): New function to register passes.
6702 * config/nds32/nds32.md (relax_group): New pattern.
6703 * config/nds32/nds32.opt (mrelax-hint): New option.
6704 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
6706 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
6708 * config/nds32/t-nds32: Modify files dependency.
6710 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6712 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
6713 (PROFILE_HOOK): Define its implementation.
6715 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6717 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
6718 type and 32-bit size.
6720 2018-04-01 Jakub Jelinek <jakub@redhat.com>
6723 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
6724 (V_128_256): New mode iterator.
6725 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
6726 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
6727 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
6729 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
6730 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
6732 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
6735 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
6736 NaN inputs correctly.
6738 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
6741 * config/rs6000/vsx.md (??r): New mode attribute.
6742 (*vsx_mov<mode>_64bit): Use it.
6743 (*vsx_mov<mode>_32bit): Likewise.
6745 2018-03-30 Martin Sebor <msebor@redhat.com>
6747 PR tree-optimization/84818
6748 * builtins.c (check_access): Use warning_n.
6750 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6753 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
6755 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
6758 2018-03-30 Julia Koval <julia.koval@intel.com>
6761 * x86-tune.def (movx, partial_reg_dependency): Enable for
6764 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
6767 * lra-constraints.c (process_alt_operands): Move setting
6768 this_alternative_matches below.
6770 2018-03-29 Martin Liska <mliska@suse.cz>
6773 * doc/invoke.texi: Document how LTO works with debug info.
6774 Describe auto-load support of binutils. Mention 'x86-64'
6775 as valid option value of -march option.
6777 2018-03-29 Jakub Jelinek <jakub@redhat.com>
6779 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
6782 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
6783 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
6784 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
6787 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
6790 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
6791 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
6792 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
6793 for RS6000_BTM_POWERPC64.
6794 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
6795 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
6796 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
6801 2018-03-28 Carl Love <cel@us.ibm.com>
6804 2017-09-27 Carl Love <cel@us.ibm.com>
6806 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
6807 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
6808 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
6811 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6813 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
6814 instead of __vector bool.
6815 (_mm_max_pu8): Likewise.
6816 (_mm_min_pi16): Likewise.
6818 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
6821 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
6822 (DIVWEUO): Likewise.
6824 (DIVDEUO): Likewise.
6825 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
6826 DIVWEUO and DIVDEUO.
6827 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
6828 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
6829 (div_extend): Likewise.
6830 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
6832 (__builtin_divweuo): Likewise.
6833 (__builtin_divdeo): Likewise.
6834 (__builtin_divdeuo): Likewise.
6836 2018-03-28 Jakub Jelinek <jakub@redhat.com>
6839 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
6840 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
6842 PR tree-optimization/82004
6843 * gimple-match-head.c (optimize_pow_to_exp): New function.
6844 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
6845 Don't fold to exp if optimize_pow_to_exp is false.
6847 2018-03-28 Martin Liska <mliska@suse.cz>
6850 * calls.c (initialize_argument_information): Fix trailing space.
6851 * common.opt: Fix typo and provide better explanation for
6852 -fsanitize-coverage option.
6853 * config/i386/i386.opt: Fix typo.
6855 2018-03-28 Jakub Jelinek <jakub@redhat.com>
6856 Martin Liska <mliska@suse.cz>
6859 * gimplify.c (asan_poison_variable): Don't do the check for
6860 gimplify_omp_ctxp here.
6861 (gimplify_decl_expr): Do it here.
6862 (gimplify_target_expr): Likewise.
6864 2018-03-28 Martin Liska <mliska@suse.cz>
6867 * config/i386/i386.c (ix86_function_arg_advance): Do not call
6868 chkp_type_bounds_count if MPX is not enabled.
6870 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
6872 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
6874 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
6877 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
6878 function to create the function decl for complex long double
6879 multiply and divide for -mabi=ieeelongdouble.
6880 (init_float128_ieee): Call it.
6882 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
6885 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
6886 -fcf-protection=branch -mibt.
6887 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
6889 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6892 * config/arm/arm.c (arm_valid_symbolic_address): Handle
6893 arm_word_relocations.
6895 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
6898 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
6899 extern array declarations.
6901 2018-03-27 Richard Biener <rguenther@suse.de>
6904 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
6905 explicit single_use checks.
6907 2018-03-27 Richard Biener <rguenther@suse.de>
6909 PR tree-optimization/85082
6910 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
6913 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6915 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
6916 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
6917 Turn on fasynchronous-unwind-tables and funwind-tables.
6919 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
6922 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
6923 (*bmi_blsr_<mode>_ccz): Ditto.
6925 2018-03-26 Tom de Vries <tom@codesourcery.com>
6927 PR tree-optimization/85063
6928 * omp-general.c (offloading_function_p): New function. Factor out
6930 * omp-offload.c (pass_omp_target_link::gate): ... here.
6931 * omp-general.h (offloading_function_p): Declare.
6932 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
6933 with attribute omp declare target for offloading functions.
6935 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
6937 PR tree-optimization/84005
6938 * tree-data-ref.h (get_base_for_alignment): Declare.
6939 * tree-data-ref.c (get_base_for_alignment_1): New function.
6940 (get_base_for_alignment): Likewise.
6941 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
6942 get_base_for_alignment to find a suitable base object, instead
6943 of always using drb->base_address.
6945 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6948 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
6949 known size by default.
6951 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
6954 * lra-constraints.c (process_alt_operands): Don't match BLKmode
6955 and non BLKmode operands.
6957 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6960 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
6961 Clean up attributes.
6963 2018-03-23 Richard Biener <rguenther@suse.de>
6966 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
6967 we are going to emit early debug for LTO.
6969 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6972 * function.c (match_asm_constraints_1): Don't optimize if input
6973 doesn't satisfy general_operand predicate for output's mode.
6976 * alias.c (write_dependence_p): Don't require for x_canonicalized
6977 non-VOIDmode if x has VOIDmode.
6980 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
6981 just don't try to optimize it rather than assert it never happens.
6983 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6985 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
6986 macro expansions for definition of ST_INTERNAL_<mode> and
6987 LD_INTERNAL_<mode> builtins.
6988 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
6990 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
6992 (altivec_expand_st_builtin): Likewise.
6993 (altivec_expand_builtin): Remove calls to deleted functions.
6994 (rs6000_address_for_altivec): Delete this function.
6995 * config/rs6000/vector.md: Remove expands for
6996 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
6998 2018-03-22 Sudakshina Das <sudi.das@arm.com>
7001 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7002 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7003 re-computing once computed.
7004 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7005 (arm_init_machine_status): Initialize
7006 machine->static_chain_stack_bytes.
7008 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7011 * doc/extend.texi: Add four new prototypes for vec_ld.
7012 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7013 definitions for more logical presentation.
7014 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7015 entries for V1TI variants of __builtin_altivec_ld builtin.
7016 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7017 handling of V1TI variant of LVX icode pattern.
7018 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7019 (rs6000_gimple_fold_builtin): Likewise.
7020 (altivec_init_builtins): Add code to define
7021 __builtin_altivec_lvx_v1ti function.
7023 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7026 * function.c (match_asm_constraints_1): Don't do the optimization
7027 if input isn't a REG, SUBREG, MEM or constant.
7029 2018-03-22 Tom de Vries <tom@codesourcery.com>
7031 PR tree-optimization/84956
7032 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7033 bb_has_abnormal_pred.
7035 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7038 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7039 DECL_INITIAL (decl) to decl at the end.
7040 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7043 2018-03-21 Joseph Myers <joseph@codesourcery.com>
7045 * doc/extend.texi (__builtin_tgmath): Document when complex
7046 integer types are treated as _Complex _Float64.
7048 2018-03-21 Tom de Vries <tom@codesourcery.com>
7050 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7052 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7054 PR tree-optimization/84960
7055 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7056 if it is ENTRY block, move them into single succ of ENTRY in that case.
7058 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
7060 PR tree-optimization/84811
7061 * poly-int.h (poly_span_traits): Remove the T3 parameter and
7062 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7063 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7064 (known_subrange_p): Update accordingly. Cast each value involved
7065 in the size comparison, rather than casting the result of the
7068 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7070 PR tree-optimization/84982
7071 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7072 by flipping the least significant bit rather than all bits from
7073 bitpos to bitpos + bitsize - 1.
7075 2018-03-21 Nathan Sidwell <nathan@acm.org>
7077 * doc/extend.texi (Deprecated Features): Remove mention of
7078 long-deleted deprecations.
7080 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7083 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7084 * configure: Regenerate.
7086 2018-03-21 Tom de Vries <tom@codesourcery.com>
7088 PR tree-optimization/83126
7089 * tree-parloops.c (num_phis): New function.
7090 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7092 2018-03-21 Nathan Sidwell <nathan@acm.org>
7094 * doc/extend.texi (Deprecated Features): Update deprecated flags,
7095 mention anon-struct/union members and trailing attributes.
7097 2018-03-21 Bin Cheng <bin.cheng@arm.com>
7099 PR tree-optimization/84969
7100 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7101 builtin memset partitions if they set different rhs values.
7103 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7105 PR rtl-optimization/84989
7106 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7107 VEC_DUPLICATE with scalar result mode.
7109 2018-03-21 Martin Liska <mliska@suse.cz>
7112 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7113 not intended return statement.
7115 2018-03-21 Martin Liska <mliska@suse.cz>
7118 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7119 (chkp_find_bound_slots_1): Limit number of iterations.
7121 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
7124 * Minor grammar fixes for x86 options.
7126 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7129 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7130 holding REG_CFA_RESTORE notes, instead turn them into a USE.
7132 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
7135 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7136 (altivec_lvx_<mode>_1op): Likewise.
7137 (altivec_stvx_<mode>_2op): Likewise.
7138 (altivec_stvx_<mode>_1op): Likewise.
7139 (altivec_lvx_<VM2:mode>): New define_expand.
7140 (altivec_stvx_<VM2:mode>): Likewise.
7141 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7142 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7143 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7144 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7145 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7146 (rs6000_gen_lvx): Likewise.
7147 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7148 (altivec_expand_stv_builtin): Likewise.
7149 (altivec_expand_builtin): Likewise.
7150 * config/rs6000/vector.md: Likewise.
7152 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7155 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7158 2018-03-20 Richard Biener <rguenther@suse.de>
7161 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
7162 sign-conversions as zero, fall back to standard scalar_stmt
7165 2018-03-20 Martin Liska <mliska@suse.cz>
7168 * predict.c (rebuild_frequencies): Handle case when we have
7169 PROFILE_ABSENT, but flag_guess_branch_prob is false.
7171 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7174 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
7175 flag_section_anchors.
7176 * varasm.c (use_blocks_for_decl_p): Remove hack for
7177 dw2_force_const_mem.
7180 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
7182 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
7183 be created, use lowpart_subreg of operands[0] rather than operands[0]
7185 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
7186 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
7187 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
7188 and n constraint instead of aarch64_shift_imm_di and Usd.
7189 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
7190 (*aarch64_<optab>_reg_minus<mode>3): ... this.
7192 2018-03-20 Sudakshina Das <sudi.das@arm.com>
7195 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
7196 to favor GPR over NEON registers.
7197 (<shift>di3_neon): Likewise.
7199 2018-03-20 Tom de Vries <tom@codesourcery.com>
7202 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
7203 (nvptx_process_pars): Emit bar.sync asap and alap.
7205 2018-03-20 Tom de Vries <tom@codesourcery.com>
7208 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
7209 seen_label if seen_label is already set.
7211 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7214 * config/i386/i386.c (fold_builtin_cpu): For features above 31
7215 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
7216 Use 1U instead of 1. Formatting fixes.
7219 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
7220 instead of TREE_TYPE (s1) for the return value.
7222 2018-03-19 Jakub Jelinek <jakub@redhat.com>
7224 PR tree-optimization/84946
7225 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
7226 bitsize + bitsize in poly_uint64 rather than poly_int64.
7229 * dwarf2asm.c: Include fold-const.c.
7230 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
7231 of decl rather than decl itself.
7233 PR rtl-optimization/84643
7234 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
7236 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
7239 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
7240 calling assemble_variable.
7242 2018-03-19 Sudakshina Das <sudi.das@arm.com>
7245 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
7246 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
7248 2018-03-19 Jim Wilson <jimw@sifive.com>
7251 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
7252 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
7253 (riscv_first_stack_step): Likewise.
7254 (riscv_option_override): Use STACK_BOUNDARY instead of
7256 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
7258 (BIGGEST_ALIGNMENT): Set to 128.
7259 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
7260 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
7263 2018-03-19 Richard Biener <rguenther@suse.de>
7265 PR tree-optimization/84933
7266 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
7267 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
7269 2018-03-19 Richard Biener <rguenther@suse.de>
7271 PR tree-optimization/84859
7272 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
7273 (cond_if_else_store_replacement): Perform sinking operation on
7274 single-store BBs regardless of MAX_STORES_TO_SINK setting.
7275 Generalize what a BB with a single eligible store is.
7277 2018-03-19 Richard Biener <rguenther@suse.de>
7279 PR tree-optimization/84929
7280 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
7281 chrec_is_positive against non-chrec arg.
7283 2018-03-19 Tamar Christina <tamar.christina@arm.com>
7286 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
7288 2018-03-18 Martin Liska <mliska@suse.cz>
7290 PR rtl-optimization/84635
7291 * regrename.c (build_def_use): Use matches_mode only when
7294 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
7296 PR tree-optimization/84913
7297 * tree-vect-loop.c (vectorizable_reduction): Don't try to
7298 vectorize chains of COND_EXPRs.
7300 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7302 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
7304 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7306 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
7308 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7310 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
7312 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7313 Kito Cheng <kito.cheng@gmail.com>
7315 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
7316 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
7317 (nds32_adjust_reg_alloc_order): New function.
7318 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
7320 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
7322 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
7323 nds32_print_operand, nds32_print_operand_address): Use
7324 HOST_WIDE_INT_PRINT_DEC instead.
7326 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7328 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
7330 2018-03-17 Jakub Jelinek <jakub@redhat.com>
7333 * config/i386/i386.c (initial_ix86_tune_features,
7334 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
7336 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
7337 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
7338 rather than 1u << ix86_tune. Formatting fix.
7339 (ix86_option_override_internal): Change ix86_arch_mask from
7340 unsigned int to unsigned HOST_WIDE_INT, initialize to
7341 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
7342 (ix86_function_specific_restore): Likewise.
7344 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7347 * postreload.c (reload_combine_recognize_pattern): Perform
7348 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
7349 truncate_int_for_mode the result for the destination's mode.
7352 * hsa-gen.c (mem_type_for_type): Fix comment typo.
7353 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
7355 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7358 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
7361 * lra-assigns.c (lra_split_hard_reg_for): Don't use
7362 regno_allocno_class_array and sorted_pseudos.
7363 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
7364 insns where regno is used.
7366 2018-03-16 Martin Liska <mliska@suse.cz>
7369 * multiple_target.c (create_dispatcher_calls): Redirect
7370 reference in the symbol table.
7372 2018-03-16 Martin Liska <mliska@suse.cz>
7375 * multiple_target.c (create_dispatcher_calls): Redirect also
7378 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7382 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
7383 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
7384 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
7386 2018-03-16 Julia Koval <julia.koval@intel.com>
7388 * doc/invoke.texi (Skylake Server): Add CLWB.
7389 Cannonlake): Remove CLWB.
7391 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7393 PR tree-optimization/84841
7394 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
7396 (FLOAT_ONE_CONST_TYPE): Define.
7397 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
7398 (sort_by_operand_rank): Put entries with higher constant_type last
7399 rather than first to match comments.
7401 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
7403 * config/nios2/nios2.md (movsi_internal): Fix thinko in
7406 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7409 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
7410 check and use address of target always.
7412 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
7415 * config/i386/i386.c (indirect_thunk_needed): Update comments.
7416 (indirect_thunk_bnd_needed): Likewise.
7417 (indirect_thunks_used): Likewise.
7418 (indirect_thunks_bnd_used): Likewise.
7419 (indirect_return_needed): New.
7420 (indirect_return_bnd_needed): Likewise.
7421 (output_indirect_thunk_function): Add a bool argument for
7423 (output_indirect_thunk_function): Don't generate alias for
7424 function return thunk.
7425 (ix86_code_end): Call output_indirect_thunk_function to generate
7426 function return thunks.
7427 (ix86_output_function_return): Set indirect_return_bnd_needed
7428 and indirect_return_needed instead of indirect_thunk_bnd_needed
7429 and indirect_thunk_needed.
7431 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
7433 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
7434 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
7435 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
7437 2018-03-15 David Malcolm <dmalcolm@redhat.com>
7438 Paul Hua <paul.hua.gm@gmail.com>
7441 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
7443 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
7445 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
7446 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
7449 2018-03-15 Tamar Christina <tamar.christina@arm.com>
7452 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7453 instead of GET_MODE_SIZE when comparing Units.
7455 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7458 * varasm.c (hash_section): Return an unchangeble hash value
7459 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
7460 Return !aarch64_can_use_per_function_literal_pools_p ().
7462 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7465 * optabs.c (emit_conditional_move): Pass address of cmode's copy
7466 rather than address of cmode as last argument to prepare_cmp_insn.
7468 2018-03-15 Julia Koval <julia.koval@intel.com>
7470 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
7471 F_AVX512VNNI, F_AVX512BITALG): New.
7473 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
7476 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
7477 insn for floating-point loads and stores.
7479 2018-03-14 Carl Love <cel@us.ibm.com>
7481 * config/rs6000/rs6000-c.c: Add macro definitions for
7482 ALTIVEC_BUILTIN_VEC_PERMXOR.
7483 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
7484 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
7485 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
7486 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
7488 * config/doc/extend.texi: Add prototypes for vec_permxor.
7490 2018-03-14 David Malcolm <dmalcolm@redhat.com>
7493 * diagnostic-show-locus.c (class layout_point): Convert m_line
7494 from int to linenum_type.
7495 (line_span::comparator): Use linenum "compare" function when
7496 comparing line numbers.
7497 (test_line_span): New function.
7498 (layout_range::contains_point): Convert param "row" from int to
7500 (layout_range::intersects_line_p): Likewise.
7501 (layout::will_show_line_p): Likewise.
7502 (layout::print_source_line): Likewise.
7503 (layout::should_print_annotation_line_p): Likewise.
7504 (layout::print_annotation_line): Likewise.
7505 (layout::print_leading_fixits): Likewise.
7506 (layout::annotation_line_showed_range_p): Likewise.
7507 (struct line_corrections): Likewise for field m_row.
7508 (line_corrections::line_corrections): Likewise for param "row".
7509 (layout::print_trailing_fixits): Likewise.
7510 (layout::get_state_at_point): Likewise.
7511 (layout::get_x_bound_for_row): Likewise.
7512 (layout::print_line): Likewise.
7513 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
7514 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
7515 * input.c (selftest::test_linenum_comparisons): New function.
7516 (selftest::input_c_tests): Call it.
7517 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
7518 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
7519 * selftest.h (ASSERT_GT): New macro.
7520 (ASSERT_GT_AT): New macro.
7521 (ASSERT_LT): New macro.
7522 (ASSERT_LT_AT): New macro.
7524 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
7526 PR rtl-optimization/84780
7527 * combine.c (distribute_links): Don't make a link based on pc_rtx.
7529 2018-03-14 Martin Liska <mliska@suse.cz>
7531 * tree.c (record_node_allocation_statistics): Use
7532 get_stats_node_kind.
7533 (get_stats_node_kind): New function extracted from
7534 record_node_allocation_statistics.
7535 (free_node): Use get_stats_node_kind.
7537 2018-03-14 Richard Biener <rguenther@suse.de>
7539 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
7540 that the value-set of ANTIC_IN doesn't grow.
7543 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7545 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7546 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7548 2018-03-14 Julia Koval <julia.koval@intel.com>
7550 * config.gcc (icelake-client, icelake-server): New.
7552 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
7553 (initial_ix86_arch_features): Ditto.
7554 (PTA_SKYLAKE): Add SGX.
7555 (PTA_ICELAKE): Remove.
7556 (PTA_ICELAKE_CLIENT): New.
7557 (PTA_ICELAKE_SERVER): New.
7558 (ix86_option_override_internal): Split up icelake on icelake client and
7560 (get_builtin_code_for_version): Ditto.
7561 (fold_builtin_cpu): Ditto.
7562 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
7563 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
7564 * config/i386/i386.h (processor_type): Ditto.
7565 * doc/invoke.texi: Ditto.
7567 2018-03-14 Jakub Jelinek <jakub@redhat.com>
7570 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
7571 INTEGER_CST offset, add it together with bitpos / 8 and
7572 sign extend based on POINTER_SIZE.
7576 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
7579 * config/i386/constraints.md (Yc): New register constraint.
7580 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
7581 Use Yc constraint for alternative 2 of operand 0. Remove
7582 preferred_for_speed attribute.
7584 2018-03-14 Richard Biener <rguenther@suse.de>
7586 PR tree-optimization/84830
7587 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
7588 with the old one to avoid oscillations.
7590 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
7593 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
7595 (assign_by_spills): Return a flag of reload assignment failure.
7596 Do not process the reload assignment failures. Do not spill other
7597 reload pseudos if they has the same reg class. Update n if
7599 (lra_assign): Add a return arg. Set up from the result of
7600 assign_by_spills call.
7601 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7602 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7603 usage_insns if it is not NULL.
7604 (spill_hard_reg_in_range): New function.
7605 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7606 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7607 function prototypes.
7608 (lra_assign): Change prototype.
7609 * lra.c (lra): Add code to deal with fails by splitting hard reg
7612 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
7614 * config/riscv/riscv.opt (mrelax): New option.
7615 * config/riscv/riscv.c (riscv_file_start): Emit ".option
7616 "norelax" when riscv_mrelax is disabled.
7617 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
7619 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7622 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
7623 reassociation for int modes.
7625 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7627 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
7628 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
7630 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
7631 * config/aarch64/aarch64-sve.md
7632 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
7633 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
7634 (*extend<mode><Vwide>2): Rename to...
7635 (aarch64_sve_extend<mode><Vwide>2): ...this.
7636 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
7637 renaming the old pattern to...
7638 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
7640 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
7641 define_expand, renaming the old pattern to...
7642 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
7643 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
7644 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
7645 account when deciding which SVE instruction the optab should use.
7646 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
7648 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7650 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
7651 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
7652 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
7653 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
7654 (tlsdesc_small_<mode>): Turn a define_expand and use
7655 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
7656 (tlsdesc_small_advsimd_<mode>): ...this.
7657 (tlsdesc_small_sve_<mode>): New pattern.
7659 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7661 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
7662 (UNSPEC_UMUL_HIGHPART): New constants.
7663 (MUL_HIGHPART): New int iteraor.
7664 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
7665 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
7667 (*<su>mul<mode>3_highpart): New define_insn.
7669 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7672 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7675 2018-03-13 Martin Liska <mliska@suse.cz>
7678 * (sem_item_optimizer::sem_item_optimizer): Initialize new
7680 (sem_item_optimizer::~sem_item_optimizer): Release it.
7681 (sem_item_optimizer::merge_classes): Register variable aliases.
7682 (sem_item_optimizer::fixup_pt_set): New function.
7683 (sem_item_optimizer::fixup_points_to_sets): Likewise.
7684 * ipa-icf.h: Declare new variables and functions.
7686 2018-03-13 Jakub Jelinek <jakub@redhat.com>
7689 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
7690 integer_pow2p@2 and test integer_pow2p in condition.
7691 (A < 0 ? C : 0): Similarly for @1.
7694 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
7695 characters starting at p contain '\0' character, don't look beyond
7699 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
7700 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
7703 * reg-stack.c (change_stack): Change update_end var from int to
7704 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
7705 also call set_block_for_insn on the newly added insns and rescan.
7708 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
7709 on the last operand.
7712 * tree.c (stabilize_reference_1): Return save_expr (e) for
7713 STATEMENT_LIST even if it doesn't have side-effects.
7715 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
7717 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
7719 2018-03-12 Renlin Li <renlin.li@arm.com>
7721 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
7722 aarch64_output_scalar_simd_mov_immediate.
7724 2018-03-12 Martin Sebor <msebor@redhat.com>
7726 PR tree-optimization/83456
7727 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
7728 for perfectly overlapping calls to memcpy.
7729 (gimple_fold_builtin_memory_chk): Same.
7730 (gimple_fold_builtin_strcpy): Handle no-warning.
7731 (gimple_fold_builtin_stxcpy_chk): Same.
7732 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
7734 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7736 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
7737 parameter. Use it for SFmode.
7738 (rs6000_function_arg_advance_1): Adjust.
7739 (rs6000_function_arg): Adjust.
7740 (rs6000_gimplify_va_arg): Pass false for that new parameter.
7742 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7744 PR rtl-optimization/84169
7745 PR rtl-optimization/84780
7746 * combine.c (can_combine_p): Check for a 2-insn combination whether
7747 the destination register is used between the two insns, too.
7749 2018-03-12 Richard Biener <rguenther@suse.de>
7751 PR tree-optimization/84803
7752 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
7753 for refs DR analysis didn't process.
7755 2018-03-12 Richard Biener <rguenther@suse.de>
7757 PR tree-optimization/84777
7758 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
7759 force-vectorize loops ignore whether we are optimizing for size.
7761 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
7763 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
7764 (TARGET_MD_ASM_ADJUST): Define.
7766 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
7767 Kito Cheng <kito.cheng@gmail.com>
7768 Chung-Ju Wu <jasonwucj@gmail.com>
7770 * config/nds32/nds32.c (nds32_compute_stack_frame,
7771 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
7772 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
7773 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
7774 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
7775 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
7776 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
7777 * config/nds32/nds32.md (prologue, epilogue): Use macro
7778 NDS32_V3PUSH_AVAILABLE_P to do checking.
7780 2018-03-11 Jakub Jelinek <jakub@redhat.com>
7783 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
7784 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
7785 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
7786 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
7787 addition of most attributes on !orig_type_die or the attribute not
7788 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
7790 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7791 Chung-Ju Wu <jasonwucj@gmail.com>
7793 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
7795 * config/nds32/nds32.opt (mvh): New option.
7797 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7798 Chung-Ju Wu <jasonwucj@gmail.com>
7800 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
7802 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
7803 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
7806 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7807 Chung-Ju Wu <jasonwucj@gmail.com>
7809 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
7811 * config/nds32/nds32-multiple.md (strlensi): New pattern.
7812 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
7814 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
7815 Kito Cheng <kito.cheng@gmail.com>
7816 Chung-Ju Wu <jasonwucj@gmail.com>
7818 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
7819 UNSPEC_FFMISM and UNSPEC_FLMISM.
7820 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
7821 for ffb, ffmism and flmism.
7822 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
7823 (unspec_ffmism): Ditto.
7824 (unspec_flmism): Ditto.
7825 (nds32_expand_builtin_impl): Check if string extension is available.
7826 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
7827 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
7829 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
7832 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
7835 * lra-assigns.c (assign_by_spills): Return a flag of reload
7836 assignment failure. Do not process the reload assignment
7837 failures. Do not spill other reload pseudos if they has the same
7839 (lra_assign): Add a return arg. Set up from the result of
7840 assign_by_spills call.
7841 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7842 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7843 usage_insns if it is not NULL.
7844 (spill_hard_reg_in_range): New function.
7845 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7846 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7847 function prototypes.
7848 (lra_assign): Change prototype.
7849 * lra.c (lra): Add code to deal with fails by splitting hard reg
7852 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
7855 * config/i386/i386.opt: Replace Enforcment with Enforcement.
7857 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
7860 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
7861 (dw_val_node): Add val_symbolic_view.
7862 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
7863 (symview_upper_bound): New.
7864 (new_line_info_table): Initialize symviews_since_reset.
7865 (dwarf2out_source_line): Count symviews_since_reset and set
7866 symview_upper_bound.
7867 (dw_val_equal_p): Handle symview.
7868 (add_AT_symview): New.
7869 (print_dw_val): Handle symview.
7870 (attr_checksum, attr_checksum_ordered): Likewise.
7871 (same_dw_val_p, size_of_die): Likewise.
7872 (value_format, output_die): Likewise.
7873 (add_high_low_attributes): Use add_AT_symview for entry_view.
7874 (dwarf2out_finish): Reset symview_upper_bound, clear
7877 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
7880 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
7881 Add strict argument and use it.
7882 (rs6000_split_multireg_move): Update for new strict argument.
7883 (mem_operand_gpr): Disallow all non-offsettable addresses.
7884 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
7886 2018-03-09 Jakub Jelinek <jakub@redhat.com>
7889 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
7890 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
7891 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
7894 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
7895 decl, use remap_type if we want to use the type.
7897 2018-03-09 Martin Sebor <msebor@redhat.com>
7899 PR tree-optimization/84526
7900 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7902 (builtin_access::generic_overlap): Be prepared to handle non-array
7905 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
7907 PR rtl-optimization/84682
7908 * lra-constraints.c (process_address_1): Check is_address flag
7909 for address constraints.
7910 (process_alt_operands): Likewise.
7911 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
7912 preprocess_constraints.
7913 * recog.h (preprocess_constraints): Add oploc parameter.
7915 * recog.c (preprocess_constraints): Test address_operand for
7916 CT_ADDRESS constraints.
7918 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
7921 * lra-assigns.c (assign_by_spills): Return a flag of reload
7922 assignment failure. Do not process the reload assignment
7923 failures. Do not spill other reload pseudos if they has the same
7925 (lra_assign): Add a return arg. Set up from the result of
7926 assign_by_spills call.
7927 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7928 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7929 usage_insns if it is not NULL.
7930 (spill_hard_reg_in_range): New function.
7931 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7932 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7933 function prototypes.
7934 (lra_assign): Change prototype.
7935 * lra.c (lra): Add code to deal with fails by splitting hard reg
7938 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7941 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
7942 Accept complain bool parameter. Only emit errors if it is true.
7943 (arm_parse_cpu_option_name): Likewise.
7944 (arm_target_thumb_only): Adjust callers of the above.
7945 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
7946 prototype to take a default true bool parameter.
7947 (arm_parse_arch_option_name): Likewise.
7949 2018-03-09 David Malcolm <dmalcolm@redhat.com>
7950 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
7954 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
7955 * configure: Regenerate.
7956 * configure.ac ("linker --version-script option"): New.
7957 ("linker soname option"): New.
7959 2018-03-09 Richard Biener <rguenther@suse.de>
7961 PR tree-optimization/84775
7962 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
7963 immediate uses of predicate stmts and mark them modified.
7966 PR tree-optimization/84178
7967 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7969 (version_loop_for_if_conversion): Delay update_ssa call.
7970 (tree_if_conversion): Delay update_ssa until after predicate
7973 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
7976 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
7977 when the function accesses prior frames.
7979 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7982 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
7983 gen_llsym, otherwise call maybe_gen_llsym.
7986 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
7987 has ',' character inside of it.
7989 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7992 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
7993 as clobbering CC_REGNUM.
7995 2018-03-08 Richard Biener <rguenther@suse.de>
7998 * tree-scalar-evolution.c: Include tree-into-ssa.h.
7999 (follow_copies_to_constant): Do not follow SSA names registered
8002 2018-03-08 Richard Biener <rguenther@suse.de>
8004 PR tree-optimization/84178
8005 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8007 (version_loop_for_if_conversion): Delay update_ssa call.
8008 (tree_if_conversion): Delay update_ssa until after predicate
8011 2018-03-08 David Malcolm <dmalcolm@redhat.com>
8013 PR tree-optimization/84178
8014 * tree-if-conv.c (release_bb_predicate): Remove the
8015 the assertion that the stmts have NULL use_ops.
8016 Discard the statements, asserting that they haven't
8017 yet been added to a BB.
8019 2018-03-08 Richard Biener <rguenther@suse.de>
8021 PR tree-optimization/84746
8022 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8023 (phi_translate): Pass in destination ANTIC_OUT set.
8024 (phi_translate_1): Likewise. For a simplified result lookup
8025 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8026 (phi_translate_set): Adjust.
8027 (do_pre_regular_insertion): Likewise.
8028 (do_pre_partial_partial_insertion): Likewise.
8030 2018-03-08 Martin Liska <mliska@suse.cz>
8032 PR gcov-profile/84735
8033 * doc/gcov.texi: Document usage of profile files.
8034 * gcov-io.h: Document changes in the format.
8036 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
8040 * dwarf2out.c (struct dw_line_info_table): Update comments for
8042 (FORCE_RESET_NEXT_VIEW): New.
8043 (FORCE_RESETTING_VIEW_P): New.
8044 (RESETTING_VIEW_P): Check for -1 too.
8045 (ZERO_VIEW_P): Likewise.
8046 (new_line_info_table): Force-reset next view.
8047 (dwarf2out_begin_function): Likewise.
8048 (dwarf2out_source_line): Simplify zero_view_p initialization.
8049 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8050 view directly. Omit view when omitting .loc at line 0.
8052 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8054 PR tree-optimization/84740
8055 * tree-switch-conversion.c (process_switch): Call build_constructors
8056 only if info.phi_count is non-zero.
8058 PR tree-optimization/84739
8059 * tree-tailcall.c (find_tail_calls): Check call arguments against
8060 DECL_ARGUMENTS (current_function_decl) rather than
8061 DECL_ARGUMENTS (func) when checking for tail recursion.
8063 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8065 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8066 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8067 Volker Reichelt's entry and add entries for people that perform
8068 GCC fuzzy testing and report numerous bugs.
8070 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
8073 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8074 readonly data in sdata, if that is disabled.
8075 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8076 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8077 -mreadonly-in-sdata option.
8079 2018-03-07 Martin Sebor <msebor@redhat.com>
8081 PR tree-optimization/84468
8082 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8083 basic block when looking for nul assignment.
8085 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
8088 * except.h (output_function_exception_table): Adjust prototype.
8089 * except.c (output_function_exception_table): Remove FNNAME parameter
8090 and add SECTION parameter. Ouput one part of the table at a time.
8091 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8092 the first part of the exception table and emit unwind directives.
8093 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8094 (i386_pe_seh_cold_init): Likewise.
8095 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8096 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8097 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8098 (ix86_output_call_insn): Emit a nop in one more case for SEH.
8099 * config/i386/winnt.c: Include except.h.
8100 (struct seh_frame_state): Add reg_offset, after_prologue and
8101 in_cold_section fields.
8102 (i386_pe_seh_end_prologue): Set seh->after_prologue.
8103 (i386_pe_seh_cold_init): New function.
8104 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8105 to seh->in_cold_section.
8106 (seh_emit_push): Record the offset of the push.
8107 (seh_emit_save): Record the offet of the save.
8108 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8109 Test seh->after_prologue to disregard the epilogue.
8110 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8111 (i386_pe_end_cold_function): New function.
8113 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8116 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8117 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8120 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8122 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8123 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8126 * multiple_target.c: Include tree-inline.h and intl.h.
8127 (expand_target_clones): Diagnose and fail if node->definition and
8128 !tree_versionable_function_p (node->decl).
8130 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
8132 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8134 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8135 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8136 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8138 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8141 * combine.c (try_combine): Use reg_or_subregno instead of handling
8142 just paradoxical SUBREGs and REGs.
8144 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
8146 * config/arc/arc.c (arc_finalize_pic): Remove function.
8147 (arc_must_save_register): We use single base PIC register, remove
8148 checks to save/restore the PIC register.
8149 (arc_expand_prologue): Likewise.
8150 * config/arc/arc-protos.h (arc_set_default_type_attributes):
8152 (arc_verify_short): Likewise.
8153 (arc_attr_type): Likewise.
8154 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8155 (walk_stores): Likewise.
8156 (arc_address_cost): Make it static.
8157 (arc_verify_short): Likewise.
8158 (branch_dest): Likewise.
8159 (arc_attr_type): Likewise.
8160 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
8161 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
8162 (arc_final_prescan_insn): Remove inserting the nops due to
8163 hardware hazards. It is done in reorg step.
8164 (insn_length_variant_t): Remove.
8165 (insn_length_parameters_t): Likewise.
8166 (arc_insn_length_parameters): Likewise.
8167 (arc_get_insn_variants): Likewise.
8168 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
8170 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8173 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
8176 PR tree-optimization/84687
8177 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
8179 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8181 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8183 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
8184 Rename to ppc_speculation_barrier.
8185 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
8186 __builtin_ppc_speculation_barrier.
8188 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8191 * combine.c (combine_simplify_rtx): Don't try to simplify if
8192 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
8195 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
8197 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
8198 to 32 bytes when compiling for POWER9.
8200 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8203 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
8204 regparm >= 3 with no arg reg available also for calls with
8205 flag_force_indirect_call. Pass decl to ix86_function_regparm.
8208 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
8210 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
8212 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
8215 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
8217 2018-03-05 Richard Biener <rguenther@suse.de>
8219 PR tree-optimization/84486
8220 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
8221 When inserting a __builtin_assume_aligned call set the LHS
8222 SSA name alignment info accordingly.
8224 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
8226 PR tree-optimization/84114
8227 * config/aarch64/aarch64.c (aarch64_reassociation_width)
8228 Avoid reassociation of FLOAT_MODE addition.
8230 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
8232 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
8233 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
8234 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
8235 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
8236 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
8237 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
8238 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
8240 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
8241 __builtin_ia32_wbinvd): New builtins.
8242 (SPECIAL_ARGS2): New.
8243 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
8244 (SPECIAL_ARGS2): New.
8245 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
8246 (ix86_valid_target_attribute_inner_p): Ditto.
8247 (ix86_init_mmx_sse_builtins): Add special_args2.
8248 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
8249 TARGET_WBNOINVD_P): New.
8250 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
8251 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
8252 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
8253 * config/i386/immintrin.h (_wbinvd): New intrinsic.
8254 * config/i386/pconfigintrin.h: New file.
8255 * config/i386/wbnoinvdintrin.h: Ditto.
8256 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
8258 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
8260 2018-03-05 Richard Biener <rguenther@suse.de>
8262 PR tree-optimization/84670
8263 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8265 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8266 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8267 (compute_antic_aux): Only assert the number of values in ANTIC_IN
8268 doesn't grow if all successors (recursively) were visited at least
8271 2018-03-05 Richard Biener <rguenther@suse.de>
8273 PR tree-optimization/84650
8274 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
8275 if executed in the loop pipeline.
8277 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
8279 * doc/configfiles.texi (Configuration Files): Move info about
8280 conditionalizing $target-protos.h to...
8281 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
8282 differs from $target-protos.h.
8284 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
8285 Chung-Ju Wu <jasonwucj@gmail.com>
8287 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
8288 * config/nds32/nds32-multiple.md (setmemsi): Define.
8289 * config/nds32/nds32-memory-manipulation.c
8290 (nds32_gen_dup_4_byte_to_word_value): New.
8291 (emit_setmem_word_loop): New.
8292 (emit_setmem_byte_loop): New.
8293 (nds32_expand_setmem_loop): New.
8294 (nds32_expand_setmem_loop_v3m): New.
8295 (nds32_expand_setmem_unroll): New.
8296 (nds32_expand_setmem): New.
8298 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8299 Chung-Ju Wu <jasonwucj@gmail.com>
8301 * config/nds32/nds32-memory-manipulation.c
8302 (nds32_emit_load_store): New.
8303 (nds32_emit_post_inc_load_store): New.
8304 (nds32_emit_mem_move): New.
8305 (nds32_emit_mem_move_block): New.
8306 (nds32_expand_movmemsi_loop_unknown_size): New.
8307 (nds32_expand_movmemsi_loop_known_size): New.
8308 (nds32_expand_movmemsi_loop): New.
8309 (nds32_expand_movmemsi_unroll): New.
8310 (nds32_expand_movmemqi): Rename ...
8311 (nds32_expand_movmemsi): ... to this.
8312 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
8313 (movmemsi): ... to this.
8314 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
8315 (nds32_expand_movmemsi): ... to this.
8317 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8318 Monk Chiang <sh.chiang04@gmail.com>
8319 Chung-Ju Wu <jasonwucj@gmail.com>
8321 * config/nds32/nds32-protos.h
8322 (nds32_expand_load_multiple): New arguments.
8323 (nds32_expand_store_multiple): Ditto.
8324 (nds32_valid_multiple_load_store): Rename ...
8325 (nds32_valid_multiple_load_store_p): ... to this.
8326 * config/nds32/nds32-memory-manipulation.c
8327 (nds32_expand_load_multiple): Refine implementation.
8328 (nds32_expand_store_multiple): Ditto.
8329 * config/nds32/nds32-multiple.md
8330 (load_multiple): Update nds32_expand_load_multiple interface.
8331 (store_multiple): Update nds32_expand_store_multiple interface.
8332 * config/nds32/nds32-predicates.c
8333 (nds32_valid_multiple_load_store): Rename ...
8334 (nds32_valid_multiple_load_store_p): ... to this and refine
8336 * config/nds32/predicates.md
8337 (nds32_load_multiple_and_update_address_operation): New predicate.
8338 (nds32_store_multiple_and_update_address_operation): New predicate.
8340 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8341 Chung-Ju Wu <jasonwucj@gmail.com>
8343 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
8344 (combo): New attribute.
8345 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
8347 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
8349 * config/nds32/nds32.opt: Change -mcmodel= default value.
8351 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
8352 Monk Chiang <sh.chiang04@gmail.com>
8353 Chung-Ju Wu <jasonwucj@gmail.com>
8355 * config/nds32/constants.md (unspec_element): New enum.
8356 * config/nds32/constraints.md (Umw): New constraint.
8357 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
8358 * config/nds32/nds32-intrinsic.md: Likewise.
8359 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
8360 (nds32_valid_smw_lwm_base_p): New.
8361 (nds32_output_smw_single_word): New.
8362 (nds32_output_lmw_single_word): New.
8363 (nds32_expand_unaligned_load): New.
8364 (nds32_expand_unaligned_store): New.
8365 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
8366 (nds32_output_smw_single_word): Declare.
8367 (nds32_output_lmw_single_word): Declare.
8368 (nds32_expand_unaligned_load): Declare.
8369 (nds32_expand_unaligned_store): Declare.
8370 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
8371 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
8372 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
8373 NDS32_BUILTIN_UASTORE_DW.
8374 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
8377 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
8378 Kito Cheng <kito.cheng@gmail.com>
8379 Chung-Ju Wu <jasonwucj@gmail.com>
8381 * config/nds32/nds32-intrinsic.c
8382 (nds32_expand_builtin_null_ftype_reg): Delete.
8383 (nds32_expand_builtin_reg_ftype_imm): Ditto.
8384 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
8385 (nds32_read_argument): New.
8386 (nds32_legitimize_target): Ditto.
8387 (nds32_legitimize_argument): Ditto.
8388 (nds32_check_constant_argument): Ditto.
8389 (nds32_expand_unop_builtin): Ditto.
8390 (nds32_expand_unopimm_builtin): Ditto.
8391 (nds32_expand_binop_builtin): Ditto.
8392 (nds32_builtin_decl_impl): Ditto.
8393 (builtin_description): Ditto.
8394 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
8395 (nds32_init_builtins_impl): Ditto.
8396 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
8397 (nds32_builtin_decl): New.
8398 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
8399 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
8401 2018-03-02 Jeff Law <law@redhat.com>
8403 * reorg.c (stop_search_p): Handle DEBUG_INSN.
8404 (redundant_insn, fill_simple_delay_slots): Likewise.
8405 (fill_slots_from_thread): Likewise.
8406 * resource.c (mark_referenced_resources): Likewise.
8407 (mark_set_resources, find_dead_or_set_registers): Likewise.
8409 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8411 * substring-locations.h (format_warning_va): Formatting fix for
8413 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
8415 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
8416 * substring-locations.c: Include intl.h.
8417 (format_warning_va): Turned into small wrapper around
8418 format_warning_n_va, renamed to ...
8419 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
8420 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
8422 (format_warning_at_substring_n): New function.
8423 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
8424 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
8425 format_warning_at_substring with just a shorter name instead of
8426 const function pointer.
8427 (fmtwarn_n): New function.
8428 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
8429 appropriate, get rid of all the fmtstr temporaries, move conditionals
8430 with G_() wrapped string literals directly into fmtwarn arguments,
8431 cast dir.len to (int), formatting fixes.
8433 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
8435 * doc/invoke.texi: Remove "Cilk Plus" references.
8437 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8438 Richard Biener <rguenther@suse.de>
8441 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
8442 for error or warning attributes if CALL_FROM_THUNK_P is set.
8445 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8448 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
8449 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
8452 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
8453 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
8455 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
8456 instead of -1U in last predictors element's probability member.
8458 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8461 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8462 arguments if they are comparable.
8464 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8466 PR tree-optimization/84634
8467 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
8468 masks and masked_loop_p with a single loop_masks, making sure it's
8469 null for bb vectorization.
8471 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8473 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8474 (vect_analyze_data_ref_access): Use loop->safe_len rather than
8475 loop->force_vectorize to check whether there is no alias.
8477 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8480 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
8482 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
8484 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
8485 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
8486 instead of a loop around prev_real_insn.
8487 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
8491 * config/i386/i386.c (ix86_print_operand): Use conditional
8492 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
8495 2018-03-02 Richard Biener <rguenther@suse.de>
8497 PR tree-optimization/84427
8498 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
8499 (bitmap_set_subtract_values): Rewrite to handle multiple
8502 (prune_clobbered_mems): Likewise.
8503 (phi_translate): Take edge instead of pred/phiblock.
8504 (phi_translate_1): Likewise.
8505 (phi_translate_set): Likewise. Insert all translated
8506 exprs for a value into the set, keeping possibly multiple
8507 expressions per value.
8508 (compute_antic_aux): Adjust for phi_translate changes.
8509 When intersecting union the expressions and prune those
8510 not in the final value set, keeping possibly multiple
8511 expressions per value. Do not use value-insertion
8512 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
8513 all expressions. Add verification that the value-sets
8514 only shrink during iteration.
8515 (compute_partial_antic_aux): Adjust for the phi_translate changes.
8516 (do_pre_regular_insertion): Likewise.
8517 (do_pre_partial_partial_insertion): Likewise.
8519 2018-03-02 Richard Biener <rguenther@suse.de>
8522 * config/darwin.c (saved_debug_info_level): New static global.
8523 (darwin_asm_lto_start): Disable debug info generation for LTO out.
8524 (darwin_asm_lto_end): Restore debug info generation settings.
8526 2018-03-01 Martin Liska <mliska@suse.cz>
8529 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
8532 2018-03-01 Richard Biener <rguenther@suse.de>
8535 * dwarf2out.c (gen_variable_die): Properly handle late VLA
8536 type annotation with LTO when debug was disabled at compile-time.
8538 2018-03-01 Matthew Fortune <mfortune@gmail.com>
8540 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
8542 (mips_final_postscan_insn): Likewise.
8544 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
8546 PR rtl-optimization/84528
8547 * alias.c (init_alias_target): Add commentary.
8548 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
8549 a unique base value if the frame pointer is not eliminated
8550 to the stack pointer.
8552 2018-03-01 Tom de Vries <tom@codesourcery.com>
8554 PR rtl-optimization/83327
8555 * lra-int.h (hard_regs_spilled_into): Declare.
8556 * lra.c (hard_regs_spilled_into): Define.
8557 (init_reg_info): Init hard_regs_spilled_into.
8558 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
8559 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
8560 (process_bb_lives): Handle hard_regs_spilled_into.
8561 (lra_create_live_ranges_1): Before doing liveness propagation, clear
8562 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
8564 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
8566 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
8567 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
8568 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
8569 * config/rs6000/aix72.h: New file.
8571 2018-02-28 Jakub Jelinek <jakub@redhat.com>
8573 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
8574 instead of warning_at with conditional singular and plural messages
8578 * stor-layout.c (update_alignment_for_field): For
8579 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
8580 && !DECL_PACKED (field), do the alignment update, just use
8581 only desired_align instead of MAX (type_align, desired_align)
8583 (place_field): Don't do known_align < desired_align handling
8584 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
8585 is non-NULL, instead do it after rli->prev_field handling and
8586 only if not within a bitfield word. For DECL_PACKED (field)
8587 use type_align of BITS_PER_UNIT.
8589 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
8591 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
8592 superfluous parentheses and trailing spaces.
8594 2018-02-28 Richard Biener <rguenther@suse.de>
8596 PR tree-optimization/84584
8597 * graphite-scop-detection.c (scop_detection::add_scop): Discard
8598 SCoPs with fake exit edge.
8600 2018-02-28 Martin Liska <mliska@suse.cz>
8603 * timevar.c (timer::print): Fix format to properly print 100%
8606 2018-02-28 Richard Biener <rguenther@suse.de>
8609 * genmatch.c (capture_info::walk_match): Do not mark
8610 captured expressions without operands as expr_p given
8611 they act more like predicates and should be subject to
8612 "lost tail" side-effect preserving.
8614 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
8616 PR rtl-optimization/81611
8617 * auto-inc-dec.c (attempt_change): Move dead note from
8618 mem_insn if it's the next use of regno
8619 (find_address): Take address use of reg holding
8620 non-incremented value. Add parm to limit search to the named
8622 (merge_in_block): Attempt to use a mem insn that is the next
8623 use of the original regno.
8625 2018-02-27 Martin Sebor <msebor@redhat.com>
8628 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
8629 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
8631 2018-02-27 Martin Sebor <msebor@redhat.com>
8633 PR translation/84207
8634 * diagnostic-core.h (warning_n, error_n, inform_n): Change
8635 n argument to unsigned HOST_WIDE_INT.
8636 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
8637 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
8638 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
8639 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
8641 2018-02-27 Richard Biener <rguenther@suse.de>
8643 PR tree-optimization/84512
8644 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
8645 Do not use the estimate returned from record_stmt_cost for
8646 the scalar iteration cost but sum properly using add_stmt_cost.
8648 2018-02-27 Richard Biener <rguenther@suse.de>
8650 PR tree-optimization/84466
8651 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8652 Adjust last change to less strictly validate use operands.
8654 2018-02-27 Martin Liska <mliska@suse.cz>
8656 PR gcov-profile/84548
8657 * gcov.c (process_file): Allow partial overlap and consider it
8658 also as group functions.
8659 (output_lines): Properly calculate range of lines for a group.
8661 2018-02-27 Martin Liska <mliska@suse.cz>
8663 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
8664 'ggc' suffixes. Change first column width.
8665 (timer::print): Fix formatting of the column.
8667 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
8669 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
8670 preserve inline entry blocks for the sake of debug inline
8671 entry point markers alone.
8672 (remove_unused_locals): Suggest in comments a better place to
8673 force the preservation of inline entry blocks that are
8674 otherwise unused, but do not preserve them.
8676 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8678 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
8680 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8683 * config/i386/constraints.md (Bs): Replace
8684 ix86_indirect_branch_register with
8685 TARGET_INDIRECT_BRANCH_REGISTER.
8687 * config/i386/i386.md (indirect_jump): Likewise.
8688 (tablejump): Likewise.
8689 (*sibcall_memory): Likewise.
8690 (*sibcall_value_memory): Likewise.
8691 Peepholes of indirect call and jump via memory: Likewise.
8692 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
8693 (*sibcall_value_GOT_32): Likewise.
8694 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8695 (GOT_memory_operand): Likewise.
8696 (call_insn_operand): Likewise.
8697 (sibcall_insn_operand): Likewise.
8698 (GOT32_symbol_operand): Likewise.
8699 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
8701 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8703 PR rtl-optimization/83496
8704 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
8705 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
8706 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
8707 redundant insn, if any.
8708 (relax_delay_slots): Likewise.
8709 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
8711 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
8713 PR tree-optimization/83965
8714 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
8715 that grouped statements are part of a reduction chain. Return
8716 true if the statement is not marked as a reduction itself but
8718 (vect_recog_dot_prod_pattern): Don't check whether the statement
8719 is part of a group here.
8720 (vect_recog_sad_pattern): Likewise.
8721 (vect_recog_widen_sum_pattern): Likewise.
8723 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8726 * final.c (rest_of_clean_state): Also look for calls inside sequences.
8728 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8731 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
8733 (ix86_output_indirect_function_return): New prototype.
8734 (ix86_split_simple_return_pop_internal): Likewise.
8735 * config/i386/i386.c (indirect_return_via_cx): New.
8736 (indirect_return_via_cx_bnd): Likewise.
8737 (indirect_thunk_name): Handle return va CX_REG.
8738 (output_indirect_thunk_function): Create alias for
8739 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
8740 (ix86_output_indirect_jmp): Remove the bool argument.
8741 (ix86_output_indirect_function_return): New function.
8742 (ix86_split_simple_return_pop_internal): Likewise.
8743 * config/i386/i386.md (*indirect_jump): Don't pass false
8744 to ix86_output_indirect_jmp.
8745 (*tablejump_1): Likewise.
8746 (simple_return_pop_internal): Change it to define_insn_and_split.
8747 Call ix86_split_simple_return_pop_internal to split it for
8749 (simple_return_indirect_internal): Call
8750 ix86_output_indirect_function_return instead of
8751 ix86_output_indirect_jmp.
8753 2018-02-26 Jakub Jelinek <jakub@redhat.com>
8756 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
8757 memset and value initialization afterwards.
8759 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
8761 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
8763 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8766 * common/config/aarch64/aarch64-common.c
8767 (aarch_option_optimization_table[]): Switch
8768 off fomit-frame-pointer
8770 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
8771 Chung-Ju Wu <jasonwucj@gmail.com>
8773 * config/nds32/nds32-multiple.md (load_multiple): Disallow
8775 (store_multiple): Ditto.
8777 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
8779 * config.gcc: Add --with-cpu support for nds32 target.
8780 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
8781 * config/nds32/nds32.opt: Add -mcpu= option.
8783 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
8785 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
8786 isel=yes): Warn for these deprecated options.
8788 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
8790 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
8791 ISA_2_5_MASKS_EMBEDDED.
8793 2018-02-23 Jakub Jelinek <jakub@redhat.com>
8795 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
8796 p->max as pointers rather than using iterative_hash_expr.
8798 2018-02-23 Carl Love <cel@us.ibm.com>
8800 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
8801 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
8803 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
8804 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
8805 P8V_BUILTIN_VEC_VUNSIGNED2.
8807 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
8810 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
8811 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
8813 * lra-constraints.c (curr_insn_transform): Set up
8814 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
8816 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
8817 * lra-eliminations.c (spill_pseudos): Ditto.
8818 (process_insn_for_elimination): Ditto.
8819 * lra-lives.c (reg_early_clobber_p): Use the new macros.
8820 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
8821 LRA_NON_CLOBBERED_ALT.
8823 2018-02-22 Martin Sebor <msebor@redhat.com>
8825 PR tree-optimization/84480
8826 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
8827 to maybe_diag_stxncpy_trunc. Call it.
8828 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
8829 from gimple_fold_builtin_strcpy. Print inlining stack.
8830 (handle_builtin_stxncpy): Print inlining stack.
8831 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
8833 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
8836 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
8837 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
8838 and -fcheck-pointer-bounds are used together.
8839 (indirect_thunk_prefix): New enum.
8840 (indirect_thunk_need_prefix): New function.
8841 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
8842 "_nt" instead of "_bnd" for NOTRACK prefix.
8843 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
8844 (output_indirect_thunk_function): Likewise.
8846 (ix86_code_end): Update output_indirect_thunk_function calls.
8847 (ix86_output_indirect_branch_via_reg): Replace
8848 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
8849 (ix86_output_indirect_branch_via_push): Likewise.
8850 (ix86_output_function_return): Likewise.
8851 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
8852 incompatible with -fcf-protection=branch and
8853 -fcheck-pointer-bounds.
8855 2018-02-22 Steve Ellcey <sellcey@cavium.com>
8858 * config/aarch64/aarch64.c (aarch64_print_address_internal):
8859 Change gcc_assert call to output_operand_lossage.
8861 2018-02-22 Steve Ellcey <sellcey@cavium.com>
8863 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
8865 2018-02-22 DJ Delorie <dj@redhat.com>
8866 Sebastian Perta <sebastian.perta@renesas.com>
8867 Oleg Endo <olegendo@gcc.gnu.org>
8869 * config/rx/rx.c (rx_rtx_costs): New function.
8870 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
8872 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
8874 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
8876 2018-02-22 Martin Liska <mliska@suse.cz>
8879 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
8880 Add "native" as a possible value.
8882 2018-02-22 Martin Liska <mliska@suse.cz>
8885 * config/i386/i386.c (ix86_option_override_internal):
8886 Add "native" as a possible value for -march and -mtune.
8888 2018-02-22 Jakub Jelinek <jakub@redhat.com>
8891 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
8892 to all type variants.
8894 PR tree-optimization/84503
8895 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
8896 width as info->bitpos + info->bitsize - start.
8897 (merged_store_group::merge_overlapping): Simplify width computation.
8898 (check_no_overlap): New function.
8899 (imm_store_chain_info::try_coalesce_bswap): Compute expected
8900 start + width and last_order of the group, fail if check_no_overlap
8902 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
8903 to group if check_no_overlap fails.
8905 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8907 * config/rs6000/altivec.md: Delete contraint arguments to
8908 define_expand, define_split, and define_peephole2, and in
8909 define_insn_and_split if always unused.
8910 * config/rs6000/darwin.md: Ditto.
8911 * config/rs6000/dfp.md: Ditto.
8912 * config/rs6000/rs6000.md: Ditto.
8913 * config/rs6000/sync.md: Ditto.
8914 * config/rs6000/vector.md: Ditto.
8915 * config/rs6000/vsx.md: Ditto.
8917 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8919 * config/rs6000/altivec.md: Write output control strings as braced
8920 blocks instead of double-quoted strings.
8921 * config/rs6000/darwin.md: Ditto.
8922 * config/rs6000/rs6000.md: Ditto.
8923 * config/rs6000/vector.md: Ditto.
8924 * config/rs6000/vsx.md: Ditto.
8926 2018-02-21 Jason Merrill <jason@redhat.com>
8928 PR c++/84314 - ICE with templates and fastcall attribute.
8929 * attribs.c (build_type_attribute_qual_variant): Remove assert.
8931 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
8933 * ipa-cp.c (determine_versionability): Fix comment typos.
8935 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
8938 * ipa-cp.c (determine_versionability): Do not version functions caling
8941 2018-02-21 Martin Liska <mliska@suse.cz>
8944 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
8945 Add "native" as a possible value.
8946 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
8947 the macro when native cpu detection is available.
8949 2018-02-21 Martin Liska <mliska@suse.cz>
8952 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
8953 Add "native" as a possible value.
8954 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
8955 when native cpu detection is available.
8957 2018-02-21 Jakub Jelinek <jakub@redhat.com>
8958 Martin Sebor <msebor@redhat.com>
8960 PR tree-optimization/84478
8961 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
8963 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
8964 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
8965 support which is conservatively correct, for 2 only stay conservative
8966 for maxlen. Formatting and comment capitalization fixes. Add STRICT
8967 argument to the 2 argument get_range_strlen, adjust 6 arg
8968 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
8970 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
8971 (gimple_fold_builtin_strlen): Pass true as last argument to
8974 2018-02-20 Martin Sebor <msebor@redhat.com>
8977 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
8978 (builtin_memref::set_base_and_offset): Same. Handle inner references.
8979 (builtin_memref::builtin_memref): Factor out parts into
8980 set_base_and_offset and call it.
8982 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8985 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
8986 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
8987 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
8988 search at the associated MODE_INT.
8990 2018-02-20 Jeff Law <law@redhat.com>
8993 PR tree-optimization/81592
8995 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
8996 for range data rather than using global data.
8997 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
8998 range data rather than using global data.
8999 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9000 pass it to children as needed.
9001 (struct directive::fmtresult): Similarly.
9002 (struct directive::set_width): Similarly.
9003 (struct directive::set_precision): Similarly.
9004 (format_integer, format_directive, parse_directive): Similarly.
9005 (format_none): Accept unnamed vr_values parameter.
9006 (format_percent, format_floating, format_character): Similarly.
9007 (format_string, format_plain): Similarly.
9008 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9009 the EVRP range analyzer for range data rather than using global data.
9010 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9011 gimple-ssa-evrp-analyze.h
9012 (class sprintf_dom_walker): Add after_dom_children member function.
9013 Add evrp_range_analyzer member.
9014 (sprintf_dom_walker::before_dom_children): Call into the EVRP
9015 range analyzer as needed.
9016 (sprintf_dom_walker::after_dom_children): New member function.
9017 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9019 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9020 (evrp_range_analyzer::pop_to_marker): Likewise.
9022 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9024 PR tree-optimization/84419
9025 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9026 with the required type if its current type is compatible but
9029 2018-02-20 Jakub Jelinek <jakub@redhat.com>
9032 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9033 after vectorization.
9035 2018-02-20 Martin Liska <mliska@suse.cz>
9038 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9039 possible values if we don't have a hint.
9041 2018-02-20 Martin Liska <mliska@suse.cz>
9045 * final.c (shorten_branches): Build align_tab array with one
9047 * opts.c (finish_options): Add alignment option limit check.
9048 (MAX_CODE_ALIGN): Likewise.
9049 (MAX_CODE_ALIGN_VALUE): Likewise.
9050 * doc/invoke.texi: Document maximum allowed option value for
9051 all -falign-* options.
9053 2018-02-19 Jakub Jelinek <jakub@redhat.com>
9056 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9057 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9058 * var-tracking.c (emit_note_insn_var_location): Remove all references
9059 to NOTE_INSN_CALL_ARG_LOCATION.
9060 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9061 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9062 Use copy_rtx_if_shared.
9063 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9064 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9065 (dwarf2out_var_location): Remove handling of
9066 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9068 * final.c (final_scan_insn): Remove all references to
9069 NOTE_INSN_CALL_ARG_LOCATION.
9070 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
9071 before dumping final insns.
9072 * except.c (emit_note_eh_region_end): Remove all references to
9073 NOTE_INSN_CALL_ARG_LOCATION.
9074 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9075 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9076 * config/arc/arc.c (hwloop_optimize): Likewise.
9077 * config/arm/arm.c (create_fix_barrier): Likewise.
9078 * config/s390/s390.c (s390_chunkify_start): Likewise.
9079 * config/sh/sh.c (find_barrier): Likewise.
9080 * config/i386/i386.c (rest_of_insert_endbranch,
9081 ix86_seh_fixup_eh_fallthru): Likewise.
9082 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9083 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9084 * config/frv/frv.c (frv_function_prologue): Likewise.
9085 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
9087 (note_outside_basic_block_p): Remove all references to
9088 NOTE_INSN_CALL_ARG_LOCATION.
9089 * gengtype.c (adjust_field_rtx_def): Likewise.
9090 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9092 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9093 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9096 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9099 PR tree-optimization/84452
9100 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9101 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9104 2018-02-19 Martin Liska <mliska@suse.cz>
9107 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9109 2018-02-19 Martin Liska <mliska@suse.cz>
9110 Richard Sandiford <richard.sandiford@linaro.org>
9112 PR tree-optimization/82491
9113 * gimple-fold.c (get_base_constructor): Make earlier bail out
9116 2018-02-19 Carl Love <cel@us.ibm.com>
9118 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9119 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9121 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9122 P8V_BUILTIN_VEC_NEG.
9124 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
9126 * config/rl78/rl78.md (movdf): New define expand.
9128 2018-02-19 Martin Liska <mliska@suse.cz>
9131 * doc/invoke.texi: Fix typo.
9132 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9134 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
9136 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9137 handle rs6000_single_float and rs6000_double_float specially for
9140 2018-02-16 Jeff Law <law@redhat.com>
9142 * config/rx/rx.c (add_pop_cfi_notes): New function.;
9145 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9148 * ipa-inline.c (inline_small_functions): Fix a typo.
9150 2018-02-16 Nathan Sidwell <nathan@acm.org>
9152 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9154 2018-02-16 Carl Love <cel@us.ibm.com>
9156 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9157 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9158 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9159 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9160 expansion to P8V_BUILTIN_VEC_FLOAT2.
9162 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
9164 PR rtl-optimization/70023
9165 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
9166 src_regno into account.
9168 2018-02-16 Carl Love <cel@us.ibm.com>
9170 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
9171 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
9172 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
9173 * config/rs6000/rs6000.c: Remove case statements for
9174 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
9175 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
9176 and P9V_BUILTIN_VEC_VINSERT4B.
9177 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
9178 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
9179 * config/rs6000/vsx.md:
9180 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
9183 2018-02-16 Carl Love <cel@us.ibm.com>
9185 * config/rs6000/altivec.h: Add builtin names vec_extract4b
9187 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
9189 * config/rs6000/rs6000-c.c: Add the definitions for
9190 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
9191 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
9192 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
9193 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
9194 definition for insert4b and define insn *insert3b_internal.
9195 * doc/extend.texi: Add documentation for vec_extract4b.
9197 2018-02-16 Nathan Sidwell <nathan@acm.org>
9199 * doc/extend.texi (Backwards Compatibility): Mention friend
9200 injection. Note for-scope is deprecated.
9201 * doc/invoke.texi (-ffriend-injection): Deprecate.
9203 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
9205 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
9206 that moved to I2, also allow destinations that are a paradoxical
9207 subreg (instead of a normal reg).
9209 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
9212 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
9215 2018-02-16 Richard Biener <rguenther@suse.de>
9217 PR tree-optimization/84037
9218 PR tree-optimization/84016
9220 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9221 Adjust vec_construct for the fact we need additional higher latency
9222 128bit inserts for AVX256 and AVX512 vector builds.
9223 (ix86_add_stmt_cost): Scale vector construction cost for
9226 2018-02-16 Richard Biener <rguenther@suse.de>
9228 PR tree-optimization/84417
9229 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
9230 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
9231 (non_rewritable_lvalue_p): Likewise, use poly-ints.
9233 2018-02-16 Martin Liska <mliska@suse.cz>
9236 * internal-fn.def (ASAN_CHECK): Set proper flags.
9237 (ASAN_MARK): Likewise.
9239 2018-02-16 Julia Koval <julia.koval@intel.com>
9241 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
9242 from PTA_CANNONLAKE.
9244 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9247 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
9248 Use ++iter rather than iter++ for std::list iterators.
9249 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
9250 defer deleting them until all nodes in the forest are processed. Do
9251 free even leaf nodes. Change to_process into auto_vec.
9254 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
9255 * vec.h (vec_default_construct): Use memset instead of placement new
9256 if BROKEN_VALUE_INITIALIZATION is defined.
9257 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
9258 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
9261 PR rtl-optimization/83723
9262 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
9263 * lra.c (lra_substitute_pseudo): Likewise. If true, use
9264 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
9266 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
9268 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
9270 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
9272 PR rtl-optimization/81443
9273 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
9274 from inner REGs to paradoxical SUBREGs.
9276 2018-02-16 Richard Biener <rguenther@suse.de>
9278 PR tree-optimization/84399
9279 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9280 For operands we can analyze at their definition make sure we can
9281 analyze them at each use as well.
9283 2018-02-16 Richard Biener <rguenther@suse.de>
9285 PR tree-optimization/84190
9286 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
9287 volatile accesses if the decl isn't volatile.
9289 2018-02-15 Jason Merrill <jason@redhat.com>
9291 PR c++/84314 - ICE with templates and fastcall attribute.
9292 * attribs.c (build_type_attribute_qual_variant): Don't clobber
9293 TYPE_CANONICAL on an existing type.
9295 2018-02-15 Jakub Jelinek <jakub@redhat.com>
9297 PR tree-optimization/84383
9298 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
9299 dstoff nor call operand_equal_p if dstbase is NULL.
9301 PR tree-optimization/84334
9302 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
9303 also a CONSTANT_CLASS_P, punt.
9305 2018-02-14 Jim Wilson <jimw@sifive.com>
9307 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
9308 first SMALL_OPERAND check. New local min_second_step. Move assert
9309 to where locals are set. Add TARGET_RVC support.
9310 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
9312 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
9314 * doc/invoke.texi: Correct -Wformat-overflow code sample.
9316 2018-02-14 Martin Sebor <msebor@redhat.com>
9318 PR tree-optimization/83698
9319 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
9320 arrays constrain the offset range to their bounds.
9321 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
9322 (builtin_access::overlap): Avoid setting the size of overlap if it's
9324 (maybe_diag_overlap): Also consider arrays when deciding what values
9325 of offsets to include in diagnostics.
9327 2018-02-14 Martin Sebor <msebor@redhat.com>
9330 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
9331 that correspond to the kind of a declaration.
9333 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
9336 * config/pa/pa.md: Load address of PIC label using the linkage table
9337 if the label is nonlocal.
9339 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
9341 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
9342 warning message if user requests -maltivec=be.
9343 * doc/invoke.texi: Document deprecation of -maltivec=be.
9345 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
9348 * config/rs6000/rs6000-c.c: Update definitions for
9349 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
9350 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
9352 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9355 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
9356 add _get_ssp intrinsics. Remove argument from
9357 __builtin_ia32_rdssp[d|q].
9358 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
9359 * config/i386/i386-builtin.def: Remove argument from
9360 __builtin_ia32_rdssp[d|q].
9361 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
9362 ix86_expand_special_args_builtin for _rdssp[d|q].
9363 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
9364 Clear register before usage.
9365 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
9366 Add documentation for new _get_ssp and _inc_ssp intrinsics.
9368 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
9370 PR tree-optimization/84357
9371 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
9372 operand 1 of an ARRAY_REF too.
9374 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
9377 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
9378 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9380 (set_of_reg): New struct.
9381 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
9382 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
9383 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9385 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
9386 Split into bitclr, bitset, bitinvert patterns if appropriate.
9387 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
9388 use rx_fuse_in_memory_bitop.
9389 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
9390 to named insn, correct maximum insn length.
9392 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
9395 * machmode.def: Define a complex mode for PARTIAL_INT.
9396 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
9398 * doc/rtl.texi: Document CSPImode.
9399 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
9401 (msp430_hard_regno_nregs_with_padding): Likewise.
9403 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
9406 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
9408 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
9410 PR rtl-optimization/84169
9411 * combine.c (try_combine): New variable split_i2i3. Set it to true if
9412 we generated a parallel as new i3 and we split that to new i2 and i3
9413 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
9414 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
9415 those to i2, not i1. Partially rewrite this scan code.
9417 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9420 * stor-layout.c (place_field): For variable length fields, adjust
9421 offset_align afterwards not just based on the field's alignment,
9422 but also on the size.
9425 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
9426 of exps and logs in the use_exp2 case.
9428 2018-02-13 Jeff Law <law@redhat.com>
9430 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
9433 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
9436 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
9440 * common.opt (gas-loc-support, gas-locview-support): New.
9441 (ginline-points, ginternal-reset-location-views): New.
9442 * doc/invoke.texi: Document them. Use @itemx where intended.
9443 (gvariable-location-views): Adjust.
9444 * target.def (reset_location_view): New.
9445 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
9446 (TARGET_RESET_LOCATION_VIEW): New.
9447 * doc/tm.texi: Rebuilt.
9448 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
9449 (dwarf2out_default_as_locview_support): New.
9450 (output_asm_line_debug_info): Use option variables.
9451 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
9452 (output_loc_list): Likewise.
9453 (add_high_low_attributes): Check option variables.
9454 Don't output entry view attribute in strict mode.
9455 (gen_inlined_subroutine_die): Check option variables.
9456 (dwarf2out_inline_entry): Likewise.
9457 (init_sections_and_labels): Likewise.
9458 (dwarf2out_early_finish): Likewise.
9459 (maybe_reset_location_view): New, from...
9460 (dwarf2out_var_location): ... here. Call it.
9461 * debug.h (dwarf2out_default_as_loc_support): Declare.
9462 (dwarf2out_default_as_locview_support): Declare.
9463 * hooks.c (hook_int_rtx_insn_0): New.
9464 * hooks.h (hook_int_rtx_insn_0): Declare.
9465 * toplev.c (process_options): Take -gas-loc-support and
9466 -gas-locview-support from dwarf2out. Enable
9467 -gvariable-location-views by default only with locview
9468 assembler support. Enable -ginternal-reset-location-views by
9469 default only if the target defines the corresponding hook.
9470 Enable -ginline-points by default if location views are
9471 enabled; force it disabled if statement frontiers are
9473 * tree-inline.c (expand_call_inline): Check option variables.
9474 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9476 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
9478 PR tree-optimization/84321
9479 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
9480 handling. Also check whether the anti-range contains any values
9481 that satisfy the mask; switch to a VR_RANGE if not.
9483 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
9486 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
9488 2018-02-13 Martin Jambor <mjambor@suse.cz>
9491 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
9492 of call statements, also set location of a load to a temporary.
9494 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
9496 * config/rl78/rl78.c (add_vector_labels): New function.
9497 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
9498 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
9499 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
9500 which checks that no arguments are passed.
9501 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
9502 * doc/extend.texi: Documentation for the new attribute.
9504 2018-02-13 Andreas Schwab <schwab@suse.de>
9506 * config/riscv/linux.h (CPP_SPEC): Define.
9508 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9511 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
9512 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
9513 OPTION_MASK_ISA_AES as first argument to def_builtin_const
9514 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
9515 instead of OPTION_MASK_ISA_PCLMUL as first argument to
9516 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
9517 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
9518 temporarily for AES and PCLMUL builtins.
9520 PR tree-optimization/84339
9521 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
9522 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
9526 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
9527 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
9528 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
9530 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
9532 * omp-simd-clone.h: New file.
9533 * omp-simd-clone.c: Include omp-simd-clone.h.
9534 (expand_simd_clones): No longer static.
9535 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
9536 cgraph.h and omp-simd-clone.h.
9537 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
9538 (vect_recog_widen_shift_pattern): Formatting fix.
9539 (vect_pattern_recog_1): Don't check optab for calls.
9542 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
9543 operands[2] into a REG before using gen_lowpart on it.
9545 2018-02-12 Jeff Law <law@redhat.com>
9548 * config/sh/sh.c (find_barrier): Consider a sibling call
9551 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
9552 successfully back substituting a reg.
9554 2018-02-12 Richard Biener <rguenther@suse.de>
9556 PR tree-optimization/84037
9557 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
9558 parameter, move visited init to caller.
9559 (vect_slp_analyze_operations): Separate cost from validity
9560 check, initialize visited once for all instances.
9561 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
9563 * tree-vect-stmts.c (vect_model_simple_cost): Make early
9565 (vect_model_promotion_demotion_cost): Likewise.
9566 (vectorizable_bswap): Guard cost modeling with !slp_node
9567 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
9569 (vectorizable_call): Likewise.
9570 (vectorizable_conversion): Likewise.
9571 (vectorizable_assignment): Likewise.
9572 (vectorizable_shift): Likewise.
9573 (vectorizable_operation): Likewise.
9574 (vectorizable_store): Likewise.
9575 (vectorizable_load): Likewise.
9576 (vectorizable_condition): Likewise.
9577 (vectorizable_comparison): Likewise.
9579 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
9582 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
9583 (ASAN_MARK): Fix fnspec to account for return value, change pointer
9584 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
9586 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9589 * params.def (inline-min-speedup): Increase from 8 to 15.
9590 (max-inline-insns-auto): Decrease from 40 to 30.
9591 * ipa-split.c (consider_split): Add some buffer for function to
9592 be considered inlining candidate.
9593 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
9596 2018-02-12 Richard Biener <rguenther@suse.de>
9598 PR tree-optimization/84037
9599 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
9600 matched stmts if we cannot swap the non-matched ones.
9602 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9604 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
9605 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
9606 _mm_maskz_scalef_round_ss): New intrinsics.
9607 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
9608 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
9609 __builtin_ia32_scalefss_round): Remove.
9610 (__builtin_ia32_scalefsd_mask_round,
9611 __builtin_ia32_scalefss_mask_round): New intrinsics.
9612 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
9613 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
9614 ((match_operand:VF_128 2 "<round_nimm_predicate>"
9615 "<round_constraint>")): Changed to ...
9616 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
9617 "<round_scalar_constraint>")): ... this.
9618 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
9619 %0, %1, %2<round_op3>}"): Changed to ...
9620 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
9621 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
9622 %2<round_scalar_mask_op3>}"): ... this.
9623 * config/i386/subst.md (round_scalar_nimm_predicate): New.
9625 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9627 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
9628 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
9629 (_mm_maskz_sqrt_round_ss): New intrinsics.
9630 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
9631 (__builtin_ia32_sqrtsd_mask_round)
9632 (__builtin_ia32_sqrtss_mask_round): New builtins.
9633 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
9634 (__builtin_ia32_sqrtss_round): Remove.
9635 (__builtin_ia32_sqrtsd_mask_round)
9636 (__builtin_ia32_sqrtss_mask_round): New builtins.
9637 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
9638 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
9639 ((match_operand:VF_128 1 "vector_operand"
9640 "xBm,<round_constraint>")): Changed to ...
9641 ((match_operand:VF_128 1 "vector_operand"
9642 "xBm,<round_scalar_constraint>")): ... this.
9643 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
9644 %0, %2, %<iptr>1<round_op3>}): Changed to ...
9645 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
9646 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
9647 %<iptr>1<round_scalar_mask_op3>}): ... this.
9648 ((set_attr "prefix" "<round_prefix>")): Changed to ...
9649 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
9651 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
9654 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
9655 Cast vec_cmpeq result to correct type.
9656 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
9657 Cast vec_cmpgt result to correct type.
9659 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
9661 * final.c (final_scan_insn_1): Renamed from...
9662 (final_scan_insn): ... this. New wrapper, to recover
9663 seen from the outermost call in recursive ones.
9664 * config/sparc/sparc.c (output_return): Drop seen from call.
9665 (output_sibcall): Likewise.
9666 * config/visium/visium.c (output_branch): Likewise.
9668 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
9670 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
9673 2018-02-10 Alan Modra <amodra@gmail.com>
9676 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
9677 Specify LR as an input.
9679 2018-02-10 Jakub Jelinek <jakub@redhat.com>
9682 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
9683 remove_member_access_dummy_vars): New functions.
9684 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
9685 lower_omp_1, execute_lower_omp): Use them.
9687 PR rtl-optimization/84308
9688 * shrink-wrap.c (spread_components): Release todo vector.
9690 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
9692 PR rtl-optimization/57193
9693 * ira-color.c (struct allocno_color_data): Add member
9694 conflict_allocno_hard_prefs.
9695 (update_conflict_allocno_hard_prefs): New.
9696 (bucket_allocno_compare_func): Add a preference based on
9697 conflict_allocno_hard_prefs.
9698 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
9699 (color_allocnos): Remove a dead code. Initiate
9700 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
9702 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9705 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
9706 constraint from =wa to wa. Avoid a subreg on the output operand,
9707 instead use a pseudo and subreg it in a move.
9708 (p9_xxbrd_<mode>): Changed to ...
9709 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
9710 (p9_xxbrd_v2df): New expander.
9711 (p9_xxbrw_<mode>): Changed to ...
9712 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
9713 (p9_xxbrw_v4sf): New expander.
9715 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9717 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
9719 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
9722 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
9723 multiply in 32-bit mode.
9724 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
9725 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
9728 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9730 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
9731 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
9732 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
9733 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
9735 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
9738 * dwarf2out.c (is_trivial_indirect_ref): New function.
9739 (dwarf2out_late_global_decl): Do not generate a location
9740 attribute for variables that have a non-trivial DECL_VALUE_EXPR
9741 and that are not defined in the current unit.
9743 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
9745 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
9746 instead of a libcall for UNORDERED.
9748 2018-02-09 Tamar Christina <tamar.christina@arm.com>
9751 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
9752 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
9754 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9757 * config/s390/s390.c (s390_set_current_function): Invoke
9758 s390_indirect_branch_settings also if fndecl didn't change.
9760 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
9762 * config/rs6000/rs6000.md (blockage): Set length to zero.
9764 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
9766 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
9768 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9771 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
9772 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
9776 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
9777 PARALLEL incoming that failed vt_get_decl_and_offset check.
9780 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
9781 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
9783 (get_variable_section): For decls in named .bss* sections pass true as
9784 second argument to bss_initializer_p.
9786 2018-02-09 Marek Polacek <polacek@redhat.com>
9787 Jakub Jelinek <jakub@redhat.com>
9790 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
9791 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
9792 Sync some changes from cxx_fold_indirect_ref.
9794 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
9796 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
9798 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
9799 (BLOCK_INLINE_ENTRY_LABEL): New.
9800 (dwarf2out_var_location): Disregard inline entry markers.
9801 (inline_entry_data): New struct.
9802 (inline_entry_data_hasher): New hashtable type.
9803 (inline_entry_data_hasher::hash): New.
9804 (inline_entry_data_hasher::equal): New.
9805 (inline_entry_data_table): New variable.
9806 (add_high_low_attributes): Add DW_AT_entry_pc and
9807 DW_AT_GNU_entry_view attributes if a pending entry is found
9808 in inline_entry_data_table. Add old entry_pc attribute only
9809 if debug nonbinding markers are disabled.
9810 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
9811 markers are enabled.
9812 (block_within_block_p, dwarf2out_inline_entry): New.
9813 (dwarf2out_finish): Check that no entries remained in
9814 inline_entry_data_table.
9815 * final.c (reemit_insn_block_notes): Handle inline entry notes.
9816 (final_scan_insn, notice_source_line): Likewise.
9817 (rest_of_clean_state): Skip inline entry markers.
9818 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
9820 * gimple.c (gimple_build_debug_inline_entry): New.
9821 * gimple.h (enum gimple_debug_subcode): Add
9822 GIMPLE_DEBUG_INLINE_ENTRY.
9823 (gimple_build_debug_inline_entry): Declare.
9824 (gimple_debug_inline_entry_p): New.
9825 (gimple_debug_nonbind_marker_p): Adjust.
9826 * insn-notes.def (INLINE_ENTRY): New.
9827 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
9828 inline entry marker notes.
9829 (print_insn): Likewise.
9830 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
9831 (INSN_DEBUG_MARKER_KIND): Likewise.
9832 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
9833 * tree-inline.c (expand_call_inline): Build and insert
9834 debug_inline_entry stmt.
9835 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
9836 inline entry blocks early, if nonbind markers are enabled.
9837 (dump_scope_block): Dump fragment info.
9838 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
9839 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
9840 (gimple_build_debug_inline_entry): New.
9841 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
9842 Enable/disable inline entry points too.
9843 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
9844 (DEBUG_INSN): Describe inline entry markers.
9846 * common.opt (gvariable-location-views): New.
9847 (gvariable-location-views=incompat5): New.
9848 * config.in: Rebuilt.
9849 * configure: Rebuilt.
9850 * configure.ac: Test assembler for view support.
9851 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
9852 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
9853 * dwarf2out.c (var_loc_view): New typedef.
9854 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
9855 (dwarf2out_locviews_in_attribute): New.
9856 (dwarf2out_locviews_in_loclist): New.
9857 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
9858 (enum dw_line_info_opcode): Add LI_adv_address.
9859 (struct dw_line_info_table): Add view.
9860 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
9861 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
9862 (zero_view_p): New variable.
9863 (ZERO_VIEW_P): New macro.
9864 (output_asm_line_debug_info): New.
9865 (struct var_loc_node): Add view.
9866 (add_AT_view_list, AT_loc_list): New.
9867 (add_var_loc_to_decl): Add view param. Test it against last.
9868 (new_loc_list): Add view params. Record them.
9869 (AT_loc_list_ptr): Handle loc and view lists.
9870 (view_list_to_loc_list_val_node): New.
9871 (print_dw_val): Handle dw_val_class_view_list.
9872 (size_of_die): Likewise.
9873 (value_format): Likewise.
9874 (loc_list_has_views): New.
9875 (gen_llsym): Set vl_symbol too.
9876 (maybe_gen_llsym, skip_loc_list_entry): New.
9877 (dwarf2out_maybe_output_loclist_view_pair): New.
9878 (output_loc_list): Output view list or entries too.
9879 (output_view_list_offset): New.
9880 (output_die): Handle dw_val_class_view_list.
9881 (output_dwarf_version): New.
9882 (output_compilation_unit_header): Use it.
9883 (output_skeleton_debug_sections): Likewise.
9884 (output_rnglists, output_line_info): Likewise.
9885 (output_pubnames, output_aranges): Update version comments.
9886 (output_one_line_info_table): Output view numbers in asm comments.
9887 (dw_loc_list): Determine current endview, pass it to new_loc_list.
9888 Call maybe_gen_llsym.
9889 (loc_list_from_tree_1): Adjust.
9890 (add_AT_location_description): Create view list attribute if
9891 needed, check it's absent otherwise.
9892 (convert_cfa_to_fb_loc_list): Adjust.
9893 (maybe_emit_file): Call output_asm_line_debug_info for test.
9894 (dwarf2out_var_location): Reset views as needed. Precompute
9895 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
9896 attribute. Set view.
9897 (new_line_info_table): Reset next view.
9898 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
9899 (dwarf2out_source_line): Likewise. Output view resets and labels to
9900 the assembler, or select appropriate line info opcodes.
9901 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
9902 (optimize_string_length): Catch it. Adjust.
9903 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
9904 dw_val_class_view_list, and remove it if no longer needed.
9905 (hash_loc_list): Hash view numbers.
9906 (loc_list_hasher::equal): Compare them.
9907 (optimize_location_lists): Check whether a view list symbol is
9908 needed, and whether the locview attribute is present, and
9909 whether they match. Remove the locview attribute if no longer
9911 (index_location_lists): Call skip_loc_list_entry for test.
9912 (dwarf2out_finish): Call output_asm_line_debug_info for test.
9913 Use output_dwarf_version.
9914 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
9915 (struct dw_val_node): Add val_view_list.
9916 * final.c (SEEN_NEXT_VIEW): New.
9917 (set_next_view_needed): New.
9918 (clear_next_view_needed): New.
9919 (maybe_output_next_view): New.
9920 (final_start_function): Rename to...
9921 (final_start_function_1): ... this. Take pointer to FIRST,
9922 add SEEN parameter. Emit param bindings in the initial view.
9923 (final_start_function): Reintroduce SEEN-less interface.
9924 (final): Rename to...
9925 (final_1): ... this. Take SEEN parameter. Output final pending
9926 next view at the end.
9927 (final): Reintroduce seen-less interface.
9928 (final_scan_insn): Output pending next view before switching
9929 sections or ending a block. Mark the next view as needed when
9930 outputting variable locations. Notify debug backend of section
9931 changes, and of location view changes.
9932 (rest_of_handle_final): Adjust.
9933 * toplev.c (process_options): Autodetect value for debug variable
9934 location views option. Warn on incompat5 without -gdwarf-5.
9935 * doc/invoke.texi (gvariable-location-views): New.
9936 (gvariable-location-views=incompat5): New.
9937 (gno-variable-location-views): New.
9939 2018-02-08 David Malcolm <dmalcolm@redhat.com>
9941 PR tree-optimization/84136
9942 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
9943 that the result of find_edge is non-NULL.
9945 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9948 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
9949 storing integer register in SImode. Fix cost of 256 and 512
9950 byte aligned SSE register store.
9952 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9954 * config/i386/i386.c (ix86_multiplication_cost): Fix
9955 multiplication cost for TARGET_AVX512DQ.
9957 2018-02-08 Marek Polacek <polacek@redhat.com>
9959 PR tree-optimization/84238
9960 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
9963 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9965 PR tree-optimization/84265
9966 * tree-vect-stmts.c (vectorizable_store): Don't treat
9967 VMAT_CONTIGUOUS accesses as grouped.
9968 (vectorizable_load): Likewise.
9970 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9972 PR tree-optimization/81635
9973 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
9974 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
9975 (test_round_for_mask): New functions.
9976 (wide_int_cc_tests): Call test_round_for_mask.
9977 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
9978 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
9979 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
9980 range returned by get_range_info.
9982 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9985 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
9986 * symtab.c: Include builtins.h
9987 (symtab_node::output_to_lto_symbol_table_p): Move here
9988 from lto-streamer-out.c:output_symbol_p.
9989 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
9990 (output_symbol_p): Move all logic to symtab.c
9991 (produce_symtab): Update.
9993 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9995 * config/s390/s390-opts.h (enum indirect_branch): Define.
9996 * config/s390/s390-protos.h (s390_return_addr_from_memory)
9997 (s390_indirect_branch_via_thunk)
9998 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
9999 (enum s390_indirect_branch_type): Define.
10000 * config/s390/s390.c (struct s390_frame_layout, struct
10001 machine_function): Remove.
10002 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10003 (indirect_branch_table_label_no, indirect_branch_table_name):
10005 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10006 (enum s390_indirect_branch_option): Define.
10007 (s390_return_addr_from_memory): New function.
10008 (s390_handle_string_attribute): New function.
10009 (s390_attribute_table): Add new attribute handler.
10010 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10011 (s390_indirect_branch_via_thunk): New function.
10012 (s390_indirect_branch_via_inline_thunk): New function.
10013 (s390_function_ok_for_sibcall): When jumping via thunk disallow
10014 sibling call optimization for non z10 compiles.
10015 (s390_emit_call): Force indirect branch target to be a single
10016 register. Add r1 clobber for non-z10 compiles.
10017 (s390_emit_epilogue): Emit return jump via return_use expander.
10018 (s390_reorg): Handle JUMP_INSNs as execute targets.
10019 (s390_option_override_internal): Perform validity checks for the
10020 new command line options.
10021 (s390_indirect_branch_attrvalue): New function.
10022 (s390_indirect_branch_settings): New function.
10023 (s390_set_current_function): Invoke s390_indirect_branch_settings.
10024 (s390_output_indirect_thunk_function): New function.
10025 (s390_code_end): Implement target hook.
10026 (s390_case_values_threshold): Implement target hook.
10027 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10029 * config/s390/s390.h (struct s390_frame_layout)
10030 (struct machine_function): Move here from s390.c.
10031 (TARGET_INDIRECT_BRANCH_NOBP_RET)
10032 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10033 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10034 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10035 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10036 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10037 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10038 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10039 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10040 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10041 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10042 (mnemonic attribute): Add values which aren't recognized
10044 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10045 pattern for branch conversion. Fix mnemonic attribute.
10046 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10047 indirect branch via thunk if requested.
10048 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10049 ("*indirect_jump"): Disable for branch conversion using out of
10051 ("indirect_jump_via_thunk<mode>_z10")
10052 ("indirect_jump_via_thunk<mode>")
10053 ("indirect_jump_via_inlinethunk<mode>_z10")
10054 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10055 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10056 ("casesi_jump_via_inlinethunk<mode>_z10")
10057 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10058 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10059 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10060 ("*indirect2_jump"): Disable for branch conversion.
10061 ("casesi_jump"): Turn into expander and expand patterns for branch
10063 ("return_use"): New expander.
10064 ("*return"): Emit return via thunk and rename it to ...
10065 ("*return<mode>"): ... this one.
10066 * config/s390/s390.opt: Add new options and and enum for the
10069 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10071 * lra-constraints.c (match_reload): Unconditionally use
10072 gen_lowpart_SUBREG, rather than selecting between that
10073 and equivalent gen_rtx_SUBREG code.
10075 2018-02-08 Richard Biener <rguenther@suse.de>
10077 PR tree-optimization/84233
10078 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10079 changed flag instead of boguously re-using phi_inserted.
10081 2018-02-08 Martin Jambor <mjambor@suse.cz>
10083 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10084 static local variables.
10086 2018-02-08 Richard Biener <rguenther@suse.de>
10088 PR tree-optimization/84278
10089 * tree-vect-stmts.c (vectorizable_store): When looking for
10090 smaller vector types to perform grouped strided loads/stores
10091 make sure the mode is supported by the target.
10092 (vectorizable_load): Likewise.
10094 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10096 * config/aarch64/aarch64.c (aarch64_components_for_bb):
10097 Increase LDP/STP opportunities by adding adjacent callee-saves.
10099 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10101 PR rtl-optimization/84068
10102 PR rtl-optimization/83459
10103 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10105 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
10107 PR tree-optimization/84224
10108 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10109 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10110 non-zero arguments.
10112 2018-02-07 Iain Sandoe <iain@codesourcery.com>
10115 * config/rs6000/altivec.md (*restore_world): Remove LR use.
10116 * config/rs6000/predicates.md (restore_world_operation): Adjust op
10117 count, remove one USE.
10119 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
10121 * doc/install.texi (Configuration): Document the
10122 --with-long-double-format={ibm,ieee} PowerPC configuration
10126 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10127 Convert from define_expand to be define_insn_and_split. Rework
10128 float/double/_Float128 conversions to QI/HI/SImode to work with
10129 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
10130 conversions to QI/HImode types did a store and then a load to
10131 truncate the value. For conversions to VSX registers, don't split
10132 the insn, instead emit the code directly. Use the code iterator
10133 any_fix to combine signed and unsigned conversions.
10134 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10135 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10136 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10137 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10138 (fix_<mode>di2_hw): Likewise.
10139 (fixuns_<mode>di2_hw): Likewise.
10140 (fix_<mode>si2_hw): Likewise.
10141 (fixuns_<mode>si2_hw): Likewise.
10142 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10143 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10144 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10145 fix<uns>_trunc<SFDF:mode>si2_p8.
10146 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10147 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10148 (fix<uns>_<mode>_mem): Likewise.
10149 (fctiw<u>z_<mode>_mem): Likewise.
10150 (fix<uns>_<mode>_mem): Likewise.
10151 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10152 the register allocator from doing a direct move to the GPRs to do
10153 a store, and instead use the ISA 3.0 store byte/half-word from
10154 vector register instruction. For IEEE 128-bit floating point,
10155 also optimize stores of 32-bit ints.
10156 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10158 2018-02-07 Alan Hayward <alan.hayward@arm.com>
10160 * genextract.c (push_pathstr_operand): New function to support
10162 (walk_rtx): Call push_pathstr_operand.
10163 (print_path): Support [a-zA-Z].
10165 2018-02-07 Richard Biener <rguenther@suse.de>
10167 PR tree-optimization/84037
10168 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
10169 (cse_and_gimplify_to_preheader): Declare.
10170 (vect_get_place_in_interleaving_chain): Likewise.
10171 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10173 (_loop_vec_info::~_loop_vec_info): Delete it.
10174 (cse_and_gimplify_to_preheader): New function.
10175 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
10176 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
10177 (vectorizable_load): Likewise. For grouped stores always base
10178 the IV on the first element.
10179 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
10180 condition before gimplifying.
10182 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10184 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
10185 *DIV_EXPR and *MOD_EXPR.
10187 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
10190 * config/i386/i386.c (ix86_option_override_internal): Mask out
10191 the CF_SET bit when checking -fcf-protection.
10193 2018-02-07 Tom de Vries <tom@codesourcery.com>
10196 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
10199 2018-02-07 Richard Biener <rguenther@suse.de>
10201 PR tree-optimization/84204
10202 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
10205 PR tree-optimization/84205
10206 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
10207 special-case isl_ast_op_zdiv_r.
10209 PR tree-optimization/84223
10210 * graphite-scop-detection.c (gather_bbs::before_dom_children):
10211 Only add conditions from within the region.
10212 (gather_bbs::after_dom_children): Adjust.
10214 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
10217 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
10218 * config/avr/avr.md: Only post-reload split REG-REG moves if
10219 either register is GENERAL_REG_P.
10221 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10223 PR tree-optimization/84235
10224 * tree-ssa-scopedtables.c
10225 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
10226 if the subtraction is performed in floating point type where NaNs are
10227 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
10228 build 1. Formatting fix.
10230 2018-02-06 Jakub Jelinek <jakub@redhat.com>
10233 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
10234 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
10235 and skip it regardless of bb boundaries. Use CALL_P macro,
10236 don't test INSN_P (insn) together with CALL_P or JUMP_P check
10237 unnecessarily, formatting fix.
10239 2018-02-06 Michael Collison <michael.collison@arm.com>
10241 * config/arm/thumb2.md:
10242 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
10243 (*thumb_mov_notscc): Ditto.
10245 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
10248 * config/rs6000/rs6000.md (su code attribute): Use "u" for
10249 unsigned_fix, not "s".
10251 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10253 * configure.ac (gcc_fn_eh_frame_ro): New function.
10254 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
10255 correct .eh_frame permissions.
10256 * configure: Regenerate.
10258 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
10260 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
10261 irrelevant options.
10263 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10265 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10266 Display warning message for -mno-speculate-indirect-jumps.
10268 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
10270 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
10272 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
10274 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
10276 PR tree-optimization/84225
10277 * tree-eh.c (find_trapping_overflow): Only call
10278 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
10280 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10283 * config/i386/i386.c: Reimplement the check of possible options
10284 -mibt/-mshstk conbination. Change error messages.
10285 * doc/invoke.texi: Fix a typo: remove extra '='.
10287 2018-02-06 Marek Polacek <polacek@redhat.com>
10289 PR tree-optimization/84228
10290 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
10292 2018-02-06 Tamar Christina <tamar.christina@arm.com>
10295 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
10296 emitted arch directives.
10297 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
10298 __ARM_FEATURE_COPROC before changing architectures.
10300 2018-02-06 Richard Biener <rguenther@suse.de>
10302 * config/i386/i386.c (print_reg): Fix typo.
10303 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
10305 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
10307 * configure: Regenerate.
10309 2018-02-05 Martin Sebor <msebor@redhat.com>
10311 PR tree-optimization/83369
10312 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
10315 2018-02-05 Martin Liska <mliska@suse.cz>
10317 * doc/invoke.texi: Cherry-pick upstream r323995.
10319 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
10321 * ira.c (ira_init_register_move_cost): Adjust comment.
10323 2018-02-05 Martin Liska <mliska@suse.cz>
10325 PR gcov-profile/84137
10326 * doc/gcov.texi: Fix typo in documentation.
10328 2018-02-05 Martin Liska <mliska@suse.cz>
10330 PR gcov-profile/83879
10331 * doc/gcov.texi: Document necessity of --dynamic-list-data when
10332 using dlopen functionality.
10334 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
10336 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
10337 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
10338 _mm_maskz_range_ss, _mm_mask_range_round_ss,
10339 _mm_maskz_range_round_ss): New intrinsics.
10340 (__builtin_ia32_rangesd128_round)
10341 (__builtin_ia32_rangess128_round): Remove.
10342 (__builtin_ia32_rangesd128_mask_round,
10343 __builtin_ia32_rangess128_mask_round): New builtins.
10344 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
10345 __builtin_ia32_rangess128_round): Remove.
10346 (__builtin_ia32_rangesd128_mask_round,
10347 __builtin_ia32_rangess128_mask_round): New builtins.
10348 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
10349 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
10350 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
10351 "<round_saeonly_constraint>")): Changed to ...
10352 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
10353 "<round_saeonly_scalar_constraint>")): ... this.
10354 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
10355 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
10356 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
10357 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
10358 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
10360 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
10362 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
10364 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
10365 Remove all values except native, 8540 and 8548.
10367 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
10369 * config/i386/i386.c (ix86_output_function_return): Pass
10370 INVALID_REGNUM, instead of -1, as invalid register number to
10371 indirect_thunk_name and output_indirect_thunk.
10373 2018-02-02 Julia Koval <julia.koval@intel.com>
10375 * config.gcc: Add -march=icelake.
10376 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
10377 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
10378 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
10379 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
10380 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
10381 (processor_target_table): Add icelake.
10382 (ix86_option_override_internal): Handle new PTAs.
10383 (get_builtin_code_for_version): Handle icelake.
10384 (M_INTEL_COREI7_ICELAKE): New.
10385 (fold_builtin_cpu): Handle icelake.
10386 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
10387 * doc/invoke.texi: Add -march=icelake.
10389 2018-02-02 Julia Koval <julia.koval@intel.com>
10391 * config/i386/i386.c (ix86_option_override_internal): Change flags type
10392 to wide_int_bitmask.
10393 * wide-int-bitmask.h: New.
10395 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10398 * config/i386/i386.md: Replace Pmode with word_mode in
10399 builtin_setjmp_setup and builtin_longjmp to support x32.
10401 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
10405 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
10407 (rs6000_supported_cpu_names): New static variable.
10408 (linux_cpu_translation_table): Likewise.
10409 (elf_platform) <cpu>: Define new static variable and use it.
10410 Translate kernel AT_PLATFORM name to canonical name if needed.
10411 Error if platform name is unknown.
10413 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
10416 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
10418 2018-02-01 Jeff Law <law@redhat.com>
10421 * config/i386/i386.c (release_scratch_register_on_entry): Add new
10422 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
10423 the scratch if RELEASE_VIA_POP is false.
10424 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
10425 If we have to save a temporary register, decrement SIZE appropriately.
10426 Pass new arguments to release_scratch_register_on_entry.
10427 (ix86_adjust_stack_and_probe): Likewise.
10428 (ix86_emit_probe_stack_range): Pass new arguments to
10429 release_scratch_register_on_entry.
10431 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
10433 PR rtl-optimization/84157
10434 * combine.c (change_zero_ext): Use REG_P predicate in
10435 front of HARD_REGISTER_P predicate.
10437 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
10439 * config/avr/avr.c (avr_option_override): Move disabling of
10440 -fdelete-null-pointer-checks to...
10441 * common/config/avr/avr-common.c (avr_option_optimization_table):
10444 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10446 PR tree-optimization/81635
10447 * tree-data-ref.c (split_constant_offset_1): For types that
10448 wrap on overflow, try to use range info to prove that wrapping
10451 2018-02-01 Renlin Li <renlin.li@arm.com>
10454 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
10455 TAILCALL_ADDR_REGS.
10456 (aarch64_register_move_cost): Likewise.
10457 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
10458 TAILCALL_ADDR_REGS.
10459 (REG_CLASS_NAMES): Likewise.
10460 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
10461 TAILCALL_ADDR_REGS. Remove IP registers.
10462 * config/aarch64/aarch64.md (Ucs): Update register constraint.
10464 2018-02-01 Richard Biener <rguenther@suse.de>
10466 * domwalk.h (dom_walker::dom_walker): Add additional constructor
10467 for specifying RPO order and allow NULL for that.
10468 * domwalk.c (dom_walker::dom_walker): Likewise.
10469 (dom_walker::walk): Handle NULL RPO order.
10470 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
10472 (rewrite_update_dom_walker): Likewise.
10473 (mark_def_dom_walker): Likewise.
10475 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10477 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
10478 (aarch64_maybe_expand_sve_subreg_move): Declare.
10479 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
10480 * config/aarch64/predicates.md (aarch64_any_register_operand): New
10482 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
10483 that are semantically a reverse operation.
10484 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
10485 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
10486 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
10488 (aarch64_can_change_mode_class): For big-endian, forbid changes
10489 between two SVE modes if they have different element sizes.
10491 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10493 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
10494 the TImode handling for big-endian targets.
10496 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10498 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
10499 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
10501 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
10502 Remove BSWAP handing for big-endian targets and use the form of
10503 LD1RQ appropariate for the mode.
10505 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10507 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
10508 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
10509 duplicated element.
10511 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10514 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
10515 check for operands that need to go through aarch64_sve_reload_be.
10517 2018-02-01 Jakub Jelinek <jakub@redhat.com>
10519 PR tree-optimization/81661
10520 PR tree-optimization/84117
10521 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
10522 * tree-eh.c: Include gimplify.h.
10523 (find_trapping_overflow, replace_trapping_overflow,
10524 rewrite_to_non_trapping_overflow): New functions.
10525 * tree-vect-loop.c: Include tree-eh.h.
10526 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
10527 * tree-data-ref.c: Include tree-eh.h.
10528 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
10530 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
10532 PR rtl-optimization/84123
10533 * combine.c (change_zero_ext): Check if hard register satisfies
10534 can_change_dest_mode before calling gen_lowpart_SUBREG.
10536 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
10539 * ira.c (ira_init_register_move_cost): Remove assert.
10541 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10543 PR rtl-optimization/84071
10544 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
10545 * doc/tm.texi: Regenerate.
10547 2018-01-31 Richard Biener <rguenther@suse.de>
10549 PR tree-optimization/84132
10550 * tree-data-ref.c (analyze_miv_subscript): Properly
10551 check whether evolution_function_is_affine_multivariate_p
10552 before calling gcd_of_steps_may_divide_p.
10554 2018-01-31 Julia Koval <julia.koval@intel.com>
10557 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
10558 * config/i386/i386.md (rdpid_rex64) New.
10559 (rdpid): Make 32bit only.
10561 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
10564 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
10565 an IDENTIFIER_NODE for FUNCTION_TYPE's.
10567 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10570 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10572 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10574 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10576 PR rtl-optimization/84071
10577 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
10578 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
10580 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10582 * config/arc/arc.c (arc_handle_aux_attribute): New function.
10583 (arc_attribute_table): Add 'aux' attribute.
10584 (arc_in_small_data_p): Consider aux like variables.
10585 (arc_is_aux_reg_p): New function.
10586 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
10587 (arc_get_aux_arg): New function.
10588 (prepare_move_operands): Handle aux-register access.
10589 (arc_handle_aux_attribute): New function.
10590 * doc/extend.texi (ARC Variable attributes): Add subsection.
10592 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10594 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
10595 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
10596 (arc_attribute_table): Add 'uncached' attribute.
10597 (arc_print_operand): Print '.di' flag for uncached memory
10599 (arc_in_small_data_p): Do not consider for small data the uncached
10601 (arc_is_uncached_mem_p): New function.
10602 * config/arc/predicates.md (compact_store_memory_operand): Check
10603 for uncached memory accesses.
10604 (nonvol_nonimm_operand): Likewise.
10605 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
10607 2018-01-31 Jakub Jelinek <jakub@redhat.com>
10610 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
10611 falign-loops=): Add Optimization flag.
10613 2018-01-30 Jeff Law <law@redhat.com>
10616 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
10617 INT_REGISTERS_SAVED. Check it prior to calling
10618 get_scratch_register_on_entry.
10619 (ix86_adjust_stack_and_probe): Similarly.
10620 (ix86_emit_probe_stack_range): Similarly.
10621 (ix86_expand_prologue): Corresponding changes.
10623 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10626 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
10627 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
10629 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
10632 * lra-constraints.c (curr_insn_transform): Process AND in the
10635 2018-01-30 Jakub Jelinek <jakub@redhat.com>
10637 PR rtl-optimization/83986
10638 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
10639 dependence against last_pending_memory_flush in addition to
10640 pending_jump_insns.
10642 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
10644 PR tree-optimization/81611
10645 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
10648 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10651 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
10654 2018-01-30 Richard Biener <rguenther@suse.de>
10655 Jakub Jelinek <jakub@redhat.com>
10657 PR tree-optimization/84111
10658 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
10659 inner loops added during recursion, as they don't have up-to-date
10662 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10665 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
10666 (can_inline_edge_by_limits_p): ... here.
10667 (can_early_inline_edge_p, check_callers,
10668 update_caller_keys, update_callee_keys, recursive_inlining,
10669 add_new_edges_to_heap, speculation_useful_p,
10670 inline_small_functions,
10671 inline_small_functions, flatten_function,
10672 inline_to_all_callers_1): Update.
10674 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10676 * profile-count.c (profile_count::combine_with_ipa_count): Handle
10679 2018-01-30 Richard Biener <rguenther@suse.de>
10681 PR tree-optimization/83008
10682 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
10683 invariant and constant vector uses in stmts when they need
10684 more than one stmt.
10686 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10689 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
10690 * configure: Regenerate.
10692 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10694 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
10696 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
10697 Use gen_rtx_REG rather than gen_lowpart.
10699 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10701 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
10702 rather than 0 when creating partial subregs.
10704 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10706 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
10709 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
10712 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
10713 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
10714 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
10715 flags. This restores the settings used before the 2017-07-24.
10716 Turning off pre increment/decrement/modify allows IVOPTS to
10717 optimize DF/SF loops where the index is an int.
10719 2018-01-29 Richard Biener <rguenther@suse.de>
10720 Kelvin Nilsen <kelvin@gcc.gnu.org>
10723 * tree-vect-stmts.c (vectorizable_call): Don't call
10724 targetm.vectorize_builtin_md_vectorized_function if callee is
10727 2018-01-22 Carl Love <cel@us.ibm.com>
10729 * doc/extend.tex: Fix typo in second arg in
10730 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
10732 2018-01-29 Richard Biener <rguenther@suse.de>
10734 PR tree-optimization/84086
10735 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
10736 (flush_ssaname_freelist): When SSA names were released reset
10737 the SCEV hash table.
10739 2018-01-29 Richard Biener <rguenther@suse.de>
10741 PR tree-optimization/84057
10742 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
10743 removed paths when removing edges.
10745 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
10747 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
10748 -mfunction-return=@var{choice}.
10750 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
10752 PR diagnostic/84034
10753 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
10754 Handle CR like TAB.
10755 (layout::print_source_line): Likewise.
10756 (test_get_line_width_without_trailing_whitespace): Add test cases.
10758 2018-01-27 Jakub Jelinek <jakub@redhat.com>
10760 PR middle-end/84040
10761 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
10764 2018-01-26 Jim Wilson <jimw@sifive.com>
10766 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
10768 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
10771 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10773 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
10774 and CMP + SUB-immediate -> SUBS.
10776 2018-01-26 Martin Sebor <msebor@redhat.com>
10778 PR tree-optimization/83896
10779 * tree-ssa-strlen.c (get_string_len): Rename...
10780 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
10781 Avoid assuming length is constant.
10782 (handle_char_store): Use HOST_WIDE_INT for string length.
10784 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
10787 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
10788 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
10790 2018-01-26 Richard Biener <rguenther@suse.de>
10792 PR rtl-optimization/84003
10793 * dse.c (record_store): Only record redundant stores when
10794 the earlier store aliases at least all accesses the later one does.
10796 2018-01-26 Jakub Jelinek <jakub@redhat.com>
10798 PR rtl-optimization/83985
10799 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
10800 REG_CFA_RESTORE insns.
10801 (delete_unmarked_insns): Don't ignore separate shrink wrapping
10802 REG_CFA_RESTORE insns here.
10805 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
10806 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
10808 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10810 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
10811 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
10812 (arc_init): Likewise.
10813 (arc_override_options): Likewise.
10814 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
10816 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
10818 * config/arc/arc.h (TARGET_DBNZ): Define.
10819 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
10820 properly set the tune attribute.
10821 (dbnz): Use TARGET_DBNZ guard.
10822 * config/arc/arc.opt (mtune): Add core3 option.
10824 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10826 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
10827 recognize new pic like addresses.
10828 (arc_delegitimize_address): Clean up.
10830 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10832 * config/arc/arc-arches.def: Option mrf16 valid for all
10834 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
10835 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
10836 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
10837 * config/arc/arc-tables.opt: Regenerate.
10838 * config/arc/arc.c (arc_conditional_register_usage): Handle
10839 reduced register file case.
10840 (arc_file_start): Set must have build attributes.
10841 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
10842 mrf16 option value.
10843 * config/arc/arc.opt (mrf16): Add new option.
10844 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
10845 * config/arc/genmultilib.awk: Handle new mrf16 option.
10846 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
10847 * config/arc/t-multilib: Regenerate.
10848 * doc/invoke.texi (ARC Options): Document mrf16 option.
10850 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10852 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
10853 * config/arc/arc.c (arc_handle_secure_attribute): New function.
10854 (arc_attribute_table): Add 'secure_call' attribute.
10855 (arc_print_operand): Print secure call operand.
10856 (arc_function_ok_for_sibcall): Don't optimize tail calls when
10858 (arc_is_secure_call_p): New function. * config/arc/arc.md
10859 (call_i): Add support for sjli instruction.
10860 (call_value_i): Likewise.
10861 * config/arc/constraints.md (Csc): New constraint.
10863 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10864 John Eric Martin <John.Martin@emmicro-us.com>
10866 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
10867 * config/arc/arc.c (_arc_jli_section): New struct.
10868 (arc_jli_section): New type.
10869 (rc_jli_sections): New static variable.
10870 (arc_handle_jli_attribute): New function.
10871 (arc_attribute_table): Add jli_always and jli_fixed attribute.
10872 (arc_file_end): New function.
10873 (TARGET_ASM_FILE_END): Define.
10874 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
10875 (arc_add_jli_section): New function.
10876 (jli_call_scan): Likewise.
10877 (arc_reorg): Call jli_call_scan.
10878 (arc_output_addsi): Remove 'S' from printing asm operand.
10879 (arc_is_jli_call_p): New function.
10880 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
10882 (movhi_insn): Likewise.
10883 (movsi_insn): Likewise.
10884 (movsi_set_cc_insn): Likewise.
10885 (loadqi_update): Likewise.
10886 (load_zeroextendqisi_update): Likewise.
10887 (load_signextendqisi_update): Likewise.
10888 (loadhi_update): Likewise.
10889 (load_zeroextendhisi_update): Likewise.
10890 (load_signextendhisi_update): Likewise.
10891 (loadsi_update): Likewise.
10892 (loadsf_update): Likewise.
10893 (movsicc_insn): Likewise.
10894 (bset_insn): Likewise.
10895 (bxor_insn): Likewise.
10896 (bclr_insn): Likewise.
10897 (bmsk_insn): Likewise.
10898 (bicsi3_insn): Likewise.
10899 (cmpsi_cc_c_insn): Likewise.
10900 (movsi_ne): Likewise.
10901 (movsi_cond_exec): Likewise.
10902 (clrsbsi2): Likewise.
10903 (norm_f): Likewise.
10913 (call_i): Remove 'S' asm letter, add jli instruction.
10914 (call_value_i): Likewise.
10915 * config/arc/arc.op (mjli-always): New option.
10916 * config/arc/constraints.md (Cji): New constraint.
10917 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
10919 (subsf3_fpx): Likewise.
10920 (mulsf3_fpx): Likewise.
10921 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
10923 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
10924 function attrbutes.
10925 * doc/invoke.texi (ARC): Document mjli-always option.
10927 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
10929 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
10930 avoid addition with 0 and use incw and decw where possible.
10932 2018-01-26 Richard Biener <rguenther@suse.de>
10934 PR tree-optimization/81082
10935 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
10936 association if it requires casting to unsigned.
10937 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
10938 from fold_plusminus_mult_expr to catch important cases late when
10939 range info is available.
10941 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10943 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
10944 * configure.ac (hidden_linkonce): New test.
10945 * configure: Regenerate.
10946 * config.in: Regenerate.
10948 2018-01-26 Julia Koval <julia.koval@intel.com>
10950 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
10951 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
10952 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
10953 _mm_mask_bitshuffle_epi64_mask): Fix type.
10954 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
10955 USI_FTYPE_V4DI_V4DI_USI): Remove.
10956 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
10957 __builtin_ia32_vpshufbitqmb256_mask,
10958 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
10959 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
10960 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
10962 2018-01-26 Alan Modra <amodra@gmail.com>
10965 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
10966 UNSPEC_VBPERMQ. Sort other unspecs.
10968 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
10970 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
10972 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
10974 PR middle-end/83055
10975 * predict.c (drop_profile): Do not push/pop cfun; update also
10977 (handle_missing_profiles): Fix logic looking for zero profiles.
10979 2018-01-25 Jakub Jelinek <jakub@redhat.com>
10981 PR middle-end/83977
10982 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
10983 on functions with #pragma omp declare simd or functions with simd
10985 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
10986 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
10987 Remove trailing \n from warning_at calls.
10989 2018-01-25 Tom de Vries <tom@codesourcery.com>
10992 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10993 for neutered workers.
10995 2018-01-24 Joseph Myers <joseph@codesourcery.com>
10998 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
10999 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11001 2018-01-24 Jeff Law <law@redhat.com>
11004 * i386.c (get_probe_interval): Move to earlier point.
11005 (ix86_compute_frame_layout): If -fstack-clash-protection and
11006 the frame is larger than the probe interval, then use pushes
11007 to save registers rather than reg->mem moves.
11008 (ix86_expand_prologue): Remove conditional for int_registers_saved
11011 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
11014 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11015 min/max for never referenced object.
11017 2018-01-24 Jakub Jelinek <jakub@redhat.com>
11019 PR middle-end/83977
11020 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11022 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11023 attributes from DECL_ATTRIBUTES (decl) without affecting
11024 DECL_ATTRIBUTES (current_function_decl).
11025 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11026 functions with non-NULL DECL_ABSTRACT_ORIGIN.
11028 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
11030 PR tree-optimization/83979
11031 * fold-const.c (fold_comparison): Use constant_boolean_node
11032 instead of boolean_{true,false}_node.
11034 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
11036 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11039 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11041 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11042 Simplify the clause that sets the length attribute.
11043 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11044 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11045 clause that sets the length attribute.
11046 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11048 2018-01-24 Tom de Vries <tom@codesourcery.com>
11051 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11052 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11053 Add strict parameter.
11054 (prevent_branch_around_nothing): Insert dummy insn between branch to
11055 label and label with no ptx insn inbetween.
11056 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11058 2018-01-24 Tom de Vries <tom@codesourcery.com>
11061 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11062 for neutered threads in warp.
11063 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11065 2018-01-24 Richard Biener <rguenther@suse.de>
11067 PR tree-optimization/83176
11068 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11071 2018-01-24 Richard Biener <rguenther@suse.de>
11073 PR tree-optimization/82819
11074 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11075 code generating pluses that are no-ops in the target precision.
11077 2018-01-24 Richard Biener <rguenther@suse.de>
11079 PR middle-end/84000
11080 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11082 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11084 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11085 to merge probabilities.
11086 * predict.c (probably_never_executed): Also mark as cold functions
11087 with global 0 profile and guessed local profile.
11088 * profile-count.c (profile_probability::combine_with_count): New
11090 * profile-count.h (profile_probability::operator*,
11091 profile_probability::operator*=, profile_probability::operator/,
11092 profile_probability::operator/=): Reduce precision to adjusted
11093 and set value to guessed on contradictory divisions.
11094 (profile_probability::combine_with_freq): Remove.
11095 (profile_probability::combine_wiht_count): Declare.
11096 (profile_count::force_nonzero):: Set to adjusted.
11097 (profile_count::probability_in):: Set quality to adjusted.
11098 * tree-ssa-tail-merge.c (replace_block_by): Use
11099 combine_with_count.
11101 2018-01-23 Andrew Waterman <andrew@sifive.com>
11102 Jim Wilson <jimw@sifive.com>
11104 * config/riscv/riscv.c (riscv_stack_boundary): New.
11105 (riscv_option_override): Set riscv_stack_boundary. Handle
11106 riscv_preferred_stack_boundary_arg.
11107 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11108 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11109 (STACK_BOUNDARY): Set to riscv_stack_boundary.
11110 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11111 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11112 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11114 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
11117 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11118 of struct ix86_frame.
11119 (ix86_expand_epilogue): Likewise. Add a local variable for
11120 the reg_save_offset field in struct ix86_frame.
11122 2018-01-23 Bin Cheng <bin.cheng@arm.com>
11124 PR tree-optimization/82604
11125 * tree-loop-distribution.c (enum partition_kind): New enum item
11126 PKIND_PARTIAL_MEMSET.
11127 (partition_builtin_p): Support above new enum item.
11128 (generate_code_for_partition): Ditto.
11129 (compute_access_range): Differentiate cases that equality can be
11130 proven at all loops, the innermost loops or no loops.
11131 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11132 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
11133 (finalize_partitions, distribute_loop): Don't fuse partition of
11134 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11135 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11136 parloop is enabled.
11138 2018-01-23 Martin Liska <mliska@suse.cz>
11140 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11141 order to ignore the predictor.
11142 (PRED_POLYMORPHIC_CALL): Likewise.
11143 (PRED_RECURSIVE_CALL): Likewise.
11145 2018-01-23 Martin Liska <mliska@suse.cz>
11147 * tree-profile.c (tree_profiling): Print function header to
11148 aware reader which function we are working on.
11149 * value-prof.c (gimple_find_values_to_profile): Do not print
11150 not interesting value histograms.
11152 2018-01-23 Martin Liska <mliska@suse.cz>
11154 * profile-count.h (enum profile_quality): Add
11155 profile_uninitialized as the first value. Do not number values
11156 as they are zero based.
11157 (profile_count::verify): Update sanity check.
11158 (profile_probability::verify): Likewise.
11160 2018-01-23 Nathan Sidwell <nathan@acm.org>
11162 * doc/invoke.texi (ffor-scope): Deprecate.
11164 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11166 PR tree-optimization/83510
11167 * domwalk.c (set_all_edges_as_executable): New function.
11168 (dom_walker::dom_walker): Convert bool param
11169 "skip_unreachable_blocks" to enum reachability. Move setup of
11170 edge flags to set_all_edges_as_executable and only do it when
11171 reachability is REACHABLE_BLOCKS.
11172 * domwalk.h (enum dom_walker::reachability): New enum.
11173 (dom_walker::dom_walker): Convert bool param
11174 "skip_unreachable_blocks" to enum reachability.
11175 (set_all_edges_as_executable): New decl.
11176 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
11177 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
11179 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
11180 but converting true to REACHABLE_BLOCKS.
11181 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
11183 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
11184 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
11185 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
11187 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
11188 if check_all_array_refs will be called.
11190 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11192 * tree.c (selftest::test_location_wrappers): Add more test
11195 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11197 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
11198 (selftest::test_bit_in_range): Likewise.
11200 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11203 * doc/sourcebuild.texi (vect_float): Say that the selector
11204 only describes the situation when -funsafe-math-optimizations is on.
11205 (vect_float_strict): Document.
11207 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11209 PR tree-optimization/83965
11210 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11211 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
11212 instead of checking only for a reduction.
11213 (vect_recog_widen_sum_pattern): Likewise.
11215 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11217 * predict.c (probably_never_executed): Only use precise profile info.
11218 (compute_function_frequency): Skip after inlining hack since we now
11219 have quality checking.
11221 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11223 * profile-count.h (profile_probability::very_unlikely,
11224 profile_probability::unlikely, profile_probability::even): Set
11225 precision to guessed.
11227 2018-01-23 Richard Biener <rguenther@suse.de>
11229 PR tree-optimization/83963
11230 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
11231 Properly terminate dominator walk when crossing the exit edge not
11232 when visiting its source block.
11234 2018-01-23 Jakub Jelinek <jakub@redhat.com>
11237 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
11238 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
11240 2018-01-22 Jakub Jelinek <jakub@redhat.com>
11242 PR tree-optimization/83957
11243 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
11244 semicolon after for body surrounded by braces.
11246 PR tree-optimization/83081
11247 * profile-count.h (profile_probability::split): New method.
11248 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
11249 Use profile_probability::split.
11250 (do_compare_rtx_and_jump): Fix adjustment of probabilities
11251 when splitting a single conditional jump into 2.
11253 2018-01-22 David Malcolm <dmalcolm@redhat.com>
11255 PR tree-optimization/69452
11256 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
11259 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11261 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
11262 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
11263 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
11265 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11267 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
11269 * config/rl78/rl78.md (movdi): New define_expand.
11270 * config/rl78/rl78.c (rl78_split_movdi): New function.
11272 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
11275 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
11277 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
11278 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
11279 128-bit to produce an UNSPEC move to get the double word with the
11280 signbit and then a shift directly to do signbit.
11281 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
11282 implementation with a new version that just does either a direct
11283 move or a regular move. Move memory interface to separate insns.
11284 Move insns so they are next to the expander.
11285 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
11286 with signbit move. Split big and little endian case.
11287 (signbit<mode>2_dm_mem_le): Likewise.
11288 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
11289 (signbit<mode>2_dm2): Likewise.
11291 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11293 * config/rl78/rl78.md (anddi3): New define_expand.
11295 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11297 * config/rl78/rl78.md (umindi3): New define_expand.
11299 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11301 * config/rl78/rl78.md (smindi3): New define_expand.
11303 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11305 * config/rl78/rl78.md (smaxdi3): New define_expand.
11307 2018-01-22 Carl Love <cel@us.ibm.com>
11309 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
11310 LVX_V1TI): Add macro expansion.
11311 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
11312 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
11313 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
11314 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
11315 Change check to determine if the instruction is a byte reversing
11316 entry. Fix typo in comment.
11317 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
11318 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
11319 Add def_builtin calls for new builtins.
11320 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
11321 Add define_insn expansion.
11323 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11325 * config/rl78/rl78.md (umaxdi3): New define_expand.
11327 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11329 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
11330 for non-QImode registers.
11332 2018-01-22 Richard Biener <rguenther@suse.de>
11334 PR tree-optimization/83963
11335 * graphite-scop-detection.c (scop_detection::get_sese): Delay
11336 including the loop exit block.
11337 (scop_detection::merge_sese): Likewise.
11338 (scop_detection::add_scop): Do it here instead.
11340 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11342 * doc/sourcebuild.texi (arm_softfloat): Document.
11344 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
11347 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
11348 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
11349 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
11351 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11352 David Edelsohn <dje.gcc@gmail.com>
11355 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11356 Change "crset eq" to "crset 2".
11357 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11358 (*call_indirect_aix<mode>_nospec): Likewise.
11359 (*call_value_indirect_aix<mode>_nospec): Likewise.
11360 (*call_indirect_elfv2<mode>_nospec): Likewise.
11361 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
11362 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
11363 change assembly output from . to $.
11364 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11365 (indirect_jump<mode>_nospec): Change assembly output from . to $.
11366 (*tablejump<mode>_internal1_nospec): Likewise.
11368 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
11371 * config/sh/sh_optimize_sett_clrt.cc:
11372 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
11374 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
11376 PR tree-optimization/83940
11377 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
11378 offset_dt to vect_constant_def rather than vect_unknown_def_type.
11379 (vect_check_load_store_mask): Add a mask_dt_out parameter and
11380 use it to pass back the definition type.
11381 (vect_check_store_rhs): Likewise rhs_dt_out.
11382 (vect_build_gather_load_calls): Add a mask_dt argument and use
11383 it instead of a call to vect_is_simple_use.
11384 (vectorizable_store): Update calls to vect_check_load_store_mask
11385 and vect_check_store_rhs. Use the dt returned by the latter instead
11386 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
11387 instead of calls to vect_is_simple_use. Pass the scalar rather
11388 than the vector operand to vect_is_simple_use when handling
11389 second and subsequent copies of an rhs value.
11390 (vectorizable_load): Update calls to vect_check_load_store_mask
11391 and vect_build_gather_load_calls. Use the cached mask_dt and
11392 gs_info.offset_dt instead of calls to vect_is_simple_use.
11394 2018-01-20 Jakub Jelinek <jakub@redhat.com>
11396 PR middle-end/83945
11397 * tree-emutls.c: Include gimplify.h.
11398 (lower_emutls_2): New function.
11399 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
11400 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
11401 it before further processing.
11404 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
11405 UINTVAL (trueop1) instead of INTVAL (op1).
11407 2018-01-19 Jakub Jelinek <jakub@redhat.com>
11411 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
11412 INCOMING_FRAME_SP_OFFSET if not defined.
11413 (scan_trace): Add ENTRY argument. If true and
11414 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
11415 emit a note to adjust the CFA offset.
11416 (create_cfi_notes): Adjust scan_trace callers.
11417 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
11418 INCOMING_FRAME_SP_OFFSET in the CIE.
11419 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
11420 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
11422 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
11423 * doc/tm.texi: Regenerated.
11425 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11427 PR rtl-optimization/83147
11428 * lra-constraints.c (remove_inheritance_pseudos): Use
11429 lra_substitute_pseudo_within_insn.
11431 2018-01-19 Tom de Vries <tom@codesourcery.com>
11432 Cesar Philippidis <cesar@codesourcery.com>
11435 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
11437 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
11440 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
11441 spaces for function labels.
11443 2018-01-19 Martin Liska <mliska@suse.cz>
11445 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
11446 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
11447 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
11448 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
11449 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
11450 (PRED_CONST_RETURN): Change from 69 to 65.
11451 (PRED_NULL_RETURN): Change from 91 to 71.
11452 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
11453 (PRED_LOOP_GUARD): Change from 66 to 73.
11455 2018-01-19 Martin Liska <mliska@suse.cz>
11457 * predict.c (predict_insn_def): Add new assert.
11458 (struct branch_predictor): Change type to signed integer.
11459 (test_prediction_value_range): Amend test to cover
11460 PROB_UNINITIALIZED.
11461 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
11462 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
11463 (PRED_LOOP_ITERATIONS_MAX): Likewise.
11464 (PRED_LOOP_IV_COMPARE): Likewise.
11465 * predict.h (PROB_UNINITIALIZED): Define new constant.
11467 2018-01-19 Martin Liska <mliska@suse.cz>
11469 * predict.c (dump_prediction): Add new format for
11470 analyze_brprob.py script which is enabled with -details
11472 * profile-count.h (precise_p): New function.
11474 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11476 PR tree-optimization/83922
11477 * tree-vect-loop.c (vect_verify_full_masking): Return false if
11478 there are no statements that need masking.
11479 (vect_active_double_reduction_p): New function.
11480 (vect_analyze_loop_operations): Use it when handling phis that
11481 are not in the loop header.
11483 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11485 PR tree-optimization/83914
11486 * tree-vect-loop.c (vectorizable_induction): Don't convert
11487 init_expr or apply the peeling adjustment for inductions
11488 that are nested within the vectorized loop.
11490 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11492 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
11495 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11499 * function.h (gimplify_parameters): Add gimple_seq * argument.
11500 * function.c: Include gimple.h and options.h.
11501 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
11502 for the added local temporaries if needed.
11503 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
11504 if there are any parameter cleanups, wrap whole body into a
11505 try/finally with the cleanups.
11507 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
11510 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
11511 Use GET_MODE_CLASS for scalar floating point.
11513 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11517 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
11518 Fix call of call_cgraph_insertion_hooks.
11520 2018-01-18 Martin Sebor <msebor@redhat.com>
11522 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
11524 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11527 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
11530 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
11533 * common.opt: (-ffile-prefix-map): New option.
11534 * opts.c (common_handle_option): Defer it.
11535 * opts-global.c (handle_common_deferred_options): Handle it.
11536 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
11537 * file-prefix-map.h: New file.
11538 (remap_debug_filename, add_debug_prefix_map): ...here.
11539 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
11540 * final.c (debug_prefix_map, add_debug_prefix_map
11541 remap_debug_filename): Move to...
11542 * file-prefix-map.c: New file.
11543 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
11544 generalize, get rid of alloca(), use strrchr() instead of strchr().
11545 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
11546 Implement in terms of add_prefix_map().
11547 (remap_macro_filename, remap_debug_filename): Implement in term of
11549 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
11550 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
11551 * dbxout.c: Include file-prefix-map.h.
11552 * varasm.c: Likewise.
11553 * vmsdbgout.c: Likewise.
11554 * xcoffout.c: Likewise.
11555 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
11556 * doc/cppopts.texi (-fmacro-prefix-map): Document.
11557 * doc/invoke.texi (-ffile-prefix-map): Document.
11558 (-fdebug-prefix-map): Update description.
11560 2018-01-18 Martin Liska <mliska@suse.cz>
11562 * config/i386/i386.c (indirect_thunk_name): Document that also
11564 (output_indirect_thunk): Document why both instructions
11565 (pause and lfence) are generated.
11567 2018-01-18 Richard Biener <rguenther@suse.de>
11569 PR tree-optimization/83887
11570 * graphite-scop-detection.c
11571 (scop_detection::get_nearest_dom_with_single_entry): Remove.
11572 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
11573 (scop_detection::merge_sese): Re-implement with a flood-fill
11574 algorithm that properly finds a SESE region if it exists.
11576 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11579 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
11580 pointer_diff optimizations use view_convert instead of convert.
11582 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11584 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11585 Generate different code for -mno-speculate-indirect-jumps.
11586 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11587 (*call_indirect_aix<mode>): Disable for
11588 -mno-speculate-indirect-jumps.
11589 (*call_indirect_aix<mode>_nospec): New define_insn.
11590 (*call_value_indirect_aix<mode>): Disable for
11591 -mno-speculate-indirect-jumps.
11592 (*call_value_indirect_aix<mode>_nospec): New define_insn.
11593 (*sibcall_nonlocal_sysv<mode>): Generate different code for
11594 -mno-speculate-indirect-jumps.
11595 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11597 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
11599 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
11600 long double type, set the flags for noting the default long double
11601 type, even if we don't pass or return a long double type.
11603 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
11606 * ipa-inline.c (flatten_function): Do not overwrite final inlining
11609 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
11611 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
11612 support for merge[hl].
11613 (fold_mergehl_helper): New helper function.
11614 (tree-vector-builder.h): New #include for tree_vector_builder usage.
11615 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
11616 (altivec_vmrglw_direct): Add xxmrglw insn.
11618 2018-01-17 Andrew Waterman <andrew@sifive.com>
11620 * config/riscv/riscv.c (riscv_conditional_register_usage): If
11621 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
11623 2018-01-17 David Malcolm <dmalcolm@redhat.com>
11626 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
11627 call the lto_location_cache before reading the
11628 DECL_SOURCE_LOCATION of the types.
11630 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
11631 Richard Sandiford <richard.sandiford@linaro.org>
11633 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
11634 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
11635 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
11636 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
11637 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
11639 * config/aarch64/constraints.md (aarch64_movti_operand):
11641 * config/aarch64/predicates.md (Uti): Add new constraint.
11643 2018-01-17 Carl Love <cel@us.ibm.com>
11645 * config/rs6000/vsx.md (define_expand xl_len_r,
11646 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
11647 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
11649 (define_expand, define_insn): Move the shift left from the
11650 define_insn to the define_expand for lxvl and stxvl instructions.
11651 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
11652 and XL_LEN_R definitions to PURE.
11654 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
11656 * config/i386/i386.c (indirect_thunk_name): Declare regno
11657 as unsigned int. Compare regno with INVALID_REGNUM.
11658 (output_indirect_thunk): Ditto.
11659 (output_indirect_thunk_function): Ditto.
11660 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
11661 in the call to output_indirect_thunk_function.
11663 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
11665 PR middle-end/83884
11666 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
11667 rather than the size of inner_type to determine the stack slot size
11668 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
11670 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
11673 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
11676 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
11678 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
11679 endian Linux systems to optionally enable multilibs for selecting
11680 the long double type if the user configured an explicit type.
11681 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
11682 have no long double multilibs if not defined.
11683 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
11684 warn if the user used -mabi={ieee,ibm}longdouble and we built
11685 multilibs for long double.
11686 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
11687 appropriate multilib option.
11688 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
11690 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
11691 for building long double multilibs.
11692 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
11694 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
11696 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
11699 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
11701 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
11704 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
11707 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
11710 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11712 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
11713 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
11715 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
11717 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
11718 different rtl trees depending on TARGET_64BIT.
11719 (rs6000_gen_lvx): Likewise.
11721 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11723 * config/visium/visium.md (nop): Tweak comment.
11724 (hazard_nop): Likewise.
11726 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11728 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
11729 -mspeculate-indirect-jumps.
11730 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
11731 for -mno-speculate-indirect-jumps.
11732 (*call_indirect_elfv2<mode>_nospec): New define_insn.
11733 (*call_value_indirect_elfv2<mode>): Disable for
11734 -mno-speculate-indirect-jumps.
11735 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
11736 (indirect_jump): Emit different RTL for
11737 -mno-speculate-indirect-jumps.
11738 (*indirect_jump<mode>): Disable for
11739 -mno-speculate-indirect-jumps.
11740 (*indirect_jump<mode>_nospec): New define_insn.
11741 (tablejump): Emit different RTL for
11742 -mno-speculate-indirect-jumps.
11743 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
11744 (tablejumpsi_nospec): New define_expand.
11745 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
11746 (tablejumpdi_nospec): New define_expand.
11747 (*tablejump<mode>_internal1): Disable for
11748 -mno-speculate-indirect-jumps.
11749 (*tablejump<mode>_internal1_nospec): New define_insn.
11750 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
11753 2018-01-16 Artyom Skrobov tyomitch@gmail.com
11755 * caller-save.c (insert_save): Drop unnecessary parameter. All
11758 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11759 Richard Biener <rguenth@suse.de>
11762 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
11763 return early, inline manually is_gimple_sizepos. Make sure if we
11764 call gimplify_expr we don't end up with a gimple constant.
11765 * tree.c (variably_modified_type_p): Don't return true for
11766 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
11767 * gimplify.h (is_gimple_sizepos): Remove.
11769 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11771 PR tree-optimization/83857
11772 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
11773 vectorizable_live_operation for pure SLP statements.
11774 (vectorizable_live_operation): Handle PHIs.
11776 2018-01-16 Richard Biener <rguenther@suse.de>
11778 PR tree-optimization/83867
11779 * tree-vect-stmts.c (vect_transform_stmt): Precompute
11780 nested_in_vect_loop_p since the scalar stmt may get invalidated.
11782 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11785 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
11786 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
11787 If off is not INTEGER_CST, issue a may not be aligned warning
11788 rather than isn't aligned. Use isn%'t rather than isn't.
11789 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
11791 <case MULT_EXPR>: Improve the case when bottom and one of the
11792 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
11793 operand, in that case check if the other operand is multiple of
11794 bottom divided by the INTEGER_CST operand.
11796 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11799 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
11800 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
11801 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
11802 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
11803 * config/pa/pa.c (pa_function_arg_advance): Likewise.
11804 (pa_function_arg, pa_arg_partial_bytes): Likewise.
11805 (pa_function_arg_size): New function.
11807 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11809 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
11810 in a separate statement.
11812 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11814 PR tree-optimization/83847
11815 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
11816 group gathers and scatters.
11818 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11820 PR rtl-optimization/86620
11821 * params.def (max-sched-ready-insns): Bump minimum value to 1.
11823 PR rtl-optimization/83213
11824 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
11825 to last if both are JUMP_INSNs.
11827 PR tree-optimization/83843
11828 * gimple-ssa-store-merging.c
11829 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
11830 store_immediate_info for bswap/nop orig_stores.
11832 2018-01-15 Andrew Waterman <andrew@sifive.com>
11834 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
11836 <UDIV>: Increase cost if !TARGET_DIV.
11838 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
11840 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
11841 (define_attr "cr_logical_3op"): New.
11842 (cceq_ior_compare): Adjust.
11843 (cceq_ior_compare_complement): Adjust.
11844 (*cceq_rev_compare): Adjust.
11845 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
11846 (is_cracked_insn): Adjust.
11847 (insn_must_be_first_in_group): Adjust.
11848 * config/rs6000/40x.md: Adjust.
11849 * config/rs6000/440.md: Adjust.
11850 * config/rs6000/476.md: Adjust.
11851 * config/rs6000/601.md: Adjust.
11852 * config/rs6000/603.md: Adjust.
11853 * config/rs6000/6xx.md: Adjust.
11854 * config/rs6000/7450.md: Adjust.
11855 * config/rs6000/7xx.md: Adjust.
11856 * config/rs6000/8540.md: Adjust.
11857 * config/rs6000/cell.md: Adjust.
11858 * config/rs6000/e300c2c3.md: Adjust.
11859 * config/rs6000/e500mc.md: Adjust.
11860 * config/rs6000/e500mc64.md: Adjust.
11861 * config/rs6000/e5500.md: Adjust.
11862 * config/rs6000/e6500.md: Adjust.
11863 * config/rs6000/mpc.md: Adjust.
11864 * config/rs6000/power4.md: Adjust.
11865 * config/rs6000/power5.md: Adjust.
11866 * config/rs6000/power6.md: Adjust.
11867 * config/rs6000/power7.md: Adjust.
11868 * config/rs6000/power8.md: Adjust.
11869 * config/rs6000/power9.md: Adjust.
11870 * config/rs6000/rs64.md: Adjust.
11871 * config/rs6000/titan.md: Adjust.
11873 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11875 * config/i386/predicates.md (indirect_branch_operand): Rewrite
11876 ix86_indirect_branch_register logic.
11878 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11880 * config/i386/constraints.md (Bs): Update
11881 ix86_indirect_branch_register check. Don't check
11882 ix86_indirect_branch_register with GOT_memory_operand.
11884 * config/i386/predicates.md (GOT_memory_operand): Don't check
11885 ix86_indirect_branch_register here.
11886 (GOT32_symbol_operand): Likewise.
11888 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11890 * config/i386/predicates.md (constant_call_address_operand):
11891 Rewrite ix86_indirect_branch_register logic.
11892 (sibcall_insn_operand): Likewise.
11894 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11896 * config/i386/constraints.md (Bs): Replace
11897 ix86_indirect_branch_thunk_register with
11898 ix86_indirect_branch_register.
11900 * config/i386/i386.md (indirect_jump): Likewise.
11901 (tablejump): Likewise.
11902 (*sibcall_memory): Likewise.
11903 (*sibcall_value_memory): Likewise.
11904 Peepholes of indirect call and jump via memory: Likewise.
11905 * config/i386/i386.opt: Likewise.
11906 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11907 (GOT_memory_operand): Likewise.
11908 (call_insn_operand): Likewise.
11909 (sibcall_insn_operand): Likewise.
11910 (GOT32_symbol_operand): Likewise.
11912 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11914 PR middle-end/83837
11915 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
11916 type rather than type addr's type points to.
11917 (expand_omp_atomic_mutex): Likewise.
11918 (expand_omp_atomic): Likewise.
11920 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11923 * config/i386/i386.c (output_indirect_thunk_function): Use
11924 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
11925 for __x86_return_thunk.
11927 2018-01-15 Richard Biener <rguenther@suse.de>
11929 PR middle-end/83850
11930 * expmed.c (extract_bit_field_1): Fix typo.
11932 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11935 * config/arm/iterators.md (VF): New mode iterator.
11936 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
11937 Remove integer-related logic from pattern.
11938 (neon_vabd<mode>_3): Likewise.
11940 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11942 PR middle-end/82694
11943 * common.opt (fstrict-overflow): No longer an alias.
11944 (fwrapv-pointer): New option.
11945 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
11946 also for pointer types based on flag_wrapv_pointer.
11947 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
11948 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
11949 opts->x_flag_wrapv got set.
11950 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
11951 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
11952 POINTER_TYPE_OVERFLOW_UNDEFINED.
11953 * match.pd: Likewise in address comparison pattern.
11954 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
11956 2018-01-15 Richard Biener <rguenther@suse.de>
11959 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
11960 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
11961 Reset type names to their identifier if their TYPE_DECL doesn't
11962 have linkage (and thus is used for ODR and devirt).
11963 (save_debug_info_for_decl): Remove.
11964 (save_debug_info_for_type): Likewise.
11965 (add_tree_to_fld_list): Adjust.
11966 * tree-pretty-print.c (dump_generic_node): Make dumping of
11967 type names more robust.
11969 2018-01-15 Richard Biener <rguenther@suse.de>
11971 * BASE-VER: Bump to 8.0.1.
11973 2018-01-14 Martin Sebor <msebor@redhat.com>
11976 * builtins.c (check_access): Avoid warning when the no-warning bit
11979 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
11981 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
11982 * ira-color (allocno_hard_regs_compare): Likewise.
11984 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
11987 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
11988 Use .pushsection/.popsection.
11990 2018-01-14 Martin Sebor <msebor@redhat.com>
11993 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
11995 2018-01-14 Jakub Jelinek <jakub@redhat.com>
11997 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
11998 entry from extra_headers.
11999 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12000 extra_headers, make the list bitwise identical to the i?86-*-* one.
12002 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12004 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12005 -mcmodel=large with -mindirect-branch=thunk,
12006 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12007 -mfunction-return=thunk-extern.
12008 * doc/invoke.texi: Document -mcmodel=large is incompatible with
12009 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12010 -mfunction-return=thunk and -mfunction-return=thunk-extern.
12012 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12014 * config/i386/i386.c (print_reg): Print the name of the full
12015 integer register without '%'.
12016 (ix86_print_operand): Handle 'V'.
12017 * doc/extend.texi: Document 'V' modifier.
12019 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12021 * config/i386/constraints.md (Bs): Disallow memory operand for
12022 -mindirect-branch-register.
12024 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12025 (GOT_memory_operand): Likewise.
12026 (call_insn_operand): Likewise.
12027 (sibcall_insn_operand): Likewise.
12028 (GOT32_symbol_operand): Likewise.
12029 * config/i386/i386.md (indirect_jump): Call convert_memory_address
12030 for -mindirect-branch-register.
12031 (tablejump): Likewise.
12032 (*sibcall_memory): Likewise.
12033 (*sibcall_value_memory): Likewise.
12034 Disallow peepholes of indirect call and jump via memory for
12035 -mindirect-branch-register.
12036 (*call_pop): Replace m with Bw.
12037 (*call_value_pop): Likewise.
12038 (*sibcall_pop_memory): Replace m with Bs.
12039 * config/i386/i386.opt (mindirect-branch-register): New option.
12040 * doc/invoke.texi: Document -mindirect-branch-register option.
12042 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12044 * config/i386/i386-protos.h (ix86_output_function_return): New.
12045 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12046 set function_return_type.
12047 (indirect_thunk_name): Add ret_p to indicate thunk for function
12049 (output_indirect_thunk_function): Pass false to
12050 indirect_thunk_name.
12051 (ix86_output_indirect_branch_via_reg): Likewise.
12052 (ix86_output_indirect_branch_via_push): Likewise.
12053 (output_indirect_thunk_function): Create alias for function
12054 return thunk if regno < 0.
12055 (ix86_output_function_return): New function.
12056 (ix86_handle_fndecl_attribute): Handle function_return.
12057 (ix86_attribute_table): Add function_return.
12058 * config/i386/i386.h (machine_function): Add
12059 function_return_type.
12060 * config/i386/i386.md (simple_return_internal): Use
12061 ix86_output_function_return.
12062 (simple_return_internal_long): Likewise.
12063 * config/i386/i386.opt (mfunction-return=): New option.
12064 (indirect_branch): Mention -mfunction-return=.
12065 * doc/extend.texi: Document function_return function attribute.
12066 * doc/invoke.texi: Document -mfunction-return= option.
12068 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12070 * config/i386/i386-opts.h (indirect_branch): New.
12071 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12072 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12073 with local indirect jump when converting indirect call and jump.
12074 (ix86_set_indirect_branch_type): New.
12075 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12076 (indirectlabelno): New.
12077 (indirect_thunk_needed): Likewise.
12078 (indirect_thunk_bnd_needed): Likewise.
12079 (indirect_thunks_used): Likewise.
12080 (indirect_thunks_bnd_used): Likewise.
12081 (INDIRECT_LABEL): Likewise.
12082 (indirect_thunk_name): Likewise.
12083 (output_indirect_thunk): Likewise.
12084 (output_indirect_thunk_function): Likewise.
12085 (ix86_output_indirect_branch_via_reg): Likewise.
12086 (ix86_output_indirect_branch_via_push): Likewise.
12087 (ix86_output_indirect_branch): Likewise.
12088 (ix86_output_indirect_jmp): Likewise.
12089 (ix86_code_end): Call output_indirect_thunk_function if needed.
12090 (ix86_output_call_insn): Call ix86_output_indirect_branch if
12092 (ix86_handle_fndecl_attribute): Handle indirect_branch.
12093 (ix86_attribute_table): Add indirect_branch.
12094 * config/i386/i386.h (machine_function): Add indirect_branch_type
12095 and has_local_indirect_jump.
12096 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12098 (tablejump): Likewise.
12099 (*indirect_jump): Use ix86_output_indirect_jmp.
12100 (*tablejump_1): Likewise.
12101 (simple_return_indirect_internal): Likewise.
12102 * config/i386/i386.opt (mindirect-branch=): New option.
12103 (indirect_branch): New.
12106 (thunk-inline): Likewise.
12107 (thunk-extern): Likewise.
12108 * doc/extend.texi: Document indirect_branch function attribute.
12109 * doc/invoke.texi: Document -mindirect-branch= option.
12111 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
12114 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12116 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
12118 * ipa-inline.c (want_inline_small_function_p): Return false if
12119 inlining has already failed with CIF_FINAL_ERROR.
12120 (update_caller_keys): Call want_inline_small_function_p before
12122 (update_callee_keys): Likewise.
12124 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12126 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12128 (rs6000_quadword_masked_address_p): Likewise.
12129 (quad_aligned_load_p): Likewise.
12130 (quad_aligned_store_p): Likewise.
12131 (const_load_sequence_p): Add comment to describe the outer-most loop.
12132 (mimic_memory_attributes_and_flags): New function.
12133 (rs6000_gen_stvx): Likewise.
12134 (replace_swapped_aligned_store): Likewise.
12135 (rs6000_gen_lvx): Likewise.
12136 (replace_swapped_aligned_load): Likewise.
12137 (replace_swapped_load_constant): Capitalize argument name in
12138 comment describing this function.
12139 (rs6000_analyze_swaps): Add a third pass to search for vector loads
12140 and stores that access quad-word aligned addresses and replace
12141 with stvx or lvx instructions when appropriate.
12142 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12143 New function prototype.
12144 (rs6000_quadword_masked_address_p): Likewise.
12145 (rs6000_gen_lvx): Likewise.
12146 (rs6000_gen_stvx): Likewise.
12147 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12148 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12149 when memory address is aligned.
12150 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12151 this split to select lvx instruction when memory address is aligned.
12152 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12153 instruction when memory address is aligned.
12154 (*vsx_le_perm_load_v16qi): Likewise.
12155 (four unnamed splitters): Modify to select the stvx instruction
12156 when memory is aligned.
12158 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
12160 * predict.c (determine_unlikely_bbs): Handle correctly BBs
12161 which appears in the queue multiple times.
12163 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12164 Alan Hayward <alan.hayward@arm.com>
12165 David Sherwood <david.sherwood@arm.com>
12167 * tree-vectorizer.h (vec_lower_bound): New structure.
12168 (_loop_vec_info): Add check_nonzero and lower_bounds.
12169 (LOOP_VINFO_CHECK_NONZERO): New macro.
12170 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
12171 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
12172 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
12173 fields. Make seg_len the distance travelled, not including the
12175 (dr_direction_indicator): Declare.
12176 (dr_zero_step_indicator): Likewise.
12177 (dr_known_forward_stride_p): Likewise.
12178 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
12180 (runtime_alias_check_p): Allow runtime alias checks with
12182 (operator ==): Compare access_size and align.
12183 (prune_runtime_alias_test_list): Rework for new distinction between
12184 the access_size and seg_len.
12185 (create_intersect_range_checks_index): Likewise. Cope with polynomial
12187 (get_segment_min_max): New function.
12188 (create_intersect_range_checks): Use it.
12189 (dr_step_indicator): New function.
12190 (dr_direction_indicator): Likewise.
12191 (dr_zero_step_indicator): Likewise.
12192 (dr_known_forward_stride_p): Likewise.
12193 * tree-loop-distribution.c (data_ref_segment_size): Return
12194 DR_STEP * (niters - 1).
12195 (compute_alias_check_pairs): Update call to the dr_with_seg_len
12197 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
12198 (vect_preserves_scalar_order_p): New function, split out from...
12199 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
12200 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
12201 (vect_vfa_access_size): New function.
12202 (vect_vfa_align): Likewise.
12203 (vect_compile_time_alias): Take access_size_a and access_b arguments.
12204 (dump_lower_bound): New function.
12205 (vect_check_lower_bound): Likewise.
12206 (vect_small_gap_p): Likewise.
12207 (vectorizable_with_step_bound_p): Likewise.
12208 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
12209 depencies if the vectorization factor is 1. Convert the checks
12210 for nonzero steps into checks on the bounds of DR_STEP. Try using
12211 a bunds check for variable steps if the minimum required step is
12212 relatively small. Update calls to the dr_with_seg_len
12213 constructor and to vect_compile_time_alias.
12214 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
12216 (vect_loop_versioning): Call it.
12217 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
12219 (vect_estimate_min_profitable_iters): Account for any bounds checks.
12221 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12222 Alan Hayward <alan.hayward@arm.com>
12223 David Sherwood <david.sherwood@arm.com>
12225 * doc/sourcebuild.texi (vect_scatter_store): Document.
12226 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
12228 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
12230 * genopinit.c (main): Add supports_vec_scatter_store and
12231 supports_vec_scatter_store_cached to target_optabs.
12232 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
12233 IFN_MASK_SCATTER_STORE.
12234 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
12236 * internal-fn.h (internal_store_fn_p): Declare.
12237 (internal_fn_stored_value_index): Likewise.
12238 * internal-fn.c (scatter_store_direct): New macro.
12239 (expand_scatter_store_optab_fn): New function.
12240 (direct_scatter_store_optab_supported_p): New macro.
12241 (internal_store_fn_p): New function.
12242 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
12243 IFN_MASK_SCATTER_STORE.
12244 (internal_fn_mask_index): Likewise.
12245 (internal_fn_stored_value_index): New function.
12246 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
12247 for scatter stores.
12248 * optabs-query.h (supports_vec_scatter_store_p): Declare.
12249 * optabs-query.c (supports_vec_scatter_store_p): New function.
12250 * tree-vectorizer.h (vect_get_store_rhs): Declare.
12251 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
12252 true for scatter stores.
12253 (vect_gather_scatter_fn_p): Handle scatter stores too.
12254 (vect_check_gather_scatter): Consider using scatter stores if
12255 supports_vec_scatter_store_p.
12256 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
12257 scatter stores too.
12258 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12259 internal_fn_stored_value_index.
12260 (check_load_store_masking): Handle scatter stores too.
12261 (vect_get_store_rhs): Make public.
12262 (vectorizable_call): Use internal_store_fn_p.
12263 (vectorizable_store): Handle scatter store internal functions.
12264 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
12265 when deciding whether the end of the group has been reached.
12266 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
12267 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
12268 (mask_scatter_store<mode>): New insns.
12270 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12271 Alan Hayward <alan.hayward@arm.com>
12272 David Sherwood <david.sherwood@arm.com>
12274 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
12275 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
12276 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
12278 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
12279 Use vect_truncate_gather_scatter_offset if we can't treat the
12280 operation as a normal gather load or scatter store.
12281 (get_group_load_store_type): Take the gather_scatter_info
12282 as argument. Try using a gather load or scatter store for
12283 single-element groups.
12284 (get_load_store_type): Update calls to get_group_load_store_type
12285 and vect_use_strided_gather_scatters_p.
12287 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12288 Alan Hayward <alan.hayward@arm.com>
12289 David Sherwood <david.sherwood@arm.com>
12291 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
12292 optional tree argument.
12293 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
12295 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
12296 but continue to use the current value as a fallback.
12297 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
12298 to compare the updates.
12299 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
12300 (get_load_store_type): Use it when handling a strided access.
12301 (vect_get_strided_load_store_ops): New function.
12302 (vect_get_data_ptr_increment): Likewise.
12303 (vectorizable_load): Handle strided gather loads. Always pass
12304 a step to vect_create_data_ref_ptr and bump_vector_ptr.
12306 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12307 Alan Hayward <alan.hayward@arm.com>
12308 David Sherwood <david.sherwood@arm.com>
12310 * doc/md.texi (gather_load@var{m}): Document.
12311 (mask_gather_load@var{m}): Likewise.
12312 * genopinit.c (main): Add supports_vec_gather_load and
12313 supports_vec_gather_load_cached to target_optabs.
12314 * optabs-tree.c (init_tree_optimization_optabs): Use
12315 ggc_cleared_alloc to allocate target_optabs.
12316 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
12317 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
12319 * internal-fn.h (internal_load_fn_p): Declare.
12320 (internal_gather_scatter_fn_p): Likewise.
12321 (internal_fn_mask_index): Likewise.
12322 (internal_gather_scatter_fn_supported_p): Likewise.
12323 * internal-fn.c (gather_load_direct): New macro.
12324 (expand_gather_load_optab_fn): New function.
12325 (direct_gather_load_optab_supported_p): New macro.
12326 (direct_internal_fn_optab): New function.
12327 (internal_load_fn_p): Likewise.
12328 (internal_gather_scatter_fn_p): Likewise.
12329 (internal_fn_mask_index): Likewise.
12330 (internal_gather_scatter_fn_supported_p): Likewise.
12331 * optabs-query.c (supports_at_least_one_mode_p): New function.
12332 (supports_vec_gather_load_p): Likewise.
12333 * optabs-query.h (supports_vec_gather_load_p): Declare.
12334 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
12335 and memory_type field.
12336 (NUM_PATTERNS): Bump to 15.
12337 * tree-vect-data-refs.c: Include internal-fn.h.
12338 (vect_gather_scatter_fn_p): New function.
12339 (vect_describe_gather_scatter_call): Likewise.
12340 (vect_check_gather_scatter): Try using internal functions for
12341 gather loads. Recognize existing calls to a gather load function.
12342 (vect_analyze_data_refs): Consider using gather loads if
12343 supports_vec_gather_load_p.
12344 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
12345 (vect_get_gather_scatter_offset_type): Likewise.
12346 (vect_convert_mask_for_vectype): Likewise.
12347 (vect_add_conversion_to_patterm): Likewise.
12348 (vect_try_gather_scatter_pattern): Likewise.
12349 (vect_recog_gather_scatter_pattern): New pattern recognizer.
12350 (vect_vect_recog_func_ptrs): Add it.
12351 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12352 internal_fn_mask_index and internal_gather_scatter_fn_p.
12353 (check_load_store_masking): Take the gather_scatter_info as an
12354 argument and handle gather loads.
12355 (vect_get_gather_scatter_ops): New function.
12356 (vectorizable_call): Check internal_load_fn_p.
12357 (vectorizable_load): Likewise. Handle gather load internal
12359 (vectorizable_store): Update call to check_load_store_masking.
12360 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
12361 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
12362 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
12363 (aarch64_gather_scale_operand_d): New predicates.
12364 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
12365 (mask_gather_load<mode>): New insns.
12367 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12368 Alan Hayward <alan.hayward@arm.com>
12369 David Sherwood <david.sherwood@arm.com>
12371 * optabs.def (fold_left_plus_optab): New optab.
12372 * doc/md.texi (fold_left_plus_@var{m}): Document.
12373 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
12374 * internal-fn.c (fold_left_direct): Define.
12375 (expand_fold_left_optab_fn): Likewise.
12376 (direct_fold_left_optab_supported_p): Likewise.
12377 * fold-const-call.c (fold_const_fold_left): New function.
12378 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
12379 * tree-parloops.c (valid_reduction_p): New function.
12380 (gather_scalar_reductions): Use it.
12381 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
12382 (vect_finish_replace_stmt): Declare.
12383 * tree-vect-loop.c (fold_left_reduction_fn): New function.
12384 (needs_fold_left_reduction_p): New function, split out from...
12385 (vect_is_simple_reduction): ...here. Accept reductions that
12386 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
12387 (vect_force_simple_reduction): Also store the reduction type in
12388 the assignment's STMT_VINFO_REDUC_TYPE.
12389 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
12390 (merge_with_identity): New function.
12391 (vect_expand_fold_left): Likewise.
12392 (vectorize_fold_left_reduction): Likewise.
12393 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
12394 scalar phi in place for it. Check for target support and reject
12395 cases that would reassociate the operation. Defer the transform
12396 phase to vectorize_fold_left_reduction.
12397 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
12398 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
12399 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
12401 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12403 * tree-if-conv.c (predicate_mem_writes): Remove redundant
12404 call to ifc_temp_var.
12406 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12407 Alan Hayward <alan.hayward@arm.com>
12408 David Sherwood <david.sherwood@arm.com>
12410 * target.def (legitimize_address_displacement): Take the original
12411 offset as a poly_int.
12412 * targhooks.h (default_legitimize_address_displacement): Update
12414 * targhooks.c (default_legitimize_address_displacement): Likewise.
12415 * doc/tm.texi: Regenerate.
12416 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
12417 as an argument, moving assert of ad->disp == ad->disp_term to...
12418 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
12419 Try calling targetm.legitimize_address_displacement before expanding
12420 the address rather than afterwards, and adjust for the new interface.
12421 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
12422 Match the new hook interface. Handle SVE addresses.
12423 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
12424 new hook interface.
12426 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12428 * Makefile.in (OBJS): Add early-remat.o.
12429 * target.def (select_early_remat_modes): New hook.
12430 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
12431 * doc/tm.texi: Regenerate.
12432 * targhooks.h (default_select_early_remat_modes): Declare.
12433 * targhooks.c (default_select_early_remat_modes): New function.
12434 * timevar.def (TV_EARLY_REMAT): New timevar.
12435 * passes.def (pass_early_remat): New pass.
12436 * tree-pass.h (make_pass_early_remat): Declare.
12437 * early-remat.c: New file.
12438 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
12440 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
12442 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12443 Alan Hayward <alan.hayward@arm.com>
12444 David Sherwood <david.sherwood@arm.com>
12446 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
12447 vfm1 with a bound_epilog parameter.
12448 (vect_do_peeling): Update calls accordingly, and move the prologue
12449 call earlier in the function. Treat the base bound_epilog as 0 for
12450 fully-masked loops and retain vf - 1 for other loops. Add 1 to
12451 this base when peeling for gaps.
12452 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
12453 with fully-masked loops.
12454 (vect_estimate_min_profitable_iters): Handle the single peeled
12455 iteration in that case.
12457 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12458 Alan Hayward <alan.hayward@arm.com>
12459 David Sherwood <david.sherwood@arm.com>
12461 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
12462 single-element interleaving even if the size is not a power of 2.
12463 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
12464 accesses for single-element interleaving if the group size is
12467 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12468 Alan Hayward <alan.hayward@arm.com>
12469 David Sherwood <david.sherwood@arm.com>
12471 * doc/md.texi (fold_extract_last_@var{m}): Document.
12472 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
12473 * optabs.def (fold_extract_last_optab): New optab.
12474 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
12475 * internal-fn.c (fold_extract_direct): New macro.
12476 (expand_fold_extract_optab_fn): Likewise.
12477 (direct_fold_extract_optab_supported_p): Likewise.
12478 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
12479 * tree-vect-loop.c (vect_model_reduction_cost): Handle
12480 EXTRACT_LAST_REDUCTION.
12481 (get_initial_def_for_reduction): Do not create an initial vector
12482 for EXTRACT_LAST_REDUCTION reductions.
12483 (vectorizable_reduction): Leave the scalar phi in place for
12484 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
12485 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
12486 epilogue code for EXTRACT_LAST_REDUCTION and defer the
12487 transform phase to vectorizable_condition.
12488 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
12490 (vect_finish_stmt_generation): ...here.
12491 (vect_finish_replace_stmt): New function.
12492 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
12493 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
12495 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
12497 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12498 Alan Hayward <alan.hayward@arm.com>
12499 David Sherwood <david.sherwood@arm.com>
12501 * doc/md.texi (extract_last_@var{m}): Document.
12502 * optabs.def (extract_last_optab): New optab.
12503 * internal-fn.def (EXTRACT_LAST): New internal function.
12504 * internal-fn.c (cond_unary_direct): New macro.
12505 (expand_cond_unary_optab_fn): Likewise.
12506 (direct_cond_unary_optab_supported_p): Likewise.
12507 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
12508 loops using EXTRACT_LAST.
12509 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
12510 (extract_last_<mode>): ...this optab.
12511 (vec_extract<mode><Vel>): Update accordingly.
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 * target.def (empty_mask_is_expensive): New hook.
12518 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
12519 * doc/tm.texi: Regenerate.
12520 * targhooks.h (default_empty_mask_is_expensive): Declare.
12521 * targhooks.c (default_empty_mask_is_expensive): New function.
12522 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
12523 if the target says that empty masks are expensive.
12524 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
12526 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
12528 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12529 Alan Hayward <alan.hayward@arm.com>
12530 David Sherwood <david.sherwood@arm.com>
12532 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
12533 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
12534 (vect_use_loop_mask_for_alignment_p): New function.
12535 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
12536 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
12537 niters_skip argument. Make sure that the first niters_skip elements
12538 of the first iteration are inactive.
12539 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
12540 Update call to vect_set_loop_masks_directly.
12541 (get_misalign_in_elems): New function, split out from...
12542 (vect_gen_prolog_loop_niters): ...here.
12543 (vect_update_init_of_dr): Take a code argument that specifies whether
12544 the adjustment should be added or subtracted.
12545 (vect_update_init_of_drs): Likewise.
12546 (vect_prepare_for_masked_peels): New function.
12547 (vect_do_peeling): Skip prologue peeling if we're using a mask
12548 instead. Update call to vect_update_inits_of_drs.
12549 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12551 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
12552 alignment. Do not include the number of peeled iterations in
12553 the minimum threshold in that case.
12554 (vectorizable_induction): Adjust the start value down by
12555 LOOP_VINFO_MASK_SKIP_NITERS iterations.
12556 (vect_transform_loop): Call vect_prepare_for_masked_peels.
12557 Take the number of skipped iterations into account when calculating
12559 * tree-vect-stmts.c (vect_gen_while_not): New function.
12561 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12562 Alan Hayward <alan.hayward@arm.com>
12563 David Sherwood <david.sherwood@arm.com>
12565 * doc/sourcebuild.texi (vect_fully_masked): Document.
12566 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
12567 default value to 0.
12568 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
12570 (vect_analyze_loop_2): ...here. Don't check the vectorization
12571 factor against the number of loop iterations if the loop is
12574 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12575 Alan Hayward <alan.hayward@arm.com>
12576 David Sherwood <david.sherwood@arm.com>
12578 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
12579 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
12580 (dump_groups): Update accordingly.
12581 (iv_use::mem_type): New member variable.
12582 (address_p): New function.
12583 (record_use): Add a mem_type argument and initialize the new
12585 (record_group_use): Add a mem_type argument. Use address_p.
12586 Remove obsolete null checks of base_object. Update call to record_use.
12587 (find_interesting_uses_op): Update call to record_group_use.
12588 (find_interesting_uses_cond): Likewise.
12589 (find_interesting_uses_address): Likewise.
12590 (get_mem_type_for_internal_fn): New function.
12591 (find_address_like_use): Likewise.
12592 (find_interesting_uses_stmt): Try find_address_like_use before
12593 calling find_interesting_uses_op.
12594 (addr_offset_valid_p): Use the iv mem_type field as the type
12595 of the addressed memory.
12596 (add_autoinc_candidates): Likewise.
12597 (get_address_cost): Likewise.
12598 (split_small_address_groups_p): Use address_p.
12599 (split_address_groups): Likewise.
12600 (add_iv_candidate_for_use): Likewise.
12601 (autoinc_possible_for_pair): Likewise.
12602 (rewrite_groups): Likewise.
12603 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
12604 (determine_group_iv_cost): Update after split of USE_ADDRESS.
12605 (get_alias_ptr_type_for_ptr_address): New function.
12606 (rewrite_use_address): Rewrite address uses in calls that were
12607 identified by find_address_like_use.
12609 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12610 Alan Hayward <alan.hayward@arm.com>
12611 David Sherwood <david.sherwood@arm.com>
12613 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
12615 * gimple-expr.h (is_gimple_addressable: Likewise.
12616 * gimple-expr.c (is_gimple_address): Likewise.
12617 * internal-fn.c (expand_call_mem_ref): New function.
12618 (expand_mask_load_optab_fn): Use it.
12619 (expand_mask_store_optab_fn): Likewise.
12621 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12622 Alan Hayward <alan.hayward@arm.com>
12623 David Sherwood <david.sherwood@arm.com>
12625 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
12626 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
12627 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
12628 (cond_umax@var{mode}): Document.
12629 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
12630 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
12631 (cond_umin_optab, cond_umax_optab): New optabs.
12632 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
12633 (COND_IOR, COND_XOR): New internal functions.
12634 * internal-fn.h (get_conditional_internal_fn): Declare.
12635 * internal-fn.c (cond_binary_direct): New macro.
12636 (expand_cond_binary_optab_fn): Likewise.
12637 (direct_cond_binary_optab_supported_p): Likewise.
12638 (get_conditional_internal_fn): New function.
12639 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
12640 Cope with reduction statements that are vectorized as calls rather
12642 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
12643 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
12644 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
12645 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12646 (UNSPEC_COND_EOR): New unspecs.
12647 (optab): Add mappings for them.
12648 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
12649 (sve_int_op, sve_fp_op): New int attributes.
12651 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12652 Alan Hayward <alan.hayward@arm.com>
12653 David Sherwood <david.sherwood@arm.com>
12655 * optabs.def (while_ult_optab): New optab.
12656 * doc/md.texi (while_ult@var{m}@var{n}): Document.
12657 * internal-fn.def (WHILE_ULT): New internal function.
12658 * internal-fn.h (direct_internal_fn_supported_p): New override
12659 that takes two types as argument.
12660 * internal-fn.c (while_direct): New macro.
12661 (expand_while_optab_fn): New function.
12662 (convert_optab_supported_p): Likewise.
12663 (direct_while_optab_supported_p): New macro.
12664 * wide-int.h (wi::udiv_ceil): New function.
12665 * tree-vectorizer.h (rgroup_masks): New structure.
12666 (vec_loop_masks): New typedef.
12667 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
12668 and fully_masked_p.
12669 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
12670 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
12671 (vect_max_vf): New function.
12672 (slpeel_make_loop_iterate_ntimes): Delete.
12673 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
12674 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
12675 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
12676 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
12677 internal-fn.h, stor-layout.h and optabs-query.h.
12678 (vect_set_loop_mask): New function.
12679 (add_preheader_seq): Likewise.
12680 (add_header_seq): Likewise.
12681 (interleave_supported_p): Likewise.
12682 (vect_maybe_permute_loop_masks): Likewise.
12683 (vect_set_loop_masks_directly): Likewise.
12684 (vect_set_loop_condition_masked): Likewise.
12685 (vect_set_loop_condition_unmasked): New function, split out from
12686 slpeel_make_loop_iterate_ntimes.
12687 (slpeel_make_loop_iterate_ntimes): Rename to..
12688 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
12689 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
12690 (vect_do_peeling): Update call accordingly.
12691 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
12693 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12694 mask_compare_type, can_fully_mask_p and fully_masked_p.
12695 (release_vec_loop_masks): New function.
12696 (_loop_vec_info): Use it to free the loop masks.
12697 (can_produce_all_loop_masks_p): New function.
12698 (vect_get_max_nscalars_per_iter): Likewise.
12699 (vect_verify_full_masking): Likewise.
12700 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
12701 retries, and free the mask rgroups before retrying. Check loop-wide
12702 reasons for disallowing fully-masked loops. Make the final decision
12703 about whether use a fully-masked loop or not.
12704 (vect_estimate_min_profitable_iters): Do not assume that peeling
12705 for the number of iterations will be needed for fully-masked loops.
12706 (vectorizable_reduction): Disable fully-masked loops.
12707 (vectorizable_live_operation): Likewise.
12708 (vect_halve_mask_nunits): New function.
12709 (vect_double_mask_nunits): Likewise.
12710 (vect_record_loop_mask): Likewise.
12711 (vect_get_loop_mask): Likewise.
12712 (vect_transform_loop): Handle the case in which the final loop
12713 iteration might handle a partial vector. Call vect_set_loop_condition
12714 instead of slpeel_make_loop_iterate_ntimes.
12715 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
12716 (check_load_store_masking): New function.
12717 (prepare_load_store_mask): Likewise.
12718 (vectorizable_store): Handle fully-masked loops.
12719 (vectorizable_load): Likewise.
12720 (supportable_widening_operation): Use vect_halve_mask_nunits for
12722 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
12723 (vect_gen_while): New function.
12724 * config/aarch64/aarch64.md (umax<mode>3): New expander.
12725 (aarch64_uqdec<mode>): New insn.
12727 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12728 Alan Hayward <alan.hayward@arm.com>
12729 David Sherwood <david.sherwood@arm.com>
12731 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
12732 (reduc_xor_scal_optab): New optabs.
12733 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
12734 (reduc_xor_scal_@var{m}): Document.
12735 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
12736 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
12737 internal functions.
12738 * fold-const-call.c (fold_const_call): Handle them.
12739 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
12740 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
12741 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
12742 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
12743 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
12744 (UNSPEC_XORV): New unspecs.
12745 (optab): Add entries for them.
12746 (BITWISEV): New int iterator.
12747 (bit_reduc_op): New int attributes.
12749 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12750 Alan Hayward <alan.hayward@arm.com>
12751 David Sherwood <david.sherwood@arm.com>
12753 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
12754 * internal-fn.def (VEC_SHL_INSERT): New internal function.
12755 * optabs.def (vec_shl_insert_optab): New optab.
12756 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
12757 (duplicate_and_interleave): Likewise.
12758 * tree-vect-loop.c: Include internal-fn.h.
12759 (neutral_op_for_slp_reduction): New function, split out from
12760 get_initial_defs_for_reduction.
12761 (get_initial_def_for_reduction): Handle option 2 for variable-length
12762 vectors by loading the neutral value into a vector and then shifting
12763 the initial value into element 0.
12764 (get_initial_defs_for_reduction): Replace the code argument with
12765 the neutral value calculated by neutral_op_for_slp_reduction.
12766 Use gimple_build_vector for constant-length vectors.
12767 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
12768 but the first group_size elements have a neutral value.
12769 Use duplicate_and_interleave otherwise.
12770 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
12771 Update call to get_initial_defs_for_reduction. Handle SLP
12772 reductions for variable-length vectors by creating one vector
12773 result for each scalar result, with the elements associated
12774 with other scalar results stubbed out with the neutral value.
12775 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
12776 Require IFN_VEC_SHL_INSERT for double reductions on
12777 variable-length vectors, or SLP reductions that have
12778 a neutral value. Require can_duplicate_and_interleave_p
12779 support for variable-length unchained SLP reductions if there
12780 is no neutral value, such as for MIN/MAX reductions. Also require
12781 the number of vector elements to be a multiple of the number of
12782 SLP statements when doing variable-length unchained SLP reductions.
12783 Update call to vect_create_epilog_for_reduction.
12784 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
12785 and remove initial values.
12786 (duplicate_and_interleave): Make public.
12787 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
12788 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
12790 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12791 Alan Hayward <alan.hayward@arm.com>
12792 David Sherwood <david.sherwood@arm.com>
12794 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
12795 (can_duplicate_and_interleave_p): New function.
12796 (vect_get_and_check_slp_defs): Take the vector of statements
12797 rather than just the current one. Remove excess parentheses.
12798 Restriction rejectinon of vect_constant_def and vect_external_def
12799 for variable-length vectors to boolean types, or types for which
12800 can_duplicate_and_interleave_p is false.
12801 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
12802 (duplicate_and_interleave): New function.
12803 (vect_get_constant_vectors): Use gimple_build_vector for
12804 constant-length vectors and suitable variable-length constant
12805 vectors. Use duplicate_and_interleave for other variable-length
12806 vectors. Don't defer the update when inserting new statements.
12808 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12809 Alan Hayward <alan.hayward@arm.com>
12810 David Sherwood <david.sherwood@arm.com>
12812 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
12813 min_profitable_iters doesn't go negative.
12815 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12816 Alan Hayward <alan.hayward@arm.com>
12817 David Sherwood <david.sherwood@arm.com>
12819 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
12820 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
12821 * optabs.def (vec_mask_load_lanes_optab): New optab.
12822 (vec_mask_store_lanes_optab): Likewise.
12823 * internal-fn.def (MASK_LOAD_LANES): New internal function.
12824 (MASK_STORE_LANES): Likewise.
12825 * internal-fn.c (mask_load_lanes_direct): New macro.
12826 (mask_store_lanes_direct): Likewise.
12827 (expand_mask_load_optab_fn): Handle masked operations.
12828 (expand_mask_load_lanes_optab_fn): New macro.
12829 (expand_mask_store_optab_fn): Handle masked operations.
12830 (expand_mask_store_lanes_optab_fn): New macro.
12831 (direct_mask_load_lanes_optab_supported_p): Likewise.
12832 (direct_mask_store_lanes_optab_supported_p): Likewise.
12833 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
12835 (vect_load_lanes_supported): Likewise.
12836 * tree-vect-data-refs.c (strip_conversion): New function.
12837 (can_group_stmts_p): Likewise.
12838 (vect_analyze_data_ref_accesses): Use it instead of checking
12839 for a pair of assignments.
12840 (vect_store_lanes_supported): Take a masked_p parameter.
12841 (vect_load_lanes_supported): Likewise.
12842 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
12843 vect_store_lanes_supported and vect_load_lanes_supported.
12844 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12845 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
12846 parameter. Don't allow gaps for masked accesses.
12847 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
12848 and vect_load_lanes_supported.
12849 (get_load_store_type): Take a masked_p parameter and update
12850 call to get_group_load_store_type.
12851 (vectorizable_store): Update call to get_load_store_type.
12852 Handle IFN_MASK_STORE_LANES.
12853 (vectorizable_load): Update call to get_load_store_type.
12854 Handle IFN_MASK_LOAD_LANES.
12856 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12857 Alan Hayward <alan.hayward@arm.com>
12858 David Sherwood <david.sherwood@arm.com>
12860 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
12862 * config/aarch64/aarch64-protos.h
12863 (aarch64_sve_struct_memory_operand_p): Declare.
12864 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
12865 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
12866 (VPRED, vpred): Handle SVE structure modes.
12867 * config/aarch64/constraints.md (Utx): New constraint.
12868 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
12869 (aarch64_sve_struct_nonimmediate_operand): New predicates.
12870 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
12871 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
12872 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
12873 structure modes. Split into pieces after RA.
12874 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
12875 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
12877 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
12878 SVE structure modes.
12879 (aarch64_classify_address): Likewise.
12880 (sizetochar): Move earlier in file.
12881 (aarch64_print_operand): Handle SVE register lists.
12882 (aarch64_array_mode): New function.
12883 (aarch64_sve_struct_memory_operand_p): Likewise.
12884 (TARGET_ARRAY_MODE): Redefine.
12886 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12887 Alan Hayward <alan.hayward@arm.com>
12888 David Sherwood <david.sherwood@arm.com>
12890 * target.def (array_mode): New target hook.
12891 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
12892 * doc/tm.texi: Regenerate.
12893 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
12894 * hooks.c (hook_optmode_mode_uhwi_none): New function.
12895 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
12896 targetm.array_mode.
12897 * stor-layout.c (mode_for_array): Likewise. Support polynomial
12900 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12901 Alan Hayward <alan.hayward@arm.com>
12902 David Sherwood <david.sherwood@arm.com>
12904 * fold-const.c (fold_binary_loc): Check the argument types
12905 rather than the result type when testing for a vector operation.
12907 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12909 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
12910 * doc/tm.texi: Regenerate.
12912 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12913 Alan Hayward <alan.hayward@arm.com>
12914 David Sherwood <david.sherwood@arm.com>
12916 * doc/invoke.texi (-msve-vector-bits=): Document new option.
12917 (sve): Document new AArch64 extension.
12918 * doc/md.texi (w): Extend the description of the AArch64
12919 constraint to include SVE vectors.
12920 (Upl, Upa): Document new AArch64 predicate constraints.
12921 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
12923 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
12924 (msve-vector-bits=): New option.
12925 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
12926 SVE when these are disabled.
12927 (sve): New extension.
12928 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
12929 modes. Adjust their number of units based on aarch64_sve_vg.
12930 (MAX_BITSIZE_MODE_ANY_MODE): Define.
12931 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
12932 aarch64_addr_query_type.
12933 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
12934 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
12935 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
12936 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
12937 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
12938 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
12939 (aarch64_simd_imm_zero_p): Delete.
12940 (aarch64_check_zero_based_sve_index_immediate): Declare.
12941 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12942 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12943 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12944 (aarch64_sve_float_mul_immediate_p): Likewise.
12945 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12946 rather than an rtx.
12947 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
12948 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
12949 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
12950 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
12951 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
12952 (aarch64_regmode_natural_size): Likewise.
12953 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
12954 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
12956 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
12957 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
12958 for VG and the SVE predicate registers.
12959 (V_ALIASES): Add a "z"-prefixed alias.
12960 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
12961 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
12962 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
12963 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
12964 (REG_CLASS_NAMES): Add entries for them.
12965 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
12966 and the predicate registers.
12967 (aarch64_sve_vg): Declare.
12968 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
12969 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
12970 (REGMODE_NATURAL_SIZE): Define.
12971 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
12973 * config/aarch64/aarch64.c: Include cfgrtl.h.
12974 (simd_immediate_info): Add a constructor for series vectors,
12975 and an associated step field.
12976 (aarch64_sve_vg): New variable.
12977 (aarch64_dbx_register_number): Handle VG and the predicate registers.
12978 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
12979 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
12980 (VEC_ANY_DATA, VEC_STRUCT): New constants.
12981 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
12982 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
12983 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
12984 (aarch64_get_mask_mode): New functions.
12985 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
12986 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
12987 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
12988 predicate modes and predicate registers. Explicitly restrict
12989 GPRs to modes of 16 bytes or smaller. Only allow FP registers
12990 to store a vector mode if it is recognized by
12991 aarch64_classify_vector_mode.
12992 (aarch64_regmode_natural_size): New function.
12993 (aarch64_hard_regno_caller_save_mode): Return the original mode
12995 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
12996 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
12997 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
12998 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
13000 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
13001 does not overlap dest if the function is frame-related. Handle
13003 (aarch64_split_add_offset): New function.
13004 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13005 them aarch64_add_offset.
13006 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13007 and update call to aarch64_sub_sp.
13008 (aarch64_add_cfa_expression): New function.
13009 (aarch64_expand_prologue): Pass extra temporary registers to the
13010 functions above. Handle the case in which we need to emit new
13011 DW_CFA_expressions for registers that were originally saved
13012 relative to the stack pointer, but now have to be expressed
13013 relative to the frame pointer.
13014 (aarch64_output_mi_thunk): Pass extra temporary registers to the
13016 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
13017 IP0 and IP1 values for SVE frames.
13018 (aarch64_expand_vec_series): New function.
13019 (aarch64_expand_sve_widened_duplicate): Likewise.
13020 (aarch64_expand_sve_const_vector): Likewise.
13021 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13022 Handle SVE constants. Use emit_move_insn to move a force_const_mem
13023 into the register, rather than emitting a SET directly.
13024 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13025 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13026 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13027 (offset_9bit_signed_scaled_p): New functions.
13028 (aarch64_replicate_bitmask_imm): New function.
13029 (aarch64_bitmask_imm): Use it.
13030 (aarch64_cannot_force_const_mem): Reject expressions involving
13031 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
13032 (aarch64_classify_index): Handle SVE indices, by requiring
13033 a plain register index with a scale that matches the element size.
13034 (aarch64_classify_address): Handle SVE addresses. Assert that
13035 the mode of the address is VOIDmode or an integer mode.
13036 Update call to aarch64_classify_symbol.
13037 (aarch64_classify_symbolic_expression): Update call to
13038 aarch64_classify_symbol.
13039 (aarch64_const_vec_all_in_range_p): New function.
13040 (aarch64_print_vector_float_operand): Likewise.
13041 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
13042 "vN" for FP registers with SVE modes. Handle (const ...) vectors
13043 and the FP immediates 1.0 and 0.5.
13044 (aarch64_print_address_internal): Handle SVE addresses.
13045 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13046 (aarch64_regno_regclass): Handle predicate registers.
13047 (aarch64_secondary_reload): Handle big-endian reloads of SVE
13049 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13050 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13051 (aarch64_convert_sve_vector_bits): New function.
13052 (aarch64_override_options): Use it to handle -msve-vector-bits=.
13053 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13054 rather than an rtx.
13055 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13056 Handle SVE vector and predicate modes. Accept VL-based constants
13057 that need only one temporary register, and VL offsets that require
13058 no temporary registers.
13059 (aarch64_conditional_register_usage): Mark the predicate registers
13060 as fixed if SVE isn't available.
13061 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13062 Return true for SVE vector and predicate modes.
13063 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13064 rather than an unsigned int. Handle SVE modes.
13065 (aarch64_preferred_simd_mode): Update call accordingly. Handle
13067 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13069 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13070 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13071 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13072 (aarch64_sve_float_mul_immediate_p): New functions.
13073 (aarch64_sve_valid_immediate): New function.
13074 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13075 Explicitly reject structure modes. Check for INDEX constants.
13076 Handle PTRUE and PFALSE constants.
13077 (aarch64_check_zero_based_sve_index_immediate): New function.
13078 (aarch64_simd_imm_zero_p): Delete.
13079 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13080 vector modes. Accept constants in the range of CNT[BHWD].
13081 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13082 ask for an Advanced SIMD mode.
13083 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13084 (aarch64_simd_vector_alignment): Handle SVE predicates.
13085 (aarch64_vectorize_preferred_vector_alignment): New function.
13086 (aarch64_simd_vector_alignment_reachable): Use it instead of
13088 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13089 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13091 (MAX_VECT_LEN): Delete.
13092 (expand_vec_perm_d): Add a vec_flags field.
13093 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13094 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13095 (aarch64_evpc_ext): Don't apply a big-endian lane correction
13097 (aarch64_evpc_rev): Rename to...
13098 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
13099 (aarch64_evpc_rev_global): New function.
13100 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13101 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13103 (aarch64_evpc_sve_tbl): New function.
13104 (aarch64_expand_vec_perm_const_1): Update after rename of
13105 aarch64_evpc_rev. Handle SVE permutes too, trying
13106 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13107 than aarch64_evpc_tbl.
13108 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13109 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13110 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13111 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13112 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13113 (aarch64_expand_sve_vcond): New functions.
13114 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13115 of aarch64_vector_mode_p.
13116 (aarch64_dwarf_poly_indeterminate_value): New function.
13117 (aarch64_compute_pressure_classes): Likewise.
13118 (aarch64_can_change_mode_class): Likewise.
13119 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13120 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13121 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13122 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13123 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13124 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13125 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13126 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13128 (Dn, Dl, Dr): Accept const as well as const_vector.
13129 (Dz): Likewise. Compare against CONST0_RTX.
13130 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13131 of "vector" where appropriate.
13132 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13133 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13134 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13135 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13136 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13137 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13138 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13139 (v_int_equiv): Extend to SVE modes.
13140 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13142 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13143 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13144 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13145 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13146 (SVE_COND_FP_CMP): New int iterators.
13147 (perm_hilo): Handle the new unpack unspecs.
13148 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13150 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13151 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13152 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13153 (aarch64_equality_operator, aarch64_constant_vector_operand)
13154 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13155 (aarch64_sve_nonimmediate_operand): Likewise.
13156 (aarch64_sve_general_operand): Likewise.
13157 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13158 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13159 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13160 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
13161 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
13162 (aarch64_sve_float_arith_immediate): Likewise.
13163 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
13164 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
13165 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
13166 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
13167 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
13168 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
13169 (aarch64_sve_float_arith_operand): Likewise.
13170 (aarch64_sve_float_arith_with_sub_operand): Likewise.
13171 (aarch64_sve_float_mul_operand): Likewise.
13172 (aarch64_sve_vec_perm_operand): Likewise.
13173 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
13174 (aarch64_mov_operand): Accept const_poly_int and const_vector.
13175 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
13176 as well as const_vector.
13177 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
13178 in file. Use CONST0_RTX and CONSTM1_RTX.
13179 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
13180 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
13181 Use aarch64_simd_imm_zero.
13182 * config/aarch64/aarch64-sve.md: New file.
13183 * config/aarch64/aarch64.md: Include it.
13184 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
13185 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
13186 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
13187 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
13188 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
13189 (sve): New attribute.
13190 (enabled): Disable instructions with the sve attribute unless
13192 (movqi, movhi): Pass CONST_POLY_INT operaneds through
13193 aarch64_expand_mov_immediate.
13194 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
13195 CNT[BHSD] immediates.
13196 (movti): Split CONST_POLY_INT moves into two halves.
13197 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
13198 Split additions that need a temporary here if the destination
13199 is the stack pointer.
13200 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
13201 (*add<mode>3_poly_1): New instruction.
13202 (set_clobber_cc): New expander.
13204 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13206 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
13207 parameter and use it instead of GET_MODE_SIZE (innermode). Use
13208 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
13209 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
13210 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
13211 Change innermode from fixed_mode_size to machine_mode.
13212 (simplify_subreg): Update call accordingly. Handle a constant-sized
13213 subreg of a variable-length CONST_VECTOR.
13215 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13216 Alan Hayward <alan.hayward@arm.com>
13217 David Sherwood <david.sherwood@arm.com>
13219 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
13220 (add_offset_to_base): New function, split out from...
13221 (create_mem_ref): ...here. When handling a scale other than 1,
13222 check first whether the address is valid without the offset.
13223 Add it into the base if so, leaving the index and scale as-is.
13225 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13228 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
13229 fold_for_warn before checking if arg2 is INTEGER_CST.
13231 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
13233 * config/rs6000/predicates.md (load_multiple_operation): Delete.
13234 (store_multiple_operation): Delete.
13235 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
13236 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
13237 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
13238 guarded by TARGET_STRING.
13239 (rs6000_output_load_multiple): Delete.
13240 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
13241 OPTION_MASK_STRING / TARGET_STRING handling.
13242 (print_operand) <'N', 'O'>: Add comment that these are unused now.
13243 (const rs6000_opt_masks) <"string">: Change mask to 0.
13244 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
13245 (MASK_STRING): Delete.
13246 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
13248 (load_multiple): Delete.
13255 (store_multiple): Delete.
13262 (movmemsi_8reg): Delete.
13263 (corresponding unnamed define_insn): Delete.
13264 (movmemsi_6reg): Delete.
13265 (corresponding unnamed define_insn): Delete.
13266 (movmemsi_4reg): Delete.
13267 (corresponding unnamed define_insn): Delete.
13268 (movmemsi_2reg): Delete.
13269 (corresponding unnamed define_insn): Delete.
13270 (movmemsi_1reg): Delete.
13271 (corresponding unnamed define_insn): Delete.
13272 * config/rs6000/rs6000.opt (mno-string): New.
13273 (mstring): Replace by deprecation warning stub.
13274 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
13276 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13278 * regrename.c (regrename_do_replace): If replacing the same
13279 reg multiple times, try to reuse last created gen_raw_REG.
13282 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
13283 main to workaround a bug in GDB.
13285 2018-01-12 Tom de Vries <tom@codesourcery.com>
13288 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
13290 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
13292 PR rtl-optimization/80481
13293 * ira-color.c (get_cap_member): New function.
13294 (allocnos_conflict_by_live_ranges_p): Use it.
13295 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
13296 (setup_slot_coalesced_allocno_live_ranges): Ditto.
13298 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
13301 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
13302 (*saddl_se_1): Ditto.
13303 (*ssubsi_1): Ditto.
13304 (*ssubl_se_1): Ditto.
13306 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13308 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
13309 rather than wi::to_widest for DR_INITs.
13310 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
13311 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
13312 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
13314 (vect_analyze_group_access_1): Note that here.
13316 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13318 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
13319 polynomial type sizes.
13321 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13323 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
13324 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
13325 (gimple_add_tmp_var): Likewise.
13327 2018-01-12 Martin Liska <mliska@suse.cz>
13329 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
13330 (gimple_alloc_sizes): Likewise.
13331 (dump_gimple_statistics): Use PRIu64 in printf format.
13332 * gimple.h: Change uint64_t to int.
13334 2018-01-12 Martin Liska <mliska@suse.cz>
13336 * tree-core.h: Use uint64_t instead of int.
13337 * tree.c (tree_node_counts): Likewise.
13338 (tree_node_sizes): Likewise.
13339 (dump_tree_statistics): Use PRIu64 in printf format.
13341 2018-01-12 Martin Liska <mliska@suse.cz>
13343 * Makefile.in: As qsort_chk is implemented in vec.c, add
13344 vec.o to linkage of gencfn-macros.
13345 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
13346 passing the info to record_node_allocation_statistics.
13347 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
13349 * ggc-common.c (struct ggc_usage): Add operator== and use
13350 it in operator< and compare function.
13351 * mem-stats.h (struct mem_usage): Likewise.
13352 * vec.c (struct vec_usage): Remove operator< and compare
13353 function. Can be simply inherited.
13355 2018-01-12 Martin Jambor <mjambor@suse.cz>
13358 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
13359 * tree-ssa-math-opts.c: Include domwalk.h.
13360 (convert_mult_to_fma_1): New function.
13361 (fma_transformation_info): New type.
13362 (fma_deferring_state): Likewise.
13363 (cancel_fma_deferring): New function.
13364 (result_of_phi): Likewise.
13365 (last_fma_candidate_feeds_initial_phi): Likewise.
13366 (convert_mult_to_fma): Added deferring logic, split actual
13367 transformation to convert_mult_to_fma_1.
13368 (math_opts_dom_walker): New type.
13369 (math_opts_dom_walker::after_dom_children): New method, body moved
13370 here from pass_optimize_widening_mul::execute, added deferring logic
13372 (pass_optimize_widening_mul::execute): Moved most of code to
13373 math_opts_dom_walker::after_dom_children.
13374 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
13375 * config/i386/i386.c (ix86_option_override_internal): Added
13376 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
13378 2018-01-12 Richard Biener <rguenther@suse.de>
13381 * dwarf2out.c (gen_variable_die): Do not reset old_die for
13382 inline instance vars.
13384 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
13387 * config/rx/rx.c (rx_is_restricted_memory_address):
13388 Handle SUBREG case.
13390 2018-01-12 Richard Biener <rguenther@suse.de>
13392 PR tree-optimization/80846
13393 * target.def (split_reduction): New target hook.
13394 * targhooks.c (default_split_reduction): New function.
13395 * targhooks.h (default_split_reduction): Declare.
13396 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
13397 target requests first reduce vectors by combining low and high
13399 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
13400 (get_vectype_for_scalar_type_and_size): Export.
13401 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
13402 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
13403 * doc/tm.texi: Regenerate.
13404 * config/i386/i386.c (ix86_split_reduction): Implement
13405 TARGET_VECTORIZE_SPLIT_REDUCTION.
13407 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13410 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
13411 in PIC mode except for TARGET_VXWORKS_RTP.
13412 * config/sparc/sparc.c: Include cfgrtl.h.
13413 (TARGET_INIT_PIC_REG): Define.
13414 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
13415 (sparc_pic_register_p): New predicate.
13416 (sparc_legitimate_address_p): Use it.
13417 (sparc_legitimize_pic_address): Likewise.
13418 (sparc_delegitimize_address): Likewise.
13419 (sparc_mode_dependent_address_p): Likewise.
13420 (gen_load_pcrel_sym): Remove 4th parameter.
13421 (load_got_register): Adjust call to above. Remove obsolete stuff.
13422 (sparc_expand_prologue): Do not call load_got_register here.
13423 (sparc_flat_expand_prologue): Likewise.
13424 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
13425 (sparc_use_pseudo_pic_reg): New function.
13426 (sparc_init_pic_reg): Likewise.
13427 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13428 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
13430 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
13432 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
13433 Add item for branch_cost.
13435 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13437 PR rtl-optimization/83565
13438 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
13439 not extend the result to a larger mode for rotate operations.
13440 (num_sign_bit_copies1): Likewise.
13442 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13445 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
13447 Use values-Xc.o for -pedantic.
13448 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
13450 2018-01-12 Martin Liska <mliska@suse.cz>
13453 * ipa-devirt.c (final_warning_record::grow_type_warnings):
13455 (possible_polymorphic_call_targets): Use it.
13456 (ipa_devirt): Likewise.
13458 2018-01-12 Martin Liska <mliska@suse.cz>
13460 * profile-count.h (enum profile_quality): Use 0 as invalid
13461 enum value of profile_quality.
13463 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
13465 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
13466 -mext-string options.
13468 2018-01-12 Richard Biener <rguenther@suse.de>
13470 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
13471 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
13472 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
13474 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
13476 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13478 * configure.ac (--with-long-double-format): Add support for the
13479 configuration option to change the default long double format on
13481 * config.gcc (powerpc*-linux*-*): Likewise.
13482 * configure: Regenerate.
13483 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
13484 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
13485 used without modification.
13487 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13489 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
13490 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
13491 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
13492 MISC_BUILTIN_SPEC_BARRIER.
13493 (rs6000_init_builtins): Likewise.
13494 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
13496 (speculation_barrier): New define_insn.
13497 * doc/extend.texi: Document __builtin_speculation_barrier.
13499 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13502 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
13503 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
13504 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
13506 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
13507 integral modes instead of "ss" and "sd".
13508 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
13509 vectors with 32-bit and 64-bit elements.
13510 (vecdupssescalarmodesuffix): New mode attribute.
13511 (vec_dup<mode>): Use it.
13513 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
13516 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
13517 frame if argument is passed on stack.
13519 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13522 * ree.c (combine_reaching_defs): Optimize also
13523 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
13524 reg2=any_extend(exp); reg1=reg2;, formatting fix.
13526 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13528 PR middle-end/83189
13529 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
13531 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13533 PR middle-end/83718
13534 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
13535 after they are computed.
13537 2018-01-11 Bin Cheng <bin.cheng@arm.com>
13539 PR tree-optimization/83695
13540 * gimple-loop-linterchange.cc
13541 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
13542 reset cached scev information after interchange.
13543 (pass_linterchange::execute): Remove call to scev_reset_htab.
13545 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13547 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
13548 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
13549 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
13550 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
13551 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
13552 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
13553 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
13554 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
13555 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
13556 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
13557 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
13558 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
13559 (V_lane_reg): Likewise.
13560 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
13562 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
13563 (vfmal_lane_low<mode>_intrinsic,
13564 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
13565 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
13566 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
13567 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
13568 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
13569 vfmsl_lane_high<mode>_intrinsic): New define_insns.
13571 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13573 * config/arm/arm-cpus.in (fp16fml): New feature.
13574 (ALL_SIMD): Add fp16fml.
13575 (armv8.2-a): Add fp16fml as an option.
13576 (armv8.3-a): Likewise.
13577 (armv8.4-a): Add fp16fml as part of fp16.
13578 * config/arm/arm.h (TARGET_FP16FML): Define.
13579 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
13581 * config/arm/arm-modes.def (V2HF): Define.
13582 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
13583 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
13584 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
13585 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
13586 vfmsl_low, vfmsl_high): New set of builtins.
13587 * config/arm/iterators.md (PLUSMINUS): New code iterator.
13588 (vfml_op): New code attribute.
13589 (VFMLHALVES): New int iterator.
13590 (VFML, VFMLSEL): New mode attributes.
13591 (V_reg): Define mapping for V2HF.
13592 (V_hi, V_lo): New mode attributes.
13593 (VF_constraint): Likewise.
13594 (vfml_half, vfml_half_selector): New int attributes.
13595 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
13597 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
13598 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
13600 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
13601 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
13602 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
13603 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
13605 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
13606 Document new effective target and option set.
13608 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13610 * config/arm/arm-cpus.in (armv8_4): New feature.
13611 (ARMv8_4a): New fgroup.
13612 (armv8.4-a): New arch.
13613 * config/arm/arm-tables.opt: Regenerate.
13614 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
13615 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
13616 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
13617 Add matching rules for -march=armv8.4-a and extensions.
13618 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
13620 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
13623 * config/rx/rx.md (BW): New mode attribute.
13624 (sync_lock_test_and_setsi): Add mode suffix to insn output.
13626 2018-01-11 Richard Biener <rguenther@suse.de>
13628 PR tree-optimization/83435
13629 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
13630 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
13631 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
13633 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13634 Alan Hayward <alan.hayward@arm.com>
13635 David Sherwood <david.sherwood@arm.com>
13637 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
13639 (aarch64_classify_address): Initialize it. Track polynomial offsets.
13640 (aarch64_print_address_internal): Use it to check for a zero offset.
13642 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13643 Alan Hayward <alan.hayward@arm.com>
13644 David Sherwood <david.sherwood@arm.com>
13646 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
13647 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
13648 Return a poly_int64 rather than a HOST_WIDE_INT.
13649 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
13650 rather than a HOST_WIDE_INT.
13651 * config/aarch64/aarch64.h (aarch64_frame): Protect with
13652 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
13653 hard_fp_offset, frame_size, initial_adjust, callee_offset and
13654 final_offset from HOST_WIDE_INT to poly_int64.
13655 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
13656 to_constant when getting the number of units in an Advanced SIMD
13658 (aarch64_builtin_vectorized_function): Check for a constant number
13660 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
13662 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
13663 attribute instead of GET_MODE_NUNITS.
13664 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
13665 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
13666 GET_MODE_SIZE for fixed-size registers.
13667 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
13668 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
13669 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
13670 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
13671 (aarch64_print_operand, aarch64_print_address_internal)
13672 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
13673 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
13674 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
13675 Handle polynomial GET_MODE_SIZE.
13676 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
13677 wider than SImode without modification.
13678 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
13679 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
13680 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
13681 passing and returning SVE modes.
13682 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
13683 rather than GEN_INT.
13684 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
13685 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
13686 (aarch64_allocate_and_probe_stack_space): Likewise.
13687 (aarch64_layout_frame): Cope with polynomial offsets.
13688 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
13689 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
13690 polynomial offsets.
13691 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
13692 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
13693 poly_int64 rather than a HOST_WIDE_INT.
13694 (aarch64_get_separate_components, aarch64_process_components)
13695 (aarch64_expand_prologue, aarch64_expand_epilogue)
13696 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
13697 (aarch64_anchor_offset): New function, split out from...
13698 (aarch64_legitimize_address): ...here.
13699 (aarch64_builtin_vectorization_cost): Handle polynomial
13700 TYPE_VECTOR_SUBPARTS.
13701 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
13703 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
13704 number of elements from the PARALLEL rather than the mode.
13705 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
13706 rather than GET_MODE_BITSIZE.
13707 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
13708 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
13709 (aarch64_expand_vec_perm_const_1): Handle polynomial
13710 d->perm.length () and d->perm elements.
13711 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
13712 Apply to_constant to d->perm elements.
13713 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
13714 polynomial CONST_VECTOR_NUNITS.
13715 (aarch64_move_pointer): Take amount as a poly_int64 rather
13717 (aarch64_progress_pointer): Avoid temporary variable.
13718 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
13719 the mode attribute instead of GET_MODE.
13721 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13722 Alan Hayward <alan.hayward@arm.com>
13723 David Sherwood <david.sherwood@arm.com>
13725 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
13726 x exists before using it.
13727 (aarch64_add_constant_internal): Rename to...
13728 (aarch64_add_offset_1): ...this. Replace regnum with separate
13729 src and dest rtxes. Handle the case in which they're different,
13730 including when the offset is zero. Replace scratchreg with an rtx.
13731 Use 2 additions if there is no spare register into which we can
13732 move a 16-bit constant.
13733 (aarch64_add_constant): Delete.
13734 (aarch64_add_offset): Replace reg with separate src and dest
13735 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
13736 Use aarch64_add_offset_1.
13737 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
13738 an rtx rather than an int. Take the delta as a poly_int64
13739 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
13740 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
13741 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
13742 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
13743 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
13744 and aarch64_add_sp.
13745 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
13746 aarch64_add_constant.
13748 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13750 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
13751 Use scalar_float_mode.
13753 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13755 * config/aarch64/aarch64-simd.md
13756 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
13757 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
13758 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
13759 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
13760 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
13761 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
13762 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
13763 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
13764 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
13765 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
13767 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13770 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
13771 targ_options->x_arm_arch_string is non NULL.
13773 2018-01-11 Tamar Christina <tamar.christina@arm.com>
13775 * config/aarch64/aarch64.h
13776 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
13778 2018-01-11 Sudakshina Das <sudi.das@arm.com>
13781 * expmed.c (emit_store_flag_force): Swap if const op0
13782 and change VOIDmode to mode of op0.
13784 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13786 PR rtl-optimization/83761
13787 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
13788 than bytes to mode_for_size.
13790 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13792 PR middle-end/83189
13793 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
13794 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
13797 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13799 PR middle-end/83575
13800 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
13801 when in layout mode.
13802 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
13803 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
13806 2018-01-10 Michael Collison <michael.collison@arm.com>
13808 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
13809 * config/aarch64/aarch64-option-extension.def: Add
13810 AARCH64_OPT_EXTENSION of 'fp16fml'.
13811 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13812 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
13813 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
13814 * config/aarch64/constraints.md (Ui7): New constraint.
13815 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
13816 (VFMLA_SEL_W): Ditto.
13819 (VFMLA16_LOW): New int iterator.
13820 (VFMLA16_HIGH): Ditto.
13821 (UNSPEC_FMLAL): New unspec.
13822 (UNSPEC_FMLSL): Ditto.
13823 (UNSPEC_FMLAL2): Ditto.
13824 (UNSPEC_FMLSL2): Ditto.
13825 (f16mac): New code attribute.
13826 * config/aarch64/aarch64-simd-builtins.def
13827 (aarch64_fmlal_lowv2sf): Ditto.
13828 (aarch64_fmlsl_lowv2sf): Ditto.
13829 (aarch64_fmlalq_lowv4sf): Ditto.
13830 (aarch64_fmlslq_lowv4sf): Ditto.
13831 (aarch64_fmlal_highv2sf): Ditto.
13832 (aarch64_fmlsl_highv2sf): Ditto.
13833 (aarch64_fmlalq_highv4sf): Ditto.
13834 (aarch64_fmlslq_highv4sf): Ditto.
13835 (aarch64_fmlal_lane_lowv2sf): Ditto.
13836 (aarch64_fmlsl_lane_lowv2sf): Ditto.
13837 (aarch64_fmlal_laneq_lowv2sf): Ditto.
13838 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
13839 (aarch64_fmlalq_lane_lowv4sf): Ditto.
13840 (aarch64_fmlsl_lane_lowv4sf): Ditto.
13841 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
13842 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
13843 (aarch64_fmlal_lane_highv2sf): Ditto.
13844 (aarch64_fmlsl_lane_highv2sf): Ditto.
13845 (aarch64_fmlal_laneq_highv2sf): Ditto.
13846 (aarch64_fmlsl_laneq_highv2sf): Ditto.
13847 (aarch64_fmlalq_lane_highv4sf): Ditto.
13848 (aarch64_fmlsl_lane_highv4sf): Ditto.
13849 (aarch64_fmlalq_laneq_highv4sf): Ditto.
13850 (aarch64_fmlsl_laneq_highv4sf): Ditto.
13851 * config/aarch64/aarch64-simd.md:
13852 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
13853 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13854 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
13855 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
13856 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
13857 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
13858 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
13859 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
13860 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
13861 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
13862 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
13863 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
13864 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
13865 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
13866 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
13867 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
13868 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
13869 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
13870 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
13871 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
13872 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
13873 (vfmlsl_low_u32): Ditto.
13874 (vfmlalq_low_u32): Ditto.
13875 (vfmlslq_low_u32): Ditto.
13876 (vfmlal_high_u32): Ditto.
13877 (vfmlsl_high_u32): Ditto.
13878 (vfmlalq_high_u32): Ditto.
13879 (vfmlslq_high_u32): Ditto.
13880 (vfmlal_lane_low_u32): Ditto.
13881 (vfmlsl_lane_low_u32): Ditto.
13882 (vfmlal_laneq_low_u32): Ditto.
13883 (vfmlsl_laneq_low_u32): Ditto.
13884 (vfmlalq_lane_low_u32): Ditto.
13885 (vfmlslq_lane_low_u32): Ditto.
13886 (vfmlalq_laneq_low_u32): Ditto.
13887 (vfmlslq_laneq_low_u32): Ditto.
13888 (vfmlal_lane_high_u32): Ditto.
13889 (vfmlsl_lane_high_u32): Ditto.
13890 (vfmlal_laneq_high_u32): Ditto.
13891 (vfmlsl_laneq_high_u32): Ditto.
13892 (vfmlalq_lane_high_u32): Ditto.
13893 (vfmlslq_lane_high_u32): Ditto.
13894 (vfmlalq_laneq_high_u32): Ditto.
13895 (vfmlslq_laneq_high_u32): Ditto.
13896 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
13897 (AARCH64_FL_FOR_ARCH8_4): New.
13898 (AARCH64_ISA_F16FML): New ISA flag.
13899 (TARGET_F16FML): New feature flag for fp16fml.
13900 (doc/invoke.texi): Document new fp16fml option.
13902 2018-01-10 Michael Collison <michael.collison@arm.com>
13904 * config/aarch64/aarch64-builtins.c:
13905 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
13906 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13907 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
13908 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
13909 (AARCH64_ISA_SHA3): New ISA flag.
13910 (TARGET_SHA3): New feature flag for sha3.
13911 * config/aarch64/iterators.md (sha512_op): New int attribute.
13912 (CRYPTO_SHA512): New int iterator.
13913 (UNSPEC_SHA512H): New unspec.
13914 (UNSPEC_SHA512H2): Ditto.
13915 (UNSPEC_SHA512SU0): Ditto.
13916 (UNSPEC_SHA512SU1): Ditto.
13917 * config/aarch64/aarch64-simd-builtins.def
13918 (aarch64_crypto_sha512hqv2di): New builtin.
13919 (aarch64_crypto_sha512h2qv2di): Ditto.
13920 (aarch64_crypto_sha512su0qv2di): Ditto.
13921 (aarch64_crypto_sha512su1qv2di): Ditto.
13922 (aarch64_eor3qv8hi): Ditto.
13923 (aarch64_rax1qv2di): Ditto.
13924 (aarch64_xarqv2di): Ditto.
13925 (aarch64_bcaxqv8hi): Ditto.
13926 * config/aarch64/aarch64-simd.md:
13927 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
13928 (aarch64_crypto_sha512su0qv2di): Ditto.
13929 (aarch64_crypto_sha512su1qv2di): Ditto.
13930 (aarch64_eor3qv8hi): Ditto.
13931 (aarch64_rax1qv2di): Ditto.
13932 (aarch64_xarqv2di): Ditto.
13933 (aarch64_bcaxqv8hi): Ditto.
13934 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
13935 (vsha512h2q_u64): Ditto.
13936 (vsha512su0q_u64): Ditto.
13937 (vsha512su1q_u64): Ditto.
13938 (veor3q_u16): Ditto.
13939 (vrax1q_u64): Ditto.
13940 (vxarq_u64): Ditto.
13941 (vbcaxq_u16): Ditto.
13942 * config/arm/types.md (crypto_sha512): New type attribute.
13943 (crypto_sha3): Ditto.
13944 (doc/invoke.texi): Document new sha3 option.
13946 2018-01-10 Michael Collison <michael.collison@arm.com>
13948 * config/aarch64/aarch64-builtins.c:
13949 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
13950 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13951 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
13952 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
13953 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
13954 (AARCH64_ISA_SM4): New ISA flag.
13955 (TARGET_SM4): New feature flag for sm4.
13956 * config/aarch64/aarch64-simd-builtins.def
13957 (aarch64_sm3ss1qv4si): Ditto.
13958 (aarch64_sm3tt1aq4si): Ditto.
13959 (aarch64_sm3tt1bq4si): Ditto.
13960 (aarch64_sm3tt2aq4si): Ditto.
13961 (aarch64_sm3tt2bq4si): Ditto.
13962 (aarch64_sm3partw1qv4si): Ditto.
13963 (aarch64_sm3partw2qv4si): Ditto.
13964 (aarch64_sm4eqv4si): Ditto.
13965 (aarch64_sm4ekeyqv4si): Ditto.
13966 * config/aarch64/aarch64-simd.md:
13967 (aarch64_sm3ss1qv4si): Ditto.
13968 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
13969 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
13970 (aarch64_sm4eqv4si): Ditto.
13971 (aarch64_sm4ekeyqv4si): Ditto.
13972 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
13973 (sm3part_op): Ditto.
13974 (CRYPTO_SM3TT): Ditto.
13975 (CRYPTO_SM3PART): Ditto.
13976 (UNSPEC_SM3SS1): New unspec.
13977 (UNSPEC_SM3TT1A): Ditto.
13978 (UNSPEC_SM3TT1B): Ditto.
13979 (UNSPEC_SM3TT2A): Ditto.
13980 (UNSPEC_SM3TT2B): Ditto.
13981 (UNSPEC_SM3PARTW1): Ditto.
13982 (UNSPEC_SM3PARTW2): Ditto.
13983 (UNSPEC_SM4E): Ditto.
13984 (UNSPEC_SM4EKEY): Ditto.
13985 * config/aarch64/constraints.md (Ui2): New constraint.
13986 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
13987 * config/arm/types.md (crypto_sm3): New type attribute.
13988 (crypto_sm4): Ditto.
13989 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
13990 (vsm3tt1aq_u32): Ditto.
13991 (vsm3tt1bq_u32): Ditto.
13992 (vsm3tt2aq_u32): Ditto.
13993 (vsm3tt2bq_u32): Ditto.
13994 (vsm3partw1q_u32): Ditto.
13995 (vsm3partw2q_u32): Ditto.
13996 (vsm4eq_u32): Ditto.
13997 (vsm4ekeyq_u32): Ditto.
13998 (doc/invoke.texi): Document new sm4 option.
14000 2018-01-10 Michael Collison <michael.collison@arm.com>
14002 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14003 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14004 (AARCH64_FL_FOR_ARCH8_4): New.
14005 (AARCH64_FL_V8_4): New flag.
14006 (doc/invoke.texi): Document new armv8.4-a option.
14008 2018-01-10 Michael Collison <michael.collison@arm.com>
14010 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14011 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14012 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14013 * config/aarch64/aarch64-option-extension.def: Add
14014 AARCH64_OPT_EXTENSION of 'sha2'.
14015 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14016 (crypto): Disable sha2 and aes if crypto disabled.
14017 (crypto): Enable aes and sha2 if enabled.
14018 (simd): Disable sha2 and aes if simd disabled.
14019 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14021 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14022 (TARGET_SHA2): New feature flag for sha2.
14023 (TARGET_AES): New feature flag for aes.
14024 * config/aarch64/aarch64-simd.md:
14025 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14026 conditional on TARGET_AES.
14027 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14028 (aarch64_crypto_sha1hsi): Make pattern conditional
14030 (aarch64_crypto_sha1hv4si): Ditto.
14031 (aarch64_be_crypto_sha1hv4si): Ditto.
14032 (aarch64_crypto_sha1su1v4si): Ditto.
14033 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14034 (aarch64_crypto_sha1su0v4si): Ditto.
14035 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14036 (aarch64_crypto_sha256su0v4si): Ditto.
14037 (aarch64_crypto_sha256su1v4si): Ditto.
14038 (doc/invoke.texi): Document new aes and sha2 options.
14040 2018-01-10 Martin Sebor <msebor@redhat.com>
14042 PR tree-optimization/83781
14043 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14046 2018-01-11 Martin Sebor <msebor@gmail.com>
14047 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14049 PR tree-optimization/83501
14050 PR tree-optimization/81703
14052 * tree-ssa-strlen.c (get_string_cst): Rename...
14053 (get_string_len): ...to this. Handle global constants.
14054 (handle_char_store): Adjust.
14056 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
14057 Jim Wilson <jimw@sifive.com>
14059 * config/riscv/riscv-protos.h (riscv_output_return): New.
14060 * config/riscv/riscv.c (struct machine_function): New naked_p field.
14061 (riscv_attribute_table, riscv_output_return),
14062 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14063 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14064 (riscv_compute_frame_info): Only compute frame->mask if not a naked
14066 (riscv_expand_prologue): Add early return for naked function.
14067 (riscv_expand_epilogue): Likewise.
14068 (riscv_function_ok_for_sibcall): Return false for naked function.
14069 (riscv_set_current_function): New.
14070 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14071 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14072 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14073 * doc/extend.texi (RISC-V Function Attributes): New.
14075 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14077 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14078 check for 128-bit long double before checking TCmode.
14079 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14080 128-bit long doubles before checking TFmode or TCmode.
14081 (FLOAT128_IBM_P): Likewise.
14083 2018-01-10 Martin Sebor <msebor@redhat.com>
14085 PR tree-optimization/83671
14086 * builtins.c (c_strlen): Unconditionally return zero for the empty
14088 Use -Warray-bounds for warnings.
14089 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14090 for non-constant array indices with COMPONENT_REF, arrays of
14091 arrays, and pointers to arrays.
14092 (gimple_fold_builtin_strlen): Determine and set length range for
14093 non-constant character arrays.
14095 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
14097 PR middle-end/81897
14098 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14101 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
14103 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14105 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14108 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14109 VECTOR_MEM_ALTIVEC_OR_VSX_P.
14110 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14111 indexed_or_indirect_operand predicate.
14112 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14113 (*vsx_le_perm_load_v8hi): Likewise.
14114 (*vsx_le_perm_load_v16qi): Likewise.
14115 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14116 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14117 (*vsx_le_perm_store_v8hi): Likewise.
14118 (*vsx_le_perm_store_v16qi): Likewise.
14119 (eight unnamed splitters): Likewise.
14121 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14123 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14124 * config/rs6000/emmintrin.h: Likewise.
14125 * config/rs6000/mmintrin.h: Likewise.
14126 * config/rs6000/xmmintrin.h: Likewise.
14128 2018-01-10 David Malcolm <dmalcolm@redhat.com>
14131 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14133 * tree.c (tree_nop_conversion): Return true for location wrapper
14135 (maybe_wrap_with_location): New function.
14136 (selftest::check_strip_nops): New function.
14137 (selftest::test_location_wrappers): New function.
14138 (selftest::tree_c_tests): Call it.
14139 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14140 (maybe_wrap_with_location): New decl.
14141 (EXPR_LOCATION_WRAPPER_P): New macro.
14142 (location_wrapper_p): New inline function.
14143 (tree_strip_any_location_wrapper): New inline function.
14145 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
14148 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14149 stack_realign_offset for the largest alignment of stack slot
14151 (ix86_find_max_used_stack_alignment): New function.
14152 (ix86_finalize_stack_frame_flags): Use it. Set
14153 max_used_stack_alignment if we don't realign stack.
14154 * config/i386/i386.h (machine_function): Add
14155 max_used_stack_alignment.
14157 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
14159 * config/arm/arm.opt (-mbranch-cost): New option.
14160 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
14163 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
14166 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
14167 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
14169 2018-01-10 Richard Biener <rguenther@suse.de>
14172 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
14173 early out so it also covers the case where we have a non-NULL
14176 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14178 PR tree-optimization/83753
14179 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
14180 for non-strided grouped accesses if the number of elements is 1.
14182 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14185 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
14186 * i386.h (TARGET_USE_GATHER): Define.
14187 * x86-tune.def (X86_TUNE_USE_GATHER): New.
14189 2018-01-10 Martin Liska <mliska@suse.cz>
14192 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
14193 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
14195 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
14196 CLEANUP_NO_PARTITIONING is not set.
14198 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14200 * doc/rtl.texi: Remove documentation of (const ...) wrappers
14201 for vectors, as a partial revert of r254296.
14202 * rtl.h (const_vec_p): Delete.
14203 (const_vec_duplicate_p): Don't test for vector CONSTs.
14204 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
14205 * expmed.c (make_tree): Likewise.
14208 * common.md (E, F): Use CONSTANT_P instead of checking for
14210 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
14211 checking for CONST_VECTOR.
14213 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14215 PR middle-end/83575
14216 * predict.c (force_edge_cold): Handle in more sane way edges
14217 with no prediction.
14219 2018-01-09 Carl Love <cel@us.ibm.com>
14221 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
14223 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
14224 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
14225 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
14226 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
14227 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
14228 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
14229 * config/rs6000/rs6000-protos.h: Add extern defition for
14230 rs6000_generate_float2_double_code.
14231 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
14233 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
14234 (float2_v2df): Add define_expand.
14236 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
14239 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
14240 op_mode in the force_to_mode call.
14242 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14244 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
14245 instead of checking each element individually.
14246 (aarch64_evpc_uzp): Likewise.
14247 (aarch64_evpc_zip): Likewise.
14248 (aarch64_evpc_ext): Likewise.
14249 (aarch64_evpc_rev): Likewise.
14250 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
14251 instead of checking each element individually. Return true without
14253 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
14254 whether all selected elements come from the same input, instead of
14255 checking each element individually. Remove calls to gen_rtx_REG,
14256 start_sequence and end_sequence and instead assert that no rtl is
14259 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14261 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
14262 order of HIGH and CONST checks.
14264 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14266 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
14267 if the destination isn't an SSA_NAME.
14269 2018-01-09 Richard Biener <rguenther@suse.de>
14271 PR tree-optimization/83668
14272 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
14274 (canonicalize_loop_form): ... here, renamed from ...
14275 (canonicalize_loop_closed_ssa_form): ... this and amended to
14276 swap successor edges for loop exit blocks to make us use
14277 the RPO order we need for initial schedule generation.
14279 2018-01-09 Joseph Myers <joseph@codesourcery.com>
14281 PR tree-optimization/64811
14282 * match.pd: When optimizing comparisons with Inf, avoid
14283 introducing or losing exceptions from comparisons with NaN.
14285 2018-01-09 Martin Liska <mliska@suse.cz>
14288 * asan.c (shadow_mem_size): Add gcc_assert.
14290 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
14292 Don't save registers in main().
14295 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
14296 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
14297 * config/avr/avr.c (avr_set_current_function): Don't error if
14298 naked, OS_task or OS_main are specified at the same time.
14299 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
14301 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
14303 * common/config/avr/avr-common.c (avr_option_optimization_table):
14304 Switch on -mmain-is-OS_task for optimizing compilations.
14306 2018-01-09 Richard Biener <rguenther@suse.de>
14308 PR tree-optimization/83572
14309 * graphite.c: Include cfganal.h.
14310 (graphite_transform_loops): Connect infinite loops to exit
14311 and remove fake edges at the end.
14313 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14315 * ipa-inline.c (edge_badness): Revert accidental checkin.
14317 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14320 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
14321 symbols; not inline clones.
14323 2018-01-09 Jakub Jelinek <jakub@redhat.com>
14326 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
14327 hard registers. Formatting fixes.
14329 PR preprocessor/83722
14330 * gcc.c (try_generate_repro): Pass
14331 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
14332 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
14335 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
14336 Kito Cheng <kito.cheng@gmail.com>
14338 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
14339 (riscv_leaf_function_p): Delete.
14340 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
14342 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14344 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
14346 (do_ifelse): New function.
14347 (do_isel): New function.
14348 (do_sub3): New function.
14349 (do_add3): New function.
14350 (do_load_mask_compare): New function.
14351 (do_overlap_load_compare): New function.
14352 (expand_compare_loop): New function.
14353 (expand_block_compare): Call expand_compare_loop() when appropriate.
14354 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
14355 option description.
14356 (-mblock-compare-inline-loop-limit): New option.
14358 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14361 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
14362 Reverse order of second and third operands in first alternative.
14363 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
14364 of first and second elements in UNSPEC_VPERMR vector.
14365 (altivec_expand_vec_perm_le): Likewise.
14367 2018-01-08 Jeff Law <law@redhat.com>
14369 PR rtl-optimizatin/81308
14370 * tree-switch-conversion.c (cfg_altered): New file scoped static.
14371 (process_switch): If group_case_labels makes a change, then set
14373 (pass_convert_switch::execute): If a switch is converted, then
14374 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
14376 PR rtl-optimization/81308
14377 * recog.c (split_all_insns): Conditionally cleanup the CFG after
14380 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
14382 PR target/83663 - Revert r255946
14383 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
14384 generation for cases where splatting a value is not useful.
14385 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
14386 across a vec_duplicate and a paradoxical subreg forming a vector
14387 mode to a vec_concat.
14389 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14391 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
14392 -march=armv8.3-a variants.
14393 * config/arm/t-multilib: Likewise.
14394 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
14396 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14398 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
14400 (cceq_ior_compare_complement): Give it a name so I can use it, and
14401 change boolean_or_operator predicate to boolean_operator so it can
14402 be used to generate a crand.
14403 (eqne): New code iterator.
14404 (bd/bd_neg): New code_attrs.
14405 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
14406 a single define_insn.
14407 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
14408 decrement (bdnzt/bdnzf/bdzt/bdzf).
14409 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
14410 with the new names of the branch decrement patterns, and added the
14411 names of the branch decrement conditional patterns.
14413 2018-01-08 Richard Biener <rguenther@suse.de>
14415 PR tree-optimization/83563
14416 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
14419 2018-01-08 Richard Biener <rguenther@suse.de>
14421 PR middle-end/83713
14422 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
14424 2018-01-08 Richard Biener <rguenther@suse.de>
14426 PR tree-optimization/83685
14427 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
14428 references to abnormals.
14430 2018-01-08 Richard Biener <rguenther@suse.de>
14433 * dwarf2out.c (output_indirect_strings): Handle empty
14434 skeleton_debug_str_hash.
14435 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
14437 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14439 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
14440 (emit_store_direct): Likewise.
14441 (arc_trampoline_adjust_address): Likewise.
14442 (arc_asm_trampoline_template): New function.
14443 (arc_initialize_trampoline): Use asm_trampoline_template.
14444 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
14445 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
14446 * config/arc/arc.md (flush_icache): Delete pattern.
14448 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14450 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
14451 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
14454 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14457 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
14458 by not USED_FOR_TARGET.
14459 (make_pass_resolve_sw_modes): Likewise.
14461 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14463 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
14466 2018-01-08 Richard Biener <rguenther@suse.de>
14468 PR middle-end/83580
14469 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
14471 2018-01-08 Richard Biener <rguenther@suse.de>
14473 PR middle-end/83517
14474 * match.pd ((t * 2) / 2) -> t): Add missing :c.
14476 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
14478 PR middle-end/81897
14479 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
14480 basic blocks with a small number of successors.
14481 (convert_control_dep_chain_into_preds): Improve handling of
14483 (dump_predicates): Split apart into...
14484 (dump_pred_chain): ...here...
14485 (dump_pred_info): ...and here.
14486 (can_one_predicate_be_invalidated_p): Add debugging printfs.
14487 (can_chain_union_be_invalidated_p): Improve check for invalidation
14489 (uninit_uses_cannot_happen): Avoid unnecessary if
14490 convert_control_dep_chain_into_preds yielded nothing.
14492 2018-01-06 Martin Sebor <msebor@redhat.com>
14494 PR tree-optimization/83640
14495 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
14496 subtracting negative offset from size.
14497 (builtin_access::overlap): Adjust offset bounds of the access to fall
14498 within the size of the object if possible.
14500 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
14502 PR rtl-optimization/83699
14503 * expmed.c (extract_bit_field_1): Restrict the vector usage of
14504 extract_bit_field_as_subreg to cases in which the extracted
14505 value is also a vector.
14507 * lra-constraints.c (process_alt_operands): Test for the equivalence
14508 substitutions when detecting a possible reload cycle.
14510 2018-01-06 Jakub Jelinek <jakub@redhat.com>
14513 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
14514 by default if flag_selective_schedling{,2}. Formatting fixes.
14516 PR rtl-optimization/83682
14517 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
14518 if it has non-VECTOR_MODE element mode.
14519 (vec_duplicate_p): Likewise.
14521 PR middle-end/83694
14522 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
14523 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
14525 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14528 * config/i386/i386-builtin.def
14529 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
14530 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
14531 Require also OPTION_MASK_ISA_AVX512F in addition to
14532 OPTION_MASK_ISA_GFNI.
14533 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
14534 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
14535 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
14536 to OPTION_MASK_ISA_GFNI.
14537 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
14538 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
14539 OPTION_MASK_ISA_AVX512BW.
14540 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
14541 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
14542 addition to OPTION_MASK_ISA_GFNI.
14543 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
14544 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
14545 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
14546 to OPTION_MASK_ISA_GFNI.
14547 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
14548 a requirement for all ISAs rather than any of them with a few
14550 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
14552 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
14553 bitmasks to be enabled with 3 exceptions, instead of requiring any
14554 enabled ISA with lots of exceptions.
14555 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
14556 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
14557 Change avx512bw in isa attribute to avx512f.
14558 * config/i386/sgxintrin.h: Add license boilerplate.
14559 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
14560 to __AVX512F__ and __AVX512VL to __AVX512VL__.
14561 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
14562 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
14564 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
14565 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
14566 temporarily sse2 rather than sse if not enabled already.
14569 * config/i386/sse.md (VI248_VLBW): Rename to ...
14570 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
14571 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
14572 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
14573 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
14574 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
14575 mode iterator instead of VI248_VLBW.
14577 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
14579 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
14580 (record_modified): Skip clobbers; add debug output.
14581 (param_change_prob): Use sreal frequencies.
14583 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14585 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
14586 punt for user-aligned variables.
14588 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14590 * tree-chrec.c (chrec_contains_symbols): Return true for
14593 2018-01-05 Sudakshina Das <sudi.das@arm.com>
14596 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
14597 of (x|y) == x for BICS pattern.
14599 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14601 PR tree-optimization/83605
14602 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
14603 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
14606 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
14608 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
14609 * config/epiphany/rtems.h: New file.
14611 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14612 Uros Bizjak <ubizjak@gmail.com>
14615 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
14616 QIreg_operand instead of register_operand predicate.
14617 * config/i386/i386.c (ix86_rop_should_change_byte_p,
14618 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
14619 comments instead of -fmitigate[-_]rop.
14621 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14624 * cgraphunit.c (symbol_table::compile): Switch to text_section
14625 before calling assembly_start debug hook.
14626 * run-rtl-passes.c (run_rtl_passes): Likewise.
14629 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14631 * tree-vrp.c (extract_range_from_binary_expr_1): Check
14632 range_int_cst_p rather than !symbolic_range_p before calling
14633 extract_range_from_multiplicative_op_1.
14635 2018-01-04 Jeff Law <law@redhat.com>
14637 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
14638 redundant test in assertion.
14640 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14642 * doc/rtl.texi: Document machine_mode wrapper classes.
14644 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14646 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
14647 using tree_to_uhwi.
14649 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14651 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
14652 the VEC_PERM_EXPR fold to fail.
14654 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14657 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
14658 to switched_sections.
14660 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14663 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
14664 test for d.testing.
14666 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
14669 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
14670 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
14672 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14675 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
14676 is BLKmode and bitpos not zero or mode change is needed.
14678 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14681 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
14684 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
14687 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
14688 instead of MULT rtx. Update all corresponding splitters.
14689 (*saddl_se): Ditto.
14690 (*ssub<modesuffix>): Ditto.
14691 (*ssubl_se): Ditto.
14692 (*cmp_sadd_di): Update split patterns.
14693 (*cmp_sadd_si): Ditto.
14694 (*cmp_sadd_sidi): Ditto.
14695 (*cmp_ssub_di): Ditto.
14696 (*cmp_ssub_si): Ditto.
14697 (*cmp_ssub_sidi): Ditto.
14698 * config/alpha/predicates.md (const23_operand): New predicate.
14699 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
14700 Look for ASHIFT, not MULT inner operand.
14701 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
14703 2018-01-04 Martin Liska <mliska@suse.cz>
14705 PR gcov-profile/83669
14706 * gcov.c (output_intermediate_file): Add version to intermediate
14708 * doc/gcov.texi: Document new field 'version' in intermediate
14709 file format. Fix location of '-k' option of gcov command.
14711 2018-01-04 Martin Liska <mliska@suse.cz>
14714 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
14716 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14718 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
14720 2018-01-03 Martin Sebor <msebor@redhat.com>
14722 PR tree-optimization/83655
14723 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
14724 checking calls with invalid arguments.
14726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14728 * tree-vect-stmts.c (vect_get_store_rhs): New function.
14729 (vectorizable_mask_load_store): Delete.
14730 (vectorizable_call): Return false for masked loads and stores.
14731 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
14732 instead of gimple_assign_rhs1.
14733 (vectorizable_load): Handle IFN_MASK_LOAD.
14734 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
14736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14738 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
14740 (vectorizable_mask_load_store): ...here.
14741 (vectorizable_load): ...and here.
14743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14745 * tree-vect-stmts.c (vect_build_all_ones_mask)
14746 (vect_build_zero_merge_argument): New functions, split out from...
14747 (vectorizable_load): ...here.
14749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14751 * tree-vect-stmts.c (vect_check_store_rhs): New function,
14753 (vectorizable_mask_load_store): ...here.
14754 (vectorizable_store): ...and here.
14756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14758 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
14760 (vectorizable_mask_load_store): ...here.
14762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14764 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
14765 (vect_model_store_cost): Take a vec_load_store_type instead of a
14767 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
14768 (vect_model_store_cost): Take a vec_load_store_type instead of a
14770 (vectorizable_mask_load_store): Update accordingly.
14771 (vectorizable_store): Likewise.
14772 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
14774 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14776 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
14777 IFN_MASK_LOAD calls here rather than...
14778 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
14780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14781 Alan Hayward <alan.hayward@arm.com>
14782 David Sherwood <david.sherwood@arm.com>
14784 * expmed.c (extract_bit_field_1): For vector extracts,
14785 fall back to extract_bit_field_as_subreg if vec_extract
14788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14789 Alan Hayward <alan.hayward@arm.com>
14790 David Sherwood <david.sherwood@arm.com>
14792 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
14793 they are variable or constant sized.
14794 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
14795 slots for constant-sized data.
14797 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14798 Alan Hayward <alan.hayward@arm.com>
14799 David Sherwood <david.sherwood@arm.com>
14801 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
14802 handling COND_EXPRs with boolean comparisons, try to find a better
14803 basis for the mask type than the boolean itself.
14805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14807 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
14808 is calculated and how it can be overridden.
14809 * genmodes.c (max_bitsize_mode_any_mode): New variable.
14810 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
14812 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
14815 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14816 Alan Hayward <alan.hayward@arm.com>
14817 David Sherwood <david.sherwood@arm.com>
14819 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
14820 Remove the mode argument.
14821 (aarch64_simd_valid_immediate): Remove the mode and inverse
14823 * config/aarch64/iterators.md (bitsize): New iterator.
14824 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
14825 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
14826 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
14827 aarch64_simd_valid_immediate.
14828 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
14829 (aarch64_reg_or_bic_imm): Likewise.
14830 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
14831 with an insn_type enum and msl with a modifier_type enum.
14832 Replace element_width with a scalar_mode. Change the shift
14833 to unsigned int. Add constructors for scalar_float_mode and
14834 scalar_int_mode elements.
14835 (aarch64_vect_float_const_representable_p): Delete.
14836 (aarch64_can_const_movi_rtx_p)
14837 (aarch64_simd_scalar_immediate_valid_for_move)
14838 (aarch64_simd_make_constant): Update call to
14839 aarch64_simd_valid_immediate.
14840 (aarch64_advsimd_valid_immediate_hs): New function.
14841 (aarch64_advsimd_valid_immediate): Likewise.
14842 (aarch64_simd_valid_immediate): Remove mode and inverse
14843 arguments. Rewrite to use the above. Use const_vec_duplicate_p
14844 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
14845 and aarch64_float_const_representable_p on the result.
14846 (aarch64_output_simd_mov_immediate): Remove mode argument.
14847 Update call to aarch64_simd_valid_immediate and use of
14848 simd_immediate_info.
14849 (aarch64_output_scalar_simd_mov_immediate): Update call
14852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14853 Alan Hayward <alan.hayward@arm.com>
14854 David Sherwood <david.sherwood@arm.com>
14856 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
14857 (mode_nunits): Likewise CONST_MODE_NUNITS.
14858 * machmode.def (ADJUST_NUNITS): Document.
14859 * genmodes.c (mode_data::need_nunits_adj): New field.
14860 (blank_mode): Update accordingly.
14861 (adj_nunits): New variable.
14862 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
14864 (emit_mode_size_inline): Set need_bytesize_adj for all modes
14865 listed in adj_nunits.
14866 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
14867 listed in adj_nunits. Don't emit case statements for such modes.
14868 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
14869 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
14870 nothing if adj_nunits is nonnull.
14871 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
14872 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
14873 (emit_mode_fbit): Update use of print_maybe_const_decl.
14874 (emit_move_size): Likewise. Treat the array as non-const
14876 (emit_mode_adjustments): Handle adj_nunits.
14878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14880 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
14881 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
14882 (VECTOR_MODES): Use it.
14883 (make_vector_modes): Take the prefix as an argument.
14885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14886 Alan Hayward <alan.hayward@arm.com>
14887 David Sherwood <david.sherwood@arm.com>
14889 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
14890 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
14891 for MODE_VECTOR_BOOL.
14892 * machmode.def (VECTOR_BOOL_MODE): Document.
14893 * genmodes.c (VECTOR_BOOL_MODE): New macro.
14894 (make_vector_bool_mode): New function.
14895 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
14897 * lto-streamer-in.c (lto_input_mode_table): Likewise.
14898 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
14900 * stor-layout.c (int_mode_for_mode): Likewise.
14901 * tree.c (build_vector_type_for_mode): Likewise.
14902 * varasm.c (output_constant_pool_2): Likewise.
14903 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
14904 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
14905 for MODE_VECTOR_BOOL.
14906 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
14907 of mode class checks.
14908 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
14909 instead of a list of mode class checks.
14910 (expand_vector_scalar_condition): Likewise.
14911 (type_for_widest_vector_mode): Handle BImode as an inner mode.
14913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14914 Alan Hayward <alan.hayward@arm.com>
14915 David Sherwood <david.sherwood@arm.com>
14917 * machmode.h (mode_size): Change from unsigned short to
14919 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
14920 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14921 or if measurement_type is not polynomial.
14922 (fixed_size_mode::includes_p): Check for constant-sized modes.
14923 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
14924 return a poly_uint16 rather than an unsigned short.
14925 (emit_mode_size): Change the type of mode_size from unsigned short
14926 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
14927 (emit_mode_adjustments): Cope with polynomial vector sizes.
14928 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14930 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14932 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
14933 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
14934 * caller-save.c (setup_save_areas): Likewise.
14935 (replace_reg_with_saved_mem): Likewise.
14936 * calls.c (emit_library_call_value_1): Likewise.
14937 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
14938 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
14939 (gen_lowpart_for_combine): Likewise.
14940 * convert.c (convert_to_integer_1): Likewise.
14941 * cse.c (equiv_constant, cse_insn): Likewise.
14942 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
14943 (cselib_subst_to_values): Likewise.
14944 * dce.c (word_dce_process_block): Likewise.
14945 * df-problems.c (df_word_lr_mark_ref): Likewise.
14946 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
14947 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
14948 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
14949 (rtl_for_decl_location): Likewise.
14950 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
14951 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
14952 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
14953 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
14954 (expand_expr_real_1): Likewise.
14955 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
14956 (pad_below): Likewise.
14957 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14958 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
14959 * ira.c (get_subreg_tracking_sizes): Likewise.
14960 * ira-build.c (ira_create_allocno_objects): Likewise.
14961 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
14962 (ira_sort_regnos_for_alter_reg): Likewise.
14963 * ira-costs.c (record_operand_costs): Likewise.
14964 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
14965 (resolve_simple_move): Likewise.
14966 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
14967 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
14968 (lra_constraints): Likewise.
14969 (CONST_POOL_OK_P): Reject variable-sized modes.
14970 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
14971 (add_pseudo_to_slot, lra_spill): Likewise.
14972 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14973 * optabs-query.c (get_best_extraction_insn): Likewise.
14974 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14975 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
14976 (expand_mult_highpart, valid_multiword_target_p): Likewise.
14977 * recog.c (offsettable_address_addr_space_p): Likewise.
14978 * regcprop.c (maybe_mode_change): Likewise.
14979 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
14980 * regrename.c (build_def_use): Likewise.
14981 * regstat.c (dump_reg_info): Likewise.
14982 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
14983 (find_reloads, find_reloads_subreg_address): Likewise.
14984 * reload1.c (eliminate_regs_1): Likewise.
14985 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
14986 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
14987 (simplify_binary_operation_1, simplify_subreg): Likewise.
14988 * targhooks.c (default_function_arg_padding): Likewise.
14989 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
14990 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
14991 (verify_gimple_assign_ternary): Likewise.
14992 * tree-inline.c (estimate_move_cost): Likewise.
14993 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14994 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
14995 (get_address_cost_ainc): Likewise.
14996 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
14997 (vect_supportable_dr_alignment): Likewise.
14998 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14999 (vectorizable_reduction): Likewise.
15000 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15001 (vectorizable_operation, vectorizable_load): Likewise.
15002 * tree.c (build_same_sized_truth_vector_type): Likewise.
15003 * valtrack.c (cleanup_auto_inc_dec): Likewise.
15004 * var-tracking.c (emit_note_insn_var_location): Likewise.
15005 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15006 (ADDR_VEC_ALIGN): Likewise.
15008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15009 Alan Hayward <alan.hayward@arm.com>
15010 David Sherwood <david.sherwood@arm.com>
15012 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
15014 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15015 or if measurement_type is polynomial.
15016 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15017 * combine.c (make_extraction): Likewise.
15018 * dse.c (find_shift_sequence): Likewise.
15019 * dwarf2out.c (mem_loc_descriptor): Likewise.
15020 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15021 (extract_bit_field, extract_low_bits): Likewise.
15022 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15023 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15024 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15025 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15026 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15027 * reload.c (find_reloads): Likewise.
15028 * reload1.c (alter_reg): Likewise.
15029 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15030 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15031 * tree-if-conv.c (predicate_mem_writes): Likewise.
15032 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15033 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15034 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15035 * valtrack.c (dead_debug_insert_temp): Likewise.
15036 * varasm.c (mergeable_constant_section): Likewise.
15037 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15040 Alan Hayward <alan.hayward@arm.com>
15041 David Sherwood <david.sherwood@arm.com>
15043 * expr.c (expand_assignment): Cope with polynomial mode sizes
15044 when assigning to a CONCAT.
15046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15047 Alan Hayward <alan.hayward@arm.com>
15048 David Sherwood <david.sherwood@arm.com>
15050 * machmode.h (mode_precision): Change from unsigned short to
15052 (mode_to_precision): Return a poly_uint16 rather than an unsigned
15054 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15055 or if measurement_type is not polynomial.
15056 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
15057 in which the mode is already known to be a scalar_int_mode.
15058 * genmodes.c (emit_mode_precision): Change the type of mode_precision
15059 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
15061 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15062 for GET_MODE_PRECISION.
15063 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15064 for GET_MODE_PRECISION.
15065 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15067 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15068 (expand_field_assignment, make_extraction): Likewise.
15069 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15070 (get_last_value): Likewise.
15071 * convert.c (convert_to_integer_1): Likewise.
15072 * cse.c (cse_insn): Likewise.
15073 * expr.c (expand_expr_real_1): Likewise.
15074 * lra-constraints.c (simplify_operand_subreg): Likewise.
15075 * optabs-query.c (can_atomic_load_p): Likewise.
15076 * optabs.c (expand_atomic_load): Likewise.
15077 (expand_atomic_store): Likewise.
15078 * ree.c (combine_reaching_defs): Likewise.
15079 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15080 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15081 * tree.h (type_has_mode_precision_p): Likewise.
15082 * ubsan.c (instrument_si_overflow): Likewise.
15084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15085 Alan Hayward <alan.hayward@arm.com>
15086 David Sherwood <david.sherwood@arm.com>
15088 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15089 polynomial numbers of units.
15090 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15091 (valid_vector_subparts_p): New function.
15092 (build_vector_type): Remove temporary shim and take the number
15093 of units as a poly_uint64 rather than an int.
15094 (build_opaque_vector_type): Take the number of units as a
15095 poly_uint64 rather than an int.
15096 * tree.c (build_vector_from_ctor): Handle polynomial
15097 TYPE_VECTOR_SUBPARTS.
15098 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15099 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15100 (build_vector_from_val): If the number of units is variable,
15101 use build_vec_duplicate_cst for constant operands and
15102 VEC_DUPLICATE_EXPR otherwise.
15103 (make_vector_type): Remove temporary is_constant ().
15104 (build_vector_type, build_opaque_vector_type): Take the number of
15105 units as a poly_uint64 rather than an int.
15106 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15108 * cfgexpand.c (expand_debug_expr): Likewise.
15109 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15110 (store_constructor, expand_expr_real_1): Likewise.
15111 (const_scalar_mask_from_tree): Likewise.
15112 * fold-const-call.c (fold_const_reduction): Likewise.
15113 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15114 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15115 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15116 (fold_relational_const): Likewise.
15117 (native_interpret_vector): Likewise. Change the size from an
15118 int to an unsigned int.
15119 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15120 TYPE_VECTOR_SUBPARTS.
15121 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15122 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15123 duplicating a non-constant operand into a variable-length vector.
15124 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15125 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15126 * ipa-icf.c (sem_variable::equals): Likewise.
15127 * match.pd: Likewise.
15128 * omp-simd-clone.c (simd_clone_subparts): Likewise.
15129 * print-tree.c (print_node): Likewise.
15130 * stor-layout.c (layout_type): Likewise.
15131 * targhooks.c (default_builtin_vectorization_cost): Likewise.
15132 * tree-cfg.c (verify_gimple_comparison): Likewise.
15133 (verify_gimple_assign_binary): Likewise.
15134 (verify_gimple_assign_ternary): Likewise.
15135 (verify_gimple_assign_single): Likewise.
15136 * tree-pretty-print.c (dump_generic_node): Likewise.
15137 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15138 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15139 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15140 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15141 (vect_shift_permute_load_chain): Likewise.
15142 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15143 (expand_vector_condition, optimize_vector_constructor): Likewise.
15144 (lower_vec_perm, get_compute_type): Likewise.
15145 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15146 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15147 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15148 (vect_recog_mask_conversion_pattern): Likewise.
15149 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15150 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15151 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15152 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15153 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15154 (vectorizable_shift, vectorizable_operation, vectorizable_store)
15155 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15156 (supportable_widening_operation): Likewise.
15157 (supportable_narrowing_operation): Likewise.
15158 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15160 * varasm.c (output_constant): Likewise.
15162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15163 Alan Hayward <alan.hayward@arm.com>
15164 David Sherwood <david.sherwood@arm.com>
15166 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
15167 so that both the length == 3 and length != 3 cases set up their
15168 own permute vectors. Add comments explaining why we know the
15169 number of elements is constant.
15170 (vect_permute_load_chain): Likewise.
15172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15173 Alan Hayward <alan.hayward@arm.com>
15174 David Sherwood <david.sherwood@arm.com>
15176 * machmode.h (mode_nunits): Change from unsigned char to
15178 (ONLY_FIXED_SIZE_MODES): New macro.
15179 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
15180 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
15181 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
15183 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
15184 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
15185 or if measurement_type is not polynomial.
15186 * genmodes.c (ZERO_COEFFS): New macro.
15187 (emit_mode_nunits_inline): Make mode_nunits_inline return a
15189 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
15190 Use ZERO_COEFFS when emitting initializers.
15191 * data-streamer.h (bp_pack_poly_value): New function.
15192 (bp_unpack_poly_value): Likewise.
15193 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15194 for GET_MODE_NUNITS.
15195 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15196 for GET_MODE_NUNITS.
15197 * tree.c (make_vector_type): Remove temporary shim and make
15198 the real function take the number of units as a poly_uint64
15199 rather than an int.
15200 (build_vector_type_for_mode): Handle polynomial nunits.
15201 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
15202 * emit-rtl.c (const_vec_series_p_1): Likewise.
15203 (gen_rtx_CONST_VECTOR): Likewise.
15204 * fold-const.c (test_vec_duplicate_folding): Likewise.
15205 * genrecog.c (validate_pattern): Likewise.
15206 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
15207 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15208 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
15209 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
15210 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
15211 * rtlanal.c (subreg_get_info): Likewise.
15212 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15213 (vect_grouped_load_supported): Likewise.
15214 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
15215 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
15216 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15217 (simplify_const_unary_operation, simplify_binary_operation_1)
15218 (simplify_const_binary_operation, simplify_ternary_operation)
15219 (test_vector_ops_duplicate, test_vector_ops): Likewise.
15220 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
15221 instead of CONST_VECTOR_NUNITS.
15222 * varasm.c (output_constant_pool_2): Likewise.
15223 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
15224 explicit-encoded elements in the XVEC for variable-length vectors.
15226 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15228 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
15230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15231 Alan Hayward <alan.hayward@arm.com>
15232 David Sherwood <david.sherwood@arm.com>
15234 * coretypes.h (fixed_size_mode): Declare.
15235 (fixed_size_mode_pod): New typedef.
15236 * builtins.h (target_builtins::x_apply_args_mode)
15237 (target_builtins::x_apply_result_mode): Change type to
15238 fixed_size_mode_pod.
15239 * builtins.c (apply_args_size, apply_result_size, result_vector)
15240 (expand_builtin_apply_args_1, expand_builtin_apply)
15241 (expand_builtin_return): Update accordingly.
15243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15245 * cse.c (hash_rtx_cb): Hash only the encoded elements.
15246 * cselib.c (cselib_hash_rtx): Likewise.
15247 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
15248 CONST_VECTOR encoding.
15250 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15251 Jeff Law <law@redhat.com>
15254 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
15255 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
15256 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
15257 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
15260 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
15261 explicitly probe *sp in a noreturn function if there were any callee
15262 register saves or frame pointer is needed.
15264 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15267 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
15268 BLKmode for ternary, binary or unary expressions.
15271 * var-tracking.c (delete_vta_debug_insn): New inline function.
15272 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
15273 insns from get_insns () to NULL instead of each bb separately.
15274 Use delete_vta_debug_insn. No longer static.
15275 (vt_debug_insns_local, variable_tracking_main_1): Adjust
15276 delete_vta_debug_insns callers.
15277 * rtl.h (delete_vta_debug_insns): Declare.
15278 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
15279 instead of variable_tracking_main.
15281 2018-01-03 Martin Sebor <msebor@redhat.com>
15283 PR tree-optimization/83603
15284 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
15285 arguments past the endof the argument list in functions declared
15286 without a prototype.
15287 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
15288 Avoid checking when arguments are null.
15290 2018-01-03 Martin Sebor <msebor@redhat.com>
15293 * doc/extend.texi (attribute const): Fix a typo.
15294 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
15295 issuing -Wsuggest-attribute for void functions.
15297 2018-01-03 Martin Sebor <msebor@redhat.com>
15299 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
15300 offset_int::from instead of wide_int::to_shwi.
15301 (maybe_diag_overlap): Remove assertion.
15302 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
15303 * gimple-ssa-sprintf.c (format_directive): Same.
15304 (parse_directive): Same.
15305 (sprintf_dom_walker::compute_format_length): Same.
15306 (try_substitute_return_value): Same.
15308 2018-01-03 Jeff Law <law@redhat.com>
15310 PR middle-end/83654
15311 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
15312 non-constant residual for zero at runtime and avoid probing in
15313 that case. Reorganize code for trailing problem to mirror handling
15316 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15318 PR tree-optimization/83501
15319 * tree-ssa-strlen.c (get_string_cst): New.
15320 (handle_char_store): Call get_string_cst.
15322 2018-01-03 Martin Liska <mliska@suse.cz>
15324 PR tree-optimization/83593
15325 * tree-ssa-strlen.c: Include tree-cfg.h.
15326 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
15327 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
15328 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
15330 (strlen_dom_walker::before_dom_children): Call
15331 gimple_purge_dead_eh_edges. Dump tranformation with details
15333 (strlen_dom_walker::before_dom_children): Update call by adding
15334 new argument cleanup_eh.
15335 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
15337 2018-01-03 Martin Liska <mliska@suse.cz>
15340 * cif-code.def (VARIADIC_THUNK): New enum value.
15341 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
15344 2018-01-03 Jan Beulich <jbeulich@suse.com>
15346 * sse.md (mov<mode>_internal): Tighten condition for when to use
15347 vmovdqu<ssescalarsize> for TI and OI modes.
15349 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15351 Update copyright years.
15353 2018-01-03 Martin Liska <mliska@suse.cz>
15356 * ipa-visibility.c (function_and_variable_visibility): Skip
15357 functions with noipa attribure.
15359 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15361 * gcc.c (process_command): Update copyright notice dates.
15362 * gcov-dump.c (print_version): Ditto.
15363 * gcov.c (print_version): Ditto.
15364 * gcov-tool.c (print_version): Ditto.
15365 * gengtype.c (create_file): Ditto.
15366 * doc/cpp.texi: Bump @copying's copyright year.
15367 * doc/cppinternals.texi: Ditto.
15368 * doc/gcc.texi: Ditto.
15369 * doc/gccint.texi: Ditto.
15370 * doc/gcov.texi: Ditto.
15371 * doc/install.texi: Ditto.
15372 * doc/invoke.texi: Ditto.
15374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15376 * vector-builder.h (vector_builder::m_full_nelts): Change from
15377 unsigned int to poly_uint64.
15378 (vector_builder::full_nelts): Update prototype accordingly.
15379 (vector_builder::new_vector): Likewise.
15380 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
15381 (vector_builder::operator ==): Likewise.
15382 (vector_builder::finalize): Likewise.
15383 * int-vector-builder.h (int_vector_builder::int_vector_builder):
15384 Take the number of elements as a poly_uint64 rather than an
15386 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
15387 from unsigned int to poly_uint64.
15388 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
15389 (vec_perm_indices::new_vector): Likewise.
15390 (vec_perm_indices::length): Likewise.
15391 (vec_perm_indices::nelts_per_input): Likewise.
15392 (vec_perm_indices::input_nelts): Likewise.
15393 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
15394 number of elements per input as a poly_uint64 rather than an
15395 unsigned int. Use the original encoding for variable-length
15396 vectors, rather than clamping each individual element.
15397 For the second and subsequent elements in each pattern,
15398 clamp the step and base before clamping their sum.
15399 (vec_perm_indices::series_p): Handle polynomial element counts.
15400 (vec_perm_indices::all_in_range_p): Likewise.
15401 (vec_perm_indices_to_tree): Likewise.
15402 (vec_perm_indices_to_rtx): Likewise.
15403 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
15404 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
15405 (tree_vector_builder::new_binary_operation): Handle polynomial
15406 element counts. Return false if we need to know the number
15407 of elements at compile time.
15408 * fold-const.c (fold_vec_perm): Punt if the number of elements
15409 isn't known at compile time.
15411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15413 * vec-perm-indices.h (vec_perm_builder): Change element type
15414 from HOST_WIDE_INT to poly_int64.
15415 (vec_perm_indices::element_type): Update accordingly.
15416 (vec_perm_indices::clamp): Handle polynomial element_types.
15417 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15418 (vec_perm_indices::all_in_range_p): Likewise.
15419 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
15421 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
15422 polynomial vec_perm_indices element types.
15423 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
15424 * fold-const.c (fold_vec_perm): Likewise.
15425 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
15426 * tree-vect-generic.c (lower_vec_perm): Likewise.
15427 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15428 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
15429 element type to HOST_WIDE_INT.
15431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15432 Alan Hayward <alan.hayward@arm.com>
15433 David Sherwood <david.sherwood@arm.com>
15435 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
15436 rather than an int. Use plus_constant.
15437 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
15438 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
15440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15441 Alan Hayward <alan.hayward@arm.com>
15442 David Sherwood <david.sherwood@arm.com>
15444 * calls.c (emit_call_1, expand_call): Change struct_value_size from
15445 a HOST_WIDE_INT to a poly_int64.
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 * calls.c (load_register_parameters): Cope with polynomial
15452 mode sizes. Require a constant size for BLKmode parameters
15453 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
15454 forces a parameter to be padded at the lsb end in order to
15455 fill a complete number of words, require the parameter size
15456 to be ordered wrt UNITS_PER_WORD.
15458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15459 Alan Hayward <alan.hayward@arm.com>
15460 David Sherwood <david.sherwood@arm.com>
15462 * reload1.c (spill_stack_slot_width): Change element type
15463 from unsigned int to poly_uint64_pod.
15464 (alter_reg): Treat mode sizes as polynomial.
15466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15467 Alan Hayward <alan.hayward@arm.com>
15468 David Sherwood <david.sherwood@arm.com>
15470 * reload.c (complex_word_subreg_p): New function.
15471 (reload_inner_reg_of_subreg, push_reload): Use it.
15473 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15474 Alan Hayward <alan.hayward@arm.com>
15475 David Sherwood <david.sherwood@arm.com>
15477 * lra-constraints.c (process_alt_operands): Reject matched
15478 operands whose sizes aren't ordered.
15479 (match_reload): Refer to this check here.
15481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15482 Alan Hayward <alan.hayward@arm.com>
15483 David Sherwood <david.sherwood@arm.com>
15485 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
15486 that the mode size is in the set {1, 2, 4, 8, 16}.
15488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15489 Alan Hayward <alan.hayward@arm.com>
15490 David Sherwood <david.sherwood@arm.com>
15492 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
15493 Use plus_constant instead of gen_rtx_PLUS.
15495 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15496 Alan Hayward <alan.hayward@arm.com>
15497 David Sherwood <david.sherwood@arm.com>
15499 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
15500 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
15501 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
15502 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
15503 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
15504 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
15505 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
15506 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
15507 * config/i386/i386.c (ix86_push_rounding): ...this new function.
15508 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
15510 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
15511 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
15512 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
15513 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
15514 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
15515 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
15516 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
15517 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
15518 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
15519 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
15521 * expr.c (emit_move_resolve_push): Treat the input and result
15522 of PUSH_ROUNDING as a poly_int64.
15523 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
15524 (emit_push_insn): Likewise.
15525 * lra-eliminations.c (mark_not_eliminable): Likewise.
15526 * recog.c (push_operand): Likewise.
15527 * reload1.c (elimination_effects): Likewise.
15528 * rtlanal.c (nonzero_bits1): Likewise.
15529 * calls.c (store_one_arg): Likewise. Require the padding to be
15530 known at compile time.
15532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15533 Alan Hayward <alan.hayward@arm.com>
15534 David Sherwood <david.sherwood@arm.com>
15536 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
15537 Use plus_constant instead of gen_rtx_PLUS.
15539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15540 Alan Hayward <alan.hayward@arm.com>
15541 David Sherwood <david.sherwood@arm.com>
15543 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
15544 rather than an int.
15546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15547 Alan Hayward <alan.hayward@arm.com>
15548 David Sherwood <david.sherwood@arm.com>
15550 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
15551 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
15552 via stack temporaries. Treat the mode size as polynomial too.
15554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15555 Alan Hayward <alan.hayward@arm.com>
15556 David Sherwood <david.sherwood@arm.com>
15558 * expr.c (expand_expr_real_2): When handling conversions involving
15559 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
15560 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
15561 as a poly_uint64 too.
15563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15564 Alan Hayward <alan.hayward@arm.com>
15565 David Sherwood <david.sherwood@arm.com>
15567 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
15569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15570 Alan Hayward <alan.hayward@arm.com>
15571 David Sherwood <david.sherwood@arm.com>
15573 * combine.c (can_change_dest_mode): Handle polynomial
15574 REGMODE_NATURAL_SIZE.
15575 * expmed.c (store_bit_field_1): Likewise.
15576 * expr.c (store_constructor): Likewise.
15577 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
15578 and polynomial REGMODE_NATURAL_SIZE.
15579 (gen_lowpart_common): Likewise.
15580 * reginfo.c (record_subregs_of_mode): Likewise.
15581 * rtlanal.c (read_modify_subreg_p): Likewise.
15583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15584 Alan Hayward <alan.hayward@arm.com>
15585 David Sherwood <david.sherwood@arm.com>
15587 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
15588 numbers of elements.
15590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15591 Alan Hayward <alan.hayward@arm.com>
15592 David Sherwood <david.sherwood@arm.com>
15594 * match.pd: Cope with polynomial numbers of vector elements.
15596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15597 Alan Hayward <alan.hayward@arm.com>
15598 David Sherwood <david.sherwood@arm.com>
15600 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
15601 in a POINTER_PLUS_EXPR.
15603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15604 Alan Hayward <alan.hayward@arm.com>
15605 David Sherwood <david.sherwood@arm.com>
15607 * omp-simd-clone.c (simd_clone_subparts): New function.
15608 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
15609 (ipa_simd_modify_function_body): Likewise.
15611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15612 Alan Hayward <alan.hayward@arm.com>
15613 David Sherwood <david.sherwood@arm.com>
15615 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
15616 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
15617 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
15618 (expand_vector_condition, vector_element): Likewise.
15619 (subparts_gt): New function.
15620 (get_compute_type): Use subparts_gt.
15621 (count_type_subparts): Delete.
15622 (expand_vector_operations_1): Use subparts_gt instead of
15623 count_type_subparts.
15625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15626 Alan Hayward <alan.hayward@arm.com>
15627 David Sherwood <david.sherwood@arm.com>
15629 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
15630 (vect_compile_time_alias): ...this new function. Do the calculation
15631 on poly_ints rather than trees.
15632 (vect_prune_runtime_alias_test_list): Update call accordingly.
15634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15635 Alan Hayward <alan.hayward@arm.com>
15636 David Sherwood <david.sherwood@arm.com>
15638 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
15640 (vect_schedule_slp_instance): Likewise.
15642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15643 Alan Hayward <alan.hayward@arm.com>
15644 David Sherwood <david.sherwood@arm.com>
15646 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
15647 constant and extern definitions for variable-length vectors.
15648 (vect_get_constant_vectors): Note that the number of units
15649 is known to be constant.
15651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15652 Alan Hayward <alan.hayward@arm.com>
15653 David Sherwood <david.sherwood@arm.com>
15655 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
15656 of units as polynomial. Choose between WIDE and NARROW based
15659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15660 Alan Hayward <alan.hayward@arm.com>
15661 David Sherwood <david.sherwood@arm.com>
15663 * tree-vect-stmts.c (simd_clone_subparts): New function.
15664 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
15666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15667 Alan Hayward <alan.hayward@arm.com>
15668 David Sherwood <david.sherwood@arm.com>
15670 * tree-vect-stmts.c (vectorizable_call): Treat the number of
15671 vectors as polynomial. Use build_index_vector for
15672 IFN_GOMP_SIMD_LANE.
15674 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15675 Alan Hayward <alan.hayward@arm.com>
15676 David Sherwood <david.sherwood@arm.com>
15678 * tree-vect-stmts.c (get_load_store_type): Treat the number of
15679 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
15680 for variable-length vectors.
15681 (vectorizable_mask_load_store): Treat the number of units as
15682 polynomial, asserting that it is constant if the condition has
15683 already been enforced.
15684 (vectorizable_store, vectorizable_load): Likewise.
15686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15687 Alan Hayward <alan.hayward@arm.com>
15688 David Sherwood <david.sherwood@arm.com>
15690 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
15691 of units as polynomial. Punt if we can't tell at compile time
15692 which vector contains the final result.
15694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15695 Alan Hayward <alan.hayward@arm.com>
15696 David Sherwood <david.sherwood@arm.com>
15698 * tree-vect-loop.c (vectorizable_induction): Treat the number
15699 of units as polynomial. Punt on SLP inductions. Use an integer
15700 VEC_SERIES_EXPR for variable-length integer reductions. Use a
15701 cast of such a series for variable-length floating-point
15704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15705 Alan Hayward <alan.hayward@arm.com>
15706 David Sherwood <david.sherwood@arm.com>
15708 * tree.h (build_index_vector): Declare.
15709 * tree.c (build_index_vector): New function.
15710 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
15711 of units as polynomial, forcibly converting it to a constant if
15712 vectorizable_reduction has already enforced the condition.
15713 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
15714 to create a {1,2,3,...} vector.
15715 (vectorizable_reduction): Treat the number of units as polynomial.
15716 Choose vectype_in based on the largest scalar element size rather
15717 than the smallest number of units. Enforce the restrictions
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-vect-data-refs.c (vector_alignment_reachable_p): Treat the
15725 number of units as polynomial.
15727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15728 Alan Hayward <alan.hayward@arm.com>
15729 David Sherwood <david.sherwood@arm.com>
15731 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
15732 * target.def (autovectorize_vector_sizes): Return the vector sizes
15733 by pointer, using vector_sizes rather than a bitmask.
15734 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
15735 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
15736 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
15738 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
15739 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
15740 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
15741 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
15742 * omp-general.c (omp_max_vf): Likewise.
15743 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15744 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
15745 * tree-vect-loop.c (vect_analyze_loop): Likewise.
15746 * tree-vect-slp.c (vect_slp_bb): Likewise.
15747 * doc/tm.texi: Regenerate.
15748 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
15750 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
15751 the vector size as a poly_uint64 rather than an unsigned int.
15752 (current_vector_size): Change from an unsigned int to a poly_uint64.
15753 (get_vectype_for_scalar_type): Update accordingly.
15754 * tree.h (build_truth_vector_type): Take the size and number of
15755 units as a poly_uint64 rather than an unsigned int.
15756 (build_vector_type): Add a temporary overload that takes
15757 the number of units as a poly_uint64 rather than an unsigned int.
15758 * tree.c (make_vector_type): Likewise.
15759 (build_truth_vector_type): Take the number of units as a poly_uint64
15760 rather than an unsigned int.
15762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15763 Alan Hayward <alan.hayward@arm.com>
15764 David Sherwood <david.sherwood@arm.com>
15766 * target.def (get_mask_mode): Take the number of units and length
15767 as poly_uint64s rather than unsigned ints.
15768 * targhooks.h (default_get_mask_mode): Update accordingly.
15769 * targhooks.c (default_get_mask_mode): Likewise.
15770 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
15771 * doc/tm.texi: Regenerate.
15773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15774 Alan Hayward <alan.hayward@arm.com>
15775 David Sherwood <david.sherwood@arm.com>
15777 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
15778 * omp-general.c (omp_max_vf): Likewise.
15779 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
15780 (expand_omp_simd): Handle polynomial safelen.
15781 * omp-low.c (omplow_simd_context): Add a default constructor.
15782 (omplow_simd_context::max_vf): Change from int to poly_uint64.
15783 (lower_rec_simd_input_clauses): Update accordingly.
15784 (lower_rec_input_clauses): Likewise.
15786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15787 Alan Hayward <alan.hayward@arm.com>
15788 David Sherwood <david.sherwood@arm.com>
15790 * tree-vectorizer.h (vect_nunits_for_cost): New function.
15791 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
15792 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
15793 (vect_analyze_slp_cost): Likewise.
15794 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
15795 (vect_model_load_cost): Likewise.
15797 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15798 Alan Hayward <alan.hayward@arm.com>
15799 David Sherwood <david.sherwood@arm.com>
15801 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
15802 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
15803 from an unsigned int * to a poly_uint64_pod *.
15804 (calculate_unrolling_factor): New function.
15805 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
15807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15808 Alan Hayward <alan.hayward@arm.com>
15809 David Sherwood <david.sherwood@arm.com>
15811 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
15812 from an unsigned int to a poly_uint64.
15813 (_loop_vec_info::slp_unrolling_factor): Likewise.
15814 (_loop_vec_info::vectorization_factor): Change from an int
15816 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
15817 (vect_get_num_vectors): New function.
15818 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
15819 (vect_get_num_copies): Use vect_get_num_vectors.
15820 (vect_analyze_data_ref_dependences): Change max_vf from an int *
15821 to an unsigned int *.
15822 (vect_analyze_data_refs): Change min_vf from an int * to a
15824 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15825 than an unsigned HOST_WIDE_INT.
15826 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
15827 (vect_analyze_data_ref_dependence): Change max_vf from an int *
15828 to an unsigned int *.
15829 (vect_analyze_data_ref_dependences): Likewise.
15830 (vect_compute_data_ref_alignment): Handle polynomial vf.
15831 (vect_enhance_data_refs_alignment): Likewise.
15832 (vect_prune_runtime_alias_test_list): Likewise.
15833 (vect_shift_permute_load_chain): Likewise.
15834 (vect_supportable_dr_alignment): Likewise.
15835 (dependence_distance_ge_vf): Take the vectorization factor as a
15836 poly_uint64 rather than an unsigned HOST_WIDE_INT.
15837 (vect_analyze_data_refs): Change min_vf from an int * to a
15839 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
15840 vfm1 as a poly_uint64 rather than an int. Make the same change
15841 for the returned bound_scalar.
15842 (vect_gen_vector_loop_niters): Handle polynomial vf.
15843 (vect_do_peeling): Likewise. Update call to
15844 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
15845 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
15847 * tree-vect-loop.c (vect_determine_vectorization_factor)
15848 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
15849 (vect_get_known_peeling_cost): Likewise.
15850 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
15851 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
15852 (vect_transform_loop): Likewise. Use the lowest possible VF when
15853 updating the upper bounds of the loop.
15854 (vect_min_worthwhile_factor): Make static. Return an unsigned int
15855 rather than an int.
15856 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
15857 polynomial unroll factors.
15858 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
15859 (vect_make_slp_decision): Likewise.
15860 (vect_supported_load_permutation_p): Likewise, and polynomial
15862 (vect_analyze_slp_cost): Handle polynomial vf.
15863 (vect_slp_analyze_node_operations): Likewise.
15864 (vect_slp_analyze_bb_1): Likewise.
15865 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15866 than an unsigned HOST_WIDE_INT.
15867 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
15868 (vectorizable_load): Handle polynomial vf.
15869 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
15871 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
15873 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15874 Alan Hayward <alan.hayward@arm.com>
15875 David Sherwood <david.sherwood@arm.com>
15877 * match.pd: Handle bit operations involving three constants
15878 and try to fold one pair.
15880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15882 * tree-vect-loop-manip.c: Include gimple-fold.h.
15883 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
15884 niters_maybe_zero parameters. Handle other cases besides a step of 1.
15885 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
15886 Add a path that uses a step of VF instead of 1, but disable it
15888 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
15889 and niters_no_overflow parameters. Update calls to
15890 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
15891 Create a new SSA name if the latter choses to use a ste other
15892 than zero, and return it via niters_vector_mult_vf_var.
15893 * tree-vect-loop.c (vect_transform_loop): Update calls to
15894 vect_do_peeling, vect_gen_vector_loop_niters and
15895 slpeel_make_loop_iterate_ntimes.
15896 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
15897 (vect_gen_vector_loop_niters): Update declarations after above changes.
15899 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
15901 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
15902 128-bit round to integer instructions.
15903 (ceil<mode>2): Likewise.
15904 (btrunc<mode>2): Likewise.
15905 (round<mode>2): Likewise.
15907 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15909 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
15910 unaligned VSX load/store on P8/P9.
15911 (expand_block_clear): Allow the use of unaligned VSX
15912 load/store on P8/P9.
15914 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15916 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
15918 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
15919 swap associated with both a load and a store.
15921 2018-01-02 Andrew Waterman <andrew@sifive.com>
15923 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
15924 * config/riscv/riscv.md (clear_cache): Use it.
15926 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
15928 * web.c: Remove out-of-date comment.
15930 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15932 * expr.c (fixup_args_size_notes): Check that any existing
15933 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
15934 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
15935 (emit_single_push_insn): ...here.
15937 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15939 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
15940 (const_vector_encoded_nelts): New function.
15941 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
15942 (const_vector_int_elt, const_vector_elt): Declare.
15943 * emit-rtl.c (const_vector_int_elt_1): New function.
15944 (const_vector_elt): Likewise.
15945 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
15946 of CONST_VECTOR_ELT.
15948 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15950 * expr.c: Include rtx-vector-builder.h.
15951 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
15952 directly on the tree encoding.
15953 (const_vector_from_tree): Likewise.
15954 * optabs.c: Include rtx-vector-builder.h.
15955 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
15956 sequence of "u" values.
15957 * vec-perm-indices.c: Include rtx-vector-builder.h.
15958 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
15959 directly on the vec_perm_indices encoding.
15961 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15963 * doc/rtl.texi (const_vector): Describe new encoding scheme.
15964 * Makefile.in (OBJS): Add rtx-vector-builder.o.
15965 * rtx-vector-builder.h: New file.
15966 * rtx-vector-builder.c: Likewise.
15967 * rtl.h (rtx_def::u2): Add a const_vector field.
15968 (CONST_VECTOR_NPATTERNS): New macro.
15969 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
15970 (CONST_VECTOR_DUPLICATE_P): Likewise.
15971 (CONST_VECTOR_STEPPED_P): Likewise.
15972 (CONST_VECTOR_ENCODED_ELT): Likewise.
15973 (const_vec_duplicate_p): Check for a duplicated vector encoding.
15974 (unwrap_const_vec_duplicate): Likewise.
15975 (const_vec_series_p): Check for a non-duplicated vector encoding.
15976 Say that the function only returns true for integer vectors.
15977 * emit-rtl.c: Include rtx-vector-builder.h.
15978 (gen_const_vec_duplicate_1): Delete.
15979 (gen_const_vector): Call gen_const_vec_duplicate instead of
15980 gen_const_vec_duplicate_1.
15981 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
15982 (gen_const_vec_duplicate): Use rtx_vector_builder.
15983 (gen_const_vec_series): Likewise.
15984 (gen_rtx_CONST_VECTOR): Likewise.
15985 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
15986 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15987 Build a new vector rather than modifying a CONST_VECTOR in-place.
15988 (handle_special_swappables): Update call accordingly.
15989 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
15990 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15991 Build a new vector rather than modifying a CONST_VECTOR in-place.
15992 (handle_special_swappables): Update call accordingly.
15994 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15996 * simplify-rtx.c (simplify_const_binary_operation): Use
15997 CONST_VECTOR_ELT instead of XVECEXP.
15999 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16001 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16002 the selector elements to be different from the data elements
16003 if the selector is a VECTOR_CST.
16004 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16005 ssizetype for the selector.
16007 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16009 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16010 before testing each element individually.
16011 * tree-vect-generic.c (lower_vec_perm): Likewise.
16013 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16015 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16016 * selftest-run-tests.c (selftest::run_tests): Call it.
16017 * vector-builder.h (vector_builder::operator ==): New function.
16018 (vector_builder::operator !=): Likewise.
16019 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16020 (vec_perm_indices::all_from_input_p): New function.
16021 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16022 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16023 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16024 instead of reading the VECTOR_CST directly. Detect whether both
16025 vector inputs are the same before constructing the vec_perm_indices,
16026 and update the number of inputs argument accordingly. Use the
16027 utility functions added above. Only construct sel2 if we need to.
16029 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16031 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16032 the broadcast of the low byte.
16033 (expand_mult_highpart): Use an explicit encoding for the permutes.
16034 * optabs-query.c (can_mult_highpart_p): Likewise.
16035 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16036 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16037 (vectorizable_bswap): Likewise.
16038 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16039 explicit encoding for the power-of-2 permutes.
16040 (vect_permute_store_chain): Likewise.
16041 (vect_grouped_load_supported): Likewise.
16042 (vect_permute_load_chain): Likewise.
16044 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16046 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16047 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16048 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16049 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16050 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16051 (vect_gen_perm_mask_any): Likewise.
16053 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16055 * int-vector-builder.h: New file.
16056 * vec-perm-indices.h: Include int-vector-builder.h.
16057 (vec_perm_indices): Redefine as an int_vector_builder.
16058 (auto_vec_perm_indices): Delete.
16059 (vec_perm_builder): Redefine as a stand-alone class.
16060 (vec_perm_indices::vec_perm_indices): New function.
16061 (vec_perm_indices::clamp): Likewise.
16062 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16063 (vec_perm_indices::new_vector): New function.
16064 (vec_perm_indices::new_expanded_vector): Update for new
16065 vec_perm_indices class.
16066 (vec_perm_indices::rotate_inputs): New function.
16067 (vec_perm_indices::all_in_range_p): Operate directly on the
16068 encoded form, without computing elided elements.
16069 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16070 encoding. Update for new vec_perm_indices class.
16071 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16072 the given vec_perm_builder.
16073 (expand_vec_perm_var): Update vec_perm_builder constructor.
16074 (expand_mult_highpart): Use vec_perm_builder instead of
16075 auto_vec_perm_indices.
16076 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16077 vec_perm_indices instead of auto_vec_perm_indices. Use a single
16078 or double series encoding as appropriate.
16079 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16080 vec_perm_indices instead of auto_vec_perm_indices.
16081 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16082 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16083 (vect_permute_store_chain): Likewise.
16084 (vect_grouped_load_supported): Likewise.
16085 (vect_permute_load_chain): Likewise.
16086 (vect_shift_permute_load_chain): Likewise.
16087 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16088 (vect_transform_slp_perm_load): Likewise.
16089 (vect_schedule_slp_instance): Likewise.
16090 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16091 (vectorizable_mask_load_store): Likewise.
16092 (vectorizable_bswap): Likewise.
16093 (vectorizable_store): Likewise.
16094 (vectorizable_load): Likewise.
16095 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16096 vec_perm_indices instead of auto_vec_perm_indices. Use
16097 tree_to_vec_perm_builder to read the vector from a tree.
16098 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16099 vec_perm_builder instead of a vec_perm_indices.
16100 (have_whole_vector_shift): Use vec_perm_builder and
16101 vec_perm_indices instead of auto_vec_perm_indices. Leave the
16102 truncation to calc_vec_perm_mask_for_shift.
16103 (vect_create_epilog_for_reduction): Likewise.
16104 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16105 from auto_vec_perm_indices to vec_perm_indices.
16106 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16107 instead of changing individual elements.
16108 (aarch64_vectorize_vec_perm_const): Use new_vector to install
16109 the vector in d.perm.
16110 * config/arm/arm.c (expand_vec_perm_d::perm): Change
16111 from auto_vec_perm_indices to vec_perm_indices.
16112 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16113 instead of changing individual elements.
16114 (arm_vectorize_vec_perm_const): Use new_vector to install
16115 the vector in d.perm.
16116 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16117 Update vec_perm_builder constructor.
16118 (rs6000_expand_interleave): Likewise.
16119 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16120 (rs6000_expand_interleave): Likewise.
16122 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16124 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16125 to qimode could truncate the indices.
16126 * optabs.c (expand_vec_perm_var): Likewise.
16128 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16130 * Makefile.in (OBJS): Add vec-perm-indices.o.
16131 * vec-perm-indices.h: New file.
16132 * vec-perm-indices.c: Likewise.
16133 * target.h (vec_perm_indices): Replace with a forward class
16135 (auto_vec_perm_indices): Move to vec-perm-indices.h.
16136 * optabs.h: Include vec-perm-indices.h.
16137 (expand_vec_perm): Delete.
16138 (selector_fits_mode_p, expand_vec_perm_var): Declare.
16139 (expand_vec_perm_const): Declare.
16140 * target.def (vec_perm_const_ok): Replace with...
16141 (vec_perm_const): ...this new hook.
16142 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16143 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16144 * doc/tm.texi: Regenerate.
16145 * optabs.def (vec_perm_const): Delete.
16146 * doc/md.texi (vec_perm_const): Likewise.
16147 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16148 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16149 expand_vec_perm for constant permutation vectors. Assert that
16150 the mode of variable permutation vectors is the integer equivalent
16151 of the mode that is being permuted.
16152 * optabs-query.h (selector_fits_mode_p): Declare.
16153 * optabs-query.c: Include vec-perm-indices.h.
16154 (selector_fits_mode_p): New function.
16155 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16156 is defined, instead of checking whether the vec_perm_const_optab
16157 exists. Use targetm.vectorize.vec_perm_const instead of
16158 targetm.vectorize.vec_perm_const_ok. Check whether the indices
16159 fit in the vector mode before using a variable permute.
16160 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
16161 vec_perm_indices instead of an rtx.
16162 (expand_vec_perm): Replace with...
16163 (expand_vec_perm_const): ...this new function. Take the selector
16164 as a vec_perm_indices rather than an rtx. Also take the mode of
16165 the selector. Update call to shift_amt_for_vec_perm_mask.
16166 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
16167 Use vec_perm_indices::new_expanded_vector to expand the original
16168 selector into bytes. Check whether the indices fit in the vector
16169 mode before using a variable permute.
16170 (expand_vec_perm_var): Make global.
16171 (expand_mult_highpart): Use expand_vec_perm_const.
16172 * fold-const.c: Includes vec-perm-indices.h.
16173 * tree-ssa-forwprop.c: Likewise.
16174 * tree-vect-data-refs.c: Likewise.
16175 * tree-vect-generic.c: Likewise.
16176 * tree-vect-loop.c: Likewise.
16177 * tree-vect-slp.c: Likewise.
16178 * tree-vect-stmts.c: Likewise.
16179 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
16181 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
16182 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
16183 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
16184 (aarch64_vectorize_vec_perm_const): ...this new function.
16185 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16186 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16187 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
16188 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
16189 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16190 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16191 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
16193 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
16194 check for NEON modes.
16195 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
16196 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
16197 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
16198 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
16200 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
16201 the old VEC_PERM_CONST conditions.
16202 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
16203 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
16204 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
16205 (ia64_vectorize_vec_perm_const_ok): Merge into...
16206 (ia64_vectorize_vec_perm_const): ...this new function.
16207 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
16208 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
16209 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
16210 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
16211 * config/mips/mips.c (mips_expand_vec_perm_const)
16212 (mips_vectorize_vec_perm_const_ok): Merge into...
16213 (mips_vectorize_vec_perm_const): ...this new function.
16214 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
16215 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
16216 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
16217 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
16218 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
16219 (rs6000_expand_vec_perm_const): Delete.
16220 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
16222 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16223 (altivec_expand_vec_perm_const_le): Take each operand individually.
16224 Operate on constant selectors rather than rtxes.
16225 (altivec_expand_vec_perm_const): Likewise. Update call to
16226 altivec_expand_vec_perm_const_le.
16227 (rs6000_expand_vec_perm_const): Delete.
16228 (rs6000_vectorize_vec_perm_const_ok): Delete.
16229 (rs6000_vectorize_vec_perm_const): New function.
16230 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16231 an element count and rtx array.
16232 (rs6000_expand_extract_even): Update call accordingly.
16233 (rs6000_expand_interleave): Likewise.
16234 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
16235 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
16236 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
16237 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
16238 (rs6000_expand_vec_perm_const): Delete.
16239 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16240 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16241 (altivec_expand_vec_perm_const_le): Take each operand individually.
16242 Operate on constant selectors rather than rtxes.
16243 (altivec_expand_vec_perm_const): Likewise. Update call to
16244 altivec_expand_vec_perm_const_le.
16245 (rs6000_expand_vec_perm_const): Delete.
16246 (rs6000_vectorize_vec_perm_const_ok): Delete.
16247 (rs6000_vectorize_vec_perm_const): New function. Remove stray
16248 reference to the SPE evmerge intructions.
16249 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16250 an element count and rtx array.
16251 (rs6000_expand_extract_even): Update call accordingly.
16252 (rs6000_expand_interleave): Likewise.
16253 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
16254 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
16256 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16258 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16260 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
16261 vector mode and that that mode matches the mode of the data
16263 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
16264 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
16265 directly using expand_vec_perm_1 when forcing selectors into
16267 (expand_vec_perm_var): New function, split out from expand_vec_perm.
16269 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16271 * optabs-query.h (can_vec_perm_p): Delete.
16272 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
16273 * optabs-query.c (can_vec_perm_p): Split into...
16274 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
16275 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
16276 particular selector is valid.
16277 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16278 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16279 (vect_grouped_load_supported): Likewise.
16280 (vect_shift_permute_load_chain): Likewise.
16281 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16282 (vect_transform_slp_perm_load): Likewise.
16283 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16284 (vectorizable_bswap): Likewise.
16285 (vect_gen_perm_mask_checked): Likewise.
16286 * fold-const.c (fold_ternary_loc): Likewise. Don't take
16287 implementations of variable permutation vectors into account
16288 when deciding which selector to use.
16289 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
16290 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
16291 with a false third argument.
16292 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
16293 to test whether the constant selector is valid and can_vec_perm_var_p
16294 to test whether a variable selector is valid.
16296 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16298 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
16299 * optabs-query.c (can_vec_perm_p): Likewise.
16300 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
16301 instead of vec_perm_indices.
16302 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
16303 (vect_gen_perm_mask_checked): Likewise,
16304 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
16305 (vect_gen_perm_mask_checked): Likewise,
16307 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16309 * optabs-query.h (qimode_for_vec_perm): Declare.
16310 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
16311 (qimode_for_vec_perm): ...this new function.
16312 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
16314 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16316 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
16317 does not have a conditional at the top.
16319 2018-01-02 Richard Biener <rguenther@suse.de>
16321 * ipa-inline.c (big_speedup_p): Fix expression.
16323 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16326 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
16329 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16333 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
16334 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
16335 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
16336 cond_taken_branch_cost 3->4.
16338 2018-01-01 Jakub Jelinek <jakub@redhat.com>
16340 PR tree-optimization/83581
16341 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
16342 TODO_cleanup_cfg if any changes have been made.
16344 PR middle-end/83608
16345 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
16346 convert_modes if target mode has the right side, but different mode
16349 PR middle-end/83609
16350 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
16351 last argument when extracting from CONCAT. If either from_real or
16352 from_imag is NULL, use expansion through memory. If result is not
16353 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
16354 the parts directly to inner mode, if even that fails, use expansion
16357 PR middle-end/83623
16358 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
16359 check for bswap in mode rather than HImode and use that in expand_unop
16362 Copyright (C) 2018 Free Software Foundation, Inc.
16364 Copying and distribution of this file, with or without modification,
16365 are permitted in any medium without royalty provided the copyright
16366 notice and this notice are preserved.